The D-serine biosynthetic enzyme serine racemase is expressed by reactive astrocytes in the amygdala of human and a mouse model of Alzheimer’s disease

Abstract

Alzheimer’s disease (AD) is characterized behaviorally by cognitive deterioration and emotional disruption, and neuropathologically by amyloid-β (Aβ) plaques, neurofibrillary tangles, and complement C3 (C3)-expressing neurotoxic, reactive astrocytes. We previously demonstrated that C3+ reactive astrocytes in the hippocampus and entorhinal cortex of AD patients express serine racemase (SR), which produces the N-methyl-D-aspartate receptor (NMDAR) co-agonist D-serine. We show here that C3+ reactive astrocytes express SR in the amygdala of AD patients and in an amyloid mouse model of familial AD (5xFAD). 5xFAD mice also have deficits in cue fear memory recall that is dependent on intact amygdala function. Our results suggest that D-serine produced by reactive astrocytes in the amygdala could contribute to glutamate excitotoxicity and neurodegeneration observed with AD progression.

INTRODUCTION

Alzheimer’s disease (AD) is the most common cause of late-onset dementia that is characterized behaviorally by cognitive deterioration and emotional disruption, and neuropathologically by the presence of extracellular amyloid beta plaques and neurofibrillary tangles. The predominant strategy for developing therapies to treat AD has focused on mechanisms that reduce amyloid burden. However, in addition to protein aggregates, reactivate glial cells are a neuropathological hallmark of AD, with large-scale genetic studies implicating perturbations in glial homeostasis as a key contributor to AD pathophysiology [16]. In AD, activated microglia can convert quiescent astrocytes into neurotoxic, reactive astrocytes that highly express complement C3 (C3) [11, 18, 19, 26].

We have shown that a majority of C3+ reactive astrocytes in the hippocampus and entorhinal cortex of AD patients, but not age-matched controls, also express serine racemase (SR), the enzyme that racemizes L-serine to D-serine [3]. Under physiological conditions, D-serine is produced by neuronal SR and is released postsynaptically to serve as the primary co-agonist at synaptic N-methyl-D-aspartate receptors (NMDARs) [5, 24, 33]. However, under inflammatory conditions, such as in AD and following traumatic brain injury (TBI), D-serine is produced by reactive glia [34, 35] and is responsible for mediating the TBI-induced deficits in synaptic plasticity and memory [25, 30]. In the current study, we showed that SR is present in reactive astrocytes of the amygdala, a brain region implicated in AD emotional dysregulation [1], in both human post-mortem brain and in a transgenic AD mouse model.

2. Materials and methods

2.1. Human post-mortem immunofluorescence labeling

Blocks of paraffin-embedded caudal amygdala sections were obtained from the Neuropathology Core of the Massachusetts Alzheimer’s Disease Research Center (ADRC) at the MassGeneral Institute for Neurodegenerative Disease. Cases were assessed by a neuropathologist and were matched for age and gender. “Control” cases (n=4–6 depending on the amygdala nucleus) were selected that had received Braak staging scores of 0/I/II, had sparse plaque pathology, and were rated as having low probability of AD by the National Institute on Aging and the Alzheimer’s Association (NIA-AA) criteria. AD cases were selected that had received Braak staging scores of V-VI (n=6–10 depending on the amygdala nucleus), had frequent plaques and were rated as having intermediate or high probability of AD. See Supplementary Table 1 for subject demographic and neuropathologic details. There were no significant differences between age (p=0.5), sex (p=0.5), and PMI (p=0.5) between controls and AD cases.

Two sections (10μm) from each case were used for triple-antigen immunofluorescent staining as described in our prior publication [3]. The primary antibodies used were mouse anti-SR 1:100 (Santa Cruz Biotechnology; RRID: AB_10847683), rabbit anti-GFAP (1:4,000 (Abcam; RRID:AB_305808), and goat anti-C3 1:150 (MP Biomedicals; RRID: 2334481). IgGk biotinylated anti-mouse secondary antibody (1:50; Santa Cruz Biotechnology), streptavidin Alexa488 (1:2500; Invitrogen), donkey-anti mouse Alex568 (1:250; Invitrogen), and donkey-anti mouse Alex647 (1:250; Invitrogen) antibodies were used for visualization. UltraCruz Blocking (Santa Cruz Biotechnology; sc-516214) was used for all blocking and antibody incubation steps.

2.2. Stereology-based quantification of glia in the human amygdala

Details of our stereology-based sampling methods used for quantification of human postmortem samples can be found as previously described [3, 4, 22]. All cell counting was performed blind to diagnosis. We optimized sampling parameters (counting frame: 150μm² for cortical amygdala and 200μm² for all other amygdala nuclei; grid size: 500μm² for cortical, medial, central, and accessory basal amygdala and 800μm² for the lateral and basal amygdala) to obtain stereological estimates that were within ~10% of the actual counts for analyzed subregions across cases and controls. Using the optical fractionator workflow in Stereoinvestigator (MBF Biosciences), we stereologically estimated the number of glial cells that were GFAP+/C3+ and GFAP+/C3+/SR+ in the amygdala (cortical, medial, central, accessory basal, lateral, and basal amygdala) using our defined sampling strategy. Cell number estimates were normalized to the area of each amygdala nucelus (mm³). No measurable thickness collapse was observed in any of our tissue sections following immunofluorescent processing based on z-axis sampling thickness used for stereological sampling in our previous studies [20, 22, 23].

2.3. Animals

Age-matched 5xFAD (B6SJL-Tg(APPSwFlLon,PSEN1*M146L*L286V)6799Vas/Mmjax; RRID:MMRRC_034840-MU)[21] and wild-type (B6SJLF1/J) littermates were bred and maintained in the animal facility at McLean Hospital. The following group sizes were used for 5xFAD immunohistochemical studies: 2-months-old: n=2/genotype, equal sexes; 4M and 2F; 10 months-old: n=3 male mice/genotype. All animals were grouped housed in polycarbonate cages and maintained on a 12-hour light/dark cycle in a temperature (22°C)- and humidity-controlled vivarium. Animals were given access to food and water ad libitum. The McLean Hospital Institutional Animal Care and Use Committee approved all animal procedures.

2.4. Immunofluorescent staining

Triple-antigen immunofluorescent staining was performed on free-floating sections as we have previously described [3]. In brief, brains were harvested from either age-matched WT and 5xFAD mice (2, 6, or 10 months old) after perfusion with 4% paraformaldehyde. 30μm coronal brain tissue sections were incubated with the same primary (rabbit anti-GFAP (1:1000), and mouse anti-SR (1:200)) and secondary antibodies used for the human amygdala tissue.

2.5. Fear conditioning

7–8 months-old male WT (n=13) and 5xFAD (n=9) mice were handled for two days in their housing room prior to fear conditioning. On Day 1, mice were placed in Med Associates conditioning chambers (context A) for 3 min, followed by the presentation of two conditioned stimulus (CS; tone)-unconditioned stimulus (US; shock) pairings that consisted of a 30s, 75dB, 6000Hz tone that co-terminated with a 2s, 0.8mA foot-shock. There was 60s intertrial interval between CS-US presentations. The mice remained in the chamber for additional 30s and then were returned to their home cages. To test contextual fear memory recall on Day 2, mice were placed back in the training context A for 3 min without tone presentations (CS) and then returned to their home cage. To test cue fear memory recall on Day 3, mice were exposed to a different context than during conditioning (smooth flooring, enclosure of the chamber to prevent animals from seeing the room, red lighting in the chamber) for a 3-minute acclimation followed by tone presentations (30s, 6000Hz) separated by a 30s intertrial interval. Freezing was measured in an automated fashion by an observed blinded to genotype using FreezeFrame software (Coulbourn Instruments).

2.6. Statistics

Statistical analyses were performed using Graphpad Prism 9.0 (La Jolla, CA) and JMP Pro v. 14 (SAS Institute, Cary, NC; analysis of covariance of human post-mortem data). Classical and robust linear regression of main outcome measures were applied on diagnostic groups, as well as illness-related and demographic covariates (e.g. sex, age, and postmortem interval [PMI]) [3, 4, 22]. Group differences relative to the main outcome measures were assessed for statistical significance using stepwise linear regression (ANCOVA). Potential confounds were tested for their effects on main outcome measures. Unpaired, two-sided t-tests and two-way ANOVA were used to analyze the fear conditioning data. Values of p < 0.05 were considered statistically significant.

3. Results

While control cases had a negligible amount of GFAP+, GFAP+/C3+, and SR astrocytes in the amygdala (Fig 1A–H), AD cases had astrocytes with robust levels of GFAP, C3, and SR (Fig 1I–P). We quantified the density of GFAP+ astrocytes that expressed C3 and SR across individual amygdala nuclei (CO, ME, CE, AB, BL, and LN) controls and AD cases (Fig. 2). In the entire amygdala, we found that AD subjects had 5-fold, 8-fold, and 31-fold increases in GFAP+, GFAP+/C3+, and GFAP+/C3+/SR+ reactive astrocytes compared to age-matched controls, respectively (Fig. 2). This pattern was consistent across all amygdala nuclei in AD cases, with the CE, AB, and BL nuclei having the largest increase in GFAP+/C3+/SR+ reactive astrocytes. Importantly, the majority of GFAP+/C3+ astrocytes also express SR. Age, sex, and PMI did not affect the statistical significance of any of our main outcome measures.

Fig. 1.

C3+ reactive astrocytes express serine racemase in the amygdala of AD subjects. Representative images showing glial fibrillary acidic protein (GFAP; magenta), complement C3 (C3; turquoise), and serine racemase (SR; green) immunostaining in amygdala from control (A-H) and AD subjects (I-P). White rectangles denote where the high-power images were taken that are shown in panels E-H and M-P. White arrows denote GFAP+/C3+/SR+ astrocytes. Scale bars = 25μm

Fig. 2.

Quantification of SR-expressing C3+ reactive astrocytes across amygdala nuclei of AD subjects. The number of GFAP+, GFAP+/C3+, and GFAP+/C3+/SR+ cells/mm³ was significantly increased in AD subjects (blue filled circles; Braak stages V/VI) compared to controls (open circles) in the whole amygdala, as well as individual amygdala nuclei. Values represent the mean ± SEM. Unpaired Student’s t-test. *p<0.05 **p<0.01. CO: cortical, CE: central, ME: medial, AB: accessory basal, BL: basal, LN: lateral nucleus.

Our prior work demonstrated that aged, but not young, TgF344-AD rats display reactive astroglial SR in the hippocampus and entorhinal cortex [3]. In this study, we determined whether astroglial SR production occurs in the amygdala of the 5xFAD mouse, an amyloidosis model of familial AD. As expected, astroglial SR was not detected in the amygdala of 2-month-old WT and 5xFAD mice (Fig 3. A–H). At 6 months of age, minimal astroglial SR expression was observed in 5xFAD mice, even though there is marked gliosis at this age (data not shown). However, at 10-months-old, there was intense astroglial SR expression in the amygdala of 5xFAD mice (Fig. 3M–P), but not age-matched WT mice (Fig. 3I–L). Interestingly, C3 expression in amygdala astrocytes of 5xFAD mice was not as widespread as in the hippocampus (Supplementary Fig. 1) or in the amygdala of human AD patients. Finally, we did not observe reductions in neuronal SR expression in 5xFAD mice across the amygdala and hippocampus at 2 months or 10 months (Supplementary Fig. 2) of age.

Fig. 3.

Serine racemase is expressed by reactive astrocytes in aged 5xFAD, but not wild-type mice. Representative immunofluorescent images of glial fibrillary associated protein (GFAP; green) and serine racemase (SR; magenta) from the amygdala of wild-type (WT: 2-months: A-D; 10-months: I-L) and 5x FAD (2-months: E-H; 10-months: M-P). White rectangles in A, E, I, M denote where the high-power images were taken that are shown in panels B-D, F-H, J-L, and N-P. * = SR+, GFAP+ astrocytes; # = GFAP+, C3+ astrocytes; arrow = SR+, GFAP+, C3+ astrocytes. Scale bars = 25μm. Identical patterns of staining shown here were observed across all mice of the same age and genotype. BA: basal amygdala; CeA: central amygdala; LA: lateral amygdala

Finally, we used a Pavlovian fear conditioning assay, one of the most widely used models for studying emotional memory and associative learning in rodents [15], to determine if amygdala function is impaired in aged 5xFAD mice. In this form of conditioning, a neutral stimulus (conditioned stimulus; CS) acquires predictive value by pairing it with an aversive, unconditioned stimulus (US; foot shock) that has an intrinsic value to the subject. After training, exposure of the animal to the CS or context alone elicits conditioned fear responses such as freezing. On day 1, both WT and 5xFAD mice were able to learn the tone-shock association as demonstrated by equal amounts of freezing behavior during the second tone presentation (Fig. 4A; tone: F_1,20=14; p=0.001; genotype: F_1,20=0.2; p=0.63). On day 2, contextual memory recall was not significantly impaired in 5xFAD mice (Fig. 4B; t(20) = 0.51; p = 0.61). However, 5xFAD mice had significantly impaired cue memory recall on day 3 (Fig 4C; genotype: F_1,20=5.36; p=0.03), a process that is amygdala dependent.

Fig. 4.

5xFAD mice have impaired fear cue memory recall. (A) Wild-type (WT; white bars; n = 13) and 5xFAD (black bars; n = 9) mice were subjected to a fear-conditioning paradigm on Day 1. The percentage of time freezing during each of the tone presentations was measured for each group. (B) On Day 2, mice were placed back in the same training context for 3min to assess contextual fear memory recall. The total percentage of time freezing during the 3min was measured for each group. (C) On Day 3, mice were placed in context B (smooth flooring, enclosure of the chamber to prevent animals from seeing the room, red lighting in the chamber) and presented with tones to assess cue memory recall. The percentage of time freezing during the tone presentations was measured for each group. Asterisk (*) indicates significant difference from the WT mice (p < 0.05). All values represent the mean ± SEM.

4. Discussion

We demonstrate robust C3+ astroglial SR expression throughout the amygdala of cases with advanced AD. We also found robust expression of SR in reactive astrocytes of the amygdala of aged 5xFAD mice that was not present in age-matched control mice. Our results demonstrate that astroglial SR production is not restricted to the hippocampus and entorhinal cortex, but can occur in other brain regions that exhibit reactive gliosis in AD.

Reactive astrogliosis is the process by which, in response to pathology, astrocytes undergo changes in transcriptional regulation, as well as biochemical, morphological, and metabolic processes that result in gain of new function(s) or loss or upregulation of homeostatic ones [8]. For example, a substate of reactive astrocytes that highly express C3 are associated with a neurotoxic function [10, 18] and are present in a range of neurodegenerative disorders, including AD [7, 9, 10, 18, 26, 36]. Furthermore, reactive astrocytes in AD are not able to effectively modulate glutamatergic synaptic transmission due to Aβ affecting both neurons and astrocytes [17, 29], leading to a buildup of extracellular glutamate that spills out of the synapse. This glutamate spillover results in chronic extrasynaptic NMDAR (eNMDAR) activation and exacerbates intracellular calcium levels, which is particularly damaging due to the proximity of eNMDARs to mitochondria [2, 27]. D-serine produced by reactive astrocytes would further exacerbate excitotoxicity by providing the co-agonist required to open eNMDARs that are already primed for activation due to excess extracellular glutamate. Thus, our results suggest that in addition to long-chain saturated lipids [10], D-serine could also contribute to the neurotoxic phenotype of C3+ reactive astrocytes. Future studies using a conditional and inducible genetic strategy to eliminate SR from reactive astrocytes in AD mouse models, as we have done in a mouse model of TBI [25, 30], would demonstrate the deleterious effects of astroglial D-serine in neurodegeneration.

The presence of astroglial SR in the amygdala of aged 5xFAD mice, but not WT mice, is consistent with our human post-mortem and TgF344-AD rat data [3]. It will be important to elucidate the molecular mechanisms and cellular cues responsible for SR production in reactive astrocytes, which could differ between brain regions given the heterogeneity of astrocytes in health and disease [13, 32]. It is also important to determine what controls the onset of astroglial SR production because reactive gliosis is well established in mice and human prior to the appearance of SR. The basal amygdala displayed more robust reactive astrogliosis and SR+ astrocytes compared to other amygdala nuclei. Future studies are needed to systemically quantify reactive astrogliosis and SR-expressing reactive astrocytes across all amygdala nuclei.

The amygdala is a central hub in the emotional learning circuit, integrating sensory information from both cortical and subcortical brain regions related to the conditioning experience [12]. We found that aged male 5xFAD mice that have reactive glial SR expression display impaired cue memory recall, which is dependent on the amygdala [31]. The observed impairment in cue memory in aged 5xFAD mice agrees with prior work [28]. We did not observe deficits in contextual memory recall in 5xFAD mice as reported in other studies. This could be due to differences in the fear conditioning protocol (e.g. contextual fear conditioning protocol), mouse background strain, and/or sex of the mice. Future work utilizing conditional genetic approaches is needed to determine the extent by which glial D-serine release in the amygdala contributes to the fear memory impairments in 5xFAD mice [6, 14, 28].

In conclusion, our data suggest that SR-expressing reactive astrocytes in the amygdala could exacerbate glutamate excitotoxicity, neurodegeneration, and emotional dysregulation observed in AD. Astroglial SR production also likely occurs in other neurodegenerative disorders that exhibit reactive gliosis. Thus, preventing D-serine production or release from reactive glia could be a strategy for developing novel therapies to treat disorders associated with chronic neuroinflammation.

Reactive astrocytes express serine racemase in the amygdala of human AD cases

SR producing reactive astrocytes express C3 in the amygdala of human AD cases

Reactive astrocytes express serine racemase in the amygdala of 5xFAD mice

Abbreviations:

AD

Alzheimer’s disease

C3

complement C3

GFAP

glial fibrillary acidic protein

NMDAR

N-methyl-D-aspartate receptor

PMI

post-mortem interval

SR

serine racemase

TBI

traumatic brain injury