Cytotoxicity of crystals involves RIPK3-MLKL-mediated necroptosis

Crystals cause injury in numerous disorders, and induce inflammation via the NLRP3 inflammasome, however, it remains unclear how crystals induce cell death. Here we report that crystals of calcium oxalate, monosodium urate, calcium pyrophosphate dihydrate and cystine trigger caspase-independent cell death in five different cell types, which is blocked by necrostatin-1. RNA interference for receptor-interacting protein kinase 3 (RIPK3) or mixed lineage kinase domain like (MLKL), two core proteins of the necroptosis pathway, blocks crystal cytotoxicity. Consistent with this, deficiency of RIPK3 or MLKL prevents oxalate crystal-induced acute kidney injury. The related tissue inflammation drives TNF-α-related necroptosis. Also in human oxalate crystal-related acute kidney injury, dying tubular cells stain positive for phosphorylated MLKL. Furthermore, necrostatin-1 and necrosulfonamide, an inhibitor for human MLKL suppress crystal-induced cell death in human renal progenitor cells. Together, TNF-α/TNFR1, RIPK1, RIPK3 and MLKL are molecular targets to limit crystal-induced cytotoxicity, tissue injury and organ failure.

C rystals are deposits of various sizes and shapes composed of atoms, ions or biomolecules, frequently with tissue injury, inflammation and remodelling. Two mechanisms may explain this association: (I) nucleation or crystal growth from a seed crystal formed on a surface medium, for example tubular epithelial cells, urolithiasis forming at Randall's plaques, calcifications in injured tendons, damaged cartilage or atheromatous vascular lesions, (II) crystal formation itself causes tissue injury and inflammation, for example in gouty arthritis, pulmonary silicosis or asbestosis, cholesterol crystals driving atherogenesis and in oxalate, cystine or urate nephropathy. Crystals trigger tissue inflammation via the NLRP3 inflammasome-and caspase-1-mediated secretion of IL-1b and IL-18 (refs 1-5). However, crystals also exert direct cytotoxic effects leading to necrotic rather than apoptotic cell death 6,7 . It is still unknown, whether crystal deposition causes necrosis in a passivemechanical or in one of the recently identified modalities of regulated cell death [8][9][10][11][12][13][14][15][16][17][18] . We hypothesized that crystal-induced tissue injury involves a regulated form of cell death and that the identification of this pathway may reveal new targets for therapeutic intervention that limit crystal-related tissue injury and organ dysfunction. Our results show that various crystals uniformly induce necroptosis, that is, a receptor-interacting serine-threonine kinase 3-(RIPK3) and mixed lineage kinase domain-like (MLKL)-dependent form of primary necrosis in vitro, in vivo and in human disease. These data identify several molecular targets for pharmacological intervention to limit tissue injury in crystallopathies.

Results
Various crystals cause primary cellular necrosis. What is the mode of crystal-induced cell death? To address this question we first studied the cytoxic effects of calcium oxalate (CaOx), monosodium urate (MSU), calcium pyrophosphate dihydrate (CPPD) and cystine crystals on kidney epithelial cells in vitro. Transmission electron microscopy images and rhodamine-sytox stains to identify living and dead cells, respectively, showed that kidney tubular epithelial cells die by primary necrosis after crystal exposure (Fig. 1a). In addition, we employed flow cytometry to better characterize the mode of CaOx-induced cell death. To this end sideward scatter, forward scatter, Hoechst 33342, annexin V, 1,1 0 -dioctadecyl-3,3,3 0 ,3 0 -tetramethyl-indocarbocyanine perchlorate, and prodidium iodide were measured. We found that CaOx crystals induced primary necrosis without signs of apoptosis (Fig. 1b,c), evidenced by pan-caspase inhibition not affecting cytotoxicity of any of the aforementioned crystals (Fig. 1d). Also the kinetics of crystal cytotoxicity were caspase independent ( Supplementary Fig. 1A) and absence of NLRP3 did not affect CaOx crystal-induced cytotoxicity ( Supplementary  Fig. 1B). This, being a caspase-independent mode of cell death, excluded NLRP3-caspase-1/11-dependent pyroptosis 10,11 . We conclude that crystal cytotoxicity does not involve apoptosis or pyroptosis but largely represents primary cellular necrosis.

Cytotoxicity of crystals involves the necroptosis pathway.
Having excluded the contribution of caspase-dependent forms of cell death, we focussed on necroptosis, a known regulated form of necrosis of non-immune cells 13 . We found the key proteins of the necroptosis pathway, that is, tumour-necrosis factor receptor-1 (TNFR1), RIPK1, RIPK3 and MLKL to be induced in tubular epithelial cells on exposure to various crystals (Fig. 2a). The RIPK1 stabilizer necrostatin-1 (ref. 19) partially (MSU, CPPD and cystine) or entirely (CaOx) prevented crystal-induced tubular epithelial cell death also in the absence of ZVAD-FMK ( Fig. 2b and Supplementary Fig. 1C), suggesting a role for RIPK1 independent of caspases. When crystals were replaced by recombinant TNF-a as a stimulus, necrostatin-1 had the same effect. However, TNFR1 was not required for crystal-induced cell death in vitro ( Supplementary Fig. 1B). Necrostatin-1 also suppressed crystal cytotoxicity in L929 cells, primary human synovial fibroblasts and HK-2 cells (Supplementary Fig. 2). The protective effect of necrostatin-1 on the cytotoxicity of crystals was confirmed by fluorescence imaging using the cell death marker prodidium iodide that enters cells only on the disruption of the plasma membrane (Fig. 2c). Both Ripk3 or Mlkl deficiency or knockdown of either RIPK3 or MLKL with specific siRNA partially abrogated, whereas knockdown of caspase-8 somewhat enhanced crystal-induced cytotoxicity in tubular epithelial cells (Fig. 3a-c and Supplementary Fig. 3). Furthermore, pre-treatment with the putative inhibitor of RIPK3 dabrafenib also protected tubular epithelial cells from crystal-induced cytotoxicity in a dose-dependent manner (Fig. 3d). Together, these data imply that crystal-induced primary necrosis may involve RIPK3/MLKLdependent necroptosis.
Ripk3 and Mlkl deficiency blocks crystalline necrosis in vivo. Crystal nephropathy (CN) is an example of crystal-induced tissue injury and organ failure 20,21 . Mice exposure to oxalate induces oxalate nephropathy, a model that mimics CN in humans, including an increase of serum creatinine and blood urea nitrogen (BUN) 5,20,22 . Acute oxalate exposure leads to CaOx crystal deposition in the mouse kidney. At the structural level CaOx crystal plugs forms within the proximal and distal tubules ( Fig. 4a-d). CaOx crystals are also found within the interstitium (Fig. 4c). Transmission electron microscopy detected luminal crystal plugs and smaller sized crystals within tubular epithelial cells ( Fig. 4e-g). Injured epithelial cells often exhibited ultrastructural signs of necrosis, such as loss of electon density and balloning of the mitochondria, nuclear chromatin irregularities, outer membrane disruption, and cytoplasm and organelle expulsion (Fig. 4h). Quick-freeze deep-etch electron microscopy illustrates crystal deposition involving cytoplasmic organelles on rupture of plasma membranes (Fig. 4i). Thus, crystal formation within the kidney is associated with necrosis of tubular epithelial cells in a temporal and spatial manner. The phenotype is associated with the induction of renal expression of TNFR1, RIPK1, RIPK3 and MLKL messenger RNA (mRNA) (Fig. 5a). We validated the in vivo contribution of this pathway by inducing oxalate nephropathy in wild type and Ripk3-and Mlkl-deficient mice. Oxalate exposure induced identical CaOx crystal deposits in both mouse strains (Fig. 5b), but all functional and structural parameters of acute CN were significantly reduced in Ripk3-or Mlkl-deficient mice (Fig. 5c-h). These include serum creatinine levels, markers of tubule necrosis and neutrophil recruitment. These results show that the necroptosis-related proteins RIPK3 and MLKL mediate oxalate crystal-induced cell necrosis in vivo.
RIPK1 does not mediate neutrophil recruitments. We performed several in vitro and in vivo experiments to test the possibility that tubule protection in Ripk3-and Mlkl-deficient mice is a secondary effect possibly due to a role of RIPK1-RIPK3-MLKL signalling in neutrophil recruitment. Intraperitoneal (i.p.) injection of various types of crystals induced neutrophil recruitment into the peritoneal cavity as a marker of crystal-induced peritonitis, which was not affected by necrostatin-1 (Fig. 6a,b). Injection of crystals into air pouches at the back of mice with and without systemic necrostatin-1 treatment gave identical results ( Fig. 6c and Supplementary Fig. 4). Importantly, we assured the plasma activity of the same charge of necrostatin-1 in vivo in a similar time frame  by testing microvascular permeability and leucocyte extravasation during postischemic cremaster muscle injury ( Fig. 6d-f). We conclude that the necrostatin-1-mediated effects on tissue injury do not involve a direct effect on neutrophil recruitment. for example, via surface TNFR1 on tubular epithelial cells. Renal mRNA and protein levels of TNF-a and TNFR1 were induced in CaOx nephropathy at 24 h (Supplementary Fig. 5A-C). A careful analysis of early time points of oxalate nephropathy revealed that kidney injury occurred several hours before intrarenal TNF expression ( Supplementary Fig. 6). Flow cytometry revealed that TNF-a was not only expressed by intrarenal mononuclear phagocytes but also by non-immune cells ( Supplementary Fig. 5D).
Immunostaining localized TNF-a expression to tubules but also interstitial cells, while TNFR1 and -2 were mostly induced in tubular epithelial cells ( Supplementary Fig. 5E). To test for a possible functional contribution of TNFR signalling we induced CN in Tnfr1-deficient and Tnfr1/2 double-deficient mice. Oxalate exposure resulted in identical CaOx crystal deposits in all mouse strains, however, all functional and structural parameters of acute CN were significantly reduced in Tnfr1-deficient mice (Fig. 7  (a-c) Mouse tubular epithelial cells were transfected with specific small inhibitor (si) RNA for RIPK3 and MLKL or a control siRNA of scrambled sequence before being exposed to crystals of CaOx (1,000 mg ml À 1 ), MSU (500 mg ml À 1 ), CPPD (500 mg ml À 1 ) and cystine (500 mg ml À 1 ). Cell viability was assessed by MTT assay (a) and cell death was assessed quantifying PI positivity (b) and C shows representative images 24 h later. Original image magnification: Â 200, scale bar, 100 mm. (d) Mouse tubular epithelial cells were pretreated with RIPK3 inhibitor dabrafenib before exposing to different type crystals. Cell viability was assessed by MTT assay 24 h later. Data are expressed as mean ± s.e.m. of three independent experiments, and was analysed using Student's t-test. Baseline viability is set as 100%. *Po0.05, **Po0.01 and ***Po0.001 either versus control siRNA. MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; PI, prodidium iodide. Phosphorylated MLKL in human acute oxalate nephropathy.
To test validate our mouse data in human disease we first analysed crystal cytotoxicity in human renal progenitor cells, that are considered to contribute to the turnover of tubular epithelial cells and their regeneration upon injury 24 . For all the crystals tested the results were identical to the other cell types examined before Oxalate exposure leads to diffuse intrarenal crystal formation as displayed by Pizzolato stain (a), scale bar, 1 mm. In contrast to healthy control mice (b) CaOx crystals form large stone-like plugs that obstruct the tubular lumen focally (c) smaller crystals attach to the luminal surface and also appear within the tubular cell cytoplasm. Scale bar, 20 mm (d) Scanning electron microscopy view of the lumen of CaOx exposed tubules confirms disruption of the proximal tubular brush boarder. Scale bar, 10 mm. Transmission electron microscopy shows intact proximal (e) and distal tubules (f) of control mice and luminal crystal plugs and intracellular crystals in mice exposed to CaOx (g). Scale bar, 5 mm. In many tubules crystal deposition is spatially associated with tubular epithelial cell necrosis, characterized by organelle swelling, cytoplasmic oedema and rupture of the plasma membrane with nuclear expulsion ((h) and all four panels of (i)). Scale bar, 2 mm. (j) Freeze fracture electron microscopy shows sheets of CaOx crystals (arrow head) in the cytoplasm (*) of tubular epithelial cells. Scale bar, 100 nm.

Discussion
Various crystals trigger cell death and inflammation. While inflammation was traditionally thought to be secondary to tissue injury, the discovery of crystals being agonists of the NLRP3 inflammasome raises the question whether crystal cytotoxicity is a consequence of NLRP3 activation or an inflammasomeindependent event with distinct signalling pathways 1 . Crystals are NLRP3 agonists, which could imply triggering also NLRP3-caspase-1/11-mediated pyroptosis, a mode of regulated cell death occurring in infected macrophages [10][11][12] . We excluded this option as Nlrp3-deficiency or caspase blockade did not affect crystal cytotoxicity. Caspase blockade virtually also excluded apoptosis as another form of regulated cell death, which was consistent with our flow cytometry results and the results of a previously published report 6 . In contrast, we found that the cytotoxicity of crystals involves necroptosis, an injury-driven   13,14,28 . As further lines of evidence, lack of Ripk3 and Mlkl also almost entirely abrogated oxalate nephropathy in vivo. These data further support the concept that cell necrosis is the initial event of crystal-induced necroinflammation and that blocking necroptosis during the early stage of tissue injury is sufficient to prevent subsequent auto-amplification loop of inflammation, immune-mediated pathology and organ failure 29 . However, RIPK3 is also involved in NLRP3 inflammasome activation 30 , which is an important trigger of tissue inflammation in oxalate nephropathy 5 . Of note NLRP3 and ASC deficiency both lacked oxalate-induced kidney injury, and was exclusively mediated by intrarenal dendritic cells, which excludes any inflammasome signalling in non-immune kidney cells. Therefore, the in vivo phenotype of the RIPK3 knockout may also involve impaired NLRP3/ASC signalling in intrarenal dendritic cells.
Providing evidence that necroptosis is also involved in human disease is difficult because the different forms of regulated necrosis cannot be distinguished by ultrastructural morphological criteria. To resolve this issue we used immunostaining for phosphorylated MLKL, which suggests MLKL-dependent regulated necrosis also in human oxalate nephropathy.
Necrostatin-1 keeps RIPK1 in an anti-necroptotic conformation and protected from crystal-induced cytotoxicity. As RIPK1 interacts with the intracellular domain of TNFR1 via its death domain 31 and TNFR1-mediated necroptosis is considered a prototype of regulated necrosis 8 , it is tempting to speculate that RIPK1 loses its protective function against crystal-induced organ failure presumably by post-translational events, such as deubiquitination on TNFR1 ligation. Consistently, lack of TNFR1 or treatment with etanercept abrogated CN in vivo, but not crystal-induced cytotoxicity in vitro. This should relate to the lack of direct interaction between crystals and TNFR1 and the absence of TNF-a in the in vitro system (Supplementary Fig. 8).
In vivo, however, TNF-a is secreted by infiltrating immune cells, which triggers necroptosis in additional cell populations secondary to the initial direct activation by crystals, thus contributing to the auto-amplification loop of necroinflammation 29 . The selective expression of TNF-a and TNFR1 in tubules of human oxalate nephropathy implies a similar role also in human disease. How exactly crystals trigger RIPK1 activation remains uncertain. MSU crystals can induce Syk signalling in dendritic cells via biding to lipid rafts 32 . MSU crystals also affect signalling by binding to the surface receptor Clec12A but this interaction does not apply to other crystals 33 . Crystals activate the NLRP3 inflammasome via lysosomal leakage of cathepsins into the cytosol 34 . Cathepsin B was reported to inhibit necroptosis by cleaving RIPK1 (ref. 35). This could also be an avenue of crystal-induced necroptosis.
In conclusion, crystal cytotoxicity involves necroptosis as a form of regulated cell death in vitro and in vivo. In vivo, also secondary TNF-driven necroptosis adds in (Fig. 10). Consequently, the therapeutic blockade of this pathway, for ARTICLE example, with soluble TNFR2-IgG fusion protein or the RIPK1 stabilizer necrostatin-1, can prevent crystal-induced tissue necrosis and organ dysfunction. These findings imply TNFR1, RIPK1, RIPK3 and MLKL being potential therapeutic targets to limit tissue injury in crystallopathies.

Methods
Animal studies. C57BL/6N mice were procured from Charles River Laboratories (Sulzfeld, Germany). Tnfr1 À / À and Tnfr2 À / À mice were originally obtained from the Jackson Laboratories (Bar Harbour, ME) and bred under specific pathogen-free conditions. Tnfr1/2 double-deficient mice (Tnfr1,2 À / À ) were generated by cross-breeding Tnfr1 À / À and Tnfr2 À / À mice. Ripk3 À / À mice were kindly provided by V. Dixit, Genentech, USA and Mlkl À / À mice were kindly provided by J. Murphy, WEHI, Australia. Mice were housed in groups of five under specific pathogen-free conditions with unlimited access to food and water. Six-to eight-week-old male mice received a single intraperitoneal (i.p.) injection of 100 mg kg À 1 of sodium oxalate (Santa Cruz Biotechnology, USA) and 3% sodium oxalate in drinking water and kidneys were harvested after 24 h. As a therapeutic strategy, mice received a single dose of either etanercept (10 mg kg À 1 i.p., Pfizer, Germany) or necrostatin-1 (1.65 mg kg À 1 i.p., Millipore, Germany) or R-7050 (18 mg kg À 1 i.p., Santa Cruz Biotechnology, USA) before sodium oxalate injections. Blood, urine and kidneys were collected at sacrifice by cervical dislocation. Kidneys were kept at À 80°C for protein isolation and in RNA later solution at À 20°C for RNA isolation. One part of the kidney was also kept in formalin to be embedded in paraffin for histological analysis. Samples for electron microscopy were fixed in glutaraldehyde. Sample sizes were determined by G*Power (Germany). All kidney disease experimental procedures were approved by the Regierung von Oberbayern, München, Germany. For peritonitis studies, peritonitis was induced by injection of 1 mg of crystals in 0.5 ml sterile PBS. Mice received a single injection of either Nec-1 (1.65 mg kg À 1 ) or vehicle half an hour before crystals injections. After 6 h, mice were killed and peritoneal cavities were washed with 5 ml of PBS. The lavage fluids were analysed for polymorphonuclear neutrophil recruitment by fluorescence-activated cell sorting using the neutrophil marker Ly-6G (BD Biosciences, Germany).
For air pouch studies, murine air pouches were generated using standard protocols in groups of 6-weeks-old Balb/c mice (n ¼ 5). In brief, we injected 3 ml of sterile air subcutaneously into the back of anaesthetized mice to form an air pouch. An additional 2 ml of sterile air was injected into pre-existent pouch 3 days after the first injection 7 . Another 2 days later, we injected 2.5 mg crystals in PBS, 2.5 mg crystals plus 1.65 mg kg À 1 necrostatin-1 into the air pouches. Crystals of CaOx, MSU, CPPD and cystine were used in each independent experiment. After 24 h the pouch fluid was collected, and neutrophils in the air lavage were analysed by fluorescence-activated cell sorting. The studies were conducted according to guidelines determined by the Law of the Ministry of Healthcare of Ukraine, No. 281 from 1 November 2011 for the care and use of laboratory animals and were approved by the Ethics Council of the Danylo Halytsky Lviv National Medical University.
For mouse cremaster muscle IR injury assay, cremaster muscle was prepared for experiment as originally described by Baez with minor modifications 36,37 . Briefly, mice were anesthetized using intraperitoneal (i.p.) injection of ketamine/xylazine mixture. The left femoral artery was cannulated for administration of microspheres and drugs. The right cremaster muscle was exposed through a ventral incision of the scrotum. Epididymis and testicle were detached from the cremaster muscle and placed into the abdominal cavity. Throughout experimental procedure the muscle was superfused with warm-buffered saline. Olympus BX 50 upright microscope (Olympus Microscopy, Germany) equipped for stroboscopic fluorescence epiillumination microscopy was used for in vivo microsopy. For measurement of centerline blood flow velocity, green fluorescent microspheres (0.96 mm diameter, (a-c) Primary renal human progenitor cells were pretreated with either ZVAD-FMK (10 mM) and Nec-1 (100 mM) or NSA (1 mM) before being exposed to CaOx (1000 mg ml À 1 ), MSU (500 mg ml À 1 ), CPPD (500 mg ml À 1 ) and cystine (500 mg ml À 1 ). Cell viability was assessed by MTT assay (a and b) and cell death was assessed quantifying PI positivity (c) 24 h later. Data are expressed as mean ± s.e.m. of three independent experiments. Baseline viability is set as 100%. Data were analysed using Student's t-test.   : 05-4397-0). An informed consent was obtained from all subjects. Paraffin sections from human native kidney biopsies were stained with 1:100 rabbit anti p-MLKL (Abcam, USA) and anti-rabbit IgG Alexa Flur 555. All assessments were performed in a blind manner.
Electron microscopy. For scanning electron microscopy, kidney tissue were rinsed with PBS and postfixed with 1% aqueous osmium tetroxide for a total of 2.5 h. Subsequently, tissues were rinsed and dehydrated through a graded series of ethanol to absolute ethanol and critical point dried using liquid CO 2 . After mounting on stubs the specimens were sputter coated with a gold-palladium alloy for 1.5 min. Images were viewed with a Hitachi 2600 electron microscope. For transmission electron microscopy, kidneys were immersed in cold modified Karnovsky's fixative containing 3% glutaraldehyde and 1% paraformaldehyde in sodium cacodylate buffer (pH 7.4) processed as follows 5 . After overnight fixation, kidneys were postfixed in 2% osmium tetraoxide, dehydrated through graded concentrations of ethanol and embedded in EMBed-812. Ultrathin sections of kidney were cut onto formvar-coated slot grids, and subsequently stained with uranyl acetate and lead citrate. The samples were viewed with a JEOL model 1200EX electron microscope (JEOL, Tokyo, Japan).
In another experiment, mouse tubular cells (MTCs) were grown on Thincerts (Greiner bio-one, Germany), and stimulated with crystals of CaOx (1,000 mg ml À 1 ), MSU (500 mg ml À 1 ), CPPD (500 mg ml À 1 ) and cysteine (500 mg ml À 1 ) for 24 h. Cells were then fixed for 1 h at room temperature by replacing the growth medium with 3% glutaraldehyde in PBS (pH 7.2). The Thincerts were then washed three times with PBS (pH 7.2) buffer and sliced into 3-mm squares. These pieces were postfixed in 2% aqueous osmium tetroxide for 2 h and then dehydrated through an ethanol series. After three washes in 100% ethanol, the pieces were transferred to acetone, then infiltrated and embedded in Epon. Sections B70 nm thick were cut using a diamond knife, collected on copper grids, and stained sequentially with uranyl acetate (saturated in 50% ethanol) and Reynolds' lead citrate. Sections were examined and photographed by a JEOL 1200 EXII transmission electron microscope at 80 kv.
For quick-freeze deep-etch electron microscopy, 1.5 mm bread slices of non-fixed kidney or kidney sections fixed overnight in 2% glutaraldehyde in 100 mM NaCl, 30 mM Hepes, 2 mM CaCl2, pH 7.2 (NaHCa) were used after rinsing in NaHCa. Tissues were then cooled to 4°K with liquid helium mounted in a Balzers 400 vacuum evaporator fractured and etched for 2.5 min at À 104°C and rotary replicated withB3 nm platinum. The process involved fracturing the kidney sample with a cryomicrotome under high vacuum, then allowed to partially freeze dry (this is called deep-etching), then covered with a thin film of platinum that forms a replica of the fractured and etched tissue. Then the sample was allowed to thaw so that the replica could be collected 40 and viewed with transmission microscope (JEOL 1400 with attached AMT digital camera).
Multi-parameter classification of cell death by flow cytometry. Cell death was induced in MTCs by irradiation with ultraviolet light type B at 1.5 mJ cm À 2 s À 1 or by treatment with CaOx crystals. Cell death was characterized by analysing six cytofluorometric parameters (size, granularity, penicillin-streptomycin exposure, plasma membrane integrity, mitochondrial membrane potential and DNA content) in a single tube measurement. This method detects eight phenotypically different subpopulations during the development of cell death in vitro. Briefly, collected cells were incubated for 30 min at room temperature with 400 ml of freshly prepared NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10274 ARTICLE NATURE COMMUNICATIONS | 7:10274 | DOI: 10.1038/ncomms10274 | www.nature.com/naturecommunications four-colour staining solution (1.8 mg ml À 1 AxA5-FITC, 100 ng ml À 1 prodidium iodide and 10 nM DiIC1(5), 1 ng ml À 1 Hoechst 33342) in Ringer's solution and subsequently analysed. Flow cytometry was performed with a Gallios cytofluorometer (Beckman Coulter, Fullerton, USA). Excitation of FITC and prodidium iodide was at 488 nm, the FITC fluorescence was detected with the FL1 sensor (525/38 nm band pass (BP) filter), the prodidium iodide fluorescence with the FL3 sensor (620/30 nm BP), the DiIC1(5) fluorescence was excited at 638 nm and detected with the FL6 sensor (675/20 nm BP), and the Hoechst 33342 fluorescence was excited at 405 nm and detected with the FL9 sensor (430/40 nm BP). Electronic compensation was applied to reduce bleed through fluorescence. Data analysis was performed with Kaluza software version 2.0 (Beckman Coulter, Fullerton, USA). Cells were classified according to their location in the forward scatter (FSc; size) versus side scatter (SSc; granularity) dot plot and their staining patterns in the FL1 versus FL3 and FL6 versus FL9 dot plots as described elsewhere 44 . In brief, cells were considered viable, if they (V1) do not expose phosphatidylserine on their surfaces (V2) display high mitochondrial membrane potential and (V3), if they exclude propidium iodide. Apoptotic cells are defined by (A1) exposure on their surfaces of phosphatidylserine (A2) and exclusion of propidium iodide. Cells are considered necrotic, if they display (N) permeability for the propidium cation of their plasma membranes. Cells are considered necrotic, if they allow the penetration of propidium iodide. Primary necrotic cells (no signs of apoptosis) do not display nuclear hypochromicity for propidium iodide (nuclear DNA is preserved). If the nuclei show nuclear hypochromicity (apoptosis has already started) the cells were considered secondary necrotic (loss of membrane selectivity of a cell executing apoptosis).
Statistical analysis. Data are presented as mean±s.e.m. A comparison of groups was performed using paired Students t-test or one-way analysis of variance with post hoc Bonferroni's correction was used for multiple comparisons. A value of Po0.05 was considered to indicate statistical significance.