Mistletoe viscotoxins increase natural killer cell-mediated cytotoxicity Julie Tabiasco 1 , Fre ´ de ´ ric Pont 2 , Jean-Jacques Fournie ´ 1 and Alain Vercellone 1 1 Institut National de la Sante ´ et de la Recherche Me ´ dicale U563 and 2 Service de spectrome ´ trie de masse de l¢ IFR 30, CHU Purpan, BP3028, Toulouse, France Mistletoe extracts have immunomodulatory activity. We show that nontoxic concentrations of Viscum album extracts increase natural killer (NK) cell-mediated killing of tumor cells but spare nontarget cells from NK lysis. The com- pounds responsible for this bioactivity were isolated from mistletoe and characterized. They have low molecular mass and are thermostable and protease-resistant. After complete purification by HPLC, they were identified by tandem MS as viscotoxins A1, A2 and A3 (VTA1, VTA2 and VTA3, respectively). Whereas micromolar concentrations of these viscotoxins are cytotoxic to the targets, the bioactivity with respect to NK lysis is within the nanomolar range and differs between viscotoxin isoforms: VTA1 (85 n M ), VTA2 (18 n M ) and VTA3 (8 n M ). Microphysiometry and assays of cell killing indicate that, within such nontoxic concentrations, viscotoxins do not activate NK cells, but act on cell conju- gates to increase the resulting lysis. Keywords: mistletoe; natural killer cells; tumor cells; visco- toxin; Viscum album. Mistletoe preparations have been used for pharmacological purposes since ancient times [1]. These days, industrially produced extracts from mistletoe are used in treatments of solid tumors [2–5]. It is thought that the molecular basis of the antitumoral activity of mistletoe lies in several distinct bioactivities. First, its lectin content is responsible for direct toxicity to tumor cells [6–9]. Secondly, the Viscum album rhamnogalacturonan oligosaccharide favors bridging of natural killer (NK)–tumor cell conjugates, enhancing effi- ciency of killing [10–15]. Thirdly, it has been found that the antitumoral human cytotoxic T lymphocytes with cd Tcell receptor are selectively activated by mistletoe ligands of phosphoantigen structure [16,17]. NK cells play an important role in antitumoral immunity as they directly kill tumor cells and regulate the adaptative immunity [18]. Activation of cytolytic functions of NK cells relies mainly on selective interactions of NK receptors with major histocompatibility complex (MHC) class I-related molecules on the tumoral cells. Distinct ligands, however, may also induce NK cell stimulation [19–23]. Hence, immunotherapeutic modulation of NK cell activation is an important issue in current antitumoral approaches. We analysed the bioactivity of V. album compounds in the in vitro killing of tumor cells by human NK cells and found that the components that mediate this bioactivity are viscotoxins. MATERIALS AND METHODS Preparation of V. album (Va) extract The green and white parts of mistletoe (V. album L., Viscaceae; 1 kg) freshly collected on Robinia pseudacacia L. were crushed and extracted twice with 5 L methanol/water (1 : 1, v/v). After filtration and volume reduction to 600 mL, the aqueous phase was successively partitioned with cyclohexane, dichloromethane and ethyl acetate. This aqueous phase was positive in tests of enhancement of NK-mediated killing of tumor cells, but was directly toxic for the tumor target cell line. Ethanol was added to the concentrated aqueous phase to achieve 85% (v/v) concen- tration. A precipitate was obtained and separated from the supernatant by centrifugation (2000 g; 10 min). The super- natant was concentrated and ethanol was added to 85% (v/v). After centrifugation, the precipitate was dissolved in water, pooled with the former precipitate, and constituted the Va extract. The final yield of Va extract was 50 g from 1 kg plant extracted. The stock solution of Va extract (237 mgÆmL )1 ) was stored at )20 °C. Viscotoxins The protocol for purifying viscotoxins was modified from a previous method [24,25]. Briefly, to fractionate Va extract, we replaced ion-exchange and exclusion chromatography with C 18 reverse-phase open column chromatography to avoid the use of salt solutions. A 4-g portion of Va extract was applied to a column (2.5 · 20 cm) of Lichroprep RP-18 (Merck, Darmstadt, Germany), which was irrigated with 300 mL 20%, 40% and 100% acetonitrile in 0.1% aqueous acetic acid. The former 40% acetonitrile eluate was freeze- dried, dissolved in 20% acetonitrile in 0.1% trifluoroacetic acid and chromatographed by C 18 reverse-phase HPLC (Nucleosil 5 lm; 300 A ˚ pore size; Bischoff, Leonberg, Germany). The column (250 · 4.6 mm) was eluted at Correspondence to: A. Vercellone, INSERM U395, CHU Purpan, BP 3028, 31024 Toulouse, France. Fax: 335 6274 8386, Tel.: 335 6274 8364, E-mail: vercello@toulouse.inserm.fr Abbreviations: MHC, major histocompatibility complex; MS n , multiple-stage MS; NK, natural killer; VTA, viscotoxin A; TNF-a, tumor necrosis factor a. (Received 28 November 2001, revised 27 March 2002, accepted 15 April 2002) Eur. J. Biochem. 269, 2591–2600 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.02932.x 1mLÆmin )1 by a linear gradient from 20% to 50% of acetonitrile in 0.1% trifluoroacetic acid over 30 min. For final purification of VTA2 and VTA3, the elution was carried out with the following gradient: 25% solvent B (acetonitrile in 0.1% trifluoroacetic acid) and 75% solvent A (0.1% aqueous trifluoroacetic acid) during the first 10 min, 1 min up to 30% solvent B, 9 min with 30% solvent B, and from 30% up to 40% over 5 min. Cation-exchange chromatography was conduced on a polypore SP 10 micron (100 · 2.1 mm) column (Applied Biosystems) using linear gradient elution at 0.3 mLÆmin )1 from 100% solvent A (50 m M phosphate buffer, pH 7) to 100% solvent B (1 M NaCl in 50 m M phosphate buffer, pH 9) in 30 min. The VTA3 standard was kindly provided by K. Urech and purified as described by Schaller et al.[24]. Chemical and enzymatic treatments Dilution in organic solvent was achieved by addition of pure acetonitrile to 80% final volume. After 2 h at room temperature, organic solvent was removed by evaporation with a Speed-vac centrifuge before further bioassays. For chemical treatments, 1 vol. Va extract or purified viscotoxin was mixed with 1 vol. 4 M NaOH or 4 M HCl and incubated for 2 h at 37 °C. After neutralization with HCl or NaOH, the samples were immediately diluted in RPMI medium supplemented with 10% human serum, and pH was adjusted to 7.0 before further bioassays. For periodate oxidation, sodium periodate (Sigma) was added (final concentration 5 m M ) to the Va extract or to purified viscotoxin, and samples were left for 2 h at room tempera- ture. Unchanged sodium periodate was further neutralized by the addition of a few drops of glycerol to the sample. Reduction and alkylation were performed as described previously [26]. Excess reagent was removed by purification on a tC 18 September–PakÒ cartridge (Walters, Milford, MA, USA) eluted successively with increasing percentages of acetonitrile. Enzymatic treatments consisted of incubating Va extract for 2 h at 37 °C with proteinase K (1.5 UÆmL )1 ; Boehringer, Mannheim, Germany), calf alkaline phosphatase (1 UÆmL )1 ; Boehringer), or sulfatase (16.8 UÆmL )1 Aeromonas aerogenes sulfohydrolase; Sigma) in 10 m M Tris/HCl, pH 7.2. Heating at 75 °C for 10 min stopped enzymatic reactions before further bioassay. SDS/PAGE SDS/PAGE analysis of Va extract and viscotoxins was performed using 7 · 10 cm gels of 20% acrylamide (37.5 : 1 ratio of acrylamide to bisacrylamide) for the resolving gel and 5% acrylamide for the stacking gel. Samples were electrophoresed at 75 V, and further stained by Coomassie Blue or silver nitrate. Mass spectrometry MS was performed using the LCQ Ion-trap mass spectro- meter (Thermo-Finnigan, San Jose ´ , CA, USA) by liquid chromatography/MS and nanospray as described [27]. The viscotoxin masses were obtained by spectral deconvolution with Bioworks software from the Xcalibur 1.2 suite (Thermo-Finnigan). Multiple stage MS (MS n ) sequencing of viscotoxins was carried out on the underivatized sample. The peak at m/z 966.5 (i.e. z ¼ +5forVTA2forwhich the mass is ¼ 4827 Da) was selected in full-scan MS and fragmented. In its MS 2 spectrum, Y ions corresponding to the disulfide-free C-terminal moiety of VTA2 were selected and fragmented by MS 3 . Fragmentation of the Y 6 ion (m/z 706.3) led to overlapping sets of Y and b ions identifying the six C-terminal amino acids of VTA2. The precision of this experiment did not allow lysine to be distinguished from glutamine. Cell lines and cultures Peripheral blood lymphocytes were obtained from hepari- nized venous blood of healthy volunteers using Ficoll-Paque (Amersham Pharmacia Biotech AB, Uppsala, Sweden). Fresh NK cell populations were obtained by peripheral blood lymphocyte depletion of non-NK cells using the NK Cell Isolation Kit (Miltenyi Biotec GmbH, Bergisch- Gladbach, Germany). The phenotype of isolated cells was analyzed by flow cytometry with CD16 mAb and CD56 mAb (clone 3G8 and N901-NKH1, respectively; Immuno- tech-Beckman-Coulter, Marseilles, France). The resulting NK population typically comprised more than 85% of CD56 cells and less than 0.5% CD3 cells. The different human cell targets (K562, an MHC class I-deficient mono- myelocytic tumor [28]; Daudi, a b2m – Burkitt’s lymphoma cell line [29]; Val, a non-Hodgkin B cell lymphoma [30]; C1R, a lymphoblastoid tumor expressing only the HLA class I allele Cw 0401 [31]; and C1R-B27 [32]) were maintained in culture in complete RPMI 1640 medium with Glutamax-I supplemented with 10% fetal calf serum (Gibco–BRL, Life Technologies, Cergy Pontoise, France). ThemurineFccR + mastocytoma P815 cell line was maintained in complete Dulbecco’s modified Eagle’s culture medium (Sigma Aldrich, St Louis, MO, USA) supplemen- ted with 5% human serum. The human interleukin-2- dependent NK cell lines NKL [33] and NK-92 [34] (kindly provided by E. Vivier, CIML, Marseilles, France) were maintained in RPMI 1640 medium with Glutamax-I (Gibco–BRL) supplemented, respectively, with 10% pooled human AB serum plus recombinant interleukin-2 (100 UÆmL )1 ; Chiron) and 10% fetal calf serum plus recombinant interleukin-2 (200 UÆmL )1 ). The cd cell line has been described previously [27]. All the culture media were complemented with 100 UÆmL )1 penicillin, 100 UÆmL )1 streptomycin and 1m M sodium pyruvate (all from Gibco–BRL). Bioactivity assays The cytolytic activity was assessed in a 4-h 51 Cr-release assay in which effector cells (NKL, NK-92, freshly isolated NK or cd cytotoxic T lymphocytes) were mixed with different target cells. Lysis assays were carried out with or without Va fractions added at different concentrations. Briefly, 10 6 target cells were labeled with 100 lCi sodium [ 51 Cr]bichro- mate (10 mCiÆmL )1 ;ICN)at37°C for 1 h. The cells were then washed three times with RPMI and added to the effector cells at 3 · 10 3 cells/well in 96-well round-bottom microplates, resulting in an effector to target cell (E/T) ratio ranging from 30 : 1 to 3 : 1 in a final volume of 0.2 mL in each well. For the redirected killing assay, target cells were 2592 J. Tabiasco et al.(Eur. J. Biochem. 269) Ó FEBS 2002 the (FccR + ) P815 cell line, and effector cells the CD16 + NKL cell line, using an E/T ratio of 3 : 1. The assays were carried out with or without 2 lgÆmL )1 anti-CD16 IgG1 (clone 3G8). After 4 h of incubation at 37 °C, 100 lL supernatant was harvested and counted in the gamma counter. The percentage specific 51 Cr release was deter- mined from the relation: [(experimental c.p.m. ) spontaneous c.p.m.)/(total c.p.m. incorporated ) spontaneous c.p.m.)] · 100 All determinations were performed in triplicate in each assay, and results shown are the mean ± SEM from triplicate determinations of a representative experiment out of five independent experiments (on average). For clarity, SEMs are not represented on most of the figures because they were below 3% of specific lysis. Release of tumor necrosis factor a (TNF-a)wasmeas- uredbyabioassayusingTNF-a-sensitive cells (WEHI- 13VAR, ATCC CRL-2148). Briefly, 5 · 10 4 NKL cells per well were incubated for 24 h at 37 °C with or without various concentrations of each viscotoxin in 100 lLculture medium. A 50-lL portion of supernatant was then added to 50 lL WEHI cells plated at 3 · 10 4 cells per well in culture medium containing actinomycin D (2 lgÆmL )1 )andLiCl (40 m M ). WEHI cells were incubated for 20 h at 37 °C. Viability of WEHI cells was then measured with a 3- (4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (Sigma) assay. Levels of TNF-a releasewerethencalculated from a standard curve obtained using purified human recombinant TNF-a (PeproTech, Inc., Rocky Hill, NJ, USA). Typically, unstimulated NKL cells release 5±2ngÆmL )1 TNF-a compared with 60 ± 5 ngÆmL )1 when induced by phorbol myristate acetate (4 ngÆmL )1 )and ionomycin (500 ngÆmL )1 ). To probe for soluble cytolytic factors released by NK cells during the lysis assay, 9 · 10 4 NK cells were mixed with 3000 K562 cells and incubated with viscotoxins for 4 h at 37 °C. A 100-lL portion of supernatant was then added to 3000 51 Cr-labeled K562 cells plated in a 96-well round- bottom microplate. After 4 h incubation at 37 °C, 100 lL supernatant was collected and counted in a gamma counter. The percentage of specific 51 Cr release was determined as for the cytotoxicity assay. RESULTS Nontoxic concentrations of Va extract enhance NK cell-mediated and antibody-redirected lysis A crude hydrosoluble extract was prepared from crushed mistletoe leaves, and added to in vitro lysis assays with K562 tumor targets and various NK effector cells. As indicated by the spontaneous release of 51 Cr from target cells alone, Va extract diluted 1 : 5000 was not directly toxic for these targets. On the other hand, Va extract increased K562 lysis by freshly isolated NK cells or by a strongly cytolytic cd T cell line [35]. NK cell lysis in the presence of Va extract was Fig. 1. V. album extract (VaE) enhances cytolytic activity of different killer cells for various target cells. (A) Lysis of K562 target cells by the following effector cells: fresh NK cells, cd cytotoxic T lymphocytes, NKL, and NK-92 titrated at different E/T ratios, in the presence (d) or absence (s)ofVaextract diluted 1 : 5000 in all assays. Shown on the right of each lysis assay are the c.p.m. of spontaneous 51 Cr release by the target cells alone (target spontaneous release) incubated under the same conditions in the presence (filled bars) or absence (empty bars) of Va extract diluted 1 : 5000 in all assays. Each result represents the mean of triplicate experi- ments. (B) NKL effectors were tested for kill- ing of the target cells lines, C1R, C1R-B27, Daudi and Val, at different E/T ratios, in the presence (d) or absence (s)ofVaextract diluted 1 : 5000. Shown on the left of each cytotoxic experiment are the c.p.m. of spon- taneous 51 Cr release by the target cells alone as above. (C) NKL cells were tested in a redi- recting killing assay against the FccR + P815 target cells. These assays were carried out in the absence or presence of CD16 mAb (2 lgÆmL )1 ) and with or without Va extract (VaE) diluted 1 : 5000. Each result represents the mean of triplicate experiments at an E/T ratio of 3 : 1. Ó FEBS 2002 Viscotoxins enhance NK cell-mediated killing (Eur. J. Biochem. 269) 2593 equivalent to four times the number of killer cells when present alone (Fig. 1A). Similar results were found with two other NK cell lines, NK-92 and NKL. Thus nontoxic concentrations of Va extract increased NK lysis. NK cells eliminate tumor cells in the absence of appropriate interactions between their inhibitory receptors and MHC class I molecules on the surface of the target cells. Thus we investigated whether Va extract would alter the lysis of target cells other than K562 by the cell line NKL. Va extract also enhanced killing of the HLA class I-deficient targets C1R and Daudi (Fig. 1B). However, Va extract did not trigger killing of the HLA class I cells Val or C1R-B27, a C1R cell line transfected with the HLA-B27 allele protecting against lysis by NKL cells [33]. NK cells may also lyse IgG-bound tumor cells [36], so we investigated whether Va extract was also bioactive in this case. Lysis of the FccRIII + cells P815 by NKL cells in the presence of CD16 mAb [37] was measured with and without Va extract. Again, Va extract was not directly toxic to P815 cells but enhanced its antibody-redirected killing (Fig. 1C). Therefore, the bioactivity of this Va extract appears to be independent of the NK-lysis activation pathway. Bioactivity-guided isolation of thermostable and low-molecular-mass proteins from Va extract Bioactivity of Va extract in NKL cells could be titrated in lysis assays by varying the E/T ratio and Va extract concentrations (Fig. 2A). This helped to guide the isolation of the molecule(s) responsible. After various treatments of Va extract, titration of the remaining bioactivity indicated that the bioactive component(s) were resistant to heating, acid, protease, phosphatase and sodium periodate (Ta- ble 1). This suggested that the molecule was thermostable, although presumably neither a protein nor a carbohydrate, despite their relative abundance (carbohydrate content 10 mgÆmL )1 ; data not shown) in Va extract revealed by silver nitrate and Coomassie Blue staining of SDS/poly- acrylamide gels (Fig. 2B). However, bioactivity in the Va extract was degraded by alkali or a reduction–alkylation treatment that targeted disulfide bridges (Table 1). Further isolation of the NK-bioactive compound(s) was based on HPLC with macroporous reverse-phase support (Fig. 2C). Bioassay-guided fractionation of the Va extract indicated that the bioactivity was due to distinct UV-absorbing peaks split into two consecutive HPLC fractions. The HPLC fractions 14–17 min and 17–20 min increased the NKL- mediated lysis of K562 cells (Fig. 2C). With regard to the specific NK lysis, no other bioactive fraction could be recovered by this procedure. Both absorbance characteris- tics of the active fractions (not shown) and their appearance on SDS/PAGE (Fig. 2B) revealed that they comprised small protein(s) with estimated molecular mass of 5 kDa. The only Va extract molecules bioactive in NK lysis are VTA1-3 On the one hand, proteins were unexpected for the unusual thermostability and protease-insensitivity of the bioactive extract, but, on the other hand, they could account for the Fig. 2. NK cell lysis bioassay-guided isolation of viscotoxins. (A) Cytolytic activity of NKL cells against K562 target cells with Va extract diluted 1 : 5000 (j), 1 : 15 000 (m), 1 : 45 000 (e) or without Va extract (s). (B) SDS/ PAGE of Va extract and fraction 17–20 min (VT)stainedwitheitherAgNO 3 or Coomassie Blue. (C) RP-HPLC separation of bioactive fractions from the Va extract monitored for total ion current and for bioactivity on NKL cells. Fractions were collected every 3 min, lyophilized, and tested in the NKL/K562 killing assay as above. Online coupling of HPLC to ion trap electrospray ionization MS was used to identify viscotoxins at the specified peaks. 2594 J. Tabiasco et al.(Eur. J. Biochem. 269) Ó FEBS 2002 sensitivity to disulfide-specific reaction (Table 1). To unam- biguously identify these molecules, we used online coupling of the above HPLC separation with ion-trap MS detection (liquid chromatography/MS). Using positive-mode electro- spray, analysis of the HPLC peaks gave a series of multicharged ion species (not shown) that were further deconvoluted in the 2000–6000 mass range. The deconvo- lution spectrum of the major peak from bioactive fraction 17–20 min gave two molecular species of 4827.0 and 4925.0 Da (Fig. 3A). They correspond, respectively, to the expected molecular mass of VTA2 (theoretical mass 4828.4) and to a noncovalent phosphate adduct of VTA2 [25]. Similar analysis of the other bioactive HPLC peaks identified VTA1 (mass 4884.0) and VTA3 (mass 4830.0) (Fig. 2C and [25]). This conclusion was checked by partial peptide sequencing using MS n . As the N-terminal moiety of native viscotoxins involves a disulfide bridge [25], sequential fragmentation of each C-terminal moiety was performed to confirm these assignments. From native VTA2, we selected byMS 2 two ions belonging to the Y series (m/z 407, 706.3) (data not shown). Its fragmentation into the expected Y-type and b-type fragments characterized the C-terminal PSDYPK hexapeptide sequence (Fig. 3B). Similar experi- ments established the identity of VTA1 and VTA3 (data not shown). The HPLC fraction eluted at 17–20 min was dried and weighed, and, although its bioactivity in NK-mediated killing was clear-cut, it still comprised a mixture of distinct viscotoxins. The same reverse-phase HPLC procedure was repeated, but using a slower elution sequence. This resulted in resolution into three peaks corresponding to each of the three distinct viscotoxins. Each of the separated viscotoxins was then individually controlled chromatographically by reverse-phase (Fig. 4A) and cation-exchange (Fig. 4B) chromatography. Parallel bioactivity monitoring of these runs revealed that NK lysis activity was recovered with each of the purified viscotoxin peaks. A molecular mass of 4884 Da identified VTA1 in the first bioactive fraction (not shown), a molecular mass of 4827 Da identified VTA2 in the second bioactive fraction (Fig. 4A), and a molecular mass of 4830 Da (and 4928 Da for the phosphate adduct) identified VTA3 in the third bioactive fraction (Fig. 4A). The bioactivities of VTA1–3 were destroyed by alkali and reduction–alkylation of disulfide bridges, but not by heat or periodate, as described above for Va extract (Table 1). So, this profile establishes the involvement of VTA1–3 as the only components of Va extract that enhance NK lysis. Moreover, the fact that polymyxin B (10 lgÆmL )1 ) did not affect the bioactivity of the purified viscotoxins ruled out any contribution of traces of lipopolysaccharide to the observed effect (not shown). Table 1. NK bioactivity of Va extract and purified viscotoxins after chemical and enzymatic treatments. The NK lysis-increasing bioactivity was normalized to 100% using the largest increase in lytic response in the presence of untreated Va extract (here + 22% of specific K562 lysis), VTA1 (here + 16%) or VTA2 (here + 20%). For comparison, the NK lysis-increasing bioactivity of chemically treated extracts is expressed as a percentage of the reference bioactivity. Va extract, VTA1 and VTA2 were used at 1 : 5000 dilution, 100 n M and 40 n M , respectively, which had no direct toxicity for K562 cells. NT, Not tested; DTT, dithiothreitol; IEt, iodoethanol. Treatment NK bioactivity (%) Va extract VTA1 VTA2 None 100 100 100 Heat (75 °C, 2 h) 91 74 88 Dilution in organic solvent (80% acetonitrile) 73 100 100 Acid (2 M HCl, 2 h, 37 °C) 67 NT NT Alkali (2 M NaOH, 2 h, 37 °C) 13 0 0 Periodate oxidation (5 m M NaIO 4 , 2 h, 20 °C) 100 91 120 Reduction + alkylation (DTT, IEt) 000 Proteinase K 90 NT NT Alkaline phosphatase 77 NT NT Sulfatase 100 NT NT Fig. 3. Biochemical identification of the visco- toxin A2. (A) Deconvoluted spectrum from the electrospray ionization MS analysis of the VTA2 peak corresponds to the expected mass of VTA2. The amino-acid sequence with disulfide bridges of VTA2 is shown on the right [25]. (B) MS 3 spectrum of the ion at m/z 706.3 generated from pentacharged VTA2 (m/z 966.6). The sequence of the C-terminal moiety and assignment of the fragmentation series are indicated. Ó FEBS 2002 Viscotoxins enhance NK cell-mediated killing (Eur. J. Biochem. 269) 2595 Using SDS/PAGE examination of serial Va extract dilutions, we estimated the total viscotoxin concentration in the Va extract to be  3m M . This estimate is in concor- dance with the respective bioactivities measured above for purified viscotoxin and Va extract. Each purified viscotoxin species was quantified by analytical HPLC, and its bioactivity in the NKL/K562 assay titrated by serial dilutions. As each of the viscotoxin samples produced similar increases in NK lysis, we compared the three viscotoxins by determining their respective EC 50 values, i.e. the VTA concentration produ- cing half-maximal increase in NK-mediated lysis. These EC 50 values are significantly different: 85 ± 6.4 n M for VTA1, 18 ± 4.2 n M forVTA2and8±3.0n M for VTA3 (two-by-two comparison, highest two-tailed P value for statistical difference equalled 0.0284). These findings matched that of an authentic VTA3 standard [24], which enhanced NK-mediated lysis with EC 50 ¼ 6.5 ± 3.5 n M without direct cytotoxicity within the 1–100 n M range (Fig. 5). This reference VTA3 EC 50 value was not signifi- cantly different from that of the VTA3 (P ¼ 0.6031) purified in this study. Together these data confirm that viscotoxins are the Va extract compounds responsible for the increase in NK cell-mediated cytotoxicity. VTA2 acts on NK–target cell conjugates The interaction between NK and target cells successively involves the formation of conjugates, the specific recogni- tion phase, and the lethal hit delivery. Afterwards, the killer cells are recycled and may serially kill several targets [38]. Because the action of viscotoxin was found to be independ- ent of the NK activation pathway (Fig. 1), we investigated Fig. 5. Effect of a VTA3 standard on NK-mediated lysis in the absence of direct cytotoxicity. Lysis of K562 target cells by NK cells at E/T ratio of 1 : 1 in the absence (0) or presence of pure VTA3 standard (kindly provided by K. Urech [24]) added in the concentration range 0–100 n M to the culture wells (up). Toxicity of VTA3 for the target cells alone was titrated over the same concentration range as above by measure- ment of the direct 51 Cr release from pulsed K562 cells (down). Each result represents the mean of triplicate experiments. Fig. 4. Biochemical characterization of the purified viscotoxins A1, A2 and A3. (A) RP-HPLC and (B) cation-exchange HPLC of purified viscotoxins A2 and A3 monitored for absorbance at 220 nm and for bioactivity on NKL cells. Fractions delimited by tick marks were collected, lyophilized, and tested in the NKL/K562 killing assay. The deconvoluted mass spectrum of each purified viscotoxin is shownintheinset. 2596 J. Tabiasco et al.(Eur. J. Biochem. 269) Ó FEBS 2002 whether a viscotoxin pulse of either NKL effector or K562 target cells before the assay could also lead to this bioactivity. Neither VTA2-pulsed killers nor VTA2-pulsed targets alone reproduced the bioactivity of viscotoxins in the usual lytic assay (Fig. 6A). Therefore, to exert its bioactiv- ity, viscotoxin must be present during the killing assay. Moreover, viscotoxins did not induce production of soluble toxic factors by NKL cells during the 4-h incubation with target cells (data not shown; see Materials and methods). These results indicate that VTA2 acts on NK–target cell conjugates. From cell mixing to final cell harvest, these cytotoxicity assays span 4 h, so we determined viscotoxin bioactivity when added at various time points. The optimum for VTA2 bioactivity was when it was added  30–45 min after cell contact (Fig. 6B). When compared with other drugs targeting NK activation [39], this time-course suggests that VTA2 affected an event subsequent to conjugate formation, in agreement with the former conclusion. VTA2 does not activate NK cells NK–target cell conjugation precedes NK cell activation, which follows a complex array of activatory and inhibitory receptor–ligand interactions. In hematopoietic cells, intra- cellular transduction of activation signals involves changes in metabolic rates, which may be detected by micro- physiometry [27]. As Va extract acts on both redirected and natural lysis by NK cells (Fig. 1), its effects appear to be independent of the NK activation pathway. Therefore, viscotoxins were not expected to directly activate NK cells. To address this issue, we tested microphysiometric responses of NKL cells to VTA2. Although a short pulse with the mitogen combination of phorbol ester and ionomycin increased the metabolic rate of NKL cells (Fig. 6C), no such change was detected in parallel assays with NK cells exposed to various concentrations of VTA2 alone (shown for 50 n M in Fig. 6C) or to phorbol and VTA2 together (data not shown). So VTA2 does not constitute an activator signal to NK cells nor bring them a cosignal additional to phorbol esters. This conclusion was also supported by the finding that NKL cells incubated with VTA2 alone did not produce any more TNF-a than unstimulated ones (data not shown). After killer–target cell binding and NK cell activation, early intracellular events involved in NK-mediated lysis comprise polarization and delivery of cytotoxic granules at the target cell synapse. We investigated whether the Fig. 6. VTA2 acts on NK–target cell conjugates without a change in extracellular acidification rate of NK cells and killing of bystander cells. (A) Enhancement of NK cytotoxicity cannot be reproduced by a VTA2 pulse of effector or target cells before the lysis assay. The effector cells or target cells were separately preincubated with 50 n M VTA2 for 4 h and washed before running the usual cytotoxic assay without VTA2. For comparison, the same assay was run without preincubation and with or without VTA2 during the killing assay. (B) Time course of VTA2 bioactivity in the killing assay. At t ¼ 0 min, effector and target cells were mixed. At the times indicated, 50 n M VTA2 (final concentration) was added, and 4 h after mixing of the cells, specific 51 Cr release was determined. For each time point, viscotoxin bioactivity was calculated as follows: (% of specific lysis with VT) ) (% of specific lysis without VTA2). For each time point, the viscotoxin bioactivity was normalized using the largest increase found with viscotoxin in the assay (+15% at t ¼ 30 min; s). For comparison, the time course of bioactivity of ion channel blockers was plotted (d)as described by Sidell et al. [39]. (C) NKL cells in microphysiometer chambers were drained with complete medium only (n), with complete medium containing phorbol myristate acetate + ionomycin (m) or with complete medium containing 50 n M VTA2 (s). For 80 min, extracellular acidification was monitored every 90 s in each chamber, and the response was normalized to the rate value before sample introduction. (D) Bioactivity of VTA2 was tested in a three-cell lysis assay involving NKL killer, K562 target and C1R-B27 bystander cells. While the presence of bystander cells did not interfere with viscotoxin bioactivity in the killing assay (left), VTA2 did not induce any detectable killing of 51 Cr-labeled bystander cells (right). Ó FEBS 2002 Viscotoxins enhance NK cell-mediated killing (Eur. J. Biochem. 269) 2597 killing-enhancing bioactivity of VTA2 resulted from altered polarization of lytic granules in such a way as to kill bystander cells in addition to bound targets. NKL cells were simultaneously exposed to K562 target cells and C1R-B27 bystander cells (one target for one bystander mixed together). We then measured the effect of VTA2 on either lysis of 51 Cr-labeled K562 target cells in the presence of bystanders, or reciprocally, on lysis of 51 Cr-labeled C1R- B27 bystanders in the presence of K562 target cells. VTA2 was devoid of direct toxicity to the bystander cells while simultaneously being bioactive against the NK targets (Fig. 6D). So, whereas VTA2 enhances killing of K562 target cells in the presence of bystanders, it does not induce further collateral lysis and hence does not affect the polarization of the cytolytic activity of NK cells. DISCUSSION The lytic activity of human NK cells is enhanced by V. album compounds [10,40,41]. Previous studies have investigated this effect using fermented Va extracts, peripheral blood lymphocytes, and cytokine-dependent and lymphokine-dependent contexts [13,15,42]. The results presented here indicate that the lysis of K562 cells by either freshly isolated NK cells or the tumoral NK cell lines NK-92 and NKL is increased by the presence of Va extract molecules in the lytic assay. The lytic potential of killer cells in the presence of the extract was almost the same as that of four times as many killer cells alone. This effect of Va extract is not restricted to NK cells, as other types of cytotoxic T lymphocyte, such as cd T cells, that exert strong NK-like activity against thesametargetsarealsoenhancedbythesemolecules. So we investigated the V. album-mediated enhancement of killing by NK cells independently of exogenous cytokine supply. In previous studies, the V. album molecules bioactive in tumor cell killing by NK cells and monocyte/macrophages were identified as oligosaccharides (arabinogalactan [11] or rhamnogalacturonan [12,43]), which link killer and target cells [43]. This report describes the isolation and identifica- tion of VTA1, VTA2 and VTA3 as the only Va extract component of NK-cell-mediated lysis. Mistletoe produces two protein families with related biological properties: lectins and viscotoxins [44,45]. Viscotoxins are a group of highly basic cysteine-rich small proteins related to the family of thionins [46] and notoriously resistant to protein- denaturing agents [45,47]. Table 1 shows that the starting bioactivity of the Va extract has such characteristics. Other workers had already reported an uncharacterized Va- derived peptide that activated NK cells in vivo [48]. Thionins are cytotoxic for a large variety of eukaryote and proka- ryote cells [49]. This toxicity relies on positive charges of thionins interacting unspecifically with phospholipids. Thus, by passive insertion into the cell membranes, thionins permeabilize and kill tumoral cells [49,50]. Few studies have in fact elucidated the basis of viscotoxin toxicity. Using the most viscotoxin-sensitive cell line [51], micromolar concen- trations of viscotoxins are toxic for the rat renal sarcoma Yoshida [24]. More recently, VTA3 has been found to induce cell membrane stiffening and destabilization [52]. Here, the action on NK-cell-mediated killing of tumor targets was analysed using nontoxic nanomolar concentra- tions of viscotoxin. The data presented lead to clearer conclusions than earlier studies of the effect of V. album on tumor cell lysis by monocytes and NK cells. The synergistic enhancement of lysis by mistletoe arabinogalactan is mediated by cytokines secreted from killer cells after its binding to NK surface receptors [13,15]. These cells pulsed with high concentrations of arabinogalactans killed more target cells [13,15]. Our study shows a different spectrum of effects of purified viscotoxins which appear to be focused on estab- lished killer–target cell conjugates. NK cell lysis results from intracellular integration of a wide array of receptor-mediated activatory and inhibitory signals. As the absence of a lytic response could result from dominance of negative signaling by cell targets, the actual activity of viscotoxins on positive signaling pathways needed to be investigated in NK cells more directly. Activation of hematopoietic cells involves increased glyco- lysis, Na + /H + antiport, and lactate and carbonic acid excretion, which together cause medium acidification [53]. Microphysiometry monitors extracellular acidification as a real-time follow-up of metabolic rate, and detected NKL response to a mitogen combination. VTA2 alone does not alter the metabolic rate of NKL cells, so they do not provide a self-sufficient positive signal for activating NK cells. Furthermore, the action of VTA2 on killer–target cell couples spares the ÔunproductiveÕ killer-bystander ones. Therefore VTA2 does not alter the polarization of cytolytic activity against the activating target, nor does it weaken the latter in such a way as to increase their lysis in the assay. Besides the lack of viscotoxin toxicity for the various targets cells in this study, this conclusion can be drawn from the several lines of evidence discussed below. First, a pulse and rinse of targets with viscotoxins did not increase their subsequent lysis by NK cells. Secondly, viscotoxins do not cause the lysis of bystander cells exposed to viscotoxins and activated NK. Investigation of the molecular basis of viscotoxin action on NK–target cell conjugates was not within the scope of this study. However, the high bioactivity of viscotoxins (i.e. in the nanomolar range) in this assay, and the differential (8–80 n M ) of bioactivity between the structurally related VTA1–3 proteins strongly suggest an interaction with an unidentified receptor involved in NK lysis. Viscotoxins are structurally related to thionins [46], which behave as selective ion channels in nanomolar concentration ranges [50]. As functional ion channels are involved in NK-mediated lysis [54], a related role for viscotoxins is hypothesized. Long-standing medicinal virtues have been imputed to mistletoe extracts, used in the treatment of some solid tumors. Although mistletoe was thought to owe much of its bioactivity to the direct toxicity of several components to tumoral cells, this report shows that highly purified viscotoxins alone enhance the efficiency of tumor cell lysis by NK cells in the absence of any direct cytotoxicity to the target. Although the mode of viscotoxin action has not been defined, we show that viscotoxins are involved in immunologically mediated tumour cell destruction. Future studies will aim to identify the viscotoxin receptors involved in this therapeutically interesting bioactivity. 2598 J. Tabiasco et al.(Eur. J. Biochem. 269) Ó FEBS 2002 ACKNOWLEDGEMENTS We wish to thank E. Vivier for helpful discussions and provision of cell lines, and E. Espinosa for the cd cell line. We also thank Dr K. Urech for the gift of the highly purified viscotoxin A3. This work was supported by institutional grants from INSERM and Programme Eureka from ARC. REFERENCES 1. Leroi, R. (1977) Nachbehandlung des operierten MammaKar- zinoms mit Viscum album. Helv. Chim. Acta 44, 403–414. 2. Friess, H., Beger, H.G., Kunz, J., Funk, N., Schilling, M. & Buchler, M.W. (1996) Treatment of advanced pancreatic cancer with mistletoe: results of a pilot trial. Anticancer Res. 16, 915–920. 3. Heiny, B.M., Albrecht, V. & Beuth, J. (1998) Correlation of immune cell activities and beta-endorphin release in breast carci- noma patients treated with galactose-specific lectin standardized mistletoe extract. Anticancer Res. 18, 583–586. 4. Zarkovic,N.,Vukovic,T.,Loncaric,I.,Miletic,M.,Zarkovic,K., Borovic,S.,Cipak,A.,Sabolovic,S.,Konitzer,M.&Mang,S. (2001) An overview on anticancer activities of the Viscum album extract Isorel. Cancer Biother. Radiopharmacol. 16, 55–62. 5. Steuer-Vogt, M.K., Bonkowsky, V., Ambrosch, P., Scholz, M., Neiss, A., Strutz, J., Hennig, M., Lenarz, T. & Arnold, W. (2001) The effect of an adjuvant mistletoe treatment programme in resected head and neck cancer patients: a randomised controlled clinical trial. Eur. J. Cancer 37, 23–31. 6. Gabius, S. & Gabius, H.J. (1999) [Immunomodulating mistletoe therapy by lectin standardization: a double-edged sword? (see comments) ]. Versicherungsmedizin 51, 128–136. 7. Gabius, H.J., Andre, S., Kaltner, H., Siebert, H.C., von der Lieth, C.W. & Gabius, S. (1996) The mistletoe myth: claims, reality and provable perspectives. Z. Arztl. Fortbild. (Jena) 90, 103–110. 8. Baxevanis, C.N., Voutsas, I.F., Soler, M.H., Gritzapis, A.D., Tsitsilonis, O.E., Stoeva, S., Voelter, W., Arsenis, P. & Papamichail, M. (1998) Mistletoe lectin I-induced effects on human cytotoxic lymphocytes. I. Synergism with IL-2 in the induction of enhanced LAK cytotoxicity. Immunopharmacol. Immunotoxicol. 20, 355–372. 9. Valentiner, U., Pfuller, U., Baum, C. & Schumacher, U. (2002) The cytotoxic effect of mistletoe lectins I, II and III on sensitive and multidrug resistant human colon cancer cell lines in vitro. Toxicology 171, 187–199. 10. Hamprecht,K.,Handgretinger,R.,Voetsch,W.&Anderer,F.A. (1987) Mediation of human NK-activity by components in extracts of Viscum album. Int. J. Immunopharmacol. 9, 199–209. 11. Wagner, H. & Jordan, E. (1988) An immunologically active ara- binogalactan from Viscum album ÔberriesÕ. Phytochemistry 27, 2511–2517. 12. Mueller, E.A., Hamprecht, K. & Anderer, F.A. (1989) Biochemical characterization of a component in extracts of Viscum album enhancing human NK cytotoxicity. Immunopharmacology 17, 11–18. 13. Mueller, E.A. & Anderer, F.A. (1990) A Viscum album oligo- saccharide activating human natural cytotoxicity is an interferon gamma inducer. Cancer Immunol. Immunother. 32, 221–227. 14. Mueller, E.A. & Anderer, F.A. (1990) Synergistic action of a plant rhamnogalacturonan enhancing antitumor cytotoxicity of human natural killer and lymphokine-activated killer cells: chemical spe- cificityoftargetcellrecognition.Cancer Res. 50, 3646–3651 (erratum appears in Cancer Res. 50, 6115). 15. Hauer, J. & Anderer, F.A. (1993) Mechanism of stimulation of human natural killer cytotoxicity by arabinogalactan from Larix occidentalis. Cancer Immunol. Immunother. 36, 237–244. 16. Fischer, S., Scheffler, A. & Kabelitz, D. (1996) Activation of human gamma delta T-cells by heat-treated mistletoe plant extracts. Immunol. Lett. 52, 69–72. 17. Fischer, S., Scheffler, A. & Kabelitz, D. (1997) Stimulation of the specific immune system by mistletoe extracts. Anticancer Drugs 8 (Suppl. 1), S33–S37. 18. Trinchieri, G. (1989) Biology of natural killer cells. Adv. Immunol. 47, 187–376. 19. Colonna, M., Moretta, A., Vely, F. & Vivier, E. (2000) A high- resolution view of NK-cell receptors: structure and function. Immunol. Today 21, 428–431. 20. Tomasello, E., Blery, M., Vely, E. & Vivier, E. (2000) Signaling pathways engaged by NK cell receptors: double concerto for activating receptors, inhibitory receptors and NK cells. Semin. Immunol. 12, 139–147. 21. Moretta, A., Biassoni, R., Bottino, C., Mingari, M.C. & Moretta, L. (2000) Natural cytotoxicity receptors that trigger human NK-cell-mediated cytolysis. Immunol. Today 21, 228–234. 22. Moretta, L., Biassoni, R., Bottino, C., Mingari, M.C. & Moretta, A. (2000) Human NK-cell receptors. Immunol. Today 21, 420–422. 23. Lanier, L.L. (2001) On guard-activating NK receptors. Nat. Immmunol. 2, 23–27. 24. Schaller, G., Urech, K. & Giannattasio, M. (1996) Cytotoxicity of different viscotoxins and extracts from the European subspecies of Viscum album L. Phytother. Res. 10, 473–477. 25. Orru,S.,Scaloni,A.,Giannattasio,M.,Urech,K.,Pucci,P.& Schaller, G. (1997) Amino acid sequence, S-S bridge arrangement and distribution in plant tissues of thionins from Viscum album. Biol. Chem. 378, 989–996. 26. Nitti,G.,Orru,S.,Bloch,C.,Morhy,L.,Marino,G.&Pucci,P. (1995) Amino acid sequence and disulphide-bridge pattern of three gamma-thionins from Sorghum bicolor. Eur. J. Biochem. 228, 250–256. 27. Espinosa, E., Belmant, C., Pont, F., Luciani, B., Poupot, R., Romagne ´ , F., Brailly, H., Bonneville, M. & Fournie ´ , J.J. (2001) Chemical synthesis and biological activity of bromohydrin pyr- ophosphate a potent phosphoantigen stimulating human cd T cells. J. Biol. Chem. 276, 18337–18344. 28. Jondal, M. & Pross, H. (1975) Surface markers on human B and T lymphocytes. VI. Cytotoxicity against cell lines as a functional marker for lymphocyte subpopulations. Int. J. Cancer 15, 596–605. 29. Klein, E., Klein, G., Nadkarni, J.S., Nadkarni, J.J., Wigzell, H. & Clifford, P. (1968) Surface IgM-kappa specificity on a Burkitt lymphoma cell in vivo and in derived culture lines. Cancer Res. 28, 1300–1310. 30. Deweindt, C., Kerckaert, J.P., Tilly, H., Quief, S., Nguyen, V.C. & Bastard, C. (1993) Cloning of a breakpoint cluster region at band 3q27 involved in human non-Hodgkin’s lymphoma. Genes Chro- mosomes Cancer 8, 149–154. 31. Litwin, V., Gumperz, J., Parham, P., Phillips, J.H. & Lanier, L.L. (1993) Specificity of HLA class I antigen recognition by human NK clones: evidence for clonal heterogeneity, protection by self and non-self alleles, and influence of the target cell type. J. Exp. Med. 178, 1321–1336. 32. DeMars, R., Rudersdorf, R.,Chang, C., Petersen, J., Strandtmann, J., Korn, N., Sidwell, B. & Orr, H.T. (1985) Mutations that impair a posttranscriptional step in expression of HLA-A and -B anti- gens. Proc. Natl Acad. Sci. USA 82, 8183–8187. 33. Robertson, M.J., Cochran, K.J., Cameron, C., Le, J.M., Tantravahi, R. & Ritz, J. (1996) Characterization of a cell line, NKL, derived from an aggressive human natural killer cell leukemia. Exp. Hematol. 24, 406–415. 34. Gong, J.H., Maki, G. & Klingemann, H.G. (1994) Characteriza- tion of a human cell line (NK-92) with phenotypical and func- tional characteristics of activated natural killer cells. Leukemia 8, 652–658. Ó FEBS 2002 Viscotoxins enhance NK cell-mediated killing (Eur. J. Biochem. 269) 2599 35. Ferrarini, M., Heltai, S., Toninelli, E., Sabbadini, M.G., Pellicciari, C. & Manfredi, A.A. (1995) Daudi lymphoma killing triggers the programmed death of cytotoxic V gamma 9/V delta 2 T lymphocytes. J. Immunol. 154, 3704–3712. 36. Perussia, B. (2000) Signaling for cytotoxicity. Nat. Immunol. 1, 372–374. 37. Perussia, B., Trinchieri, G. & Cerottini, J.C. (1979) Functional studies of Fc receptor-bearing human lymphocytes: effect of treatment with proteolytic enzymes. J. Immunol. 123, 681–687. 38. Kupfer, A. & Singer, S.J. (1989) Cell biology of cytotoxic and helper T cell functions: immunofluorescence microscopic studies of single cells and cell couples. Annu. Rev. Immunol. 7, 309–337. 39. Sidell, N., Schlichter, L.C., Wright, S.C., Hagiwara, S. & Golub, S.H. (1986) Potassium channels in human NK cells are involved in discrete stages of the killing process. J. Immunol. 137, 1650–1658. 40. Hajto, T. (1986) Immunomodulatory effects of iscador: a Viscum album preparation. Oncology 43, 51–65. 41. Schink, M. (1997) Mistletoe therapy for human cancer: the role of the natural killer cells. Anticancer Drugs 8 (Suppl. 1), S47–S51. 42. Stein, G.M. & Berg, P.A. (1997) Mistletoe extract-induced effects on immunocompetent cells: in vitro studies. Anticancer Drugs 8 (Suppl. 1), S39–S42. 43. Mueller, E.A. & Anderer, F.A. (1990) Chemical specificity of effector cell/tumor cell bridging by a Viscum album rhamno- galacturonan enhancing cytotoxicity of human NK cells. Immunopharmacology 19, 69–77. 44. Jung, M.L., Baudino, S., Ribereau-Gayon, G. & Beck, J.P. (1990) Characterization of cytotoxic proteins from mistletoe (Viscum album L.). Cancer Lett. 51, 103–108. 45. Park, J.H., Hyun, C.K. & Shin, H.K. (1999) Cytotoxic effects of the components in heat-treated mistletoe (Viscum album). Cancer Lett. 139, 207–213. 46. Florack, D.E. & Stiekema, W.J. (1994) Thionins: properties, possible biological roles and mechanisms of action. Plant Mol. Biol. 26, 25–37. 47. Bruneton, J. (1999) Pharmacognosie.3rdedn.EditionsMe ´ dicales Internationales, Paris. 48. Kuttan, G. & Kuttan, R. (1992) Immunomodulatory activity of a peptide isolated from Viscum album extract (NSC 635, 089). Immunol. Invest. 21, 285–296. 49. Carrasco, L., Vazquez, D., Hernandez-Lucas, C., Carbonero, P. & Garcia-Olmedo, F. (1981) Thionins: plant peptides that modify membrane permeability in cultured mammalian cells. Eur. J. Biochem. 116, 185–189. 50. Hughes, P., Dennis, E., Whitecross, M., Llewellyn, D. & Gage, P. (2000) The cytotoxic plant protein, beta-purothionin, forms ion channels in lipid membranes. J. Biol. Chem. 275, 823–827. 51.Urech,K.,Schaller,G.,Ziska,P.&Giannasttasio,M.(1995) Comparative study on the cytotoxic effect of viscotoxin and mistletoe lectin on tumor cells in culture. Phytother. Res. 9, 49–55. 52. Coulon, A., Berkane, E., Sautereau, A.M., Urech, K., Rouge, P. & Lopez, A. (2002) Modes of membrane interaction of a natural cysteine-rich peptide: viscotoxin A3. Biochim. Biophys. Acta 1559, 145–159. 53. McConnell, H.M., Owicki, J.C., Parce, J.W., Miller, D.L., Baxter, G.T., Wada, H.G. & Pitchford, S. (1992) The cytosensor micro- physiometer: biological applications of silicon technology. Science 257, 1906–1912. 54. Schlichter, L., Sidell, N. & Hagiwara, S. (1986) Potassium chan- nels mediate killing by human natural killer cells. Proc. Natl. Acad. Sci. USA 83, 451–455. 2600 J. Tabiasco et al.(Eur. J. Biochem. 269) Ó FEBS 2002 . extract increased K562 lysis by freshly isolated NK cells or by a strongly cytolytic cd T cell line [35]. NK cell lysis in the presence of Va extract was Fig. 1. V. album extract (VaE) enhances cytolytic. of V. album on tumor cell lysis by monocytes and NK cells. The synergistic enhancement of lysis by mistletoe arabinogalactan is mediated by cytokines secreted from killer cells after its binding. Mistletoe viscotoxins increase natural killer cell-mediated cytotoxicity Julie Tabiasco 1 , Fre ´ de ´ ric Pont 2 , Jean-Jacques Fournie ´ 1 and