Insulin/PI3K signaling protects dentate neurons from oxygen–glucose deprivation in organotypic slice cultures

Organotypic hippocampal slice cultures

Animals were handled in accordance with local guidelines for ethical use of animals in research at the University of California San Diego, which is in accordance with international guidelines. All possible efforts were made to minimize animal suffering and the number of animals used. Mouse hippocampal slice cultures were prepared and grown on semiporous membranes, according to the standard interface method (Stoppini et al. 1991) as previously described (Yao et al. 2007) with some modifications. In brief, 8-day-old CD1 mice (Charles River Laboratories, Raleigh, NC, USA) were decapitated and the dorsal hippocampi quickly isolated under aseptic conditions. The hippocampi were cut into 400 μm thick, transverse slices by a McIlwain tissue chopper, and the slices transferred to Gey’s balanced salt solution (Gibco, Paisley, Scotland) for separation and trimming of the tissue slices. Tissue slices from the middle part of the dorsal hippocampi were obtained and transferred to membrane inserts (Millicell, Bedford, MA, USA) placed in 35 mm culture dishes with 1.2 mL of serum-containing culture medium composed of basal medium Eagle’s medium (50%), Earle’s balanced salt solution (25%), horse serum (25%), and l-glutamine (1 mmol/L), supplemented with 50 U/mL penicillin/streptomycin and 10 mmol/L glucose. Cultures were maintained in a 5% CO₂, 37°C incubator for 12–14 days before experiments were performed. Culture medium was half-replaced the second day after plating and twice a week thereafter until the day of treatment.

Oxygen–glucose deprivation

Oxygen–glucose deprivation was used as an in vitro model of cerebral ischemia. The inserts with slice cultures were placed into 1 mL of sucrose-containing artificial CSF (ACSF) solution composed of (in mmol/L): NaCl 129, KCl 5, CaCl₂ 1.3, MgCl₂ 1.5, NaHCO₃ 21, and sucrose 10 (315 mOsm, pH 7.4) following three washes using the same solution. The cultures were then placed into an airtight chamber (Billups and Rothenberg, Del Mar, CA, USA) and exposed to 5 min of 95% N₂/5% CO₂ gas flow. After that, the chamber was sealed for 25–60 min at 37°C. Control cultures were maintained for the same time under normoxic atmosphere in glucose-containing ACSF solution. After OGD, slice cultures were returned to their original culture conditions for 48–72 h.

Propidium iodide uptake and quantification of cell death

Cell damage was monitored and quantified by densitometric measurements of the cellular uptake of propidium iodide (PI; Sigma Chemical, St Louis, MO, USA), a fluorescent dye that intercalates into the DNA of necrotic cells that have lost plasma membrane integrity and does not penetrate living or early apoptotic cell membranes. 5 μg/mL PI was added into the culture medium 24 h before treatments and kept at the same concentration throughout the experiment. The PI uptake was recorded before any pharmacological treatments and OGD, and then at different time points after treatments. All cells were killed after experiments by submerging the cultures in sucrose (10 mM) containing ACSF solution at 4°C for 24 h, and the maximum PI uptake served as positive controls for each slice. The death of most of neurons was confirmed in our preliminary experiments by using high glutamate (up to 20 mM), or in each experiment by extending exposure time up to 72 h until no further PI uptake in the neuronal layers were detected.

The PI uptake was quantified densitometrically using Axiovision Image software (Zeiss, Yena, Germany), delineating the hippocampal subfields CA1, CA2, CA3, and dentate gyrus (DG). From the PI uptake in these subfields in each culture, measured at the different time points, the basal PI uptake recorded in the same subfields before treatments was subtracted. Maximum PI uptake after all the cells were killed served as positive controls. Cell injury was expressed as either means of PI uptake for each subregions or percentage cell death (PI uptake/maximum PI uptake). For detection of PI intensity, slices were excited with a 510–560 nm light and the emitted fluorescence was acquired at 610 nm using a rhodamine filter on an inverted fluorescence microscope (Zeiss Axiovert 200 M microscope). Images were taken using a Charged coupled device (CCD) camera and analyzed on a PC with Axiovision imaging-processing software (Zeiss). The fluorescence intensity was measured offline and damage was given as the integral density of PI intensity in CA1, CA2, CA3, and DG regions.

Pharmacological agents and glutamate treatments

In the interface culture method used in this study, glial cells redistribute into a thin layer at the bottom of the culture. This layer of glial cells forms an interface between tissue and culture medium (Buchs et al. 1993), and may function as a diffusion barrier and prevent access of some pharmacological agents from getting to neurons. To facilitate access of glutamate to neurons, slice cultures were treated by submersion. Other agents were treated by applying a thin layer of drug-containing solutions above the top of slice cultures and letting it diffuse through the slice culture. Control cultures were treated in the same manner with drug-free solutions, to prove that the procedure itself was harmless. Doses used in this study were chosen according to the literature, and also based on our preliminary studies. Insulin, IGF-1, and signaling pathway blockers were applied 24 h before and during OGD exposure, and glutamate receptor blockers were applied 1 h before and during OGD.

Immunofluorescence staining in slice cultures

The slice cultures (at days in vitro 12–14) were fixed in 4%p-formaldehyde for 6 h at 4°C and then permeabilized with 1% Triton X-100 in phosphate-buffered saline (PBS) overnight at 4°C. The slices were incubated in 10% goat serum for 1 h before being treated for 3 days with either a rabbit anti-IRβ (C-terminal) polyclonal antibody (1 : 100; Santa Cruz Biotechnology, Santa Cruz, CA, USA) or a rabbit anti-phospho-Akt (Ser473) monoclonal antibody (1 : 100; Cell Signaling, Danvers, MA, USA) in antibody diluent (Invitrogen, Carlsbad, CA, USA) at 4°C. After thorough washes, slices were incubated overnight at 4°C with Alexa 488-labeled anti-rabbit IgG (1 : 500; Invitrogen). After several PBS washes, slices were mounted on slides with Prolong antifade reagent with 4′,6-diamidino-2-phenylindole (DAPI) (Invitrogen). Immunofluorescent images were obtained and analyzed using confocal microscopy (Olympus FV1000, Olympus America Inc., Center Valley, PA, USA). In control experiments, primary antibodies were omitted and no fluorescence signals were detected.

Distinctive sensitivity of hippocampal subregions to OGD

Among hippocampal subregions, the CA2 subfield has not been studied as much, especially in slice culture preparations, because CA2 is a rather small part and more difficult to identify than CA1 and CA3. In our experimental conditions, however, we identified a distinct group of cells in CA2 region that had a higher incorporation of PI 3 h after exposure to OGD for 45 min or longer (Fig. 1a and d). This is distinctively different from the relative resistance that CA2 neurons exhibit to glutamate excitotoxicity (Fig. 1b and d). These and other properties we later characterized not only helped to identify the CA2 subregion in our study (Fig. 1c), but also provided evidence that CA2 neurons are functionally distinctly different from their neighboring CA1 and CA3 neurons.

To further determine subregion-, time-, and duration-dependent sensitivity of hippocampal neurons to OGD, we subjected slice cultures to 25, 40, and 45 min OGD and measured PI uptake in CA1, CA2, CA3, and DG before, 3 h, and 24 h after OGD exposure. Slice cultures exposed to OGD for 25 min exhibited little neuronal damage after 3 h reperfusion, and there was only very mild PI uptake in CA1 subregion after 24 h (Fig. 2a and d). However, 40 min OGD induced marked PI fluorescence uptake in the CA1 and CA3 subregions (Fig. 2b, d, and e), with less neuronal damage detected in DG granule neuronal layer and CA2 subregions. Significant injury was detected in all subregions including CA2 and DG after 24 h reperfusion following a further increase in duration of OGD exposure (Fig. 2c, d, and e). There was marked PI uptake in all subregions after OGD for 45 min or longer (data not shown). These data suggest that CA1 and CA3 neurons are more vulnerable to OGD than CA2 neurons and DG granule neurons. Relative resistance of DG was exhibited in two ways: (i) only more severe insults (longer duration of OGD) induced cell injury and (ii) a lower percentage of cell death was seen compared with other subregions subjected to the same milder insults. As DG has higher cell density compared with CA subregions, a comparable PI uptake value represents a smaller percentage of cell death than CA subregions (Fig. 2).

Insulin/IGF protects against cell injury induced by OGD

The hippocampus has a high number of IRs (Steen et al. 2005). Insulin/IGF-1 has been shown to have neuroprotective effects against memory loss in humans (Benedict et al. 2007), and plays a role in promoting survival of newly formed neurons in DG in diabetic animal models (Lang et al. 2009). Thus, we next examined the effects of insulin and IGF-1 on responses of hippocampal subregions to OGD. We chose a 45 min OGD for further study as there was marked PI uptake in DG after this exposure time (Fig. 2). Quantification of PI uptake showed that pre-treatment for 24 h with insulin (1–10 μM) or IGF-1 (0.1–1 μM) dose-dependently reduced OGD-induced injury in all subregions with differential subregional sensitivity (Fig. 3a–c). Both insulin and IGF treatments protected neurons with similar subregional selectivity. At lower concentrations, insulin/IGF markedly reduced injury in DG and CA2 subregions, but not CA1 and CA3 (Fig. 3). A reduction in PI uptake in CA1 and CA3 was observed at higher concentrations of insulin (10 μM) and IGF (1 μM), although only the latter reached statistical significance (Fig. 3c). The lack of effect of insulin in the CA1 subregion is not because of a more severe injury in CA1 than the DG subregion, because insulin did not protect CA1 even with less severe insults while there was little injury in DG (50 ± 5% and 53 ± 5% cell death induced by OGD in CA1 in the absence and presence of insulin, respectively). Thus, sensitivity of hippocampal subregions to insulin/IGF-mediated neuroprotection is higher in DG and CA2 than CA1 and CA3, which is opposite to the order of their sensitivity to OGD insult (Fig. 2). As 1 μm insulin-induced submaximal effects in the DG subregion (Fig. 3d), we chose the dose of 1 μM insulin to further investigate a probable mechanism by which insulin exerts its subregional selective neuroprotective effects.

The neuroprotective effect of insulin involves PI3K/Akt pathway

Insulin can activate several signaling pathways, including PI3K/Akt pathway. We next carried out experiments using inhibitors of PI3K signaling such as wortmannin (Wortm; a potent specific PI3K inhibitor), and triciribine (a specific inhibitor of Akt). Under basal conditions, wortmannin (0.1–1 μM, 24 h) did not have any effects on the PI uptake. However, at higher concentrations, Wortmannin (10 μM) selectively induced cell injury in the DG subregion, which was attenuated in the presence of insulin (Fig. 4a and b). Inhibition of PI3K by wortmannin preferentially induced cell injury in the inner DG granule cell layer where newly formed neurons are found (Fig. 4a; Parent 2003). These data suggest that under normal conditions DG neuronal survival is more critically dependent on PI3K signaling than CA regions, possibly via neurogenesis-associated survival mechanisms. Although wortmannin alone did not appear to alter OGD-induced neuronal injury in CA and DG, the protective effects of insulin were prevented in the presence of Wortmannin, even at concentrations that had no effects on basal conditions (Fig. 4c). These data indicated that insulin protects neurons from OGD-induced injury through activating PI3K signaling. Similarly, triciribine selectively induced cell injury in the DG subregion at a higher concentration (10 μM, 24 h) under basal conditions which was prevented by insulin (Fig. 5a and b). However, triciribine (1 μM) prevented insulin-induced reduction in OGD-induced injury (Fig. 5c). CA1 and CA3 subregions were less sensitive to basal PI3K/Akt inhibition as well as protective effects of insulin against OGD. The CA2 subregion is unique in that it was less sensitive to endogenous PI3K/Akt inhibition compared with DG, but more sensitive to insulin protection than CA1 and CA3 (4, 5). Taken together, these data suggested that dentate neurons are more dependent on endogenous PI3K/Akt signaling pathways than CA neurons in the maintenance of neuronal survival/death homeostasis, and the action of insulin in protecting dentate neurons from OGD-induced injury is critically dependent on PI3K/Akt signaling pathway.

Insulin neuroprotection is not mediated by MAPK pathway

As insulin was shown to activate both PI3K and mitogen-activated protein kinase (MAPK) pathways, we next studied the role of MAPK pathway in insulin action using PD98059 (10 μM), an inhibitor of Mitogen-activated protein kinase kinase (MEK)/extracellular signal-regulated kinase pathway, U0126 (1–10 μM), a potent and selective MEK inhibitor, and SP600125 (1 μM), a c-Jun N-terminal kinases (JNK) inhibitor. We found that none of the three inhibitors had any effects on basal cell survival in CA and DG subregions at the concentrations used (data not shown), nor did they prevent insulin-induced reduction in PI uptake against OGD-induced cell injury in DG and CA2 (Fig. 6). These data suggested that the neuroprotective effects of insulin are not mediated by activating MAPK pathway. Furthermore, it appeared that blocking MAPK pathway further protected CA1 neurons in the presence of insulin (Fig. 6a), suggesting that activation of MAPK signaling by insulin may contribute to insensitivity to insulin/PI3K/Akt-mediated protection. However, we did not see a consistent effect of MAPK inhibitors alone on OGD-induced neuronal injury, either pre-treated for 24 h or 1 h prior to OGD (data not shown). As PI3K and MAPK signaling has been shown to act antagonistically (van der Heide et al. 2003; Hui et al. 2005), we hypothesized that the effects of MAPK signaling may be enhanced by blocking PI3K. Indeed, compared with wortmannin (10 μM) alone, U0126 (10 μM) in the presence of wortmannin markedly reduced OGD-induced cell injury in CA subregions but to a lesser extent in DG (PI uptake was 310 ± 19 vs. 187 ± 45 in CA1, 310 ± 22 before vs. 260 ± 29 in DG, wortmannin alone vs. wortmannin + U0126, respectively, n = 12). These data suggested that the insulin/MAPK pathway may act to antagonize insulin/PI3K/Akt and contribute to the relative resistance of CA1 and CA3 to insulin/IGF protection, as well as their relative vulnerability to OGD.

NMDA receptor-dependent OGD-induced cell injury

It was shown that insulin modulates CA1 neuronal activity in a NMDA receptor-dependent manner (van der Heide et al. 2005). Thus, its action may involve direct (synaptic activity independent, IR mediated) or indirect (synaptic activity dependent, NMDA mediated) actions. Thus, we next used NMDA receptor blocker 3-(z-carboxypiperazin-4-yl) propyl-1-phosphonic acid (CPP) to examine the mechanisms of cell injury induced by OGD in insulin pre-treated slices. We found that CPP (10 μM) in the presence of insulin (1 μM) nearly completely blocked 45 min OGD-induced PI uptake in all subregions (Fig. 7a–c). We also tested protective effects of insulin and CPP against 50 min OGD. We found that 10 μM CPP alone did not significantly reduce PI uptake induced by OGD (PI uptake was 276 ± 22 vs. 264 ± 47 in the absence vs. presence of CPP, n = 11). At a higher concentration, CPP (30 μM) induced partial neuroprotection, as was seen with 1 μM insulin, which selectively protected DG and CA2 neurons (Fig. 7d). However, a combination of CPP (10–30 μM) and insulin (1 μM) nearly completely blocked the cell injury induced by OGD in all CA subregions (Fig. 7). These data suggested that with a more severe OGD insult, multiple mechanisms may need to be targeted to effectively prevent neuronal death.

Insulin receptor β and basal phospho-Akt immunoreactivity in hippocampal slice cultures

Insulin receptor mRNA and protein have been found in brain regions including hippocampus from in vivo animal models (Marks et al. 1990; Zhao et al. 1999; Park et al. 2009). We next tested the distribution of IRs and endogenous phospho-Akt (p-Akt) in our slice culture preparations, using antibodies against IRβ subunit (C-terminal) and p-Akt (Ser473) in untreated slice cultures. We found that IRβ immunoreaction was detected in all subregions of hippocampus, with higher expression in the neuropil than in cell body layers (Fig. 8a, n = 4). A strong signal was detected in the molecular and polymorphic layers of the DG, in the alveus and stratum lacunosum-moleculare, the mossy fibers and Schaffer collaterals in the CA regions (Fig. 8a). In contrast, higher levels of endogenous p-Akt were detected in the pyramidal cell layer of CA subregions and the molecular layer of the DG (Fig. 8b, n = 14), with lower expression in the polymorphic layer and the inner granule cell layer of DG, where neurons were most sensitive to blockade of PI3K/Akt (4, 5, 8). Our data confirmed that insulin signaling is detected in our slice cultures and its activation may protect neurons from OGD-induced injury.