Induction of serine racemase expression and D-serine release from microglia by amyloid β-peptide

As a first test of the role D-serine might play in Aβ-stimulated microglial neurotoxicity, we measured D-serine levels in microglia-conditioned medium. Reverse-phase HPLC was performed on media samples, and conditions were determined under which D-serine could be quantified. Treatment of primary microglia with Aβ_1–42 for 20–24 h resulted in a large increase in D-serine in the medium (Fig. 1). The maximal D-serine concentration varied between experiments, ranging from 115 to 660 μM. LPS also evoked an increase in D-serine levels. Neither Aβ nor LPS caused an elevation of glycine levels, which typically approximated the resting levels of D-serine (e.g., Fig. 1B). MTT assays were also performed, excluding any artifacts of cell number or lysis (not shown).

D-serine is produced primarily by conversion from L-serine by serine racemase. This racemase is known to be expressed in protoplasmic astrocytes in vivo. To confirm its expression in microglia, semi-quantitative RT-PCR was performed on mRNA isolated from several culture types. Serine racemase expression was detected in cultures of primary rat microglia and appeared to increase after activation with Aβ (Fig. 2A). We also surveyed microglial cell lines to exclude possible astrocyte contamination. The HAPI rat microglial line contained serine racemase mRNA after treatment with Aβ (Fig. 2B), as did the N9 mouse microglial line in the presence two other proinflammatory stimuli: sAPPα and LPS (Fig. 2C). The mouse sequence was subcloned and sequenced to confirm identity.

The presence of serine racemase mRNA in activated microglia raised the possibility that increases in expression of this enzyme were responsible for the apparent elevations of D-serine release by Aβ, so western blot analysis was performed on cell lysates from primary microglia. In both cell lysates and positive control samples, the serine racemase antibody detected monomeric protein (~37 kD) and an apparent dimer (~74 kD) (Fig. 3); specificity of the detection was confirmed by a preabsorption control (Fig. 3A). Such oligomers of the enzyme have been described recently and appear to include its soluble, active forms [20]; as reported in that study, we found the serine racemase dimer to be insensitive to reducing agents. Exposure of primary microglia to Aβ had little or no effect on monomeric serine racemase but resulted in significantly higher levels of the apparent dimer (299% of control) (Fig. 3B). Similar inductions were observed in the HAPI microglial cell line.

To address the possibility of a transcriptional induction of serine racemase, a 1.5 kb sequence 5' to the luciferase coding region was cloned from human genomic DNA. This sequence was placed in the pGL3-basic plasmid for luciferase reporter assays. HAPI microglial cells were transfected with this construct and treated with either Aβ or LPS. After one day of treatment, luciferase levels indicated an induction of the presumptive serine racemase promoter by both stimuli (Fig. 4).

Previous experiments demonstrated a release of glutamate by microglia activated with sAPP and Aβ. Useful in those studies were bioassays in which hippocampal neurons were monitored for intracellular ionic calcium concentration ([Ca²⁺]_i) during application of conditioned medium collected from control or activated microglia [12]. As an initial step to determine if proinflammatory activation of microglia might evoke release of NMDA-R agonists other than glutamate, we sought conditions suitable for detecting ligands of the glycine/D-serine site of the NMDA receptor. With no other manipulations, application of glutamate to hippocampal neurons elevated [Ca²⁺]_i to levels that were partially inhibited by a glycine/D-serine site antagonist, 5,7-dicholorokynurenic (DCKA), suggesting synaptic release of endogenous glycine. To circumvent this effect in bioassays of conditioned medium, tetrodotoxin (TTX) was employed. This intervention resulted in a [Ca²⁺]_i response to glutamate that was smaller and insensitive to DCKA (data not shown). Therefore, 800 nM TTX was included in subsequent bioassays of microglia conditioned medium. Under these conditions, graded responses to D-serine could be detected at concentrations from 3–100 μM (data not shown).

Bioassays were performed on conditioned medium from primary microglia activated with either Aβ_1–42 or LPS. Hippocampal neurons responded to such media with a rapid increase in [Ca²⁺]_i (Fig. 5). Conditioned medium from Aβ-treated microglia evoked a modest response at a dilution of 1:100 into the imaging buffer; a 1:18 dilution elevated [Ca²⁺]_i dramatically. Medium from LPS-treated microglia had a similar effect (Fig. 5B). By contrast, the conditioned medium from unactivated sister cultures showed no effect on neuronal [Ca²⁺]_i at ratios up to 1:18 (Fig. 5A) and evoked only a modest increase at 1:10 (Fig. 5B). Acute treatment of neurons with equivalent amounts of Aβ or LPS had no significant effect on [Ca²⁺]_i. DCKA (100 μM) reversed the [Ca²⁺]_i response to microglia-conditioned medium. The elevation was also sensitive to more general antagonists of the NMDA receptor (data not shown). As an independent test of the role of D-serine in the calcium responses evoked by microglia-conditioned medium, samples of media were incubated with D-amino acid oxidase (DAAOx) to remove D-serine; catalase was also added to the imaging buffer to obviate effects of H₂O₂ produced by the DAAOx. Treatment with DAAOx dramatically lowered the ability of microglia-conditioned medium to evoke a [Ca²⁺]_i response (Fig. 5C). Similar elevations of a DAAOx-sensitive NMDA agonist were observed in medium conditioned by the HAPI microglial cell line. Several controls for the specificity of the DAAOx treatments were performed (data not shown): i. exogenous glycine was able to overcome the effect of DAAOx, confirming independence from hydrogen peroxide or similar artifacts of the DAAOx treatment; ii. when samples of conditioned media were treated with DAAOx that had been heat-inactivated, there was little difference from untreated media samples; iii. DAAOx was shown to be specific for D-serine under the conditions of incubation by tests in which the enzyme was incubated with [¹⁴C]glycine or [³H]D-serine, subsequently analyzed by thin-layer chromatography.

To explore the ramifications of D-serine release, we tested the influence of D-serine on neuronal health. Primary hippocampal neurons were exposed to 1 or 3 μM D-serine, and effects on metabolic activity were monitored by MTT assay the following day. Treatment with 1 μM D-serine lowered MTT values to 71.1% of control (±8.08), and 3 μM D-serine resulted in a value that was 11.54% of control (±7.50). D-serine also generally potentiated the toxicity of low levels of glutamate (data not shown).

We next tested whether D-serine played a requisite role in the neurotoxicity exhibited by Aβ-treated microglia. Primary microglia were left untreated or were exposed to Aβ overnight. Conditioned medium from these cells was applied to cultures of primary hippocampal neurons (Fig. 6). A higher rate of LDH release was observed in the presence of conditioned medium from Aβ-treated microglia compared to that obtained from untreated microglia. As a control for the potential neurotoxicity of residual Aβ carried over with the conditioned medium, an aliquot of Aβ was diluted into culture medium in a cell culture dish lacking cells and incubated under identical conditions; treatment of neurons with medium thus prepared showed no significant toxicity under the conditions of our assay (not shown). The neurotoxicity resulting from Aβ-activated microglia was partially reversed by inclusion of 1 or 10 μM DCKA (Fig. 6). Furthermore, pretreatment of the conditioned medium from Aβ-treated microglia with DAAOx also partially reversed its neurotoxicity. Because DAAOx can generate hydrogen peroxide, a separate treatment was tested utilizing catalase, but this did not alter the effect of DAAOx (data not shown).

Based on the inductions by Aβ and other proinflammatory stimuli, the levels of expression of serine racemase in AD brain tissue were examined. RT-PCR of mRNA isolated from hippocampus of AD indicated a significant elevation of serine racemase expression compared to age-matched controls (AMC) (Fig. 7). Within the AD group alone, the female subjects showed a nonsignificant trend towards higher levels than the male subjects. The AD pool contained a higher percentage of females, creating the possibility that gender contributed to the difference observed. However, the difference between AD and AMC was significant within males alone (ratios of racemase:GAPDH signals, AD: 1.62 ± 0.342; AMC: 0.447 ± 0.024; p < 0.05).

The studies presented here document the capacity of activated microglia to express serine racemase and release D-serine, thereby implicating D-serine as a contributor to the neurotoxicity exhibited by inflammatory situations in the CNS. Microglia stimulated with Aβ or LPS released D-serine, a potent NMDA receptor coagonist. The conditioned medium from such microglia elevated [Ca²⁺]_i in cultured hippocampal neurons in a manner that was largely reversed by D-serine/glycine-site antagonists, as well as more general antagonists of NMDA receptors. Pretreatment of the conditioned medium with DAAOx also blocked the effects on neuronal [Ca²⁺]_i. Aβ treatment elevated the steady-state levels of serine racemase mRNA and protein, suggesting that increased synthesis may be involved in the release of D-serine observed under these conditions. Finally, a potential role for D-serine in Alzheimer's disease was further implicated by the observation that serine racemase mRNA is elevated in Alzheimer's brain tissue.

D-serine has gained increased scrutiny as a NMDA receptor agonist that may be more important than glycine in vivo, at least in specific regions or developmental stages. However, the potential contribution of D-serine in excitotoxic pathologies has not received much attention. Damage resulting from intracortical infusion of NMDA is attenuated by an inhibitor of poly-ADP ribose polymerase (PARP), and this neuroprotection was associated with a depression in the levels of D-serine but not glycine [21]. Conventional wisdom held that D-serine/glycine sites on NMDA receptors are typically saturated in vivo, making elevations in their agonists irrelevant. However, this idea has been refuted for over a decade now by observations of responsiveness to infused glycine [22]. Similarly, applications of D-serine have shown dramatic physiological effects [9, 10, 23, 24]. One set of results indicates that much of the biological action of exogenous D-serine may come from stimulation of extrasynaptic NMDA receptors [8]. To the extent that the actions of D-serine on NMDA receptors replicate those of glycine (perhaps, even more potently), the vast literature on exogenous glycine in excitotoxicity paradigms can be translated to D-serine. But issues of production, release, uptake, and catabolism appear distinct for these two glutamate co-agonists, making studies of D-serine a distinct priority.

Release of D-serine from astrocytes can be stimulated by non-NMDA, ionotropic glutamate receptor agonists [25]. This fact has led to the hypothesis that the synaptic elements of astrocytes may contribute to synaptic efficacy by participating in a positive-feedback loop whereby neuronal release of glutamate stimulates astrocytes to release D-serine and further amplify NMDA receptor activation [26]. Recently, data were published consistent with the possibility that serine racemase is activated by direct binding of calcium [20]; notably, AMPA/kainate receptor activation elevates intracellular calcium levels in astrocytes [27]. Therefore, a global release of glutamate – or, in fact, any strong calcium agonist – may lead to extrasynaptic release of D-serine, from both astrocytes and microglia. Similar to its effects on neurons, Aβ can elevate [Ca²⁺]_i in microglia [28], as can LPS [29]. To wit, the degree of elevation in D-serine released into medium was surprisingly high given the changes in serine racemase protein levels, suggesting that some of the Aβ-evoked increase in D-serine release may have come from stimulation of enzyme activity, in addition to expression levels. Detailed time-course analyses of protein levels and D-serine release may provide some insight into this question.

In the experimental paradigms applied here, the presumptive actions of D-serine in microglia-conditioned medium were attenuated by DAAOx. Under the normal conditions of neurotransmission, glutamate concentrations at the synapse are reduced primarily by astrocyte uptake [30]. The mechanisms controlling D-serine concentrations are less clear; the relative contributions of degradation (e.g., by DAAOx), glial uptake, or diffusion out of the synaptic cleft are topics of ongoing research. There appears to be a transporter for D-serine at the synapse [31], but it is incompletely characterized. Neuronal uptake, perhaps to replenish presynaptic stores, would be consistent with the finding that some pyramidal neurons in the cerebral cortex and neurons in the nucleus of the trapezoid body contain D-serine [32]. Degradation of D-serine by DAAOx produces hydrogen peroxide, creating potential for additional harm. However, the concentrations of peroxide predicted from this reaction would be make a relatively minor contribution to neuropathology compared to the potent synergistic activation of NMDA receptors by D-serine and glutamate.

A role for excitotoxicity in CNS inflammation is becoming well established. One of the first analyses of the relative roles of various neurotoxins released by activated microglia found that NMDA receptor antagonists were the most efficacious neuroprotectants [33]. Subsequently, Giulian et al. [34] described an excitotoxin released from microglia exposed to amyloid plaques. Excitotoxicity appears to contribute to neuronal damage in more general models of inflammation as well, such as intracerebroventricular LPS infusion [35]. Several studies have concluded that nitric oxide (NO) mediates microglial neurotoxicity because inducible NO synthase (iNOS) responds to proinflammatory stimuli and general NOS inhibitors can be protective in micoglia-neuron cocultures [36]. However, most such experiments cannot distinguish between NO generated by microglia versus that generated by the neurons themselves through classic excitotoxic mechanisms [37]. When corrected for K_i, inhibitors selective for neuronal NOS are more potent protective agents than are iNOS-selective compounds [12], suggesting that the primary neurotoxic agents microglia produce are excitotoxins that activate nNOS to produce NO within the neurons themselves.

Previously, Li et al. [38] showed that microglial cells synthesize and release IL-1 in response to conditioned media obtained from glutamate-stressed neurons. The neurons respond with an increase in expression and processing of βAPP. Secreted APP and Aβ can stimulate proinflammatory activation in microglia [13, 18], including the release glutamate [12, 39]. These data are consistent with the plethora of evidence linking inflammatory mechanisms to AD pathogenesis [40]. Together with the potential for such stimuli to also trigger release of D-serine, these findings suggest that a vicious circle of inflammation and excitotoxicity may be important in AD pathogenesis. Excitotoxic events are a common aspect of many forms of neurodegeneration, even when they occur secondarily to ischemia or trauma, and considerable evidence suggests that excessive stimulation of glutamate receptors occurs in AD [1‐3].

Free D-serine concentrations are reported to be unaltered in the Alzheimer brain [41, 42]. However, one study found an elevation of overall serine levels in AD CSF per unit volume, but when normalized to protein concentration, the serine levels were similar between AD and controls [43], suggesting that elevated protein levels in AD CSF could confound analyses and interpretations. Our initial analysis here indicates that there is an elevated steady-state level of serine racemase mRNA in AD hippocampus versus age-matched controls. Nevertheless, the elevation of D-serine itself might be expected to occur early in the disease progression; thus, any elevation might be difficult to detect after the disease has progressed to its final stages.

Our semi-quantitative analysis showed that the mRNA for serine racemase was increased in AD brain nearly three-fold relative to age-matched controls. It is possible that a portion of this difference can be accounted for by the hypothetical increase in numbers of astrocytes in AD. However, a similar analysis of GFAP mRNA in AD reported levels to be only 57% higher in AD than in controls [44], and this effect includes an augmented expression per cell [45]. A recent microarray analysis of AD concluded that GFAP mRNA could not be compared to controls reliably due to variability across post-mortem interval, agonal state, etc [46].

In conclusion, Aβ and other AD-relevant proinflammatory stimuli are capable of stimulating release of D-serine from microglia. Together with the release of glutamate evoked by similar conditions, a cooperative activation of NMDA receptors could be anticipated. In addition to delineating details of the mechanisms by which CNS inflammation harms neuronal elements, this line of evidence may be relevant to the development of therapies. If approaches targeting the general inflammatory system or glutamatergic neurotransmission are accompanied by unacceptable contraindications, a more specific interference with D-serine production or release may be more useful in AD and other neurodegenerative conditions. For this reason and others, it will be important to elucidate the mechanisms controlling D-serine synthesis, degradation, and transport under normal and pathological situations.