OUP user menu

Tumour necrosis factor-induced death of adult human oligodendrocytes is mediated by apoptosis inducing factor

Anna Jurewicz , Mariola Matysiak , Krzysztof Tybor , Lukasz Kilianek , Cedric S. Raine , Krzysztof Selmaj
DOI: http://dx.doi.org/10.1093/brain/awh627 2675-2688 First published online: 11 October 2005


Tumour necrosis factor (TNF)-induced death of oligodendrocytes, the cell type targeted in multiple sclerosis, is mediated by TNF receptor p55 (TNFR-p55). The ligation of TNFR-p55 induces several signal transduction pathways; however, the precise mechanism involved in human oligodendrocyte (hOL) death is unknown. We defined that TNF-induced death of hOLs is non-caspase dependent, as evidenced by lack of generation of caspases 8, 1 and 3 active subunits; lack of cleavage of caspases 1 and 3 fluorogenic substrates; and lack of hOL death inhibition by the general caspase inhibitor, ZVAD.FMK. Electrophoresis of TNF-exposed hOL DNA revealed large-scale DNA fragmentation characteristic of apoptosis-inducing factor (AIF)-mediated cell death, and co-localization experiments showed that AIF translocation to the nucleus occurred upon exposure to TNF. AIF depletion by an antisense strategy prevented TNF-induced hOL death. These results indicate that TNF-induced death of hOLs is dependent on AIF, information of significance for the design strategies to protect hOLs during immune-mediated demyelination.

  • oligodendrocytes
  • apoptosis inducing factor
  • multiple sclerosis
  • cell death
  • AIF = apoptosis-inducing factor
  • anti-GFAP = anti-glial fibrillary acidic protein
  • anti-MBP = anti-myelin basic protein
  • asAIF = antisense AIF
  • FACS = fluorescence-activated cell sorter
  • FADD = Fas-associated protein with death domain
  • hOL = human oligodendrocyte
  • OL = oligodendrocyte
  • PI = propidium iodide
  • TNF = tumour necrosis factor
  • TNFR-p55 = TNF receptor p55
  • TPCK = N-tosyl-l-phenylalanine chloromethyl ketone
  • TRADD = TNF-R1-associated death domain


Several immune effector mechanisms have been implicated in oligodendrocyte (OL) death in vitro and in vivo. OLs appeared to be relatively sensitive to stress-induced and inflammatory related mediators. Non-major histocompatibility complex restricted injury of OLs by soluble factors, including tumour necrosis factor (TNF) family members, has been described in several experimental models (Selmaj, 1992; Zajicek et al., 1992). TNF has been shown to induce delayed and highly specific patterns of OL injury. Other glial cells, including astrocytes and microglia, are not injured by TNF and can even respond by proliferating (Selmaj, 1992). However, the molecular mechanisms of TNF-induced OL death is not well characterized and downstream effector molecules are not known. The peculiar morphology of OLs exposed to TNF, comprising slow retraction of cell processes, watery cytoplasm and only partially degraded nuclei, might suggest involvement of a process different from classical death receptor-induced mechanism of cell death.

TNF elicits a wide spectrum of cellular responses governed by interactions with two distinct cell surface receptors, TNF–Rp55 (TNFR1) and TNF–Rp75 (TNFR2) (Vandanabeele et al., 1995). However, the paradigm of TNF activity involves cell death which is mediated mainly by ligation of TNFR-p55. In many cell types, TNF-induced cell death can be prevented by inhibition of the caspase cascade (Salvesen and Dixit, 1997). Activation of TNF–Rp55 results in recruitment of the TNF-R1-associated death domain (TRADD) protein (Hsu et al., 1995). Subsequently, TRADD interacts with TNF receptor associated factor-2 (TRAF-2), or with Fas-associated protein with death domain (FADD) (Hsu et al., 1996). FADD interacts with procaspase 8 which appears to be a direct activator of the apoptotic protease cascade (Boldin et al., 1996). This caspase is considered an initiator of the caspase cascade, leading to the activation of caspase 1, whereas caspases 3 and 7 are activated at a later phase of apoptosis and are effectors acting on a large number of substrates (Green, 2000). The alternative intracellular apoptotic pathway appeared to involve mitochondrial dysfunction by releasing cytochrome c which triggers activation of caspase-9 and then caspase-3 (Kluck et al., 1997). However, more recently, non-caspase dependent mechanisms of TNF-induced cell apoptosis have been proposed. These mechanisms mainly involve mitochondria-derived factors such as apoptosis inducing factor (AIF) (Daugas et al., 2000), endonuclease G (Li et al., 2001; Parrish et al., 2001), the induction of reactive oxygen species (ROS) (Jackobson, 1996), calpain (Villa et al., 1998), and serine protease activation (Wright et al., 1997). One feature distinguishing these mechanisms of cell death from caspase-dependent mechanisms is the lack of involvement of a caspase-dependent DNase, resulting in a different pattern of DNA cleavage.

The objective of this study was to define the effector mechanism involved in TNF-induced human oligodendrocyte (hOL) death. Our data suggest that death of mature hOLs induced by TNF does not require caspase activation but rather depends on AIF translocation from mitochondria to nuclei and the induction of large-scale DNA degradation.

Materials and methods


TNF was purchased from Endogen (Woburn, MA); caspase inhibitor ZVAD.FMK, calpain inhibitor ZLLY.FMK and control peptide FA.FMK, from Enzyme Systems Products (Livermore, CA); serine protease inhibitor N-tosyl-l-phenylalanine chloromethyl ketone (TPCK), Hoechst 33342, propidium iodide (PI), and staurosporine, mouse monoclonal anti-human fibroblast surface protein antibody (Ab) from Sigma (Poznan, Poland); annexin V conjugated with fluoroisothiocyanate (FITC), mouse anti-CD-11b Ab PE conjugated from Pharmingen, BD Biosciences (San Diego, CA); and goat anti-AIF Ab, goat anti-FLIP Ab, goat anti-IAP Ab and chicken anti-goat rhodamine conjugated Ab, from Santa Cruz Biotechnology (Santa Cruz, CA), rabbit antiglial fibrillary acidic protein (anti-GFAP) Ab, goat anti-rabbit rhodamine conjugated Ab, rat antimyelin basic protein (anti-MBP) Ab, goat anti-rat FITC conjugated Ab, mouse anti-MBP Ab, goat anti-mouse FITC conjugated Ab, mouse antigalactocerebroside (anti-GalC) (01) Ab, goat anti-mouse IgM rhodamine conjugated Ab and goat anti-mouse IgM FITC conjugated Ab from Chemicon (Chemicon Europe, UK). Rabbit complement was purchased from Biotest AG (Dreieich, Germany). Cathepsin inhibitors, ZFL and ca-074-Me, were kindly provided by Dr Marja Jaattela (Institute of Cancer Biology, Copenhagen, Denmark).

Target cell populations

WEHI 1640 cells, a murine fibroblast cell line, were prepared and maintained according to ATCC protocols. OLs were prepared from adult human brain tissue resected during neurosurgical procedures, as previously described (Jurewicz et al., 1998). In brief, the tissue was trypsinized, rubbed through mesh and spun down on a 30% Percoll gradient. Following this, the dissociated cells were suspended in MEM (GIBCO, Gaithersburg, MD), with 5% FCS (GIBCO), streptomycin (50 µg/ml) and penicillin (50 U/ml), and then cultured for 48 h in culture flasks. This step enables the separation of adherent cells (e.g. microglia), from non-adherent cells (OLs). The non-adherent OL fraction was plated onto poly-l-lysine-coated 96-well microtitre plates at 5 × 104 cells/well and cultured for 2 weeks. To assess purity of hOL cultures, the immunostaining was performed with rat anti-MBP Ab, followed by anti-rat FITC conjugated Ab and counterstaining with anti-CD11b PE conjugated, rabbit anti-GFAP Ab, followed by anti-rabbit rhodamine conjugated Ab. In parallel these same hOL cultures were immunostained with monoclonal anti-human fibroblast surface protein Ab, followed by anti-mouse rhodamine conjugated Ab. The purity of hOLs was ∼95%, as described previously (Jurewicz et al., 1998). Astrocytes and microglial cell were not detected in OL cultures as assessed by fluorescence-activated cell sorter (FACS) analysis (Fig. 1A) and microscope analysis (Fig. 1B). The remaining minor cell population was identified as fibroblasts (Fig. 1C). To confirm that TNF-induced hOL death was not affected by fibroblasts, the hOL cultures were deprived of fibroblasts by complement lysis. For these, cultured cells were first incubated with monoclonal anti-human fibroblast surface protein (Sigma) for 1 h at room temperature and after washing, then complement (Biotest) was added for another 1 h at 37°C. Microglial cells were cultured in flasks for 1.5 weeks, and a few days before testing were collected, spun down and plated onto culture plates. The purity of microglial cells was assessed by staining with Ab against Mac-1 (CD11b) (Pharmingen), and was >90%. For each experiment, cells were used at the same timepoint after removal from brain tissue.

Fig. 1

FACS analysis of cells from OL cultures. (A) The cells were stained with anti-GalC (OL marker), anti-CD11b (microglial cells marker), anti-GFAP intracellular staining (astrocyte marker). The positive cells for staining by the fluorescence shift in Fl1 (yellow peaks), compared with the control histogram (isotype control) (shaded peaks). Histograms represent fluorescence intensity on the horizontal axis and relative cell number on the vertical axis. Data are representative of three identical experiments. (B) The lack of microglia and astrocytes in hOL cultures was confirmed by immunostaining with anti-GalC Ab, followed by FITC-coupled anti-mouse Ab (green, left panel) and counterstaining with anti-GFAP (middle panel) or anti-CD11b (right panel) Ab and microscope assessment. (C) The same cultured hOLs were stained with anti-MBP Ab, followed by FITC-coupled anti-rat Ab (green, left panel) and counterstained anti-fibroblast Ab, followed by rhodamine coupled anti-mouse Ab (red, right middle panel). Both stainings were superimposed (merge, right panel). (D) TNF-induced hOL death is dose-dependent as determined by a percentage of PI-positive cells. Cell death was assessed after 72 h of TNF stimulation. Each point represents a mean result from 3 to 5 experiments. (E) Fibroblast deprivation of OL cultures did not change TNF-induced OL death. (F) TNF induces large-scale (lower panel), but not oligonucleosomal (upper panel), DNA fragmentation of hOL. After TNF treatment, AIF translocates to nuclei (G), hOLs were labelled with anti-GalC Ab followed by FITC-coupled anti-mouse IgM Ab (green, left panel) and anti-AIF Ab followed by Rh-coupled anti-goat IgG Ab (red; middle panel), and with Hoechst (blue; right panel). Data are representative of three independent experiments.

Flow cytometric analysis of cell death

To assess cell death, before detachment from poly-l-lysine coated wells, cells were stained simultaneously with FITC-conjugated annexin V (Pharmingen) to detect exposure of phosphatidylserine on the cell surface, and with PI (Sigma) to detect permeable dead cells, and the staining was measured by flow cytometry (Becton-Dickinson, San José, CA). For this, cells were washed in phosphate-buffered saline and 10 µl of FITC-conjugated annexin V and PI (5 µg/ml) in calcium-containing buffer, according to the manufacturer's protocol (Pharmingen). After cell detachment, we assessed the Fl1 intensity shift for FITC-conjugated annexin V and the Fl2 intensity shift for PI staining. According to previous reports (Vermes et al., 1995), apoptotic cells were defined as those showing annexin V staining before the appearance of staining for PI.

OL immunofluorescent staining and confocal microscopy for apoptosis-related molecules

OLs, unstimulated and stimulated with TNF (1000 U/ml for 72 h) were grown on glass coverslips. hOLs were fixed with 4% paraformaldehyde and permeabilized Triton X-100 for Ab staining. To visualize AIF localization, cells were first stained with anti-AIF Ab, followed by staining with secondary rhodamine-conjugated anti-goat Ab (Santa Cruz). To label nuclei, hOLs were stained with Hoechst 33342 for 5 min. Confocal microscopy was performed using the Confocal Laser Scanning Microscope TCS SP2 (Leica, Lasertechnik GmbH, Heidelberg, Germany). To definitely confirm that AIF translocation occurred in OLs, the cells were counterstained with anti-galactocerebroside Ab. For each experiment, cells stained with fluorophore alone were imaged to ensure that fluorescence emission did not bleed between channels. Parallel cell cultures were stained with anti-IAP1 Ab or anti-FLIP Ab and counterstained with anti-GalC Ab. The cells illustrated for each experiment are representative of at least three independent experiments.

DNA gel electrophoresis

For detection of oligonucleosomal DNA fragmentation, nuclear DNA from lysed cells (treated with RNase and proteinase K, according to standard protocols), was subjected to conventional horizontal 1% agarose gel electrophoresis followed by ethidium bromide staining. For pulse field gel electrophoresis (PAGE), DNA was prepared from agarose plugs (2 × 106 cells), digested twice with proteinase K (1 mg/ml; 50°C; 12 h) in NDS buffer (0.5 M EDTA, 10 mg/ml lauroyl sarcosine) (Bio-Rad), and washed in TBE 3 0.5, followed by electrophoresis in a Bio-Rad CHEF-DR II (Richmond, CA) equipment (1% agarose; TBE 3 0.5; 200 V; 24 h; pulse wave 60 s; 120° angle). Molecular weight standards were from Bio-Rad.

Mitochondrial function assay

An ApoAlert Mitochondrial Membrane Sensor Kit (Clontech) and assessment by FACS analysis were used to detect changes of mitochondrial membrane potential, according to the manufacturer's protocol. Mitosensor is a cationic dye aggregated in mitochondria and exhibits red fluorescence. However, in cells which have lost the ability to maintain the proper mitochondrial membrane permeability, the cationic dye cannot accumulate in mitochondria and remains in monomeric form in the cytoplasm where it exhibits green fluorescence. Cells were rinsed gently with serum-free medium and Mitosensor was added, incubated at 37°C in 5% CO2 for 20 min and subjected to FACS analysis.

Fluorogenic peptide substrate caspase activation assays

Fluorogenic peptide DABCYL-Tyr-Val-Ala-Asp-Ala-Pro-Val-EDANS (Bachem, Blubendorf, Switzerland), with restriction sequence for caspase 1 at a concentration of 20 µM, was added to 5 × 105 cells previously permeabilized with digitonin (Sigma), 10 µg/ml at for 10 min. In the presence of active caspase 1, the peptide is processed into two fragments which generate green fluorescence which can be measured by flow cytometry (Becton-Dickinson). The caspase 1 activation was measured before and after 30 min, 2, 24, 48, 72, 96 h of TNF stimulation at a concentration of 1000 U/ml. For the WEHI cell line, caspase 1 activation was assessed before and after 30 min, 2, 24 h of TNF stimulation; later time points were not accessible because of earlier cell death.

For caspase 3 activation, fluorogenic peptide carbobenzoxy-Asp-Glu-Val-Asp-7-amino-4-trifluoromethyl coumarine (Bio-Rad, Hercules, CA), containing restriction sequence for caspase 3, was added to cell lysates obtained from 1 × 107 cells at a concentration of 40 μg/ml. Fluorescence was measured immediately and after 30 min of incubation at room temperature, using a spectrofluorimeter (Perkin-Elmer, Norwalk, CT), with an excitation wavelength of 375 nm and an emission wavelength of 530 nm, according to the manufacturer's protocol. Caspase 3 activation was assessed before and after 30 min, 2, 15, 2, 36, 48, 72 h of TNF stimulation (1000 U/ml), for both OLs and microglia. For the WEHI cell line, caspase 3 activation was evaluated before and after 30 min, 2, 15, and 24 h of TNF stimulation. After subtraction of fluorescence in samples lacking protein, the amount of liberated fluorophore was determined by comparison with a standard curve.

Western blot analysis for caspases, PARP, DISC proteins, IAP proteins, FLIP and AIF

The levels of death-inducing signalling complex (DISC) proteins and caspases 8, 1 and 3, poly(ADP)ribose polymerase (PARP), and their cleavage products, were determined by Western and immunoblotting. The cells (1 × 106) were pelleted and lysed in lysing buffer (20 mM Tris–HCl, pH 7.4, 0.15 M NaCl, 1% Triton, 2.5 mM sodium pyrophosphate, 1 mM EDTA, 1 mM beta-glycerophosphate, 1 mM NA3VO4, 1 mM phenylmethylsulphonyl fluoride, 1 µg/ml aprotinin and leupeptin). Lysates were centrifuged at 14 000 r.p.m. for 5 min. An equal amount of protein from each cell lysate was separated on SDS–PAGE and electrophoretically transferred to PVDF membranes (Immobilon, Millipore). The membranes were blocked with 5% dried milk in Tris-buffer saline overnight, incubated for 1–2 h with primary antibodies (1 µg/ml in TTBS) for 1.5 h, and then with peroxidase-conjugated secondary antibodies: donkey anti-goat IgG (Santa Cruz) (1:20 000); goat anti-mouse IgG (Sigma) (1:15 000) and goat anti-rabbit IgG (Santa Cruz) (1:20 000), depending on the species of the primary Ab, and developed with chemiluminescence. Proteins were detected using an Amersham enhanced chemiluminescence system. The primary antibodies used in Western blots were mouse anti-caspase 1 and anti-caspase 8 IgG; rabbit anti-caspase 3 IgG, and mouse anti-PARP IgG (Pharmingen); goat anti-TRADD, anti-FADD, anti-RIP, anti-IAP1, anti-IAP2, anti-FLIP, anti-AIF and rabbit anti-TRAF2 IgG (Santa Cruz).


To confirm the lack of cleaved fragments of caspases 8, 1, 3 and PARP on Western blots of hOL, cell lysates were introduced into the primary step of immunoprecipitation. Cell lysates obtained from 5 × 106 cells were immunoprecipitated with primary Ab for 2 h at 4°C and then with goat anti-rabbit IgG1-agarose or goat anti-mouse IgG-agarose (Sigma), for 2 h, at 4°C. Subsequently, immunoprecipitates were analysed as described above for Western blot.

Antisense AIF construct

To determine the function of AIF in OL death, a pcDNA3.1 plasmid containing the AIF sequence in an antisense orientation was constructed. The human AIF sequence was taken from GenBank, accession number AF 100928. The AIF sequences were copied from a human cDNA library (Clontech), using the primers 5′-CCTTGCGGCCGCATGTTCCGGTGTGGAGGCCTGGCG-3′ and 5′-CCTTGGTACCGCAGTGTCTTTGTGACATTTGGGT-3′, giving the PCR product a molecular weight of 574 bp. PCR conditions were as follows: 95°C, 4 min; 94°C, 1 min; 55°C, 1 min; 72°C, 1 min; 35 cycles, 72°C, 10 min. The first PCR product was extracted from agarose gel and was used as a template for the second PCR with the same primers. Re-PCR conditions were as follows: 95°C, 4 min; 94°C, 1 min; 55°C, 45 s; 72°C, 1 min; 30 cycles; 72°C, 10 min. After purification, the second PCR product was cloned into a pcDNA3.1 vector in an antisense orientation. The restriction sites/enzymes were KpnI and NotI. Vectors containing proper fragments were cloned using Escherichia coli TOP10F′ strain of competent cells (Stratagene). Plasmids were purified on Qiagen columns according to the manufacturer's procedures (Qiagen). The deletion efficiency was assessed on Western blot by protein expression decrease.

Cell transfection

A group of 5 × 106 OLs, microglial cells and WEHI cells were transfected with pcDNA3.1 vector alone or containing AIF fragment in an antisense orientation (asAIF), or GFP insert (Promega). Cells were incubated with 2 µg of pcDNA3.1 vectors and 1 µl lipofectin (Life Technologies, Inc., Gaithersburg, MD), according to previous protocols (Skov et al., 1997). Cells were grown for 48 h before experimentation. The level of cell transfection was checked by pcDNA3.1 containing GFP insert, thus enabling assessment of living cells in a fluorescence microscope. In addition, the efficacy of cell transfection was confirmed by flow cytometry of cells transfected with pcDNA3.1-GFP and was >80%. The level of decrease of protein expression was assessed by Western blot and densitometry.

Immunostaining of the brain sections

Brain tissue was obtained at early autopsy (<8 h post-mortem) from two patients with multiple sclerosis. Autopsy tissue was obtained under IRB approval. Histology was performed and the lesions used for the study were of chronic active type. For immunostaining, frozen sections of brain tissue were used whereby acetone-fixed frozen sections were stained using anti-AIF Ab, anti-MBP Ab (to visualize OLs) and counterstained with Hoechst 33342 (to visualize nuclei). Sections were air-dried for 60 min, fixed in acetone for 10 min at 4°C and then permeabilized with Triton X-100 for Ab staining. After blocking with serum, sections were incubated with anti-AIF primary goat Ab (Santa Cruz) for 1.5 h at room temperature. Secondary rhodamine conjugated Ab (anti-goat IgG) was applied for 60 min at room temperature. To visualize OLs, the brain sections were stained with anti-MBP mouse Ab (Chemicon) for 1.5 h at room temperature and after that with secondary anti-mouse FITC conjugated Ab (Santa Cruz) for 60 min at room temperature. To label nuclei, the brain sections were stained with Hoechst 33342 for 5 min at room temperature.

Microscopy of the brain sections was performed using the Zeiss Axiovert 200 Microscope (Zeiss, Germany) and the pictures were taken with CoolSnapp HQ (Photometrics, USA).


TNF induces human adult OL death

As described earlier (Selmaj and Raine, 1988; Jurewicz et al., 2003), TNF added to hOLs at a concentration of 1000 U/ml induced hOL death. TNF-induced hOL death was dose-dependent (Fig. 1D). The morphology of dying OLs, which displayed with patchy chromatin aggregation and a watery cytoplasm, suggested a non-classical apoptotic mechanism of cell death, although annexin-V cell surface staining of hOLs on TNF treatment preceded PI staining (Vermes et al., 1995). Morphological features and the sequence of membrane molecular events, together with time-delayed onset of TNF-induced hOL death, suggested an apoptotic-like mechanism of cell death rather than classical apoptosis. The contribution of microglia and astrocytes to the observed TNF-induced hOL death was excluded since these cells were absent in hOL cultures (Fig. 1A and B). The potential role of fibroblast, which were occasionally seen at very low numbers in hOL cultures, was excluded by demonstration of no change of the rate of hOL death in cultures deprived of fibroblasts by specific complement lysis (Fig. 1E).

TNF induces large scale DNA fragmentation

To assess the mechanism of the TNF-induced hOL death pathway, two types of DNA electrophoresis were performed to evaluate the pattern of DNA cleavage. In TNF-treated hOLs large-scale DNA fragmentation was detected but not oligonucleosomal DNA fragmentation (Fig. 1F). Large-scale DNA fragmentation was characterized by DNA fragments of ∼50 kb. This type of DNA fragmentation is typically mediated by AIF released from mitochondria (Susin et al., 2000). Large-scale DNA fragmentation occurred in native hOLs in the absence of caspase inhibition. In control WEHI cells, large-scale DNA fragmentation was seen only when caspases were efficiently inhibited (Fig. 1F).

Upon TNF treatment, AIF translocates to the nuclei of hOLs

To demonstrate direct involvement of AIF in TNF-induced hOL death, we performed co-localization experiments which showed that after TNF exposure, AIF was translocated to the nucleus of GalC-positive cells (Fig. 1G). Prior to TNF stimulation of hOLs, AIF was clearly localized outside the nucleus, whereas 72 h after the addition of TNF, anti-AIF Ab reactivity was partially superimposed on the Hoechst 33342 staining, indicating translocation of AIF to the nucleus (Fig. 1G).

AIF inhibition prevents TNF-induced death

To confirm that AIF is involved in TNF-induced hOL death, AIF inhibition was performed using the asAIF. hOLs transfected with asAIF showed significant depletion of AIF protein (∼60%), as evidenced by Western blot and densitometry (Fig. 2A). The depletion was transient with maximal effect at 48 h after transfection, the timepoint when hOLs were used for TNF-induced cell death assays. AIF depletion in hOLs correlated with diminished TNF-induced death (Fig. 2B). However, asAIF exerted no effect on TNF-induced death of WEHI cells, which are mainly caspase-dependent (Fig. 2B). asAIF transfection of hOLs without TNF exposure (Fig. 2B) and transfection of microglial cells (data not shown) did not affect cell survival, except for some non-specific asAIF activity.

Fig. 2

asAIF inhibits AIF expression in hOL as demonstrated by Western blot after 48 h of exposure to asAIF (A) and prevents TNF-induced hOL death assessed by annexin V-FITC and PI staining and FACS analysis (B). The positive cells for annexin V-FITC and PI are demonstrated by the fluorescence shift in Fl1 and Fl2 (yellow peaks), compared with the control histogram (no TNF stimulation, no asAIF transfection) (shaded peaks). Histograms represent fluorescence intensity on the horizontal axis and relative cell number on the vertical axis. asAIF alone did not affect cell survival (B). Data are representative of three identical experiments.

AIF inhibition did not affect changes in mitochondrial membrane potential after TNF stimulation

TNF stimulation induced changes in mitochondrial membrane potential after 4 h of exposure (Fig. 3), and in our earlier studies, we showed that the mitochondrial membrane change was an essential component of TNF-induced hOL death (Jurewicz et al., 2003). asAIF transfection reduced AIF protein expression and prevented hOL death. However, it did not change the pattern of mitochondrial activation (Fig. 3), thus excluding a role for other mitochondria-derived apoptotic factors.

Fig. 3

asAIF did not prevent TNF-induced change in mitochondrial membrane permeability as assessed by a mitochondrial membrane integration test and FACS analysis. The positive cells (TNF-stimulated) (yellow peaks), are demonstrated by the fluorescence shift in Fl1 compared with the control histogram (non-stimulated cells) (shadowed peaks). Histograms represent fluorescence intensity on the horizontal axis and relative cell number on the vertical axis. Data are representative of four identical experiments.

TNF-induced death of hOLs does not involve caspase activation

To assess the role of caspases in hOL death induced by TNF, we analysed the presence and activation of caspases 8, 1 and 3 in hOLs. Western blot analysis revealed that all three caspases could be detected in OLs in proenzyme form (Fig. 4A). However, in response to stimulation with TNF, caspases 8, 1 and 3 were not cleaved into active subunits although the cells were driven into cell death. Lack of caspases 8, 1 and 3 activation in hOLs was also demonstrated by the lack of generation of fluorogenic derivates from caspase substrates (Fig. 4B). Caspase activation was not detected in hOL even at very late timepoints. In the control WEHI cell line, TNF-induced caspases 8, 1 and 3 activation occurred within 24 h post-TNF exposure and preceded apoptotic cell death. As expected, microglia, which were resistant to TNF-induced death, did not show caspases 8, 1 and 3 activation (Fig. 4B). To further assess the lack of caspase involvement in TNF-induced hOL death and to exclude any alternate signalling leading to caspase 3 activation, we measured the degradation of PARP, the final substrate of caspase 3. hOLs stimulated with TNF did not reveal PARP cleavage, in contrast to WEHI cells (Fig. 4C).

Fig. 4

(A) TNF did not induce caspases 8, 1 and 3 cleavage into active subunits in hOL at a concentration up to 1000 U/ml and 72 h incubation time, as assessed by Western blot. (B) TNF did not induce generation of fluorogenic derivates from caspases 1 and 3 substrates as assessed by flow cytometry (caspase 1) and fluorimetric analysis (caspase 3) in hOLs. The positive cells (TNF-stimulated) (yellow peaks) are demonstrated by the fluorescence shift in Fl1 compared with the control histogram (non-stimulated cells) (shaded peaks). Histograms represent fluorescence intensity on the horizontal axis and relative cell number on the vertical axis (caspase 1) (B). (C) TNF did not induce PARP cleavage in hOL as assessed by Western blot. Data are representative of five identical experiments.

Inhibition of caspase activation does not prevent TNF-induced OL apoptosis

To further analyse the role of caspase cascades in hOL death induced by TNF, we used a caspase inhibitor. The general caspase inhibitor, ZVAD.FMK, added at a concentration of 20 µM, 2 h before stimulation with TNF, totally suppressed TNF-induced apoptosis in the WEHI cell line (Fig. 5), whereas ZVAD.FMK at a concentration of 10–50 µM and preincubated up to 4 h, did not affect TNF-induced death of hOLs. These results confirmed that the mechanism of hOL death induced with TNF was caspase-independent.

Fig. 5

TNF-induced hOL death was not prevented by pan-caspase inhibitor ZVAD.FMK, calpain inhibitor ZLLY.FMK, serine proteinase inhibitor TPCK and cathepsin inhibitor ZLY as assessed by annexinV-FITC and PI staining and flow cytometry. The cells positive for annexin V-FITC and PI are demonstrated by the fluorescence shift in Fl1 and Fl2 (yellow peaks) compared with the control histogram (no TNF stimulation) (shadow peaks). Histograms represent fluorescence intensity on the horizontal axis and relative cell number on the vertical axis. Data are representative of four identical experiments.

Calpains, serine proteases and cathepsins are not involved in TNF-induced hOL death

Another non-caspase dependent mechanism of TNF-induced cell death involves activation of calpain and serine protease. Calpains have been shown to mediate apoptosis by cleaving cytoskeletal proteins and the proapoptotic protein, Bax (Villa et al., 1998). To assess the potential role of calpains in TNF-induced hOL death we pretreated hOLs with a calpain inhibitor, ZLLY.FMK, at a concentration of 40 µM for 2 h. However, ZLLY.FMK did not diminish TNF-induced hOL apoptotic death (Fig. 5).

Serine proteases, particularly AP24 protease, acting downstream of caspase 3, have also been implicated in cell death mechanisms (Wright et al., 1997). Therefore, to test whether serine proteases were involved in TNF-induced hOL death, we used TPCK, a general inhibitor of serine proteases, which has been previously reported to block apoptotic DNA fragmentation (Wright et al., 1997). TPCK added at a concentration of 0.1–5 µM for 2 h prior to TNF treatment, did not prevent TNF-induced hOL death (Fig. 5). In addition, cathepsin inhibitors (ZFL, at a concentration of 1–5 µM, and ca-074-Me, at a concentration of 5–50 µM), were also ineffective in preventing TNF-induced hOL death (Fig. 5).

DISC molecules in hOLs

Since the lack of caspases 8, 1 and 3 activation may be the result of insufficient upstream signalling from TNFR-p55 activation, we assessed the presence of TRADD, FADD and TRAF-2 in hOLs, microglia and the WEHI cell line. All adaptor molecules were detected by Western blot in all cell types studied, including hOLs (Fig. 6). These results proved that all caspase adaptor molecules required for signal transduction from TNFR-p55 were present in hOLs; thus the lack of caspase activation during TNF-induced hOL apoptosis was not related to their absence.

Fig. 6

hOLs, control and TNF-stimulated, express DISC molecules (TRADD, FADD and TRAF2), as assessed by Western blot. Data are representative of three independent experiments.

Caspase endogenous inhibitors, FLIP and IAP-1, are present in hOLs

To further analyse the mechanisms underlying the lack of caspase activation in TNF-induced hOL death, the expression of two endogenous caspase inhibitors, FLIP and IAP-1, was assessed. The FLIP and IAP-1 expression in hOLs was demonstrated by immunostaining of GalC-positive cells with anti-IAP-1 and anti-FLIP Ab (Fig. 7A and B). In addition, by using Western blot analysis, it was shown that FLIP and IAP-1 were present in unstimulated hOLs (Fig. 7C and D). In addition, TNF stimulation increased FLIP expression in hOL by 50%, as evidenced by Western blot and densitometry (Fig. 7C). An increase in FLIP expression was detected after 24 h of exposure to TNF. In microglia and WEHI cells, TNF had no effect on FLIP expression. IAP-1 expression remained stable after exposure of hOLs to TNF (Fig. 7D). These results suggest that enhancement of FLIP expression might correlate to lack of caspase activation after TNFR-p55 ligation.

Fig. 7

FLIP and IAP-1 are expressed in hOLs. (A) Cultured hOLs were stained with anti-GalC Ab followed by FITC-coupled anti-mouse IgM Ab (green, left panel) and with anti-FLIP Ab followed by Rh-coupled anti-goat IgG Ab (red, middle panel). Both stainings were superimposed (merge, right panel). (B) Cultured hOLs were stained with anti-GalC Ab followed by FITC-coupled anti-mouse IgM Ab (green, left panel) and with anti-IAP Ab followed by Rh-coupled anti-goat IgG Ab (red, middle panel). Both stainings were superimposed (merge, right panel). (C) TNF enhanced expression of the long form of FLIP on hOL treatment at a concentration of 1000 U/ml and incubation times up to 24 h as assessed by Western blot, but not IAP-1 expression (D). Data are representative of three identical experiments.

Staurosporine-induced hOL apoptosis requires caspase-3 activation and PARP cleavage

To assess whether other apoptosis inducers might activate a caspase cascade in hOL, we used staurosporine, a protein kinase C inhibitor, which has been shown to be one of the strongest and most general apoptotic cell death inducers. As expected, staurosporine treatment at a concentration of 1 µM for 72 h induced hOL death (data not shown). However, staurosporine-induced apoptosis of hOL involved caspase 3 activation and PARP cleavage, as evidenced by the presence of caspase 3 and PARP cleavage subunits (Fig. 8).

Fig. 8

Staurosporine-induced PARP cleavage and caspase 3 cleavage in hOLs, as assessed by Western blot. Data are representative of three identical experiments.

Presence of nuclear AIF in brain tissue

In brain sections, within the edge of multiple sclerosis lesion, some OLs defined by anti-MBP staining showed nuclei positive for AIF. In co-localized experiments, nuclei staining with Hoechst overlapped with staining for AIF (Fig. 9) proving that AIF was translocated into the nuclei. Outside the multiple sclerosis lesion we were not able to demonstrate costaining of Hoechst and AIF, indicating that translocation of AIF was discriminative for OL within the edge of multiple sclerosis lesion.

Fig. 9

At the edge of multiple sclerosis lesion AIF co-localized with Hoechst in MBP-positive OLs. Frozen sections were simultaneously stained with anti-AIF Ab (red), Hoechst (blue), and anti-MBP Ab (green). The co-localization of AIF and Hoechst is demonstrated as violet in the merge panel (arrows).


In this report, we have investigated a variety of intracellular transduction pathways involved in TNF-induced death in mature adult human OLs, the putative cell target in multiple sclerosis. In this regard, we present evidence that TNF-induced death of adult hOLs is not caspase-dependent and involves AIF as an alternative death molecule.

The results of our experiments strongly suggest that caspases were not activated during TNF-induced death of mature adult hOLs. The general caspase inhibitor, ZVAD.FMK, did not prevent TNF-induced hOL death and accordingly, upon TNF treatment, we were unable to detect proteolytic activation of caspases 8, 1 and 3, as indicated by the lack of generation of active subunits. These findings are consistent with the absence of PARP cleavage, a substrate for activated caspase 3, and the lack of generation of fluorogenic products derived from caspase 1 and caspase 3 specific substrates. Caspases 8, 1 and 3 are expressed in OLs (Gu et al., 1999), but their role in TNF-induced death of adult mature hOLs has not been defined. In embryonic mouse mixed glial cell cultures enriched for OLs Hisahara et al. (1997) showed caspase involvement in TNF-induced apoptosis. Andrews et al. (1998) reported that the general caspase inhibitor, ZVAD.FMK, partially suppressed apoptotic changes elicited by the combined activity of IFN-γ and TNF only in CG4-positive OL progenitor cells but not in differentiated OLs. Also, nerve growth factor, interacting with p75 receptor of TNF family, induced caspase 3-dependent apoptosis of post-natal rat OLs (Casaccia-Bonnefil et al., 1996), but not of adult human mature OLs (Ladiwala et al., 1998). Together with the present findings, these results demonstrate significant caspase dependence of TNF-induced cell death on selective developmental stages of the OL cell lineage. It is possible that end-stage, non-dividing cells, such as mature adult hOLs, may possess mechanisms which protect them from the highly efficient death-inducing machinery which uses the caspase cascade. Staurosporine, a protein kinase inhibitor, which acts as a strong apoptotic cell death inducer in many cell types, including OLs (Tang et al., 2000), induced caspase 3 and PARP cleavage, indicating that caspase processing can occur in hOLs but is not initiated by physiological factors, such as TNF. Consistent with the absence of caspase activation during TNF-induced apoptosis of hOL is the lack of features typical of apoptosis in these cells on Fas ligation (D'Souza et al., 1996). Together, these findings suggest that mature adult hOLs have an indolent caspase-dependent apoptotic programme which is not activated by death domain receptors. The lack of caspase activation by mature hOLs on TNF treatment might be the result of caspase inactivation by endogenous caspase inhibitors. Several caspase inhibitors, such as FLIP (Kataoka et al., 1998) and IAP family (Garcia-Calvo et al., 1998), have been characterized. We have found that, on treatment of hOLs with TNF, FLIP expression was upregulated, indicating ligand-specific induction of caspase 8 inactivation. Recently, it was suggested that TNF-induced FLIP activation and promotion of cell survival depends on the generation of complex I and the rapid activation of NF-kB (Micheau and Tschopp, 2003).

Although caspase-mediated apoptosis of different cell types is the principal programme of cell death induced by TNF (Juo et al.,1998; Yeh et al., 1998), recent data suggest that TNF family members can also induce cell death independently of caspase activation (Holler et al., 2000). On the basis of morphological features of dying cells, three subclasses of TNF-induced death can be distinguished: classical apoptosis, which is dependent on caspase activation; apoptosis-like cell death; and necrosis-like cell death (Jaattela and Tschopp, 2003). Apoptosis-like cell death is mediated by AIF, endonuclease G or cathepsins (Susin et al., 1999, 2000; Foghsgaard et al., 2001; Wang et al., 2002), whereas necrosis-like cell death depends on RIP activation and ROS (Schultze-Osthoff et al., 1992; Vercammen et al., 1998; Goossens et al., 1999; Holler et al., 2000). In our studies, although TNF-treated hOLs showed early externalization of phosphatidylserine followed by increased PI uptake [suggesting features characteristic for apoptotic cell death (Vermes et al., 1995)], the molecular signalling pathways involved in apoptosis, apoptosis-like and necrosis-like cell death overlapped significantly and might be shared (Leist and Jaattela, 2001). Apoptosis-like death is defined by a less-compact, lumpy condensation of chromatin, in contrast to the compact and spherical masses at the nucleoli periphery seen during caspase-dependent cell death (Susin et al., 1999; Leist and Jaattela, 2001). In our earlier studies (Selmaj et al., 1991a), using light and electron microscopy, we described OLs dying in response to TNF as having pycnotic nuclei and a watery and degenerate cytoplasm (Selmaj et al., 1991b). Such a pattern of TNF-induced morphology suggested an apoptosis-like mechanism of hOL death, possibly with some elements of necrosis. The apoptosis-like death pathway is mediated by AIF and/or endonuclease G. AIF is a novel apoptotic effector molecule belonging to a group of mitochondrion-derived apoptotic factors (Susin et al., 1999). AIF is released from mitochondria on apoptotic stimulation and contrary to caspase 3-activated DNase, mediates large-scale DNA fragmentation of 50 kb fragments. AIF-driven proapoptotic activity on most occasions does not involve caspase activation (Susin et al., 2000), although it was recently suggested that endonuclease G and AIF release from mitochondria might require caspase activation downstream of Bax/Bak-mediated mitochondrial permeabilization (Arnoult et al., 2003). However, since the upstream signalling pathways to Bcl-2 members are numerous and largely unknown, it is not clear at this moment whether caspase-dependent AIF release can be coupled to death receptor activation. By demonstrating the characteristic large-scale DNA fragmentation and translocation of AIF into nucleus in dying hOLs on treatment with TNF, and by preventing TNF-induced hOL death with the inhibition of AIF using an antisense construct, we have proven that AIF is critically involved in TNF-induced hOL death. Similar to CAD, endonuclease G can digest nuclear chromatin to produce oligonucleosomal fragments, thus morphologically distinguishing its activity from AIF (Samejima et al., 2001). Since AIF also mediates delayed neuronal death induced by ischaemia (Cregan et al., 2002), and excitotoxicity (Yu et al., 2002), it may be an important participant in cell death within the CNS. It was also shown recently that AIF, but not caspases, is involved in NO-induced death of developing OLs (Baud et al., 2004). Using multiple sclerosis brain tissue we were able to demonstrate that some OLs within the edge of multiple sclerosis lesions showed AIF translocation into nuclei. Thus these results provide relevance of our in vitro findings to the mechanism of OL demise in multiple sclerosis.

Caspase-independent apoptosis may also be mediated by the calpain family of calcium-dependent proteases (Villa et al., 1998), and serine protease, AP 24, which has the capacity to activate DNA fragmentation (Wright et al., 1997). It has been shown that calpains and serine proteases may be involved in apoptosis induced by TNF which is not caspase dependent. However, in the present work, calpain inhibitor and a serine protease inhibitor were not able to prevent TNF-induced hOL death. In addition, cathepsins, derived from lysosome, can participate in caspase-independent cell death induced by ligation of death receptors. Using two inhibitors of cathepsin activity, we have clearly shown that these molecules are not involved in TNF-induced hOL death.

In a previous study (Jurewicz et al., 2003), we showed that an isoform of c-Jun, NH2-terminal kinase 3 (JNK3), is critically involved in TNF-induced hOL death. We demonstrated that JNK involvement in TNF-induced death of hOLs was dependent on mitochondrial dysfunction by demonstrating that an MKK4 dominant-negative mutant, MKK4-DN, suppressed mitochondrial membrane permeability. Consistent with this conclusion were findings showing that downregulation of JNK by NF-kB, involving induction of Gadd45b, inhibits mitochondrial depolarization in mouse embryonal fibroblasts (De Smaele et al., 2001). Thus, TNF-induced JNK3 activation occurs prior to an increase in mitochondrial membrane permeability and might contribute to the release of mitochondria-derived apoptotic factors, including AIF. Ongoing studies will directly address potential interactions between JNK3 activation and AIF release from mitochondria.

In conclusion, we have shown that TNF-induced death of mature adult hOLs is not caspase dependent, is an apoptotic-like process and appears to require AIF activation. These novel findings might be of importance in the design of new molecular tools to prevent OL death in CNS demyelinating diseases, such as multiple sclerosis.


We thank Prof. Janusz Blasiak and Dr Ewa Gloc, University of Lodz, Poland, for their extended help with pulsed field electrophoresis. We thank Dr Marja Jaattela for cathepsin inhibitors, Dr Grazyna Galazka and Dr Bozena Szymanska, Laboratory of Neuroimmunology, Medical University of Lodz, Poland, for help with plasmid construction. This work was supported by KBN grants 3P05A 03123, 3PO5A 04425, MU grant 502-11-259, NS 11920, NS 08952 and NMSS 1001-J-10 and was also supported by Center of Excellence.


View Abstract