Lysosomal release of cathepsins causes ischemic damage in the rat hippocampal slice and depends on NMDA-mediated calcium influx, arachidonic acid metabolism, and free radical production


  • James A. Windelborn,

    1. Neuroscience Training Program, University of Wisconsin, Madison, Wisconsin, USA
    2. Departments of Comparative Biosciences, University of Wisconsin, Madison, Wisconsin, USA
    Search for more papers by this author
  • Peter Lipton

    1. Neuroscience Training Program, University of Wisconsin, Madison, Wisconsin, USA
    2. Departments of Physiology, University of Wisconsin, Madison, Wisconsin, USA
    Search for more papers by this author

Address correspondence and reprint requests to Peter Lipton, Professor, Department of Physiology, University of Wisconsin, 1300 University Avenue, Madison, WI 53706, USA. E-mail:


NMDA-mediated calcium entry and reactive oxygen species (ROS) production are well-recognized perpetrators of ischemic neuronal damage. The current studies show that these events lead to the release of the protein hydrolase, cathepsin B, from lysosomes 2 h following 5-min oxygen–glucose deprivation in the rat hippocampal slice. This release reflects a lysosomal membrane permeabilization (LMP) and was measured as the appearance of diffuse immunolabeled cathepsin B in the cytosol of CA1 pyramidal neurons. Necrotic neuronal damage begins after the release of cathepsins and is prevented by inhibitors of either cathepsin B or D indicating that the release of cathepsins is an important mediator of severe damage. There was an increase in superoxide levels, measured by dihydroethidium fluorescence, at the same time as LMP and reducing ROS levels with antioxidants, Trolox or N-tert-butyl-α-phenyl nitrone, blocked LMP. Both LMP and ROS production were blocked by an NMDA channel blocker (MK-801) and by inhibitors of mitogen-activated protein kinase kinase (U0126), calcium-dependent/independent phospholipases A2 (methyl arachidonyl fluorophosphonate) but not calcium-independent phospholipases A2 (bromoenol lactone) and cyclooxygenase-2 (NS398). A cell-permeant specific inhibitor of calpain (PD150606) prevented LMP, but not ROS production. It is concluded that LMP results in part from calcium-initiated and extracellular signal-regulated kinase-initiated arachidonic acid metabolism, which produces free radicals; it also requires the action of calpain.

Abbreviations used

artificial CSF


bromoenol lactone




calcium-dependent phospholipases A2




extracellular signal-regulated kinase


ischemic ACSF


calcium-independent phospholipases A2


lysosomal membrane permeabilization


methyl arachidonyl fluorophosphonate


mitogen-activated protein kinase kinase


oxygen–glucose deprivation


pre-incubation ACSF


N-tert-butyl-α-phenyl nitrone


phosphate-buffered saline


region of interest


reactive oxygen species

A key feature of cerebral ischemia is the delayed neuronal death caused by the initial insult (Lipton 1999). Large increases in calcium influx (Picconi et al. 2006) and free radical production (Solenski et al. 1997; Monje et al. 2000; Aarts et al. 2002; McManus et al. 2004) are observed during ischemia and reperfusion in many systems, including the acute rat hippocampal slice, which was used in this study. While blocking these events with NMDA receptor antagonists or antioxidants attenuates delayed neuronal damage, the downstream targets of calcium and free radicals are not well known.

Lysosomal cathepsins, including cathepsin B (EC, are translated as inactive pro-enzymes and mature to active enzymes via proteolytic cleavage only when exposed to the acidic environments of late endosomes and lysosomes (Gieselmann et al. 1985; Mach et al. 1993; Sleat et al. 2006). This process ensures that these proteases are separated from essential cellular proteins and contributes to a lysosomal environment necessary for the homeostatic turnover of proteins and other cellular materials (de Duve 2005).

The cytosolic accumulation of cathepsin activity is associated with a variety of pathologies. Mutations of cytosolic cathepsin inhibitors have been implicated in some forms of epilepsy (Pennacchio et al. 1996), and overwhelming the capacity of these endogenous inhibitors by microinjection of active cathepsin D (EC into cultured fibroblasts causes caspase-dependent apoptosis (Roberg et al. 2002). Furthermore, release of cathepsins B and D precedes death following various insults such as tumor necrosis factor-α or staurosporine treatments, UV irradiation, or oxidative stress (Olsson et al. 1989; Ollinger 2000; Foghsgaard et al. 2001; Kagedal et al. 2001; Werneburg et al. 2002; Johansson et al. 2003; Bivik et al. 2006; Kurz et al. 2006; Nilsson et al. 2006). In each of these cases, inhibition of cathepsin B or D attenuates damage. Thus, it is very likely that the release of lysosomal cathepsins into the cytosol causes cellular damage following a wide variety of insults, but these same proteases normally play a vital homeostatic role within the lysosomal lumen (de Duve 2005).

Recent evidence implicates movement of the lysosomal cysteine protease, cathepsin B, into the cytosol in neuronal death following ischemia in primates and rodents (Seyfried et al. 1997; Yamashima et al. 1998; Yoshida et al. 2002; Tsubokawa et al. 2006). The current studies were undertaken in the hippocampal slice model to allow determination of the biochemical steps leading to the ischemia-induced release of proteases from lysosomes and the subsequent damage.

Materials and methods

Buffers and chemicals

Dihydroethidium (DHE) and PD150606 were obtained from Calbiochem (San Diego, CA, USA). A polyclonal antibody (S-12) raised in goat against cathepsin B was obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Alexa Fluor 488-conjugated chicken anti-goat IgG and Neurotrace 530/615 were obtained from Molecular Probes (Eugene, OR, USA). CA074-ME and Pepstatin A were from Peptides International (Louisville, KY, USA). Methyl arachidonyl fluorophosphonate (MAFP) and bromoenol lactone (BEL) were obtained from Cayman Chemicals (Ann Arbor, MI, USA). Nitro-l-arginine methyl ester hydrochloride was from Alexis (San Diego, CA, USA). All other chemicals were obtained from Sigma-Aldrich (St Louis, MO, USA).

Three different buffer compositions were used for hippocampal slice incubations. Artificial CSF (ACSF): 124 mM NaCl, 3 mM KCl, 26 mM NaHCO3, 1.4 mM KH2PO4, 1.3 mM MgSO4, 10 mM glucose, and 1.2 mM CaCl2. Pre-incubation ACSF (pACSF): same as ACSF, except with 10 mM MgSO4 and 0.5 mM CaCl2. Ischemic ACSF (iACSF): same as ACSF, except with 0 mM glucose. ACSF and pACSF were equilibrated for 30 min prior to tissue incubations with 95% oxygen and 5% carbon dioxide. iACSF was equilibrated for 30 min prior to incubations with 95% nitrogen and 5% carbon dioxide. Buffers were also bubbled gently throughout each experiment with the same gas mixtures.

Animals and slice preparation

Adult male Sprague–Dawley rats (175–199 g) were obtained from Harlan (Indianapolis, IN, USA). All procedures were carried out according to NIH Guidelines and with the approval of the University of Wisconsin Medical School Animal Care and Use Committee. Rats were decapitated using a guillotine and brains were removed to ice-cold pACSF. Hippocampi were dissected out and sliced transversely to 500 μm on a Vibratome. Slices were placed on nylon bolting cloth platforms submerged in pACSF. After incubating for 45 min in this buffer (Feig and Lipton 1990), platforms with slices were submerged in ACSF. All experiments were performed at 35°C.

Oxygen–glucose deprivation and reoxygenation

Slices were incubated in ACSF for 2 h prior to initiation of oxygen–glucose deprivation (OGD). Drugs were applied directly to the bath buffer where noted. Following a quick rinse with iACSF, slices were incubated in iACSF for 5 min, unless otherwise noted. Following OGD, platforms with slices were returned to ACSF for the appropriate period of reoxygenation.

Measurement of cathepsin leakage from lysosomes in slices via fluorescence immunohistochemistry

Slices were fixed and sectioned to 60 μm, as described above. Antigen retrieval was performed on the sections by submerging them in phosphate-buffered saline (PBS) at pH 1.9 with 0.05% pepsin for 30 min. Sections were rinsed in PBS at pH 7.3 before incubating overnight in PBS with 5% normal chicken serum and antibody raised against cathepsin B (4 μg/mL). Sections were exposed to a fluorescent Nissl stain, Neurotrace 530/615 (1 : 100 dilution), rinsed, and then incubated in 5% normal chicken serum and 1 : 100 Alexa Fluor-488 conjugated chicken anti-goat IgG. Sections were mounted on Fisher Plus slides (Fisher Scientific, Pittsburgh, PA, USA) with VectaShield Hardset mounting medium (Vector Laboratories, Burlingame, CA, USA). Sections were viewed on a Bio-Rad Radiance 2100 confocal imaging system (Bio-Rad, Hercules, CA, USA) using LaserSharp 2000 software in the W. M. Keck Laboratory for Biological Imaging (Madison, WI, USA). Alexa 488 is excited by a laser line at 488 nm and signal is collected at 520 ± 10 nm. Neurotrace 530/615 is excited by a laser line at 543 nm and signal is collected at > 600 nm. Images were collected from the pyramidal cell layer of area CA1. A z-pass was performed through each section to find its optical center. All images were collected from this optical layer.

Measurement of relative permeability of isolated lysosomes

Effects of slice manipulations on the permeability of isolated lysosomes were determined by measuring distribution of acid phosphatase, a specific lysosomal marker enzyme (Piqueras et al., 1994). Eighteen slices were homogenized in 200 μL 10 mM HEPES, pH 7.4, with 250 mM sucrose, 2 mM MgCl2, and 1 mM Mg-ATP (homogenizing buffer). Homogenates were centrifuged at 700 g for 10 min at 4°C using an Eppendorf Minispin Plus (Eppendorf, Westbury, NY, USA). Supernatants were removed and centrifuged at 14 100 g for 20 min. The 14 100 g precipitate, containing primarily lysosomes and mitochondria, was resuspended in 75 μL homogenizing buffer and incubated at 37°C for 60 min. Intact organelles were then precipitated by centrifugation of the suspension at 14 100 g for 20 min. The amount of acid phosphatase activity in supernatant and precipitate was measured. To do this, both precipitate and soluble fractions were exposed to the substrate 4-nitrophenyl phosphate (Sigma-Aldrich), a colorimetric indicator of phosphatase activity, for 30 min at pH 4.5. The reaction was stopped with 0.5 N NaOH, and the presence of acid phosphatase activity was detected by absorbance at 405 nm in a plate reader. Total acid phosphatase activity was computed and the percentage present in the soluble fraction was calculated. This assay was chosen for its sensitivity and long history in the study of lysosomes.

Measurement of superoxide in the slice

Slices were exposed to 50 μM DHE during the final 15 min of an experimental incubation. DHE is converted to a charged, fluorescent derivative upon exposure to superoxide (Benov et al. 1998). This derivative is fluorescent without binding to DNA or RNA, and thus is observed mainly in the cytoplasm. Slices were submerged in cold 4%p-formaldehyde in 0.1 M phosphate buffer and 5% sucrose, and fixed overnight at 4°C before cryopreserving and sectioning to 60 μm, as described in the Immunohistochemistry procedures. Sections were then mounted on Fisher Plus slides and coverslipped with VectaShield Hardset. DHE fluorescence was observed on a Bio-Rad Radiance 2100 confocal imaging system with excitation using a laser line at 488 nm and emission detected at > 600 nm.

Measurement of neuronal damage

Changes in neuronal morphology were used as a measure of damage. Following an experiment, slices were fixed overnight at 4°C in 4%p-formaldehyde dissolved in 0.1 M phosphate buffer at pH 7.3 with 5% sucrose. Slices were cryopreserved at pH 7.3 with increasing concentrations of sucrose/glycerol (10/5%, 15/10%, and 20/10%). Slices were then mounted on a freezing microtome in 0.1 M phosphate buffer with 30% sucrose and 10% glycerol; 120 μm of tissue was removed from the surface of a slice before taking three sections of 60-μm each. This was performed to eliminate any tissue on the surface of each slice that might have been damaged during the experiment and post-processing (Feig and Lipton 1990). Sections were stained with cresyl violet, and the CA1 pyramidal cell layer was viewed with a light microscope. Cells were defined as healthy or damaged based on specific morphological characteristics that were defined previously (Feig and Lipton 1990). Briefly, category 1 cells are healthy, with round soma, pale nucleus, and slightly darker cytoplasm. Category 2 cells show a slight general darkening compared with category 1 and are somewhat polygonal, but plasma and nuclear membranes are still well defined. Category 3 cells, which are considered severely damaged, lack obvious soma boundaries, the nucleus is darkened and Nissl substance has become aggregated in the darkened cytoplasm and nucleus. Category 4 cells are swollen, with small dense nuclei, and have lost most Nissl staining. The CA1 pyramidal layer cells within a field were categorized. The percentage of cells exhibiting categories 3–4 morphologies was computed, and an increase in this percentage represented damage when compared with normoxic tissue.

Image analysis


Images collected were analyzed using ImageJ software from the NIH (Bethesda, MD, USA). Computations were run in Microsoft Excel XP (Redmond, WA, USA).

Immunohistochemistry of cathepsin B

Cathepsin B movement into the cytosol was used as a measure of lysosomal membrane permeabilization (LMP) in the slice. Permeabilization was measured as increased homogenous staining, as opposed to punctate staining, in the CA1 pyramidal neurons. The following procedure was developed to measure the homogenous staining. A first region of interest (ROI) was drawn around a small acellular area in the image, where staining was very weak. Pixel intensities less than the brightest 1% in this region were assumed to represent regions without cathepsins and were excluded from analysis of homogeneous cellular staining by using a high-pass cut-off. A new ROI was then drawn around a contiguous region of pyramidal neurons in the cell layer, being careful to avoid large acellular areas. In this ROI, puncta are formed by high concentrations of cathepsin B within the lysosome. These are eliminated from the analysis by a low-pass cut-off. Once the acellular and punctate pixels have been eliminated from the analysis, the remaining pixels, with intensities between the high- and low-pass cut-offs, represent what we term homogenous cathepsin B staining. The percentage of the ROI occupied by these pixels was measured and was the percentage of homogenous cathepsin B that we used to estimate LMP (% homogenous cathepsin B).


Dihydroethidium is converted to a charged fluorescent product, distinct from ethidium, by superoxide (Benov et al. 1998). The intensity of oxidized DHE fluorescence in CA1 pyramidal neurons reflects the concentration of superoxide present during the 15-min exposure period. A ROI is drawn around a contiguous region of neurons in the pyramidal cell layer of CA1. The total fluorescent output from the ROI is measured and divided by the number of cells. This reflects the average cellular superoxide level in the ROI.


Statistics were applied to data using Graphpad InStat 3 (GraphPad Software Inc., San Diego, CA, USA). Data were subjected to repeated measures anova with Tukey’s post-test; one-way anova with Student–Newman–Keuls post-test was applied where appropriate. In graphs, significance is indicated by plus-signs (+) or asterisks (*), with ***p < 0.001, **p < 0.01, and *p < 0.05. In all cases, + is versus OGD and reoxygenation without drug and * is versus normoxia without drug as discussed in the text.


Cathepsin redistribution in response to OGD and reoxygenation

In situ staining of cathepsin B appeared mostly punctate in freshly sliced and incubated normoxic tissues, indicating its concentration within lysosomes (Fig. 1b and c). Slices were exposed to OGD for 5 min, and cathepsin distribution was measured at various times between 0 and 120 min of reoxygenation (Fig. 1a). Staining in the neuronal cytosol appeared between 1 and 2 h of reoxygenation (Fig. 1d and e), suggesting a delayed LMP. This homogenous cytosolic cathepsin B staining appeared as a ‘film’ throughout much of the cell body, as opposed to lysosomal staining that appeared as puncta. The cytosolic accumulation of cathepsin B was not blocked by cycloheximide, at a concentration at which 98% of all brain slice protein synthesis is blocked (Feig and Lipton 1993) (Fig. 1f). Thus, the cytosolic cathepsin does not result from de novo synthesis of cathepsin B.

Figure 1.

 OGD and reoxygenation cause lysosomal membrane permeabilization, as measured by immunohistochemical redistribution of cathepsin B to the cytosol. (a) In CA1 pyramidal neurons of the normoxic slice (dark diamonds, n = 5 per time point) cathepsin B staining appears punctate. Following 5-min OGD (light squares), cathepsin B accumulates in the cytosol after 60–120 min of reoxygenation. (b–d) Neurons are stained with a fluorescent Nissl dye (red, b and d), and labeled with antibodies raised against cathepsin B (green, b and d). The areas outlined in panels (b and d) are magnified in panels (c and e), respectively, with Nissl signal omitted for observation of cathepsin B distribution. In normoxic tissue (b and c), staining appears punctate, while cytosolic cathepsin B increases following OGD and 2 h of reoxygenation (d and e). The arrow in panel (e) indicates an area of homogenous staining in a CA1 pyramidal neuron. (f) The increase in homogenous staining was not because of de novo synthesis of cathepsin B. Little homogenous cathepsin B staining was measured in normoxic tissue (dark bars, n = 3); significantly more homogenous cathepsin B was measured in tissue exposed to OGD and 2 h of reoxygenation (light bars) in the presence or absence of 1 μM cycloheximide. ***p < 0.001 and **< 0.01 and error bars are SEM. Scale bar, 20 μm.

The stability of isolated lysosomes following OGD and reoxygenation was compared with that of lysosomes isolated from normoxic tissue (Fig. 2). After exposure to normoxia or to OGD and reoxygenation, slices were homogenized and lysosomal/mitochondrial fractions were obtained by centrifugation. These fractions were incubated in vitro for 60 min. At the end of this time the percentage of total lysosomal acid phosphatase that had leaked from the lysosome was calculated. Lysosomes from ischemic tissue released about twice as much acid phosphatase as those from normoxic tissue. Thus, 5-min OGD causes a permeabilization of the lysosomal membrane that persists even when lysosomes are removed from their cellular environment.

Figure 2.

 Hippocampal lysosomes isolated from slices previously exposed to OGD are more permeable than those obtained from normoxic tissue. Enriched lysosomal/mitochondrial fractions were incubated at 37°C for 60 min after exposing experimental slices to 5-min OGD and 2 h of reoxygenation in situ. More acid phosphatase activity was released from lysosomes obtained from ischemic tissue than from lysosomes from normoxic tissue, indicating that the former were more permeant (n = 6). **p < 0.01 and error bars are SEM.

Cathepsin redistribution depends on NMDA-mediated calcium influx and free radical production

Calcium influx through the NMDA receptor and free radical production initiate damaging processes following ischemia (Lipton 1999). The role of these processes in LMP was determined by exposing slices to OGD and reoxygenation in the presence and absence of either dizocilpine hydrogen maleate (MK-801), a specific antagonist of NMDA receptors, or an antioxidant (Fig. 3); 10 μM MK-801 blocked redistribution of cathepsin B to the cytosol following 5-min OGD and 2 h of reoxygenation (Fig. 3a and c–e), indicating that calcium influx through NMDA sensitive channels plays a critical role in LMP. To test whether or not reactive oxygen species (ROS) are involved in LMP, slices were exposed to 5-min OGD and 2-h reoxygenation in the presence or absence of 2 mM N-tert-butyl-α-phenyl nitrone (PBN), a spin-trapping agent, or 1 μM (±)-6-hydroxy-2,5,7,8-tetramethylchromane-2-carboxylic acid (Trolox), a water-soluble analog of α-tocopherol. Exposure to either antioxidant during OGD and reoxygenation blocked LMP, confirming the involvement of ROS in LMP (Fig. 3b and f–g).

Figure 3.

 Reactive oxygen species and calcium influx through NMDA channels contribute to LMP following OGD and reoxygenation. (a and b) CA1 neurons in normoxic tissue (dark bars) showed little homogenous cathepsin B staining. Exposure of slices to 5-min OGD and 2 h of reoxygenation (light bars) caused accumulation of cathepsin B in the cytosol, indicating that LMP had occurred (**p < 0.01 vs. time-matched normoxic control). (a) 10 μM MK-801, an NMDA receptor antagonist present during OGD and reoxygenation, blocked the cytosolic accumulation of cathepsin B (+p < 0.05 vs. reoxygenation/no drug, n = 10). (b) Two antioxidants, 2 mM N-tert-butyl-α-phenyl nitrone (PBN) or 1 μM 6-hydroxy-2,5,7,8-tetramethylchromane-2-carboxylic acid (Trolox), also blocked the cytosolic accumulation of cathepsin B following OGD and reoxygenation (n = 6). Representative CA1 pyramidal cells from the indicated slice treatments (Reox = 5-min OGD and 2 h of reoxygenation) are shown in panels (c–g). Scale bar, 20 μm and error bars are SEM.

OGD-induced superoxide production is mediated by the NMDA receptor

Superoxide was measured using DHE. The specificity of DHE for superoxide was tested in vitro and, as reported by Benov et al. (1998), we found that DHE was oxidized by superoxide, but not hydrogen peroxide, nitric oxide, or peroxynitrite (data not shown). There was a significant increase in superoxide concentration in CA1 pyramidal neurons after OGD and 2 h of reoxygenation (Fig. 4a–d); 2 mM PBN and 1 μM Trolox blocked DHE oxidation, confirming that these antioxidants attenuate superoxide production at concentrations that also block LMP (Fig. 4b and f–g). Superoxide production was blocked by 10 μM MK-801 (Fig. 4a and c–e), indicating that OGD-induced superoxide production is initiated by calcium entry.

Figure 4.

 An NMDA-dependent increase of superoxide occurs in CA1 pyramidal cells after OGD and reoxygenation. (a) The intensity of dihydroethidium (DHE) fluorescence increased following 5-min OGD and 2 h of reoxygenation (**p < 0.01 and ***p < 0.001 vs. time-matched normoxic control), indicating increased levels of superoxide. (b) 10 μM MK-801, present during OGD and reoxygenation, blocked the increased production of superoxide (++p < 0.01 and +++p < 0.001 vs. reoxygenation without drug). (c) Both antioxidants, 2 mM PBN and 1 μM Trolox, also blocked superoxide production following OGD and reoxygenation. Representative CA1 pyramidal cells from the indicated slice treatments are shown in panels. Scale bar, 20 μm and error bars are SEM.

Arachidonic acid metabolism by cPLA2 and cyclooxygenase-2 mediates superoxide production

Arachidonic acid metabolism is increased during ischemia and reperfusion/reoxygenation in many systems. It is liberated from membrane lipids by calcium-dependent or -independent phospholipases A2 (cPLA2 and iPLA2, respectively). Slices were exposed to 5-min OGD and 2 h of reoxygenation in the presence of 10 μM BEL, an antagonist of iPLA2 (EC, or in the presence of 1 μM MAFP, an antagonist of both iPLA2 and cPLA2 (Lio et al. 1996). MAFP, but not BEL, blocked increased DHE oxidation following OGD and reoxygenation, indicating that cPLA2 (EC activity leads to the superoxide production (Fig. 5a–b). Phosphorylation by the Ser/Thr extracellular signal-regulated kinase (ERK; EC is also necessary for activation of cPLA2 (Gijon et al. 1999; Xu et al. 2002). Inhibition of ERK signaling with the specific mitogen-activated protein kinase kinase (MEK; EC antagonist, 1 μM U0126, blocked DHE oxidation following OGD and reoxygenation (Fig. 5d), further implicating cPLA2 in superoxide production.

Figure 5.

 Arachidonic acid metabolism by cPLA2 and COX-2, and ERK activity, produces superoxide following OGD and reoxygenation in CA1 pyramidal neurons. The intensity of DHE fluorescence increased following 5-min OGD and 2 h of reoxygenation (light bars, *p < 0.05, **p < 0.01 and ***p < 0.001 vs. time-matched normoxic control). Dark bars are time-matched normoxic controls. (a) The iPLA2 antagonist, 10 μM BEL, did not attenuate the increase of superoxide levels following OGD and reoxygenation (n = 6). (b) 1 μM MAFP, an antagonist that targets both iPLA2 and cPLA2, however, blocked increased DHE fluorescence following OGD and reoxygenation (n = 12, + vs. OGD and reoxygenation without drug). (c) The COX-2 antagonist, 5 μM NS398, also blocked increased production of superoxide following OGD and reoxygenation (n = 12). (d) Inhibition of ERK activity with the specific MEK antagonist, 1 μM U0126, blocked increased superoxide production following OGD and reoxygenation (n = 11). Error bars are SEM.

One pathway of arachidonic acid metabolism that leads to free radical production is its metabolism to prostaglandin by cyclooxygenase (COX). Slices were exposed to OGD and reoxygenation in the presence of 50 μM N-[2-(cyclohexyloxy)-4-nitrophenyl] methanesulfonamide (NS398), a specific inhibitor of COX-2 (EC NS398 blocked the increased oxidation of DHE following OGD and reoxygenation, indicating that COX-2 metabolism of arachidonic acid to prostaglandin is the major source of superoxide following that insult (Fig. 5c).

The steps involved in free radical formation were then tested as to their roles in LMP. Slices were exposed to 5-min OGD and 2 h of reoxygenation in the presence of either 10 μM BEL or 1 μM MAFP (Fig. 6a and b). Inhibition of iPLA2 alone, with BEL (Fig. 6a), did not attenuate cathepsin release, but inhibition of both iPLA2 and cPLA2 with MAFP (Fig. 6b) blocked redistribution of cathepsin B to the cytosol. The MEK antagonist, U0126, also blocked OGD-induced cathepsin redistribution (Fig. 6d). Finally, inhibition of COX-2 with NS398 also blocked cathepsin release to the cytosol (Fig. 6c). Together, these data strongly suggest that LMP is a consequence of OGD-induced arachidonic acid production and metabolism by cPLA2 and COX-2, respectively. Furthermore, the same pathway that leads to superoxide production also leads to LMP.

Figure 6.

 Arachidonic acid metabolism by cPLA2 and COX-2, and ERK activity, contributes to LMP following OGD and reoxygenation in CA1 pyramidal neurons, as measured by cytosolic redistribution of cathepsin B. Homogenous cathepsin B antigenicity increased following 5-min OGD and 2 h of reoxygenation (a–d; light bars, **p < 0.01 and ***p < 0.001 vs. time-matched normoxic control). Dark bars are time-matched normoxic controls. (a) The iPLA2 antagonist, 10 μM BEL, did not attenuate redistribution of cathepsin B following OGD and reoxygenation (n = 5). (b) 1 μM MAFP, an antagonist that targets both iPLA2 and cPLA2, however, blocked cathepsin B redistribution following OGD (n = 14, +p < 0.05 vs. OGD and reoxygenation without drug). (c) The COX-2 antagonist, 5 μM NS398, also blocked cathepsin B redistribution to a homogenous pattern (n = 8). (d) Inhibition of ERK activity with 1 μM U0126 blocked cathepsin B redistribution as well (n = 7). Error bars are SEM.

Calpain mediates LMP independently of superoxide production

Previous studies have shown that calpain (EC associates with neuronal lysosomes following ischemia over a time course that would be consistent with a causative role for this calcium-dependent cysteine protease in LMP (Yamashima et al. 2003). As predicted if calpain contributes to LMP following 5-min OGD and 2 h of reoxygenation, the specific calpain antagonist, 50 μM PD150606, blocked release of cathepsin B to the cytosol (Fig. 7a). However, PD150606 did not attenuate DHE oxidation following OGD and reoxygenation (Fig. 7b), indicating that calpain does not contribute to LMP by modulating superoxide production.

Figure 7.

 Calpain is necessary for LMP after OGD and reoxygenation independent of any effect on superoxide production. (a) LMP was measured as an increase of homogenous cathepsin B in CA1 pyramidal cells following 5-min OGD and 2 h of reoxygenation (light bars, *p < 0.05 and **p < 0.01 vs. time-matched normoxic control). The specific calpain antagonist, 50 μM PD150606, blocked redistribution of cathepsin B to the cytosol following OGD and reoxygenation (n = 5, +p < 0.05 vs. reoxygenation without drug). (b) Changes in superoxide levels in CA1 pyramidal cells were measured as changes in the fluorescence intensity of DHE. PD150606 did not block increased DHE fluorescence following OGD and reoxygenation (n = 5), indicating that inhibition of calpain did not attenuate OGD-induced superoxide production.

LMP following 10 min OGD differs from that following 5 min OGD

The rates of ischemic changes and severity of damage in vivo increases with the duration of insult and the reduction of blood flow to a given area (Lipton 1999). Unlike neuronal damage in the ischemic penumbra, which occurs within hours to days following the initial insult and is attenuated by free radical scavengers, damage in the core occurs more rapidly, without a detectable change in ROS, and is not attenuated by free radical scavengers (Solenski et al. 1997). The rat hippocampal slice was exposed to a longer duration of OGD to model a more severe insult. The cytosolic accumulation of cathepsin B after 10-min OGD occurred immediately following the insult, with no need for reoxygenation (Fig. 8a–c). PBN did not attenuate this change after 10-min OGD (Fig. 8d). This indicates that LMP after 10-min OGD does not depend on the production of ROS. Other steps in the cascade that is involved in LMP after 5 min were then tested. Neither MAFP nor PD150606 attenuated the cytosolic accumulation of cathepsin B after 10-min OGD (Fig. 8d), indicating that cPLA2 and calpain are not necessary for LMP after the more severe insult. The effect of 10-min OGD on lysosomes appeared much more severe than the effect of 5-min. The number of puncta per pyramidal neuron decreased after 10-min OGD, indicating drastic loss of cathepsin from lysosomes; in contrast there was no change in the number of puncta after 5-min OGD, indicating a relatively small loss of cathepsins (Fig. 8e). This, along with the very different pharmacology of LMP following the two insults, suggests that the permeabilization mechanisms may be qualitatively different in the two cases. Unlike the delayed increase of ROS seen after 5-min OGD in CA1 pyramidal cells of the rat hippocampal slice, production of ROS did not change during or following 10-min OGD (Fig. 9) paralleling the absence of a free radical increase in the core of the infarct.

Figure 8.

 LMP is more rapid and severe after 10-min OGD than following 5-min OGD and does not depend on free radical production, cPLA2, or calpain. (a and b) In CA1 pyramidal neurons of the normoxic slice (dark diamonds, n = 5 per time point) cathepsin B staining appears punctate. (a and c) Following 10-min OGD (light squares, **p < 0.01 and ***p < 0.001 vs. time-matched normoxic control), cathepsin B accumulates in the cytosol immediately (n = 5). (b and c) Neurons are stained with a fluorescent Nissl dye (red), and labeled with antibodies raised against cathepsin B (green). (d) Accumulation of cathepsin B in the cytosol after 10-min OGD and 15-min reoxygenation (light gray bars) is not blocked by 2 mM PBN (n = 4), 1 μM MAFP (n = 5), or 50 μM PD150606 (n = 5). (e) The number of puncta per pyramidal cell does not change following 5-min OGD and 120-min reoxygenation, although cytosolic cathepsin B does increase (Fig. 1). In contrast, the number of puncta per cell decreases significantly versus rat-matched normoxic controls after 10-min OGD and 15-min reoxygenation (n = 5 matched experiments). Scale bar, 20 μm and error bars are SEM.

Figure 9.

 Increased superoxide production does not occur following 10-min OGD. No increase of superoxide concentrations in CA1 pyramidal cells was observed at any time during or after 10-min OGD, as measured by DHE fluorescence up to 2 h after OGD. Light squares are with 10-min OGD and dark diamonds are time-matched normoxic control. Each point is the average of five independent experiments and error bars are SEM.

Majority of morphological damage occurs after cathepsin redistribution and is blocked by cathepsin inhibitors

Nissl substance condensation is an often-used marker of neuronal damage in ischemia (Smith et al. 1984; Raley-Susman and Lipton 1990; McBean et al. 1995; Chen et al. 1998). Evidence points to proteolytic degradation of cytoskeletal proteins as a major cause of the ER collapse that underlies condensation (Ogata et al. 1989; Inuzuka et al. 1990; Kaku et al. 1993; Minger et al. 1998; Banik et al. 2000; Ogburn and Figueiredo-Pereira 2006; Zhang et al. 2006a). Therefore, we chose to examine damage following ischemia using Nissl substance condensation as a marker. The arrow in Fig. 10e points to one of many cells exhibiting a damaged morphology, while the arrowhead in Fig. 10c points to a healthy neuron. In normoxic tissue, the number of damaged neurons was low (Fig. 10a); 120 min following 5-min OGD, a significant increase in the number of damaged neurons appeared (Fig. 10a and d). At 240 min of reoxygenation, 2 h after the appearance of LMP, the number of damaged cells had increased further, to about 400% the number in normoxic controls (Fig. 10e).

Figure 10.

 CA1 pyramidal neurons exhibit delayed morphological damage following OGD and reoxygenation that depends on cathepsin activity. Neuronal health was assessed morphologically following staining of tissue with cresyl violet, with the number of neurons exhibiting condensed nuclei (morphology 3 or 4) reported as a measure of irreversible damage (see Materials and methods). (a) Exposure of slices to 5-min OGD and the indicated periods of reoxygenation caused significant damage to CA1 pyramidal cells only after 2 h of reoxygenation (**p < 0.01 and ***p < 0.001 vs. time-matched normoxic control with unpaired two-tailed t-test, n = 8–16 rats per time point). Although many neurons exhibit damage morphology 2 after 120 min of reoxygenation, these neurons were not included in analysis of severe damage, as it is uncertain that neurons exhibiting this morphology are irreversibly damaged. (b) In contrast to the delayed appearance of damage observed after 5-min OGD, damage appeared within 15 min after 10-min OGD. The appearance of damage after 5-min OGD and 240 min of reoxygenation (dark bars) was blocked by treatment, at the initiation of reoxygenation, of either a cathepsin B antagonist (7.5 μM CA074-ME) or a cathepsin D antagonist (7.5 μM Pepstatin A). The appearance of damage after 10-min OGD and 15 min of reoxygenation (light bars) was not blocked by either CA074-ME or Pepstatin A. (***p < 0.001 vs. time-matched normoxic control and +p < 0.05 and ++p < 0.01 vs. reoxygenation/no drug with one-way anova and Tukey’s post-test, n = 7). Representative CA1 pyramidal layer fields from the indicated slice treatments are shown in panels (c–g). Scale bar, 50 μm and error bars are SEM.

If LMP damages neurons following 5-min OGD, then inhibition of enzymes released to the cytosol during LMP should attenuate damage. Two lysosomal proteases, cathepsin B, a cysteine protease, and cathepsin D, an aspartic protease, are likely perpetrators of cell damage. Exposure of hippocampal slices to 5-min OGD and 4 h of reoxygenation in the presence of either CA074-ME or Pepstatin A attenuated delayed damage (Fig. 10b and f–g). The effectiveness of either inhibitor suggests that both cathepsins B and D are necessary for neuronal damage following OGD.

Condensation of Nissl substance in pyramidal neurons after 10-min OGD occurred within 15 min after starting reoxygenation. Unlike the damage after 5-min OGD this damage was not attenuated by CA074-ME or Pepstatin A (Fig. 10b). Presumably, the increased severity of the insult recruits additional damaging processes.


Previous in vivo studies have implicated extralysosomal cathepsins in ischemic brain damage in vivo by demonstrating shifts of cathepsin antigenicity from lysosomes into the cytosol and by showing that cathepsin blockers reduce ischemic damage (e.g. Benchoua et al. 2004). The current studies, in the hippocampal slice, extend these observations by describing a biochemical cascade that leads from the ischemic insult to the LMP. The results also show that the timing of LMP is consistent with its having a causative role in ischemic cell damage. The results also show limits on the damaging role of cathepsins. They play a critical role in damage following a relatively mild insult (5-min OGD) but not after a more severe insult (10-min OGD). This is despite their far more massive release after 10-min OGD. Presumably, other processes become even more damaging than cathepsin proteolysis after the more severe insult. These could be actions of other lysosomal enzymes or be completely independent processes.

Biochemical processes leading to LMP

The effects of inhibitors revealed a likely set of reactions leading from the 5-min ischemic insult to the delayed LMP (Fig. 11).

Figure 11.

 A model describing calcium-dependent lysosomal membrane permeabilization. Factors indicated in grey ovals were shown by the current studies to be necessary for LMP. Furthermore, the signaling cascade enclosed by the black rectangle was shown to be responsible for essentially all the superoxide production after OGD. Dashed lines indicate uncertainties. It is uncertain whether NMDA-mediated calcium, or calcium from some other source, activates calpain. It is also uncertain whether ERK is activated by calcium or by some other mechanism. The mechanism by which the two parallel pathways cause LMP was not addressed here, as indicated by the black star at the lysosomal membrane, but likely possibilities are found in the Discussion.

Calcium entry within the first 5 min of OGD is strongly attenuated by 10 μM MK-801 (Lobner and Lipton 1993; Zhang and Lipton 1999). The block of LMP by MK-801, a specific inhibitor of the NMDA receptor suggests that NMDA-mediated calcium entry is an initiating step.

The importance of cPLA2 for LMP, and its calcium dependence, suggests that this entering calcium activates the enzyme in the next step of the process. This is consistent with data showing that increases in arachidonic acid following ischemia are mediated by the NMDA receptor (Lazarewicz et al. 1992; Mrsic-Pelcic et al. 2002) and that NMDA-mediated calcium entry increases cPLA2 activity (Lucas and Dennis 2005; Shen et al. 2007). The calcium binds to the enzyme, activating its movement to the membrane (Schaloske and Dennis 2006). This calcium may also play another key role by increasing levels of phospho-ERK (Krapivinsky et al. 2003; Ivanov et al. 2006). ERK-mediated phosphorylation is necessary for cPLA2 activation (Gijon et al. 1999) and inhibition of MEK by the specific inhibitor U0126 prevented LMP.

Arachidonic acid does permeabilize rat liver lysosomes to proteins (Zhang et al. 2006b). However the blockade of LMP by the COX-2 inhibitor NS398, and by free radical scavengers, shows that metabolism of arachidonic acid by COX-2 with concomitant production of superoxide (Kukreja et al. 1986) is a necessary next step in the process. A similar cascade, resulting from direct NMDA-activation, leads to prostaglandin E and free radical production in hippocampus in vivo (Pepicelli et al. 2002).

Thus, the data implicate NMDA-mediated calcium entry, followed by metabolism of phospholipids to arachidonic acid, followed by metabolism of arachidonic acid by COX-2 and the resulting production of superoxide as a key pathway in LMP. This is illustrated in Fig. 11. The finding that superoxide production is solely dependent on COX-2 metabolism of arachidonic acid supports the activation of this sequence following OGD. The only reasonable way to incorporate the complete dependence of superoxide production on the NMDA receptor, ERK and cPLA2, as shown in this study, is to conclude that the postulated sequence occurs and is responsible for superoxide production.

A direct action of ROS on lysosomes may contribute to LMP. In support of this, rat liver lysosomes are permeabilized upon exposure to hydroxyl radical-generating systems (Mak et al. 1983; Olsson et al. 1989) and H2O2 acts on several tissue culture systems to permeabilize lysosomes (Zdolsek et al. 1993; Zhao et al. 2001). However, neurons have not been tested in this regard and it is unknown whether the concentrations of ROS following ischemia reach sufficient levels to cause LMP in this manner. ROS may well be necessary but not sufficient. Alternatively, other products of the sequence, shown within the box in Fig. 11, could act on the lysosome. Superoxide may be necessary because it augments their production via positive feedback (Rashba-Step et al. 1997; Seo et al. 2001).

Calpain activity is also critical for LMP. The fact that PD150606 did not prevent ROS formation shows that calpain does not act on the pathway causing superoxide formation. As illustrated in Fig. 11 it must act in parallel to this pathway. Calpain may be acting directly on the lysosome, as has been suggested (Yamashima et al. 2003); it is notable that both caspase 8 and calpain release cathepsins from purified liver lysosomes (Guicciardi et al. 2000). These authors suggest that calpain-cleaved Bid initiates pore formation in the lysosomal membrane. The requirement for both calpain and the superoxide cascade suggests that the two act in concert to cause LMP following 5-min OGD, which is illustrated by black star symbol in Fig. 11.

Processes causing LMP also cause ischemic neuronal damage

Genetic and pharmacological manipulations have shown that many processes are critical to ischemic damage. It is notable that all the steps in the cascade described here, as well as calpain-mediated proteolysis, are reactions that are critical to ischemic cell death. Manipulations that prevent any one of them strongly attenuate ischemic cell death in mild ischemia models. These manipulations include inhibition of cPLA2 (Arai et al. 2001; Tabuchi et al. 2003), of the MEK/ERK pathway (Chu et al. 2004), of COX-2 (Nagayama et al. 1999; Iadecola et al. 2001), of calpain (Hong et al. 1994; Li et al. 1998), and manipulations that reduce superoxide accumulation (Chan et al. 1998) or NMDA receptor activation (Dogan et al. 1997; Aarts et al. 2002). While these processes may well have damaging effects that are independent of LMP, the strong overlap is consistent with a major role for LMP in ischemic cell damage following relatively mild insults.

LMP mediates neuronal damage in the slice following 5-min OGD

The data strongly implicate LMP in damage to the slice. Development of morphological damage following 5-min OGD is far more rapid in the hippocampal slice than in the in situ hippocampus (Kirino et al. 1984; Lipton 1999) but the cellular changes, as observed in this study and elsewhere, are morphologically similar to in situ changes (Martin et al. 1998; Lipton 1999; Yonekura et al. 2004). An early phase with moderate cytoplasmic darkening and cell layer disorganization appeared immediately after 5-min OGD. This was followed by a second phase which appeared similar to ischemic cell change observed in situ, in which the nucleus was darkened, the cytoplasm was heavily Nissl-stained and the cells were shrunken. A third phase, which appeared later, was marked by ghost-like cell cytoplasm and small, extremely dark nuclei, similar to edematous cell change observed in situ. The second and third phases did not occur until after the appearance of LMP, and inhibition of cathepsin B or D prevented the transition to these severe forms of damage. Cells appeared near normal for at least 4 h following OGD. Thus, cathepsins B- and D-mediated proteolysis play major roles in cell damage. These results are consistent with several reports from in vivo tissue showing that cathepsins are released from lysosomes soon after in vivo ischemia and that cathepsin inhibition attenuates damage (Yamashima et al. 1998; Benchoua et al. 2004). Those in vivo studies did not address the biochemical steps causing the LMP.

These results are also consistent with a large number of studies implicating cathepsin release from lysosomes in cell death in tissues other than brain. While in many cases the cathepsins lead to apoptotic death (Foghsgaard et al. 2001; Werneburg et al. 2002; Bivik et al. 2006) it has recently become apparent that they also cause necrotic cell death (Zong and Thompson 2006; Goldstein and Kroemer 2007; Luke et al. 2007), as they appear to do in the current studies. The basis for the two options has not been addressed, but certainly reduced ATP during and after ischemia (Lobner and Lipton 1993) will shift towards necrotic death (Goldstein and Kroemer 2007).

Basis of free radical production during OGD

In this quite mild (though very damaging) insult, superoxide did not accumulate until 90–120 min following OGD. All of this accumulation resulted from COX-2 metabolism of arachidonic acid. Free radicals have been known to participate in ischemic brain damage, but the source of the free radicals has, until recently, remained undetermined.. Prominent possibilities have included mitochondrial dysfunction (Frantseva et al. 2001), phospholipid metabolism (Muralikrishna and Hatcher 2006), and NADPH oxidase (Kunz et al. 2006). Recently, Kunz et al. found that COX-2 inhibition during focal ischemia in the mouse did not reduce ROS production, while inhibition of NADH oxidase did. It is quite reasonable that different pathways might dominate in different insults and different species.


A 5-min OGD leads to relatively mild LMP and release of cathepsins, which play a critical role in necrotic morphological changes. LMP results from a biochemical cascade initiated by NMDA receptor-mediated calcium entry and requires the production of superoxide via the COX-2 mediated oxidation of arachidonic acid. There are strong parallels between processes involved in LMP in vitro and in ischemic cell damage in vivo.


This work was supported by the American Heart Association (PL – AHA 03506572) and the National Institutes of Health (JAW – NIH 2T32GM007507-26). Special thanks to Rachel Booth, Lance Rodenkirch, Michael Hendrickson, Sherry Feig, and Inge Sigglekow for helpful advice and technical assistance.