Deoxyglucose prevents neuronal loss mediated by microglia
To test whether DOG was neurotoxic or neuroprotective, we added 10 mM DOG to primary co-cultures of neurons and glia from rat cerebellum for 96 hours (from 7 to 11 DIV), and observed that there was an apparent increase in the density of live neurons compared with the untreated condition (Figure 1A,D). The density of live microglia substantially decreased over 96 hours of DOG treatment (Figure 1B) but there was no effect on astrocytes (Figure 1C). Thus, in response to DOG, the density of neurons apparently increased by about 30% (relative to untreated cultures) and microglia were almost eliminated, while astrocytes were unaffected. The effect on neuronal density was dependent on the time of incubation with DOG (Figure 2B), with more live neurons surviving with longer treatments with DOG.
The apparent increase in neuronal density induced by DOG could be due to preventing any spontaneous neuronal loss occurring in the untreated cultures, because long-term culture of neurons in the presence of microglia can result in spontaneous neuronal loss [20, 21]. To test whether spontaneous neuronal loss was occurring in the untreated cultures, we quantified neuronal density between 7 and 11 DIV, and confirmed that neuronal loss was occurring in the untreated cultures (from 407.5 ± 43.5 to 293.5 ± 4.5). We imaged the same cells at 7 DIV and at 11 DIV, confirming that there is a loss of neurons (Figure 1E).
To determine whether microglia were involved in this neuronal loss over time, we treated the neuronal–glial cultures with 50 mM LME for 4 hours to specifically eliminate microglia [25]. We observed that when microglia were eliminated at 7 DIV, there were more live neurons by 11 DIV, and that addition of DOG from 7 to 11 DIV no longer increased neuronal density in the absence of microglia (Figure 2A). Elimination of microglia is thus sufficient to prevent spontaneous neuronal loss in culture and to prevent the protection of those neurons by DOG. As DOG also depletes microglia from the cultures (Figure 1B), the neuronal protection by DOG appears to be due to DOG depleting the microglia and thereby preventing the spontaneous neuronal loss.
Deoxyglucose prevents inflammatory neuronal loss induced by amyloid beta or trauma, but exacerbates hypoxic neuronal death
To test whether DOG may be neuroprotective or neurotoxic in particular neurological diseases and conditions, we set up three different pathological models in culture that relate to Alzheimer’s disease, brain trauma and stroke.
We have previously reported that nanomolar levels of Aβ induce neuronal loss over 3 days in neuronal–glial co-cultures that is prevented if microglia are eliminated by LME [17, 26]. We thus tested whether DOG could prevent this neuronal loss, and indeed found that 10 mM DOG prevented the neuronal loss induced by 250 nM Aβ (Figure 3A). Aβ-induced neuronal loss, when mediated by microglia, can thus be prevented by DOG, presumably by eliminating the microglia.
Brain trauma can cause delayed neuronal damage mediated by microglia [18, 19]. We modeled brain trauma in vitro by adding disrupted neurons to live neuronal–glial co-cultures, and found that this increased the loss of live neurons over 96 hours (Figure 3B). Disrupted neurons were prepared by scraping and homogenizing neuronal–glial cultures (which are 85% neurons), resulting in cellular debris and released content. Addition of 10 mM DOG prevented the neuronal loss induced by disrupted neurons (Figure 3B). DOG might thus be neuroprotective in brain trauma by depleting microglia.
A number of brain pathologies, such as stroke, vascular dementia and trauma, are associated with hypoxia, so we tested whether DOG could prevent or exacerbate hypoxia-induced neuronal death. Moreover, stroke can be ischemic (decrease of blood flow), which can be mimicked in vitro by combining hypoxia and glucose deprivation with DOG. Neuronal–glial cultures were exposed to hypoxia (2% oxygen for 12 hours) ± DOG (10 mM) plus 4 days of reoxygenation. As previously, there were significantly more live neurons when DOG was added in normoxic (21% oxygen) conditions (Figure 4A) associated with microglial depletion (Figure 4B). In the absence of DOG, hypoxia induced a small loss of live neurons (Figure 4A). However, in the presence of DOG, hypoxia induced massive neuronal death (Figure 4A), presumably because during hypoxia neurons are more reliant on glycolysis for energy production. DOG can thus be neuroprotective or neurotoxic depending on conditions.
Although DOG decreased microglial levels at both 21% and 2% oxygen, the decrease was less in the latter, hypoxic conditions (Figure 4B). We do not know the reason for this – perhaps oxygen-derived radicals are involved in the cell death, or hypoxia upregulates remaining glycolysis, or the neuronal death occurring in these conditions causes microglial proliferation.
Microglia are killed by deoxyglucose via ATP depletion and phagocytosis
DOG depleted microglia in neuronal–glial cultures over 96 hours (Figure 1B), so we tested whether DOG would kill or deplete microglia in a glial culture (Figure 5A), a pure microglial culture (Figure 5B) or the microglial cell line BV-2 (Figure 6). Twenty-four hours of DOG treatment caused a significant decrease in microglial density in all three cultures and significant increases in microglial necrosis (measured by propidium iodide uptake) in glial cultures and the BV-2 cell line (Figures 5 and 6). The decrease in microglial density in BV-2 cultures was substantially greater with 10 mM DOG than with 1 or 5 mM DOG (Additional file 1: Figure S1), so we used 10 mM DOG in subsequent experiments. The proportion of microglia that were chromatin condensed but had not taken up propidium iodide (a measure of apoptosis) was low in all conditions. The proportion of DOG-treated BV-2 microglia that exposed phosphatidylserine (measured by Annexin V binding; Figure 6C) increased with time, preceding the cells becoming necrotic (measured by propidium iodide uptake; Figure 6D), which could be due to apoptosis or energy depletion-induced phosphatidylserine exposure.
We tested whether the broad-spectrum caspase inhibitors ZVal-Ala-d,l-Asp(OMe)-fluoromethylketone and bocaspartyl (OMe)-fluoromethylketone could prevent DOG-induced death of microglia. However, there was no such protection (Additional file 2: Table S1), suggesting that apoptosis is not involved.
DOG can kill cancer cells by interfering with N-linked glycosylation, which is reversible by exogenous addition of mannose [27, 28]. We tested whether addition of 2 mM mannose for 48 hours could prevent DOG-induced microglial loss in primary glial cultures, but found no effect of mannose on microglia loss (Additional file 2: Table S1).
DOG can activate sirtuins, such as Sirtuin 1, in cells via inhibiting glycolysis and increasing NAD+ levels. Sirtuin 1 is a NAD+-dependent histone deacetylase that is inhibited by nicotinamide. However, we found that nicotinamide (at 100, 300 or 600 μM) was not able to prevent DOG-induced microglial death (Additional file 2: Table S1), suggesting that sirtuins are not involved in this death.
Autophagy can be activated by DOG in some cells [29]. We tested whether the autophagy inhibitor chloroquine (10 or 25 μM) could prevent DOG-induced microglia death, but found no effect (Additional file 2: Table S1).
DOG can deplete cellular ATP by inhibition of glycolysis [1, 2]. We therefore tested whether DOG could deplete ATP in microglia prior to any cell death or loss. Indeed, we found that DOG caused marked ATP depletion after just 1 hour of DOG treatment of pure, primary microglia (Additional file 3: Figure S2A) or BV-2 microglia (Figure 6B) prior to any increase in cell death (Figure 6D). This suggests that DOG induces death of microglia by ATP depletion, and that microglia are dependent on glycolysis for energy generation.
To test whether DOG could stimulate or inhibit the activation of microglia prior to any significant microglial death, we measured TNFα levels in the culture medium of pure, primary microglia at 3 hours after addition of 10 mM DOG. We observed a small increase in TNFα production/release (Additional file 3: Figure S2B) prior to any increase in cell death (Figure 5B).
As the loss of microglia could be due to phagocytosis by other microglia, we tested whether deoxyglucose could stimulate or inhibit microglial phagocytosis. Pure microglia were treated with 10 mM deoxyglucose for 21 hours and then 5 μm carboxylate-modified latex microspheres were added for 2 hours. We observed a very small increase in phagocytosis of microspheres with DOG treatment (Figure 7A).
Dead and dying cells can be rapidly removed by phagocytosis, and microglia are particularly active in this type of phagocytosis [15]. To test whether phagocytosis was involved in the microglial loss induced by DOG, we added an inhibitor of phagocytosis: cytochalasin D (1 μM) together with 10 mM DOG for 6 hours. Cytochalasin D delayed the loss of microglia induced by DOG treatment (Figure 7B), indicating that phagocytosis may be involved in the microglial loss. Annexin V is a phosphatidylserine-binding protein, which when added extracellularly can block phosphatidylserine-mediated phagocytosis. However, addition of Annexin V (at concentrations that we have previously found to block microglial phagocytosis: 100 nM) could not prevent the microglial loss induced by DOG (Additional file 2: Table S1), suggesting that eat-me signals other than phosphatidylserine are involved in the phagocytosis of dying microglia by microglia.