Human frozen tissues and formalin-fixed sections were obtained from the Alzheimer's Disease Research Center, Washington University School of Medicine and the Alzheimer's Disease Research Center, Case Western Reserve University. Experimental procedures involving these samples were approved by the IRB of above two institutions and the Rutgers University IRB. All animal procedures carried out in this study were in accordance with Rutgers University IACUC standards, approval ID: 06-027.

Frozen brain tissues or formalin-fixed/paraffin-embedded 10 µm sections from AD and age-matched controls were obtained from the Alzheimer's Disease Research Centers at Case Western Reserve University and Washington University School of Medicine. Some of the cases used for this study had been clinically diagnosed with AD and were subsequently confirmed with standard pathological examination. Other cases died with no explicit diagnosis of their dementia, but were scored on neuropathological examinations as being either Braak stage V or VI. For the purposes of this study, we defined these cases as AD.

Two AD mouse models were examined ^[21] in this study. Both carry the entire human APP gene inserted into the mouse genome from a microinjected yeast artificial chromosome. Mice from the R1.40 line (B6•129-Tg(APPSw)40Btla/J) carry the human APP gene with the Swedish mutation for familial AD – K670N/M671L. Mice from the 8.9 line (B6•129S2-Tg(APP)8.9Btla/J) are similar to R1.40 animals but carry a wild type APP gene. Both transgenic lines express all mRNA and protein isoforms of the human gene in a correct spatiotemporal pattern.

Embryonic cortical neurons were isolated by standard procedures. E16.5 embryonic cerebral cortices were treated with 0.25% Trypsin-EDTA and dissociated into single cells by gentle trituration. Cells were suspended in Neurobasal medium supplemented with B27 and 2 mM glutamine, then plated on coverslips or dishes coated with poly-L-Lysine (0.05 mg/mL). All cultures were grown for a minimum of 5 days in vitro (DIV) before any treatment. Murine neuroblastoma N2a cells were purchased from ATCC (Manassas, VA, USA) and cultured in standard DMEM media supplemented with FBS (2% for differentiation, 10% for routine passage) or Neurobasal+B27 with or without glutamine. Synthetic amyloid-β peptide (Aβ1–42) was purchased from AnaSpec, Inc. (San Jose, CA). Peroxide oxidized Aβ1–42 dimer was prepared from synthetic human Aβ1–42 (5 µM) by incubation in PBS in the presence of hydrogen peroxide plus Cu2+ as previously described ^[22].

An expression construct encoding human GS was purchased from Origene (Cat# SC118847, pCMV6-XL5-GLUL, GenBank accession No.: NM_002065). The disease-related mutation (1021C→T) was introduced by means of site-directed mutagenesis using the system developed by Mutagenex, Inc. The success of the mutagenesis was confirmed by DNA sequencing using a GS internal forward primer (Flk-F): [gene: 5′-GACCCCTTCCGTAAGGACCC] and a vector reverse primer (pCMV6 Seq-R): [gene: 5′-TTAGGACAAGGCTGGTGGGCAC]. Sub-confluent N2a cells were transfected with either Lipofectamine™ 2000 or Lipofectamine™ LTX with PLUS™ Reagent (Invitrogen) following the manufacturer's suggested procedure.

For Western blots, protein extracts from tissues or cultures were made in RIPA buffer with protease inhibitors and phosphatase inhibitors, and then separated in 4–20% SDS–PAGE gel.

All antibodies used are commercially available. Antibodies against GS, 53BP1, ATM, phospho-S1981-ATM (^PS1981-ATM), cleaved caspase 6, LC3, LC3B,, PCNA, cyclin A, PSD-95 and VAMP2 were all purchased from Abcam. Antibodies against cleaved caspase 3, ATR, phospho-S428-ATR (^PS428-ATR), and phospho-S807/811-retinoblastoma (^PS807/811-RB) were from Cell Signaling. The Map2 antibody was purchased from Sigma. Antibodies against γH2AX, Tau3R, Tau4R and phospho-tau (AT8) antibodies are all mouse monoclonals purchased from Millipore or Thermo. Synaptophysin antibody was from Invitrogen. β-actin antibodies were from Santa Cruz Biotechnology. Iba1 antibody was from Wako (Japan). The specificity of these antibodies has been established previously and confirmed by us through the use of Western blots.

The comet assay procedure was carried out using OxiSelect ™ Comet Assay Kit from Cell Biolabs following the procedure suggested by the manufacturer. Briefly, cultured cells were detached from the culturing vessels by trypsinization. Cells were then mounted in soft gels and electrophoresed for 15 minutes at 1 V/cm. After electrophoresis, slides were washed in water and dried at room temperature. Finally, slides were stained with Vista Green DNA dye (Cell Biolabs, Inc.) and visualized under a Leica DM5000B fluorescent microscope or a Zeiss LSM 510 confocal microscope using the FITC filter.

Each mouse was transcardially perfused with 4% paraformaldehyde under deep Avertin anesthesia, after which the brain was dissected from the skull and post-fixed overnight. For most material, the brain was embedded in paraffin and sectioned at 10 µm on a rotary microtome. The human materials were formalin-fixed, paraffin-embedded and sectioned at 10 µm. These were treated as described previously ^[23]. For immunohistochemistry, all paraffin sections underwent antigen retrieval with high temperature citrate buffer for 20 min, and then soaked in 0.3% hydrogen peroxide to remove endogenous peroxidase activity. Primary antibody was diluted in 10% goat serum with 0.5% Tween-20. Primary antibodies were detected using biotinylated goat anti-rabbit or anti-mouse secondary antibody (1∶400), avidin-biotin complex horseradish peroxidase and peroxidase substrates including DAB, DAB+Ni and VIP (Vector Laboratories). Some sections were counterstained with hematoxylin QS. For immunofluorescence, mouse brains were embedded in OCT (Tissue-Tek) and sectioned on a cryostat at 10 µm. Sections were pretreated with citrate buffer for 10 minutes, and then incubated with primary antibody. Alexa linked secondary antibodies were used to detect the presence of the antigens. Stained sections were photographed and viewed at a final magnification of 200 using Leica Application Suite/Leica DM5000B.

For both LDH assay and MTT assay, cells were cultured in media without phenol red. The LDH activity of the supernatant was measured by Promega CytoTox 96® Non-Radioactive Cytotoxicity Assay, according to the manufacturer's instructions, recording absorbance at 490 nm using absorbance at 650 nm as reference. Cell viability (MTT assay) was evaluated using a Promega CellTiter 96® Non-Radioactive Cell Proliferation Assay kit following the manufacturer's instructions, recording absorbance at 570 nm using absorbance at 650 nm as reference.

Neurons take up exogenous glutamine as part of the glutamate/glutamine cycle. Brain astrocytes clear glutamate from the synaptic cleft, convert it glutamine and secrete it to the extracellular space. Here it is picked up by neurons and converted back into glutamate for use as a transmitter or used as a precursor for other cellular components. Astrocytes typically maintain GS levels that are readily visualized by immunocytochemistry; in neurons under normal physiological conditions, GS is nearly undetectable. When exogenous glutamine levels drop, however, neurons can be induced to express GS ^[11]. Thus, although the levels of exogenous glutamine can be easily regulated, the levels of endogenous glutamine are a more complex function of uptake plus the activity of GS.

We found that in cultures of primary cortical neurons, reduction of exogenous glutamine had a small negative effect on neuronal viability (Figure 1A, C; quantified in Figure 1E). As the full effects of low media glutamine on the levels of intracellular glutamine might be offset by GS-induced glutamine synthesis, we tested the effects of the GS-inhibitor, methionine sulfoximine (MSO). MSO is the best known inhibitor of GS. It was originally isolated from nitrogenchloride-treated zein as the toxin responsible for the induction of convulsions, hysteria and epileptic fits in a number of animals (see reviews by Eisenberg et al. 2000) ^[24]. When added to neuronal culture media in the presence of 2 mM glutamine, MSO alone also caused the death of only a small number of neurons (Figure 1B, E). By contrast, when MSO was added to cells grown in glutamine-free medium (thus reducing both endogenous and exogenous glutamine), massive neuronal cell death was observed (Figure 1D, E). The reductions in exogenous and endogenous glutamine were synergistic. In normal medium there is a baseline level of caspase-3 labeling of about 7–8% of the total number of cells. MSO alone increased the rate of cell death by ∼3-fold; glutamine removal from the medium alone increased cell death by ∼2-fold. Applying MSO in the absence of glutamine, however, increased cell death by ∼8-fold. Thus, a healthy neuron requires glutamine both in the media and synthesized internally by the actions of GS.

In normal human hippocampus, astrocytes were the only cell type with detectable levels of GS by immunocytochemistry (Figure 2A). In AD patient samples, consistent with previous reports ^[25], we found numerous hippocampal neurons with elevated levels of GS (arrows, Figure 2A). Astrocyte staining was not noticeably altered in these AD samples. The percentage of GS-positive AD pyramidal neurons varied from subject to subject and from region to region, ranging from 1–50% of the total population. In the 4 control cases we examined, GS-positive neurons were rare (<1%). We validated these findings by Western blot using frozen brain samples from prefrontal cortex (Brodmann area 9 – BA9) of 20 additional subjects (12 AD; 8 control). As with the hippocampal immunohistochemistry, there was variability among the samples. These blots clearly indicate, however, that GS levels were elevated in frontal cortex of AD patients compared with controls (Figure 2B). When these results were averaged across all subjects, the increases in GS monomer and a higher molecular weight species that is likely to be a GS dimer were both highly significant (Figure 2C).

To verify that up-regulation of GS in neurons is triggered by glutamine deprivation, we cultured mouse primary cortical neurons (DIV 7–8) in Neurobasal media with or without glutamine, in the presence or absence of 5 mM MSO, for 3 days before collecting samples for Western blotting and immunocytochemistry. In normal Neurobasal medium with 2 mM glutamine, neurons express only trace amount of GS. As predicted by the in vivo work, GS expression increased robustly when primary neurons were deprived of glutamine either by removing exogenous glutamine from the media (Figure 2D, lane 1) or by blocking endogenous GS activity by MSO (Figure 2D, lane 4). The same relationships were found in N2a cells, a murine neuroblastoma cell line (Figure 2E).

Deprivation of exogenous glutamine can cause higher GS expression in neurons. The question we next addressed was whether this de novo expression of GS in neurons is protective or detrimental. Our results further underscore the complex nature of glutamine regulation in cells. We tested two GS constructs expressing the human GS transcript 1: wild type GS (GS^WT) and an activity-deficient mutant, GS^R341C. The GS^R341C construct harbors a point mutation, R341C (a T→C change at nucleotide 1021 of human GS transcript 1), and encodes a peptide with reduced GS activity ^[8]. N2a cells were transfected with either construct or treated with only transfection reagents as a negative control. After 24 hours, culture media were replaced with differentiation media (Neurobasal without glutamine). After another 24 hours, cells were treated with three different stressors – 100 µM H₂O₂, 100 nM oxidized Aβ_1–42 (oxyAβ), or 5 µM etoposide – for 24 hours in the same media. Then cells were processed for Western blotting or immunocytochemistry. Extensive degeneration of both processes and cell bodies was observed in GS^WT but not GS^R341C -transfected cells (Figure 3A). The levels of cleaved caspase-3 and caspase-6 were substantially higher in cells with high GS activity (GS^WT transfection) than in cells expressing the GS^R341C mutant (Figure 3B), even though comparable amount of the two proteins were present. We repeated the transfections using cells grown in medium containing 2 mM glutamine. Intriguingly, with ample exogenous glutamine, high GS expression reduced caspase-3 activity – as measured by the cleavage of key cellular proteins such as retinoblastoma (RB, Figure 3C, D). To directly evaluate cell death we treated N2a cells with H₂O₂ and measured the lactate dehydrogenase (LDH) activity in culture media, The LDH results confirmed the caspase 3 and caspase 6 data (Figure 3E). First, low glutamine increases cell death and second, the difference between wild type and mutant GS is only seen in low glutamine. Viewed together, our data point to the conclusion that high levels of exogenous glutamine are highly beneficial to neuronal cells whereas treatments that lead to high GS activity may be useful only in certain conditions.

The results in Figure 3 show that cells in low glutamine are more sensitive to the effects of a variety of different stressors. To explore this effect further, we treated N2a cells grown in normal or reduced glutamine with H₂O₂ (1 mM – a model of oxidative stress); ZnSO₄ (0.2 mM – a model of heavy metal toxicity); and etoposide (20 µM – a model of DNA damage). Cells were stressed for 30 minutes then harvested for Western blot analysis. The results in Figure 4A illustrate that cells grown in any medium can respond to each of these conditions. They up-regulate their levels of several stress response proteins such as 53BP1 and appropriately modify other proteins either by phosphorylation (ATM) or cleavage (caspase-3). The magnitude of all of these responses, however, is considerably diminished in low glutamine. For example, in all conditions, the levels of 53BP1 protein increased in response to stress. In low glutamine, however, the final levels of protein were significantly less than in normal glutamine. Activation of ATM and ATR by phosphorylation revealed a similar picture. Under oxidative stress or etoposide-induced DNA damage, the activating phosphorylation of both proteins increased. Etoposide treatment drove robust ATM auto-phosphorylation, this change in phosphorylation led to an actual change in ATM kinase activity can be seen in the increased levels of γH2AX, while the phosphorylation of ATR was driven preferentially by ZnSO₄. For each of these measures of stress response, however, the magnitude of the response in low glutamine was significantly less than in normal exogenous glutamine. One likely reason for this difference is that the levels of total ATM and total 53BP1 are reduced in low glutamine (Figure 4A–B). Viewed as a whole, the data in Figure 4A suggest that in many different domains, cells in low glutamine are not well prepared for stress.

We presumed that the findings in Figure 4A reflected the reduced levels of the various damage response proteins. This assumes, however, that the levels of damage induced by the various stressors are equivalent in low glutamine. To address this concern, we examined the response of the cells to etoposide since DNA damage can be easily monitored by Comet assay. In the absence of etoposide, DNA damage is minimal, even in low glutamine (Figure 4C). After 30 min in 5 µM etoposide, DNA fragmentation is easily revealed by the appearance of significant comet tails (top row, Figure 4C). We then removed the etoposide and followed the cells' ability to repair their DNA. Within 4 hours, the Comet tails found on cells in both culture conditions disappeared, indicating the cells were able to join the majority of their DNA breaks, with or without glutamine (Figure 4C). Yet while the DNA might have been “repaired” in both situations, there was significantly more cell death in the glutamine-deprived cultures as assessed by the number of pycnotic nuclei (Figure 4D). The majority of this death is likely necrosis not apoptosis because markers of apoptosis (activated caspase-3) and autophagy (LC3-II) were higher in cultures with glutamine than in those without glutamine (Figure 4A, B). To directly evaluate cell death and viability, we also performed LDH assay after a 6 hour recovery (Figure 4E) and MTT assay after a 24 hour recovery (Figure 4F) from a 30 minutes H₂O₂ challenge. Collectively, glutamine is essential for the stress response and post damage survival of cells.

N2a cells grown in low glutamine are sensitive to toxic effect of DNA damage and oxidative stress, likely due to low expression of stress response proteins such as ATM and 53BP1(Figure 4A–B). However, this impaired response capability is not unique to the N2a cells; similar reductions of ATM, ATR and 53BP1 are detected in primary neurons deprived of glutamine (Figure 5A). Like N2a cells, neurons grown in low glutamine were more sensitive to DNA damage as well (Figure 5B). The importance of glutamine to the survival of neuronal cells prompted us to examine the effect of glutamine supplement on Aβ-induced toxicity. Primary neurons were cultured in medium with 2 mM glutamine (replaced every 2–3 days) until DIV 10–15. Cells were then re-fed with fresh medium with or without glutamine. Twenty-four hours later, the cells were treated with 100 nM oxyAβ for additional 24 hours; oxyAβ has been shown to more readily form neurotoxic oligomers and is more potent than un-oxidized Aβ in its ability to induce cellular stress in neurons ^[22]. We monitored the neuronal response using the AT8 phospho-tau antibody. Although oxyAβ induced tau phosphorylation in cultures with or without glutamine, the levels of phospho-tau were significantly reduced in the presence glutamine (Figure 5C). The effects of glutamine were dose dependent as can be seen from the effects on tau phosphorylation (Figure 5D). Even neurons in long-term established cultures responded to 3 days of glutamine deprivation and Aβ treatment by showing more processes beading (Figure 5E) and less synaptophysin staining (Figure 5F) suggesting reduced synaptic density.

These in vitro findings provide multiple pieces of evidence that under conditions of neuronal stress such as is found in neurodegenerative disease, glutamine might have significant clinical value. The Aβ results further suggest that Alzheimer's disease might be a prime candidate for intervention. To test this concept in a preclinical disease model, we examined the effects of dietary glutamine on the phenotypes of the R1.40 and 8.9 transgenic mouse strains. Both lines carry an entire genomic copy of the region of human chromosome 21 with the APP gene ^[21]. The R1.40, mice develop amyloid plaques by 13 months (on a C57BL/6J background); 8.9 mice remain plaque-free throughout their lives on this background. Both lines develop neuronal cell cycle events in brain regions homologous to those affected in humans ^[26], ^[27]. These cell cycle events are robust outcome measures that are useful in tracking impending neurodegenerative events. They are induced by the AD-related brain inflammation as they can be prevented by NSAID treatment or advanced by an additional immune challenge such as LPS injection ^[28]. Given that LPS infusion in healthy human subjects causes a dramatic depletion of glutamine from their brains ^[29], we speculated glutamine may have a significant effect on LPS-treated R1.40 and 8.9 mice.

One-year-old R1.40 and APP8.9 mice were fed 4% glutamine in their drinking water for 10 days. The following day they were injected with a low dose of LPS to induce systemic inflammation ^[28], and sacrificed 2 days later for analysis. Control mice were treated the same way except that they were kept on regular drinking water. We had 3 mice in the glutamine supplement group and 4 mice in the control group. We did extensive comparison of all the markers we used and found no differences between APP8.9 and R1.40. The only exception was the transgene encoded APP; R1.40 mice had higher APP expression than APP8.9 mice, as expected. Instead, we found a significant difference between mice with or without glutamine in their drinking water. As expected, in mice fed with normal drinking water (Gln−), LPS treatment was accompanied by microglial activation, as can be seen from the thickening and shortening of their Iba-1-stained processes (Figure 6A). In mice supplemented with glutamine, however, LPS failed to induce activation. Microglia in these mice (Gln+) showed the normal resting phenotype with thin and long Iba-1-positive processes. Glutamine supplementation also led to a reduced level of tau phosphorylation, as well as ATM phosphorylation (Figure 6B–C). The expression of the cell cycle maker, PCNA (Figure 6B–C) was also reduced with glutamine supplementation. Immunostaining using cyclin A antibody corroborated the PCNA results (Figure 6D, arrows). VAMP2, one of the synaptic vesicle SNARE proteins that is reduced both in AD models and in AD patients ^[29], was restored by glutamine supplementation, as was synaptophysin (Figure 6B–C). The levels of PSD95, however, remained unchanged. Thus, a 10-day regimen of glutamine supplementation was sufficient to block the expression of several indices of neurodegenerative disease in two AD mouse models.