The 55-kDa Tumor Necrosis Factor Receptor Induces Clustering of Mitochondria through Its Membrane-proximal Region*

The cytokine tumor necrosis factor (TNF) activates diverse signaling molecules resulting in gene expression, differentiation, and/or cell death. Here we report a novel feature induced by TNF, namely translocation of mitochondria from a dispersed distribution to a perinuclear cluster. Mitochondrial translocation correlated with sensitivity to the cell death-inducing activity of TNF and was mediated by the 55-kDa TNF receptor (TNF-R55), but not by Fas, indicating that the signaling pathway requires a TNF-R55-specific but death domain-independent signal. Indeed, using L929 cells that ex-press mutant TNF-R55, we showed that the membrane-proximal region of TNF-R55 was essential for signaling to mitochondrial translocation. In the absence of translocation, the cell death response was markedly delayed, pointing to a cooperative effect on cell death. Translocation of mitochondria, although dependent on the mi-crotubules, was not imposed by the latter and was equally induced by TNF-independent immunoinhibition of the motor protein kinesin. Additionally, immunoinhibition with antibody directed against the tail domain of kinesin synergized with TNF-induced cell death. Based on this functional mimicry, we propose that a TNF-R55 membrane-proximal region-dependent signal impedes mitochondria-associated kinesin, resulting in coopera-tion with the TNF-R55 death domain-induced cytotoxic response membrane

The cytokine tumor necrosis factor (TNF) activates diverse signaling molecules resulting in gene expression, differentiation, and/or cell death. Here we report a novel feature induced by TNF, namely translocation of mitochondria from a dispersed distribution to a perinuclear cluster. Mitochondrial translocation correlated with sensitivity to the cell death-inducing activity of TNF and was mediated by the 55-kDa TNF receptor (TNF-R55), but not by Fas, indicating that the signaling pathway requires a TNF-R55-specific but death domainindependent signal. Indeed, using L929 cells that express mutant TNF-R55, we showed that the membraneproximal region of TNF-R55 was essential for signaling to mitochondrial translocation. In the absence of translocation, the cell death response was markedly delayed, pointing to a cooperative effect on cell death. Translocation of mitochondria, although dependent on the microtubules, was not imposed by the latter and was equally induced by TNF-independent immunoinhibition of the motor protein kinesin. Additionally, immunoinhibition with antibody directed against the tail domain of kinesin synergized with TNF-induced cell death. Based on this functional mimicry, we propose that a TNF-R55 membrane-proximal region-dependent signal impedes mitochondria-associated kinesin, resulting in cooperation with the TNF-R55 death domain-induced cytotoxic response and causing the observed clustering of mitochondria.
Mitochondria are the energy-providing organelles in eukaryotic cells. However, accumulating evidence shows that these organelles also have an active function in cell death. Disruption of mitochondrial transmembrane potential (⌬⌿ m ) 1 during apo-ptosis induced by various stimuli in diverse cell types represents an irreversible commitment to cell death, preceding the late characteristics of apoptosis, such as DNA condensation and degradation as well as formation of apoptotic bodies (1,2). A causative link between ⌬⌿ m and nuclear apoptosis is supported by the release of a caspase-like, apoptosis-inducing factor from the mitochondrial intermembrane space after permeability transition, a condition leading to ⌬⌿ m disruption (3). In addition to apoptosis-inducing factor, cytochrome c induces apoptosis in a cell-free system in the presence of dATP and cytosolic extracts (4). The release of cytochrome c from mitochondria is independent of permeability transition and ⌬⌿ m disruption, suggesting a possible role in apoptosis in cell types that do not exhibit disruption of ⌬⌿ m (5,6). Moreover, Bcl-2 prevents release of apoptosis-inducing factor and cytochrome c from the mitochondria (3,5,6), and Bcl-x L inhibits the accumulation of cytochrome c in the cytosol during apoptosis possibly by binding to it and thus blocking its availability (7). Therefore, the antiapoptotic role of Bcl-2 and family members could be based on counteracting mitochondrial dysfunction and subsequent release of apoptogenic factors.
Mitochondrial dysfunction also plays a crucial role in cell types that exhibit necrosis-like cell death after activation of their death program. Disruption of plasma membrane integrity by the proinflammatory cytokine tumor necrosis factor (TNF) represents an irreversible commitment to cell death in the murine fibrosarcoma cell line L929 (8) and depends on the production of reactive oxygen species (ROS) by the mitochondria (9 -11). Also, in drug-resistant leukemia cells, inhibition of the mitochondrial respiratory chain by TNF is, at least partially, responsible for cytotoxicity (12), while in human ovarian carcinoma cell lines there is evidence for an involvement of mitochondrial ROS in TNF-mediated cell death (13).
The 55-kDa TNF receptor (TNF-R55) and Fas (APO-1 or CD95), both members of the TNF receptor family, can trigger the cell death program of several cell types. Additionally, TNF-R55 mediates gene induction by activation of the nuclear factor B (NF-B) and the transcription factors AP-1 and ATF-2 (14) and triggers phosphorylation/dephosphorylation cascades by activation of p38/RK mitogen-activated protein kinase (15). The clustered intracellular domains of TNF-R55 bind, directly or indirectly, a variety of signaling molecules responsible for the multiple signals emanating from the activated receptor. Evidence for the role of TNF-R55-associated (but also of Fasassociated) proteins in the cytocidal function of both receptors emerged from the identification of the so-called "death domain" (DD) located in the C-terminal region of the intracellular part of the receptors (16,17). The DD mediates protein-protein * This work was supported in part by the Interuniversitaire Attractiepolen and the Vlaams Actiecomité voor Biotechnologie. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ interactions of the receptor with other DD-containing proteins that act as initiating centers for different signaling cascades (18). Thus, Fas-associated, recruited directly by Fas (19,20) and indirectly by TNF-R55 through prior binding of TNF receptor-associated DD (21), links the receptors to caspase-8 and the cell death program (22,23). Also, receptor-interacting protein binds to TNF receptor-associated DD and may be involved in binding and/or function of TRAF2 (18,24).
Although it seems that the DD of TNF-R55 elicits most of the known TNF-induced phenomena, there is accumulating evidence for a role of the membrane-proximal region of the intracellular part of the receptor. Induction of nitric-oxide synthase requires this region in addition to the DD (16). Activation of c-Raf-1 kinase and phospholipase A 2 occurs through the membrane-proximal half of TNF-R55 (25,26). Furthermore, the activation of membrane-associated neutral sphingomyelinase is linked to this region by the protein FAN (27), while also phosphatidylinositol-4-phosphate 5-kinase can interact with TNF-R55 through its juxtamembrane region (28).
In this report, we describe a novel activity of TNF, namely the induction of an altered spatial distribution of mitochondria in TNF-sensitive cells. This translocation of mitochondria depends on a signal from the membrane-proximal region of TNF-R55 and thus represents a novel function of this region. The driving force of this translocation was not the microtubule (MT) cytoskeleton, but rather a loss of outward directed movement of mitochondria caused by an impaired activity of the molecular motor kinesin. Functional studies as well as the kinetics of mitochondrial translocation and its correlation with the deathinducing activity of TNF, but not with gene induction, implicate this response and its apparent molecular counterpart, namely inactivation of mitochondria-associated kinesin, in the cell death response.

MATERIALS AND METHODS
Cell Lines and Cultures-L929, L929r2, L929hFas, L929hR55wt, and L929hR55⌬MPR cells (29) 2 were cultured in Dulbecco's modified Eagle's medium supplemented with 5% (v/v) heat-inactivated fetal calf serum, 5% (v/v) heat-inactivated newborn calf serum, 100 units/ml penicillin, 0.1 mg/ml streptomycin, and 2 mM L-glutamine, at 37°C in a humidified 5% CO 2 , 95% air incubator. U937 human histiocytic lymphoma cells were maintained in RPMI 1640, supplemented with 10% fetal calf serum, 0.4 mM sodium pyruvate, 50 M ␤-mercaptoethanol, 100 units/ml penicillin, 0.1 mg/ml streptomycin, and 2 mM L-glutamine. SUK4 and SUK5 hybridomas (30) were purchased from the Developmental Studies Hybridoma Bank maintained by the Department of Pharmacology and Molecular Sciences (Johns Hopkins University School of Medicine, Baltimore, MD) and the Department of Biological Sciences (University of Iowa, Iowa City, IA) under contract N01-HD-6 -2915 from the NICHD, National Institutes of Health; they were cultured in RPMI 1640 supplemented as above. All cell lines were mycoplasma-free as judged from an enzyme immunoassay (Boehringer Mannheim).
Reagents-Murine TNF (mTNF) was produced in Escherichia coli and purified to at least 99% homogeneity in our laboratory. It had a specific activity of 1.9 ϫ 10 8 IU/mg of protein (National Institute for Biological Standards and Control, Potters Bar, UK), contained 4 ng of endotoxin/mg of protein, and was used at 1000 IU/ml. Propidium iodide (PI), cycloheximide (CHX), H 2 O 2 , menadione, tert-butyl hydroperoxide, and nocodazole (all from Sigma) were used at final concentrations of 30 M, 50 g/ml, 2.5 mM, 1 mM, 2.5 mM, and 10 M, respectively. Rhodamine 123 (R123), MitoTracker CMTMRos, and paclitaxel (all from Molecular Probes, Eugene, OR) were used at final concentrations of 1 M. Staurosporine (Boehringer Mannheim) was used at 10 M.
SUK4 antibody (Ab) and SUK5 Ab were purified by means of a Protein G column (Amersham Pharmacia Biotech) from ascites fluid and culture supernatant, respectively. Ab labeling with Cy5 fluorochrome was performed using an Ab labeling kit from Amersham Pharmacia Biotech according to the manufacturer's instructions.
Mitochondrial Translocation Assay-Cells were seeded in cham-bered coverslips (2 cm 2 , 10 5 cells/well) and preincubated overnight at 37°C in a humidified 5% CO 2 , 95% air incubator. Mitochondria were stained with R123 or MitoTracker CMTMRos for 30 min at 37°C before analysis. L929 cells and L929-derived cells were washed three times with Hepes-buffered MEM, supplemented with the same additives as Dulbecco's modified Eagle's medium (MEM-Hepes), to remove excess R123 or MitoTracker CMTMRos and kept in MEM-Hepes. MEM-Hepes was preferred over Dulbecco's modified Eagle's medium because of the CO 2 independence and the fluorescent background problems generated by Dulbecco's modified Eagle's medium. U937 cells were maintained in RPMI 1640 medium during the measurements. The distribution of mitochondria was analyzed with a Zeiss LSM 410 confocal microscope. R123 and MitoTracker CMTMRos were excited and detected as shown in Table I. The time point at which TNF was added is referred to as 0 h. At regular times after TNF administration, fluorescence confocal laser scanning microscopy (CLSM) images from four randomly chosen microscopic fields, each containing approximately 50 cells, were recorded. Translocation was represented as the percentage of viable cells exhibiting the clustered mitochondrial distribution. Cytotoxicity Assay-Apoptosis induced by TNF and anti-human Fas mAb (500 ng/ml; ImmunoTech, Alston, MA) in U937 cells was quantified in a flow cytometric assay based on DNA degradation using PI. After intercalation in DNA, PI generates a fluorescent emission light, the intensity of which correlates with the amount of DNA. In viable, permeabilized cells, this results in fluorescence intensity peaks corresponding to the G 1 (2n), S and G 2 /M (4n) phases of the cell cycle. Cells dying by apoptosis contain degraded DNA, resulting in hypoploidy and hence reduced PI fluorescence. U937 cells were permeabilized by freezing/thawing in the presence of PI. The number of hypoploid cells was determined with an Epics 753 flow cytometer (Coulter, Hialeah, FL), using the same excitation and emission wavelengths as in CLSM (Table  I), as a measure for TNF-induced apoptosis.
Cell death of L929 and L929-derived cell lines was measured by quantification of PI-positive cells by CLSM and was induced by incubation with either TNF in L929, TNF and CHX in L929r2, htr-1 mAb (100 ng/ml; a gift of Dr. M. Brockhaus, F. Hoffmann-La Roche, Basel, Switzerland), or anti-mTNF-R55 (500 ng/ml; Genzyme Corporation, Boston, MA) in L929hR55wt and L929hR55⌬MPR, and anti-human Fas mAb (500 ng/ml) in L929hFas cells. Cell death is presented as the percentage of PI-positive cells in the whole cell population.
Measurement of ROS Formation by Flow Cytometry-Dihydrorhodamine 123 was added to suspension cultures at the same time as TNF. Cell samples were taken at regular time intervals and analyzed on an Epics 753 flow cytometer. R123 fluorescence resulting from dihydrorhodamine 123 oxidation was excited and detected (Table I). 3000 viable cells were measured per sample. Cell debris and multicell aggregates were gated out electronically. The variation observed in ROS measurements was Ͻ10%.
Interleukin-6 (IL-6) Bioassay-The presence of IL-6 bioactivity in the culture supernatant of 3 ϫ 10 5 cells/ml was determined after 5 h of TNF stimulation, using the proliferative response of the IL-6-responsive murine plasmacytoma cell line 7TD1 (31).
Immunofluorescence CLSM-L929 cells were grown on glass coverslips. Where necessary, MitoTracker CMTMRos was added to the culture medium for 30 min to stain mitochondria before fixation. Coverslips, fixed at Ϫ20°C with methanol, were stained for MTs by incubation with 10-fold diluted rat anti-tubulin hybridoma YOL1/34 supernatant (Harlan Sera-Lab, Crawley Down, UK) at room temperature for 1 h, followed by extensive washing with phosphate-buffered saline containing 1% bovine serum albumin and a 30-min incubation with FITC-conjugated goat anti-rat IgG Ab (1:100; Harlan Sera-Lab). Staining of the kinesin heavy chain (KHC) was performed by incubation of coverslips with 10 g/ml Cy5-labeled SUK4 mAb. Coverslips were mounted in VectaShield (Vector Laboratories, Burlingame, CA). Cells were observed with a Zeiss LSM410 confocal microscope, typically with a scanning time of 1 s and 8ϫ line averaging. The excitation wave-2 D. Vercammen and E. Boone, unpublished data. lengths and emission filters used for the different fluorochromes are shown in Table I. Syringe Loading-Syringe loading of L929 cells with mAb was performed as described previously (32). Briefly, L929 cells were harvested from adherent, subconfluent cultures by treatment with enzyme-free cell dissociation solution (Life Technologies, Paisley, UK). 5 ϫ 10 5 cells were transferred to a 1.5-ml reaction tube in a volume of 0.1 ml of medium. mAb was added to a final concentration of 0.48 mg/ml, with an unlabeled/fluorescent labeled ratio of 15:1, followed by the addition of Pluronic F-68 (Sigma; 2% (v/v)). This solution was transferred 20 times through a 30-G injection needle. Excess Ab and pluronic F-68 were washed away, and the cells were seeded in chambered coverslips for CLSM analysis. The cells were allowed to recover for 12 h before the experiment. There was typically 10% cell death resulting from loading; dead cells were removed by washing after the recovery period. Loaded cells were identified by the presence of fluorescence in the cytosol and represented typically 50 -60% of the total cell population.
Flow Cytometry of TNF Receptor Expression Levels-L929hR55wt and L929hR55⌬MPR cells were harvested from adherent, subconfluent cultures by treatment with enzyme-free cell dissociation solution. NF-B Activity-Activation of NF-B was measured in an electrophoretic mobility shift assay. Subconfluent monolayers of L929r2 cells were treated with TNF for 1 h. Nuclear protein and binding reactions were performed as described previously (33).

RESULTS
TNF Causes Mitochondria to Translocate before Cell Death-TNF treatment of the murine fibrosarcoma cell line L929 results in cell death by an atypical, necrosis-like process, which is characterized by cell swelling, disruption of plasma membrane integrity, and subsequently cellular collapse (8). Mitochondrial dysfunction is crucial in this TNF-induced cytotoxic process; ROS are formed in the mitochondria, and interference with the generation or scavenging of these ROS arrests cell death (9 -11). Using CLSM, we observed that within 1 h of treatment of L929 cells with TNF the spatial distribution of mitochondria evolved in the majority of the cells from an originally scattered, bipolar or nearly symmetric distribution to an asymmetric, clustered distribution (Fig. 1). Analysis of the kinetics of this mitochondrial translocation revealed that the response preceded mitochondrial ROS production and subsequent cell death by several hours (Fig. 2A). For these experiments, the mitochondria-specific dye R123 was preferred because preliminary experiments had shown that this dye did not interfere with cell viability or responsiveness to TNF. However, since the specificity of R123 derives from its ⌬⌿ m -dependent accumulation in the mitochondria, a disruption of ⌬⌿ m in a fraction of the mitochondria could be perceived as a translocation. Yet, revealing the spatial distribution of mitochondria with MitoTracker, a mitochondria-specific dye that is retained in the mitochondria by the thiol reactivity of its chloromethyl moiety (34), yielded an altered distribution identical to the one observed after treatment with TNF (data not shown). Consequently, the modified distribution of R123 and MitoTracker fluorescence in TNF-treated L929 cells reflects the modified spatial distribution of mitochondria and not a loss of ⌬⌿ m in a fraction of the organelles.
In contrast to L929, cell death induced by TNF in the human histiocytic lymphoma cell line U937 is accompanied by plasma membrane blebbing, DNA degradation, and subsequent formation of apoptotic bodies (data not shown), features characteristic of apoptosis. To verify whether also in these cells mitochondria converge in response to treatment with TNF, TNF-treated U937 cells were stained with R123 and analyzed by CLSM. As shown in Fig. 2B, cells exhibiting a clustered spatial distribution of mitochondria became apparent within 1 h of TNF treatment, preceding the onset of the apoptotic response by several hours. Clearly, translocation of mitochondria occurs independently of the mechanism of cell death, viz. necrosis-like or apoptotic, and is an early feature relative to the implementation of the cell death program.
Cell death is accompanied by morphological changes that occur independently of the type of signal triggering the process. Since the observed translocation of mitochondria could be such a phenomenon, we induced cell death in L929 cells by administration of the cytotoxic agent menadione, tert-butyl hydroperoxide, H 2 O 2 , or staurosporine, after which the distribution of mitochondria was analyzed (Fig. 2C). Despite clear cytotoxic- ity, no changes in the spatial distribution of mitochondria were observed. This negative result clearly demonstrates that induction of mitochondrial translocation is an intrinsic feature of the TNF signal transduction pathway.
Translocation of Mitochondria Correlates with TNF Signaling to Cell Death-To establish the relationship between the observed clustering of mitochondria and the cytocidal versus noncytocidal activities of TNF, we studied the occurrence of mitochondrial translocation, cell death, activation of NF-B, and gene induction in the L929 variant L929r2 (29). L929r2 cells are resistant to the cytocidal activity of TNF but retain responsiveness to its noncytocidal activities exemplified by activation of NF-B and subsequent expression of IL-6. However, the addition of the RNA or protein synthesis inhibitor actino-mycin D or CHX, respectively, fully restores the cytocidal response to TNF. As shown in Fig. 3A, mitochondria converged after TNF treatment in the presence of CHX at a rate similar to that previously observed in the parental L929 cell line. However, under TNF-resistant conditions, i.e. in the absence of CHX, mitochondrial translocation did not occur (Fig. 3A) despite the typical activation by TNF of NF-B and IL-6 gene expression (Fig. 3B). This correlation between translocation and cell death, but not gene expression, indicates that the TNF-induced clustering of mitochondria is linked to the cell death-inducing activity of the cytokine.
The Membrane-proximal Region of TNF-R55 Is Required for Translocation of Mitochondria-In L929 cells, mitochondrial translocation was triggered to the same extent by mTNF and hTNF (data not shown). However, the latter binds to TNF-R55 but not to TNF-R75 (35), indicating that translocation is a TNF-R55-mediated function. Furthermore, since the TNF-R55 DD is necessary and sufficient to induce cell death (17,36), the link observed between mitochondrial translocation and cell death points to a DD-mediated signal. Accordingly, we verified the involvement of the DD by analyzing the distribution of mitochondria during Fas-induced cell death because, similar to TNF-R55, Fas induces cell death by interaction of its intracellular DD with other DD-containing mediators (37). As shown in Fig. 4A, clustering of Fas by anti-Fas mAb readily induced apoptosis in U937 cells. However, the mitochondria did not converge into a cluster. Likewise, in L929hFas cells that express a transfected human Fas gene, 3 clustering of Fas did not 3 D. Vercammen, unpublished data.

FIG. 2. TNF-induced translocation of mitochondria precedes cell death.
A, the increment of cells exhibiting translocated mitochondria (line), of dead cells (bar) and of ROS production (inset) was measured following the addition of TNF. Translocation was analyzed by CLSM in approximately 150 -200 cells using the mitochondria-specific probe R123. Cell death was quantified simultaneously by PI uptake. The production of mitochondrial ROS was measured in a parallel flowcytometric assay as described previously (11)  modify the spatial distribution of mitochondria despite the occurrence of cell death. In contrast, the same L929hFas cells triggered by TNF showed extensive translocation of mitochondria before cell death. Obviously, the induction of mitochondrial translocation is a feature specific for TNF-R55 and not shared by Fas.
Since the membrane-proximal region of TNF-R55 has almost no counterpart in Fas, we studied its involvement in the transmittal of the translocation signal with two L929 transfectants, viz. L929hR55wt and L929hR55⌬MPR. The first expresses the full-length hTNF-R55; the latter expresses the hTNF-R55 deletion mutant hTNF-R55⌬MPR, which lacks the membraneproximal region ⌬202-304 (numbering according to Ref. 16). 4 Fig. 4B shows that aggregation of hTNF-R55 by the agonistic anti-hTNF-R55 mAb htr-1 resulted in clustering of mitochondria and cell death. Both responses were slightly faster compared with those induced by endogenous mTNF-R55, clustered by agonistic anti-mTNF-R55 mAb, probably due to the higher expression level of the transfected receptor (Fig. 4C). However, the truncated hTNF-R55⌬MPR did not induce clustering of mitochondria (Fig. 4B). This loss of responsiveness was not due to clonal variation, since clustering of the endogenous mTNF-R55 restored the translocation of mitochondria. Remarkably, this induction of clustering was accompanied by a pronounced increase in the rate of cell death, despite the considerably lower expression levels of the endogenous receptor (Fig. 4, B and C). Similar results were obtained using additional, independently derived transfectants (data not shown). Also, the absence of mitochondrial translocation and the delayed cell death after clustering of hTNF-R55⌬MPR was not caused by a reduced DD activity as indicated by the normal levels of NF-B activation, IL-6 induction, and activation of p38/RK mitogen-activated protein kinase (data not shown). Apparently, the TNF-R55 membrane-proximal region is necessary to transduce the signal for translocation as well as to enhance the cytocidal response. The Mitochondrial Translocation Response Is MT-dependent-Mitochondria and other subcellular organelles are associated with the MT cytoskeleton. Consequently, the MT organization defines the dispersal of the organelles in the intracellular space. The transition of mitochondrial distribution in TNF-treated cells from an evenly spread, peripheral location to a clustered, perinuclear one may thus be a repercussion of a similar transition of the MT cytoskeleton. This possibility was supported by the observation that the reversible MT-disrupting drug nocodazole completely prevented TNF-R55-mediated clustering of mitochondria (Fig. 5A). Subsequent removal of nocodazole restored the MTs and reinstated the translocated distribution of mitochondria, confirming that an intact MT system is required for this cellular response. In agreement with previous data (38), disruption of the MT cytoskeleton also resulted in TNF receptor down-modulation (data not shown), making it impossible to assess the effect of nocodazole on TNF-induced cell death. In contrast to nocodazole, the MT-stabilizing drug paclitaxel did not affect TNFinduced mitochondrial translocation or cell death (Fig. 5A, and data not shown). Since the MT-stabilizing effect of paclitaxel prevents extensive rearrangement of the MT cytoskeleton, this negative result makes it unlikely that such a rearrangement is responsible for the observed clustering of mitochondria. This was further supported by immunofluorescent visualization of the MTs in combination with mitochondria in TNF-treated and untreated cells (Fig. 5B). Thus, the MTs radiate outward to the plasma membrane in both TNF-treated and untreated cells, while in TNF-treated cells the mitochondria are no longer scattered throughout the cell but have withdrawn to the perinuclear region. Therefore, a retraction of the MT cytoskeleton can be excluded as the driving force behind the TNF-induced translocation of mitochondria. Nevertheless, the inhibitory effect of nocodazole indicates that the process of mitochondrial translocation requires intact MTs. tochondria-The spatial distribution of mitochondria, as well as of other membranous organelles, is directed not only by the ultrastructure of the cytoskeleton but also by the transport of the organelles along the MTs by molecular motors such as kinesin (39,40) and dynein (41). The majority of known kinesin motor proteins are plus-end directed. In fibroblast-like cell types, kinesin-mediated organelle transport is predominantly directed to the cell periphery. Accordingly, loss of kinesinmediated transport would result in an accumulation of mitochondria in the perinuclear region due to the remaining minusend-directed motor activity of dynein. Indeed, it has been reported previously that inactivation of kinesin in human fibroblasts, obtained by microinjection of antagonistic anti-kinesin Ab, results in clustering of mitochondria in the perinuclear region (42). Since TNF induces a similar phenomenon, we investigated the involvement of kinesin in this TNF response. To this end, SUK4 mAb against KHC was introduced into the cytoplasm of L929 cells using the syringe loading technique. SUK4 mAb recognizes an epitope in the conserved motor domain of KHC and as a consequence blocks the MT-dependent movement of kinesin (30). Western blot analysis of L929 cell extracts showed that SUK4 also recognizes murine kinesin molecules (Fig. 6A). Localization studies using SUK4 mAb and CLSM showed that kinesin partially co-localized with mitochondria in both untreated and TNF-treated L929 cells (Fig.  6B). This result excludes the possibility of a TNF-induced dissociation of kinesin from mitochondria as the molecular basis for mitochondrial translocation. Finally, CLSM analysis of SUK4 mAb syringe-loaded L929 cells showed a translocation of mitochondria that was phenotypically similar to the one induced by TNF (Fig. 6, C and D). In control experiments using an irrelevant Ab, the mitochondria did not translocate. Thus, loss of kinesin motor activity mimics TNF-induced clustering of mitochondria, suggesting that the latter is caused by a similar inactivation of kinesin-mediated organelle transport. However, in contrast to TNF, the SUK4 mAb-induced translocation of mitochondria was not accompanied by cell death, nor did it interfere positively or negatively with TNF-induced cell death (data not shown). Clearly, clustering of mitochondria by itself is not harmful to the cell. However, this does not necessarily exclude the possibility of the underlying molecular mechanism participating in the cell death process. In this case, the presumed inactivation of kinesin by TNF would not occur through the motor domain but instead through other regulatory sites of the kinesin molecule. Likely candidates are the tail domain of KHC and the associated light chain, which together form the organelle-binding site, located at the opposite end of the molecule. This site has been reported to be involved in the regulation of the ATPase and motor activity of the kinesin motor domain (43)(44)(45)(46)(47). To verify the possibility of a contribution to cell death from these regulatory sites, L929 cells were syringeloaded with a mAb directed against the tail domain of KHC (SUK5; Ref. 30). As with SUK4 mAb, this treatment resulted in translocation of mitochondria (data not shown), but in addition, it synergized with the induction of cell death by TNF (Fig. 7). Again, the treatment was not cytocidal per se. Thus, presumably, an analogue TNF-R55-triggered blocking of regulatory sites in the KHC tail domain enhances the sensitivity of the cell to the cell death-inducing activity of TNF and at the same time inactivates kinesin-mediated organelle transport, causing mitochondria to translocate into a perinuclear cluster. This concept reconciles the delayed cell death observed in L929hR55⌬MPR cells with the absence of mitochondrial translocation in these cells. DISCUSSION Recently, it became clear that mitochondria, in addition to their established function as energy-providing organelles, also play an important role in cell death triggered by several stimuli in various cell types. Accordingly, mitochondrial dysfunction is an essential part of TNF-induced cell death (9 -11). Here we Although the latter has a more compact appearance in the TNF-treated cell, the structural organization of the MTs in both cells shows no major differences. In both cases, the MTs are undamaged and typically radiate out from the perinuclear region to the periphery of the cytoplasm. Scale bars, 15 m. report that TNF treatment induces a modification of the spatial distribution of mitochondria; they translocate from an initially scattered, bipolar or nearly symmetric distribution in untreated cells to a cluster near the nucleus in TNF-treated cells. This clustering was specifically induced by TNF and correlated with the death-inducing activity of the cytokine, independently of the nature of the activated cell death program, viz. apoptotic or necrosis-like. Hence, this modified distribution of mitochondria in response to TNF treatment represents a novel activity of the cytokine in TNF-sensitive cells.

Involvement of Kinesin in TNF-induced Translocation of Mi-
The correlation of TNF-induced clustering of mitochondria with cell death suggests that TNF-R55 and not TNF-R75 mediates this response. Indeed, specific triggering of TNF-R55 by hTNF, which is a TNF-R55-specific ligand for murine cells (35), induced translocation of mitochondria in L929 cells. The DD of both TNF-R55 and Fas is believed to be the main mediator of the cell death signal of these death receptors. However, no mitochondrial translocation was observed during Fas-induced cell death, suggesting that the required signal is TNF-R55 DD-independent. Using L929 cells expressing mutant TNF-R55, we showed that the membraneproximal domain of this receptor is essential for signaling to mitochondrial translocation. These results assign a new function to the membrane-proximal region, namely induction of translocation of mitochondria to the perinuclear cytoplasm, and explain the specific induction of the response by TNF-R55 but not by Fas. Intriguingly, we found that in the absence of the membrane-proximal domain also the cell death response was markedly delayed. This apparently reflects a cooperative effect on the DD-triggered cytocidal response of the signal from the membrane-proximal region. To date, none of the known functions of the TNF-R55 membrane-proximal region, viz. induction of nitric-oxide synthase, c-Raf-1 and phospholipase A 2 , appears to be involved in the signal transduction cascade leading to cell death. Furthermore, since the FAN-binding site is still present in the deletion mutant hTNFR55⌬MPR and since TNF does not induce nitric-oxide (red). In the SUK4 mAb-loaded cell, both mitochondria and kinesin show a clustered distribution, similar to the one observed after TNF treatment. In contrast, syringe loading of irrelevant Ab did not influence the dispersal of mitochondria. D, L929 cells were left unloaded (open bar) or were syringe-loaded with an irrelevant Ab (filled bar) or SUK4 mAb (hatched bar). After overnight recovery, the percentage of cells exhibiting translocation of mitochondria was analyzed in the population that took up Ab. Scale bars, 15 m.

FIG. 6. Involvement of kinesin in TNF-induced translocation.
A, Western blot analysis of L929 cell lysates with SUK4 mAb revealed two specific protein bands of approximately 130 kDa, corresponding to the known molecular mass of KHC (arrowheads). Irr., irrelevant Ab. B, CLSM fluorescence images of L929 cells stained for kinesin (SUK4 mAb; red) and mitochondria (green). In untreated cells (Ϫ) as well as in cells treated for 3 h with TNF (ϩ), kinesin partially co-localized with the mitochondria, as apparent from the yellow/orange color in the overlay. This result indicates that TNF does not affect the binding of kinesin to mitochondria. In TNF-treated cells, the TNF-induced translocation of mitochondria is apparent as opposed to their bipolar distribution in untreated cells. C, L929 cells were syringe-loaded with Cy5-conjugated SUK4 mAb or irrelevant Ab. After a 12-h recovery period, the distribution of mitochondria (MitoTracker; green) was analyzed in the cells that had taken up Ab, as apparent from the presence of Cy5 fluorochrome synthesis in L929 cells, 5 both signaling pathways can be excluded as mediators not only of the cell death-enhancing activity of the membrane-proximal region but also of its mitochondrial translocation-inducing activity. It may be assumed that both activities rely on other signaling pathways; considering the correlation of translocation with cell death, these pathways might be identical. Clearly, additional functional studies on the TNF-R55 membrane-proximal region will be necessary to identify the downstream signaling events involved.
Using immunofluorescence microscopy and MT inhibitors, we showed that intact MTs are required for translocation, although the MT cytoskeleton does not impose an altered distribution of mitochondria. This points to a motor protein-dependent molecular mechanism. Indeed, we found that immunoinhibition of kinesin resulted in a translocation phenotype identical to that observed by us after TNF treatment. Thus, TNF seems to inactivate kinesin-mediated transport of mitochondria. This inactivation might be mediated by caspases that are known to be activated during TNF-induced apoptosis. However, in L929 cells no caspase activation is apparent. Furthermore, in these cells the broad spectrum caspase inhibitor zVAD.fmk did not block mitochondrial translocation (data not shown), excluding the possibility of an involvement of caspases in this response. Alternative candidates are MT-associated proteins, such as MAP2 and MAP4, which constrain the kinesinmediated transport of membranous organelles after overexpression (48 -50). However, the sensitizing effect on the cytocidal activity of TNF by immunoinhibition of the KHC tail domain indicates that kinesin is a direct target of TNF. Thus, inactivation of kinesin resulted in a functional behavior similar to that generated by the TNF-R55 membrane-proximal region, namely translocation of mitochondria and enhancement of cell death. This similarity shows that kinesin is a direct target of TNF-R55-triggered signal transduction. A direct effect on kinesin is further supported by TNF-induced changes in phosphorylation of kinesin immunoprecipitates. 6 This is consistent with reports showing that kinesin-mediated transport is regulated by phosphorylation/dephosphorylation of kinesin heavy and light chains and/or associated phosphoproteins. Also, the presence of phosphotransferases in kinesin immunoprecipitates supports a role of phosphorylation events in controlling kinesin (43)(44)(45)(46)(47). On the whole, our data clearly suggest that kinesin may exceed its established function as molecular motor and that it exerts additional activities contributing to cell death. Further analysis of kinesin and its regulation should address this intriguing issue.