Hyperthermia restores apoptosis induced by death receptors through aggregation-induced c-FLIP cytosolic depletion

TRAIL is involved in immune tumor surveillance and is considered a promising anti-cancer agent owing to its limited side effects on healthy cells. However, some cancer cells display resistance, or become resistant to TRAIL-induced cell death. Hyperthermia can enhance sensitivity to TRAIL-induced cell death in various resistant cancer cell lines, including lung, breast, colon or prostate carcinomas. Mild heat shock treatment has been proposed to restore Fas ligand or TRAIL-induced apoptosis through c-FLIP degradation or the mitochondrial pathway. We demonstrate here that neither the mitochondria nor c-FLIP degradation are required for TRAIL-induced cell death restoration during hyperthermia. Our data provide evidence that insolubilization of c-FLIP, alone, is sufficient to enhance apoptosis induced by death receptors. Hyperthermia induced c-FLIP depletion from the cytosolic fraction, without apparent degradation, thereby preventing c-FLIP recruitment to the TRAIL DISC and allowing efficient caspase-8 cleavage and apoptosis. Hyperthermia-induced c-FLIP depletion was independent of c-FLIP DED2 FL chain assembly motif or ubiquitination-mediated c-FLIP degradation, as assessed using c-FLIP point mutants on lysine 167 and 195 or threonine 166, a phosphorylation site known to regulate ubiquitination of c-FLIP. Rather, c-FLIP depletion was associated with aggregation, because addition of glycerol not only prevented the loss of c-FLIP from the cytosol but also enabled c-FLIP recruitment within the TRAIL DISC, thus inhibiting TRAIL-induced apoptosis during hyperthermia. Altogether our results demonstrate that c-FLIP is a thermosensitive protein whose targeting by hyperthermia allows restoration of apoptosis induced by TNF ligands, including TRAIL. Our findings suggest that combining TRAIL agonists with whole-body or localized hyperthermia may be an interesting approach in cancer therapy.

demonstrating that a large number of stimuli lead to c-FLIP degradation and restoration of apoptosis induced by death receptors, 7 hyperthermia has recently been proposed to restore TRAIL pro-apoptotic signaling through ubiquitination of c-FLIP on K195. 17 Herein, we provide evidence that proteosomal-mediated degradation of c-FLIP, albeit induced during hyperthermia, is not required for sensitization or restoration of TRAIL-induced cell death. Instead, our findings demonstrate that both c-FLIP isoforms are thermolabile proteins that aggregate during hyperthermia. As a consequence, c-FLIP proteins are not available in the cytosol and their recruitment within the TRAIL DISC is impaired, which allows efficient initiator caspase activation.

Results
Hyperthermia restores TRAIL-induced apoptosis in a mitochondrial-independent manner. Hyperthermia restores TRAIL-induced apoptosis in tumor cells 17,23,24 but not in normal cells. 25 In line with these findings, incubating resistant cancer cell lines of various origin for 1 h at 42°C (HS) in the presence of TRAIL followed by subsequent incubation at 37°C for 5 h (Figure 1a), significantly increased apoptosis triggered by TRAIL as compared with a 6 h incubation time at 37°C (Figure 1b). Incubation of the cells during the first hour at milder temperature, that is 39°C, or at 0°C failed to do so ( Supplementary Figures 1a and b). As expected, incubating cells for 1 h at 42°C was sufficient to induce a time-dependent upregulation of inducible HSPs including HSP27, αB-crystallin, HSP70 or HSP110 (Supplementary Figure 1c), phosphorylation of HSP27 on serine 15, 78 and 82, translocation of HSP27 into the non-ionic detergent insoluble fraction and to increase protein ubiquitination (Supplementary Figure 1d). However, restoration of TRAIL pro-apoptotic activity by HS (Figure 1b) was not correlated with steady state differential expression levels of HSPs (Supplementary Figure 2b) or TRAIL receptors but to some extent with c-FLIP expression levels ( Supplementary Figures 2a and b). Likewise, regardless of their initial sensitivity to TRAIL (Supplementary Figure 2c), cells expressing high amounts of c-FLIP were more responsive to TRAIL during HS (Figure 1b) than cells expressing low levels (Supplementary Figure 2).
To understand the molecular mechanisms underlying the gain of function during hyperthermia, we decided to use the resistant mammary carcinoma cell line MDA-MB-231 as a model cell line. Cell death induced by TRAIL after a HS, was mainly driven by caspases as the pan-caspase inhibitor z-VAD totally abrogated apoptosis induced by TRAIL ( Figure 1c) and rescued MDA-MB-231 clonogenic growth after TRAIL stimulation ( Figure 1d). As evidenced by immunoblotting, hyperthermia enhanced initiator caspase-8, -9 and -10 processing, as compared with control conditions (37°C), both in the cytosolic and the membrane-enriched fractions (Figure 1e). Accordingly, higher amounts of RIP-and PARP-cleaved products were detected in HS samples (Figure 1e). Caspase-3, caspase-8 and caspase-9 activation was increased by more than twofold after hyperthemia (Figure 1f). Consistent with this increase, hyperthermia was also able to enhance Fas ligand-induced apoptosis, but failed to increase apoptosis induced by the PKC-inhibitor staurosporine (Supplementary Figure 3a). This result suggests that the mitochondrial pathway is likely dispensable to restore TRAIL-induced cell death during HS. To test this hypothesis, caspase-9 or BID were silenced. Although BID silencing attenuated TRAILinduced apoptosis in MDA-MB-231 cells at 37°C, it failed to inhibit apoptosis induced by TRAIL after a HS (Figure 1g). Loss of caspase-9 had no effect on these cells, irrespective of the temperature (Figure 1g). Consistent with a lack of requirement for mitochondrial activation, ectopic expression of Bcl-xL only slightly attenuated TRAIL-induced apoptosis during HS as compared with Mock-infected cells (Figure 1h). Moreover, restoration of TRAIL-induced apoptosis by HS was as efficient in the Bax-deficient prostate carcinoma cell line DU145 (Figure 1b), as in parental MDA-MB-231 cells pre-incubated with Bax channel blockers (Supplementary Figure 3b). By contrast and as expected, caspase-8 silencing completely abrogated apoptosis induced by TRAIL both at 37°C and after HS (Figure 1i), indicating that reactivation of the mitochondrial pathway is dispensable for TRAIL signaling during HS.
Hyperthermia impedes c-FLIP recruitment to the TRAIL DISC. As the mitochondrial pathway is not a prerequisite for HS and TRAIL synergistic effect, and because it has been proposed that HS may act at the level of plasma membrane, 26 we decided to analyze TRAIL DISC composition during and after HS. As shown Figure 2a, TRAIL DISC formation and composition were significantly different in cells incubated at 37°C as compared with cells incubated at 42°C. Strikingly, although c-FLIP was co-recruited with caspase-8 within the TRAIL DISC at 37°C, it was not recruited at 42°C, even when cells were allowed to recover from the HS at 37°C (Figure 2a). Consistently, a loss of c-FLIP/caspase-8 interaction was detected after caspase-8 immunoprecipitation in cells stimulated with TRAIL in HS condition ( Figure 2b). In line with the gain of caspase activation and apoptosis induced by TRAIL after a HS, we observed an increase in caspase-8 and caspase-10 processing within the DISC, 120 min after TRAIL stimulation, and the cleavage of RIP1 ( Figure 2).
HSPs inhibit apoptosis through their ability to interfere with mitochondria. 27 However, a report suggested that HSP90 confers resistance to TRAIL through its ability to interact with c-FLIP and to increase c-FLIP recruitment within the DISC. 28 HSP90 could be detected at 37°C consistent with c-FLIP recruitment ( Figure 2a). However, at 42°C and therefore in the absence of c-FLIP, albeit to a lesser extent, HSP90 was still detected in the DISC (Figure 2a). Moreover, HSP90 was not observed at 37°C in the caspase-8 pull-down despite the presence of c-FLIP, but was detected 5 and 15 min after TRAIL stimulation at 42°C, a condition in which little c-FLIP remained associated with caspase-8 ( Figure 2b). Contrary to HSP90, however, recruitment of HSP27 was consistently found in the TRAIL DISC at 42°C but not at 37°C (Figure 2b), and no HSP70 could be detected in the DISC, irrespective of the temperature (not shown).
To sort out whether the presence of HSP27 or HSP90 within the DISC might be relevant, these inducible HSPs were silenced. Although the silencing of HSP27 enhanced TRAIL both at 37°C and 42°C, silencing of HSP70, HSP90α or HSP90β had no significant effect, irrespective of the Figure 1 Hyperthermia increases TRAIL-induced cell death in a caspase-dependent but mitochondrial-independent manner. (a) Schematic representation of the protocol used to stimulate cells with TRAIL. Cells were either stimulated at 37°C for 6 h, or incubated in the presence (T+HS) or absence (HS) of His-TRAIL for 1 h at 42°C followed by a 5 h additional incubation at 37°C. (b) Indicated cancer cell lines were stimulated with 500 ng/ml His-TRAIL (TRAIL) or TRAIL and hyperthermia (T+HS) as described above and apoptosis was measured after 6 h after the onset of the stimulation by Hoechst staining. (c) MDA-MB-231 cells were pre-incubated or not 30 min with 5 μM caspase inhibitor z-VAD and stimulated with 50 ng/ml His-TRAIL. Apoptosis was measured after 6 h by Hoechst staining. (d) Five hundred MDA-MB-231 cells, plated overnight in a 6-well plate, were pre-incubated or not for 30 min with 20 μM z-VAD prior stimulation or not with 500 ng/ml His-TRAIL at 37°C or in hyperthermic condition (HS) for 1 h and allowed to recover for a week at 37°C before staining with methylene blue. (e) MDA-MB-231 cells were stimulated or not with 500 ng/ml His-TRAIL as indicated for 2, 4 or 6 h and cytosolic or membrane fractions were isolated (see methods). Expression levels of indicated proteins were detected by immunoblotting. One representative blot is shown (n = 3). (f) MDA-MB-231 cells were stimulated as described above with 50 ng/ml His-TRAIL and caspase activities were measured by luminometry 1-8 h after stimulation using caspase-3/7 (DEVD), caspase-8/10 (IETD) or caspase-9 (LEHD) luminogenic substrates. (g) MDA-MB-231 cells were transfected with non-targeting (si-Nt), caspase-9 (si-C9) or BID (si-BID) targeting siRNAs. Seventy-two hours after transfection, cells were stimulated or not with 100 ng/ml His-TRAIL and apoptosis was analyzed after 6 h by Annexin V staining and flow cytometry. Caspase-9 and BID expression levels are shown on the right. Silencing simultaneously both isoforms of HSP90, however, attenuated apoptosis induced by TRAIL during HS but not at 37°C ( Supplementary Figures 3c and d). HSP90 is a chaperone of RIP1, and is known to stabilize this kinase. 29 To exclude the possibility that the effects of HSP90α/β silencing might require RIP1, the kinase was silenced. Unlike HSP90α/β, silencing of RIP1 was unable to alter TRAIL-induced cell death during HS (Supplementary Figure 3d), suggesting that HSP90's regulatory properties are independent of RIP1. Analysis of caspase activation by immunoblotting in MDA-MB-231 cells silenced for HSP27, HSP90α/β or RIP1 demonstrated that these proteins are unable to regulate early events of TRAIL-induced cell death during HS (Supplementary Figure 3e). Neither caspase-8 nor c-FLIP cleavage was altered in cells lacking HSPs or RIP1, irrespective of the temperature. However, as expected, in cells silenced for HSP27, activation of caspase-9 and caspase-3 was significantly increased upon TRAIL stimulation, both at 37 and 42°C (Supplementary Figure 3e), indicating that HSP27 mainly inhibits TRAIL signaling at the mitochondrial level. Our results show that HSPs are not directly involved in regulating HS-induced TRAIL sensitization at the DISC level.
Hyperthermia induced c-FLIP insolubilization occurs in an ubiquitination-and phosphorylation-independent manner and is not regulated by c-FLIP DED2 chain assembly domain. Hyperthermia has been proposed to induce ubiquitination-dependent c-FLIP degradation through the proteasome, allowing restoration of TRAIL-and mapatumumab-induced cell death. 17 In MDA-MB-231 stimulated with TRAIL during HS, c-FLIP isoforms remain highly expressed for up to 2 h as evidenced in whole-cell lysates ( Figure 3a). We thus hypothesized that the impairment of c-FLIP recruitment to TRAIL DISC during HS ( Figure 2) is unlikely due to its mere degradation by the proteasome. In line with our hypothesis, although c-FLIP levels dropped after TRAIL treatment during HS in the detergent-soluble fraction (Figure 3b), c-FLIP rapidly accumulated in the insoluble fraction ( Figure 3b). Interestingly, caspase-8, caspase-10 and FADD were also found in the insoluble fraction. Yet, contrary to c-FLIP, a large proportion of caspase-10 and FADD, and to a lesser extent of caspase-8, remained in the soluble fraction 2 h after TRAIL stimulation ( Figure 3b). Its high susceptibility to insolubility, most likely rendered c-FLIP unavailable for DISC recruitment and thus contributed to TRAIL-induced apoptosis restoration during HS. Accordingly, silencing of c-FLIP alone was sufficient to restore TRAIL sensitivity to similar extent as cells exposed to a HS (Figure 3c Figure 4d). However, inhibition of the proteasome by MG132 led to the accumulation of c-FLIP in the insoluble fraction, demonstrating that c-FLIP degradation occurs after its depletion from the cytosol. Insolubilization of c-FLIP was also independent of c-FLIP DED2 chain assembly motif, as the two chain assembly motif mutants (F114G and F114G/ L115G) 31,32 were as efficiently depleted from the cytosolic fraction as WT c-FLIP ( Supplementary Figure 5a). At the contrary, depletion of FL114/115G c-FLIP mutant was even more pronounced than WT c-FLIP during HS. Interestingly, this mutant was less efficient than WT c-FLIP in inhibiting TRAIL-induced cell death upon HS (Supplementary Figure 5b), again indicating that depletion of c-FLIP from the cytosol, alone, is sufficient to restore TRAIL-induced cell death. Restoration of c-FLIP in the cytosol enables c-FLIP recruitment to TRAIL DISC and inhibits apoptosis induced by TRAIL during HS. Translocation of c-FLIP during HS was also found in other tumor cell lines of various origin ( Figure 4a). As observed in MDA-MB-231, disappearance of c-FLIP from the soluble fraction in these cells was always more efficient than depletion of FADD, RIPK1, caspase-8 or caspase-10, suggesting that enough DISC component remains in the cytosol to allow efficient TRAILinduced cell death in the absence of c-FLIP. In order to determine whether the mere increase in temperature, but not sequestration of c-FLIP into subcellular compartments induced or not through HS-mediated MAPK signalling, is sufficient to induce c-FLIP aggregation and loss from the cytosol, cell lysates obtained from unstimulated MDA-MB-231 cells were incubated at 0, 37 or 42°C for the indicated period of time, then centrifuged to separate NP40 soluble and insoluble fractions (Figure 4b) and samples were analyzed by immunoblotting. In cell extracts incubated at 42°C, c-FLIP content decreased in the soluble fraction in a timedependent manner to accumulate in the insoluble fraction ( Figure 4c). As expected, incubation of cell lysates at 0 or 37°C for 60 min failed to induce c-FLIP depletion from the cytosolic fraction. Insolubilization of c-FLIP, but also of initiator caspases, was most likely triggered through protein aggregation as addition of increasing amounts of glycerol before incubation at 42°C reduced, in a dose-dependent manner, the amount of c-FLIP and caspase-8 present in the insoluble fraction and restored significant c-FLIP protein content in the cytosolic fraction (Figure 4d). Addition of glycerol on intact MDA-MB-231 or HCT116 cells prior to incubation at 42°C also inhibited HS-induced c-FLIP aggregation as evidenced by the increase of c-FLIP L in the cytosolic fraction (Figures 4e  and f). As glycerol is able to prevent the loss of c-FLIP from the cytosol, we speculated that it might protect, at least partially, tumor cells from TRAIL-induced cell death after a HS. To address this question, MDA-MB-231 cells were incubated for 60 min in the presence of increasing amounts of glycerol before stimulation with TRAIL at 42°C. Analysis of apoptosis induced in these conditions indicates that addition of glycerol prior stimulation protected cells from TRAILinduced apoptosis during HS (Figure 4g) and restored c-FLIP recruitment to the TRAIL DISC (Figure 4h). Our results thus provide evidence that restoration of death receptor-induced cell death by hyperthermia is essentially mediated through c-FLIP aggregation.

Discussion
Hyperthermia was first demonstrated to restore cell death induced by TNFα in the late 80s. 33,34 Only recently has hyperthermia been successfully exploited in the clinic with TNFα to treat limb soft tissue sarcomas with high response rates, 35 or locally advanced cancers. 36 Besides TNFα, hyperthermia can also promote Fas ligand and TRAIL-induced apoptosis. 25,37,38 Accordingly, we show here that hyperthermia restores TRAIL pro-apoptotic signaling in a large panel of tumor cell lines. Yet, contrary to previous demonstrations pointing to a contribution of mitochondria, 37,39-41 our findings clearly demonstrate that the mitochondrial pathway is not a prerequisite because neither caspase-9-, Bid-or baxdeficiency nor Bcl-xL overexpression compromised TRAILinduced cell death during HS. Lack of apparent contribution of caspase-9, in our settings, as opposed to Bid at 37°C is most likely due to differential contribution of other mitochondrial proapoptogenic factors such as Smac/Diablo, whose release in the cytosol was shown to play an important contribution for TRAIL-induced cell death in MDA-MB-231 cells. 42 Likewise, XIAP but not caspase-9 inhibition has recently been demonstrated to play a role in regulating TRAIL-induced apoptosis in HCT116 cells, described as type II cells. 43 Irrespective of mitochondrial pro-apoptogenic factors, our results demonstrate that inhibition of c-FLIP recruitment within the TRAIL DISC is the main mechanism through which hyperthermia restores TRAIL-induced cell death. Accordingly, the loss of c-FLIP recruitment within the DISC was always associated with increased activation of initiator caspases. Overexpression of c-FLIP alone or inhibition of c-FLIP aggregation, using glycerol, was sufficient to restore c-FLIP recruitment within the DISC and to compromise TRAILinduced apoptosis during HS. Moreover, c-FLIP silencing, alone, was sufficient to phenocopy the effects of HS, and combining c-FLIP silencing and HS induced no additional gain of function.
c-FLIP is probably the most important regulator of apoptosis induced by death receptors. 6 This caspase-8 inhibitor is highly regulated in normal and tumor cells, 44,45 and highly susceptible to a plethora of compounds, rendering c-FLIP an interesting target for cancer therapy. 7,46 Hyperthermia has recently been proposed to enhance TRAIL-and mapatumumab-induced cell death through FLIP degradation. 17 Although our results are in full agreement with the finding that c-FLIP is the main regulator targeted by hyperthermia allowing restoration of TRAIL sensitivity in resistant tumor cells, our data suggest that loss of c-FLIP within the TRAIL DISC is, however, not a direct consequence of its degradation, but rather of its aggregation and thereby its disappearance from the cytosol. Loss of c-FLIP from the cytosol was evidenced almost as early as 5 min after HS, much earlier than the onset of c-FLIP cleavage or c-FLIP degradation that was almost not observed in our settings. Depletion of c-FLIP from the cytosol also coincided with the loss of c-FLIP binding to caspase-8 after TRAIL DISC formation. Contrary to c-FLIP, however, caspase-8 was still recruited at 42°C, consistent with the gain of pro-apoptotic function afforded by short incubation of the cancer cells at 42°C in the presence of TRAIL.
Moreover, despite the fact that the ubiquitin-proteasomal pathway emerges as an important regulator of c-FLIP expression in tumor cells, 47 our results demonstrate that neither phosphorylation or ubiquitination of c-FLIP nor inhibition of the proteasome inhibited c-FLIP depletion from the cytosol during hyperthermia, indicating that ubiquitinationmediated proteosomal degradation of c-FLIP is not required for its depletion from the cytosol.
Our results show that the depletion of c-FLIP from the cytosol and its recovery in an insoluble cellular fraction after a heat shock is most likely triggered by its aggregation. Hyperthermia is known to induce protein aggregation, leading eventually to cell death. 48,49 In agreement with these findings, not only the disappearance of c-FLIP but also, to a lesser extent, of initiator caspases including caspase-8 and caspase-10 from the non-ionic detergent soluble fraction after hyperthermia was detected in all the tumor cell lines studied. However, contrary to the caspase-8 or the caspase-10, insolubilization of c-FLIP led to full disappearance of the protein from the cytosol suggesting that c-FLIP may be more thermolabile than caspase-8 or caspase-10, or that complete depletion induced by HS is due to lower expression levels of c-FLIP, as compared to caspase-8 or FADD. 32 Preferential depletion of c-FLIP was not due to its DED2 chain assembly motif, two amino acids found to be essential for FADD binding and recruitment to the TRAIL DISC. 31,32 Keeping in mind that the amounts of c-FLIPs are most of the time much lower than caspase-8 32 or that c-FLIP recruitment within the TRAIL DISC is 5 to 10 times lower than caspase-8, 31,50 and that at equivalent concentration, caspase-8, like FADD, as shown by the use of recombinant GST-fused proteins (Supplementary Figure 6a), are found as efficiently as c-FLIP in the insoluble fraction of the bacterial cell extracts after a heat shock, preferential loss of c-FLIP from the cytosolic fraction is likely to be due, at least in part, to its low intracellular steady state level. Yet, it can't be excluded that in addition, its higher thermolability, as predicted from its amino acid sequence 51 (Supplementary Figure 6b). might also play a role.
Whatsoever, the addition of glycerol, a compound shown to inhibit protein aggregation, 49 before hyperthermia prevented, not only the loss of capase-8 and FADD from the cytosolic fraction but also the loss of c-FLIP, enabling its recruitment within the DISC and conferring partial protection to TRAILinduced cell death.
Altogether our results provide the first demonstration that aggregation of c-FLIP induced by hyperthermia, but not degradation, impairs c-FLIP recruitment to TRAIL DISC and thus enhances or restores TRAIL-induced cell death in resistant cells through depletion of cytosolic c-FLIP reservoir. Keeping in mind that extensive research is being pursued worldwide to use TRAIL or TRAIL derivatives in the clinic and that c-FLIP isoforms are often highly expressed in tumor cells, 7 inhibiting c-FLIP solubility with locally applied or whole-body hyperthermia could be relevant to cancer TRAIL-based therapies. 2,17 Materials and Methods Ligand production and chemicals. His-tagged TRAIL and FasL were produced and used as described previously. 52 Staurosporine, glycerol, nonidet P-40 (NP40) and sodium dodecyl sulfate (SDS) were purchased from Sigma-Aldrich (Lyon, France). Bax channel blocker was from Santa Cruz Biotechnology (Tebu-bio, Le Perray en Yvelines, France). MG-132 (Cat# 1748) was from Tocis (Bristol, United Kingdom). The pan-caspase inhibitor z-VAD-fmk and cycloheximide (Cat# ALX-380-269) were from Enzo Life Science (Villeurbanne, France).
Treatments with hyperthermia and TRAIL. Cells were treated with the indicated concentration of His-TRAIL in supplemented DMEM and heated at 42°C for 1 h or the indicated time in a preheated water bath or at 37°C in an incubator. Cells were then incubated for 5 h or the indicated time with 5% CO 2 at 37°C before analysis.
Hoechst analysis. Apoptosis was assessed by Hoechst staining (20 μg/ml) and determination of the percentage of condensed and/or fragmented nuclei from at least 300 cells per conditions on three different fields. Experiments were repeated at least three times.
Annexin V analysis. Annexin V-FITC staining kit was purschased from Miltenyi Biotec (Miltenyi Biotec, Paris, France) and used according to the manufacturer's instructions. Stained cells were analyzed with a BD LSR2 flow cytometer (BD Biosciences). 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.
Caspases activity analysis. Cells (10 3 ) were implanted in 96-well plates. Twenty-four hours after, cells were stimulated with 50 ng/ml His-TRAIL at 37°C for the indicated time. Alternatively, cells were stimulated with TRAIL at 42°C for 1 h then were incubated or not at 37°C for the remaining time (1-7 h), before caspase activity analysis. Caspases-3/-7 (DEVDase, Cat# G8090), Caspase-8 (IETDase, Cat# G8200) and Caspase-9 (LHDase, Cat# G8210) activities were measured by luminometry with a commercial kit obtained from Promega (Promega France, Charbonnière, France) according to the manufacturer's protocol. Experiments were repeated at least three times.
Lysates and fractionation. Cells were treated as indicated and washed in cold PBS. Whole-cell lysis was performed using the SDS gel loading buffer. Samples were sonicated and boiled for 5 min before loading and immunoblotting analysis. For soluble/insoluble fractionation experiments, cells were lysed for 20 min on ice in a non-ionic detergent containing buffer composed of 1% NP40, 20 mM Tris-HCl pH 7.5, 150 mM NaCl and 10% glycerol. Lysates were then centrifuged 12 min at 10 000 r.c.f. and supernatants (soluble fraction) or the pellet (insoluble fraction) were recovered in 1 × final concentration of SDS gel loading buffer and processed as above for immunoblot analysis. Alternatively, for glycerol studies, NP40 containing lysis buffer was used without or with increasing concentrations of glycerol as indicated.
Western blotting. Immunoprecipitates or cell lysates obtained from WCL or cell fractionation were resolved by SDS-PAGE and transferred to PVDF membranes (Amersham, Biosciences, Les Ullis, France). Nonspecific binding sites were blocked by incubating membranes for 1 h in PBS containing 0.01% of Tween 20 and 5% (w/v) dried skimmed milk (PBS-Tm). Immunoblots were then incubated with specific primary antibody diluted in PBS-Tm for 2 h at room temperature or overnight at 4°C, washed three times in PBS-T for 10 min. Membranes were then incubated with corresponding HRP-conjugated secondary antibody in PBS-T for 1 h and washed three times followed by a chemiluminescence detection with the Western Bright Quantum kit (Advansta, Menlo Park, CA, USA). Protein expression levels were detected with ChemiDoc MP gel imager (Bio-Rad) or using X-ray films.

Conflict of Interest
The authors declare no conflict of interest.