Hippocampal alterations in glutamatergic signaling during amyloid progression in AβPP/PS1 mice

Our previous research demonstrated that soluble amyloid-β (Aβ)42, elicits presynaptic glutamate release. We hypothesized that accumulation and deposition of Aβ altered glutamatergic neurotransmission in a temporally and spatially dependent manner. To test this hypothesis, a glutamate selective microelectrode array (MEA) was used to monitor dentate (DG), CA3, and CA1 hippocampal extracellular glutamate levels in 2–4, 6–8, and 18–20 month-old male AβPP/PS1 and age-matched C57BL/6J control mice. Starting at 6 months of age, AβPP/PS1 basal glutamate levels are elevated in all three hippocampal subregions that becomes more pronounced at the oldest age group. Evoked glutamate release was elevated in all three age groups in the DG, but temporally delayed to 18–20 months in the CA3 of AβPP/PS1 mice. However, CA1 evoked glutamate release in AβPP/PS1 mice was elevated at 2–4 months of age and declined with age. Plaque deposition was anatomically aligned (but temporally delayed) with elevated glutamate levels; whereby accumulation was first observed in the CA1 and DG starting at 6–8 months that progressed throughout all hippocampal subregions by 18–20 months of age. The temporal hippocampal glutamate changes observed in this study may serve as a biomarker allowing for time point specific therapeutic interventions in Alzheimer’s disease patients.


Results
Learning and memory retrieval. Cognitive performance was assessed using the Morris water maze (MWM) learning and memory recall behavioral paradigm. No differences in learning were observed between 2-4 month and 6-8 month old AβPP/PS1 and age-matched C57BL/6J mice (Fig. 1A-F). At the 18-20 month age range AβPP/PS1 were slower to learn the location of the hidden escape platform compared to age-matched C57BL/6J mice as indicated by the cumulative distance from the platform (F 1,24 = 5.111, P = 0.33) and the area under the curve (AUC) of this parameter (t 24 = 2.488, p = 0.02) as shown in Fig. 1G. This age group of AβPP/PS1 mice also spent less time searching the target quadrant for the escape platform (F 1,24 = 13.10, P = 0.0014) over the five training sessions (t 24 = 3.824, p = 0.0008 as shown in Fig. 1H. Additionally, AβPP/PS1 mice spent more time navigating the periphery of the maze during the training sessions as indicated by percentage of time in the thigmotaxic zone (F 1,24 = 7.284, P = 0.01) and the AUC of this parameter (t 24 = 3.016, p = 0.006) as shown in In vivo DG glutamate dynamics. Representative glutamate release traces from the DG are presented in  In vivo CA1 glutamate dynamics. Figure 4A depicts CA1 representative glutamate release traces for both C57BL/6J and AβPP/PS1 mice across the three age groups tested. At 2-4 months of age no differences in CA1 basal glutamate are observed (Fig. 4B). AβPP/PS1 CA1 basal glutamate increases with disease progression and is elevated at 6-8 (t 20 = 2.618, p = 0.02) and 18-20 (t 22 = 3.946, p = 0.001) months of age compared to age-matched C57BL/6J mice. An opposite effect with disease progression is observed with CA1 stimulus-evoked glutamate release (Fig. 4C). The younger age groups studied, 2-4 (t 20 = 4.834, p = 0.0001) and 6-8 (t 20 = 2.142, p = 0.04) months of age have elevated glutamate release, but is similar to C57BL/6J mice at 18-20 months of age. AβPP/ PS1 CA1 release rate (Fig. 4D) tended to be elevated at the youngest age group only (t 20 = 1.800, p = 0.08). No differences between genotypes are observed at the other age groups. The clearance of CA1 evoked glutamate ( Fig. 4E) is elevated in AβPP/PS1 mice early in disease progression at the 2-4 (t 20 = 2.078, p = 0.05) and 6-8 (t 20 1.612, p = 0.12) month time points, but no difference is observed at 18-20 months of age. CA1 glutamate dynamics are inverted with basal levels increasing and evoked-release decreasing with respect to disease progression.
Amyloid plaque staining. Amylo-Glo RTD plaque staining reagent was used to measure changes in amyloid plaque pathology throughout the hippocampus. Representative 10 × magnification images form the DG and CA1/CA3 hippocampal subregions from all mice studied are shown in Fig

Discussion
One of the first preclinical symptoms associated with AD is attributed to hyperactive hippocampal neuronal networks. Hippocampal hyperactivity is observed in Aβ positive MCI patients, which persists with disease progression despite increasing rates of hippocampal atrophy and dementia scale ratings 9 . Elevated hippocampal activity has also been reported in several mouse models of AD with progressive amyloidosis [24][25][26][27][28] . These changes in neuronal networks are initially observed in mice before plaque deposition 29,30 , supporting a role for soluble Aβ in neuronal hyperactivity. Endogenous Aβ was shown to increase the release probability at excitatory synapses 29,31,32 as well as stimulate synaptic glutamate release 20,21 . However, as Aβ 42 accumulation leads to aggregation, the neuronal proximity to the surrounding plaques determines their activity states. Neurons closest to the plaques develop hyperactive phenotypes while those further from plaques are markedly silenced 33 . This may partially be explained by the accumulation of soluble Aβ isoforms surrounding plaques that colocalize with postsynaptic densities and cause synapse loss 34 . Furthermore, this would create a localized area of intense synaptic activity that can propagate Aβ pathology 35 www.nature.com/scientificreports/ whereby intensified glutamatergic activity and amyloid accumulation extend throughout cortical areas, contribute to seizure activity 38 , and cascade into the degenerative processes observed in AD 39 .
In the present study, learning and memory recall deficits in AβPP/PS1 mice were only observed at the oldest age group studied. Discordant results exist in the literature regarding when AβPP/PS1 mice begin to experience MWM cognitive deficits. Differences in cognition can begin by 6 months of age 40 with others reporting first appearance at 10 months that progresses with age 41 that are in line with the present study and our prior work 22 . Increased thigmotaxic behavior was also observed in the oldest AβPP/PS1 age group studied, similar to previous www.nature.com/scientificreports/ reports in this transgenic AD model 42 . When first navigating the MWM, mice tend to remain close to the wall until they learn to search the middle of the pool for an escape route. While learning the MWM, the thigmotaxic behavior in 18-20 month-old AβPP/PS1 subsided slower and was more prevalent during the probe challenge compared to age-matched C57BL/6J mice. This may be indicative of sensorimotor impairments and anxiety that affected learning and memory in AD mice at 18-20 months of age. AβPP/PS1 mice progressively accumulate soluble Aβ 42 starting as early as 3 months of age 43 . This accumulation eventually determines amyloid burden that becomes visible by 6 months of age and increases throughout the lifespan of these transgenic mice 44,45 . Likewise, the present study indicated no plaque deposition at 3 months of age, but by 6 months of age, plaque accumulation was most prominent in the CA1 followed by the DG with little to none observed in the CA3. As AβPP/PS1 mice reach 18 months of age, the magnitude of plaque deposition increases throughout the hippocampus, and this subregion distribution pattern continues including observable plaque pathology in the CA3.
Because plaque burden is a result of Aβ 42 accumulation in AβPP/PS1 mice 44 , the subregion distribution pattern of plaque deposition is indicative of soluble Aβ 42 concentration. This disease stage dependent progression of amyloid accumulation and deposition explains the basal and evoked glutamate release data reported in this manuscript. Elevated evoked glutamate release was prominent in the CA1 and DG at 2-4 and 6-8 months of age, but this was not observed until 18-20 months of age in the CA3. Elevated basal glutamate follows a similar hippocampal subregion distribution pattern that becomes more pronounced with age. However, this is temporally delayed to 6-8 months which may indicate amyloid accumulation first sensitizes neurons for increased release, then progresses to consistently elevated circulating levels of glutamate. The CA1 was the only subregion where evoked glutamate release declines with age despite increasing plaque deposition. A reduction of CA1 dendritic architecture was linked to enhanced cellular excitability at an age when plaque deposition was present 26 . However, as Aβ 42 accumulation progresses, a threshold may be passed where neurons become hypoactive or synapse loss is too pronounced 30 that diminishes glutamate release. Further studies are required to understand the how hippocampal soluble Aβ 42 levels contribute to changes in glutamatergic neurotransmission.
The pattern of amyloid deposition presented here is similar to the Braak neuropathological staging of hippocampal amyloid progression in AD 5,6 . This staging also coincides with CA1 neuronal loss occurring before other hippocampal subregions 46,47 . It is known that CA1 neurons are more vulnerable to global cerebral ischemia 48 and degenerate faster in epileptic patients 49 . The underlying cause of selective CA1 neuronal degeneration is a result of glutamate-mediated excitotoxic mechanisms involving excessive calcium influx through NMDAR activation, mitochondrial dysfunction, and reactive oxygen species. These events culminate in necrotic cell loss that releases more glutamate into the extracellular space thus propagating damage to surrounding neurons. This process supports our discordant CA1 glutamate observations with increasing basal but decreasing evoked glutamate release with age in AβPP/PS1 mice.
This research builds upon a growing body of literature indicating temporally altered hippocampal glutamatergic signaling during the progression of AD pathology. Our previous studies support hippocampal glutamate levels are still elevated at 12 months of age in male AβPP/PS1 mice 22,23 . Others have shown this is not a sex specific characteristic since female AβPP/PS1 mice have elevated CA1 dialysate glutamate levels at 7 months of age that also decline by 17 months of age 41 . Noninvasive techniques such as glutamate chemical exchange saturation transfer GluCEST also indicate a decrease in hippocampal glutamate in 18-20 months old AβPP/PS1 mice 50 . Interestingly these observations are not specific for progression of amyloid pathology. Electrochemical studies in the 5-8 month old tau mouse model of AD, P301L, develop elevated hippocampal glutamate 51,52 . A separate tau mouse model, P301S, also has elevated hippocampal glutamate at 3 months old, that declines at 18-20 months of age 53,54 as measured by GluCEST techniques. While the amyloid and tau pathology are likely acting through different mechanisms to elicit glutamate release, these studies show a concomitant change in vesicular glutamate transporter 1 that corresponds to the elevated glutamate levels regardless of AD pathology. When considered with the present research, these studies support temporal changes in hippocampal glutamate during AD progression with elevated levels early in pathology that decline in later disease stages. Understanding these types of regional differences may help to refine severity and progression of AD and tailor appropriate treatment options.
In early AD stages, before overt atrophy, overactivation of the NMDAR is hypothesized to impede detection of physiological signals leading to the cognitive impairment observed in AD 39,55 . Accordingly, meta-analysis of memantine treatment, a partial NMDAR antagonist, ameliorates cognitive and functional performance in mild-to-moderate AD patients when administered as monotherapy or in combination with anticholinesterase inhibitors 56 . This treatment only delays cognitive decline and does not have disease modifying benefits 57 . Since memantine modulates glutamate signaling rather than attenuating the glutamatergic tone, the persistently elevated glutamate levels during AD progression may induce excitotoxic effects that accounts for the neuronal, cognitive, and functional loss. As such, drugs that attenuate glutamate release or enhance clearance may provide long-term therapeutic benefits if initiated before signs of cognitive impairment.

conclusion
These data support a growing body of literature indicating hyperactive hippocampal glutamate signaling contributes to AD pathogenesis. The temporal hippocampal glutamate changes observed in this study may serve as a biomarker allowing for time point specific therapeutic interventions that can be tailored for maximal efficacy. Simultaneously monitoring changes in hippocampal glutamate with plaque and tangle pathology may further refine stages of AD progression. Morris water maze training and probe challenge. The MWM was used to assess spatial learning and memory recall. Mice were trained to utilize visual cues placed around the room to repeatedly swim to a static, hidden escape platform (submerged 1 cm below the opaque water surface) regardless of starting quadrant 23,58 .
The MWM paradigm consisted of 5 consecutive training days with three, 90 s trials/day and a minimum intertrial-interval of 20 min. Starting quadrant was randomized for each trial. After two days without testing, the escape platform was removed and all mice entered the pool of water from the same starting position for a single, 60 s probe challenge to test long-term memory recall. The ANY-maze video tracking system (Stoelting Co., Wood Dale, IL; RRID:SCR_014289) was used to record mouse navigation during the training and probe challenge. The three trials for each training day were averaged for each mouse.
Enzyme-based microelectrode arrays. Enzyme-based MEAs with platinum (Pt) recording surfaces ( Fig. 6A) were fabricated, assembled, coated (Fig. 6B), and calibrated for in vivo mouse glutamate measurements as previously described [59][60][61] . One microliter of glutamate oxidase stock solution (1 U/µl) was added to 9 µl of a 1.0% BSA and 0.125% glutaraldehyde w/v solution and applied dropwise to a Pt recording surface. This preparation aides in enzyme adhesion to the Pt recording surface for enzymatic degradation of glutamate to α-ketoglutarate and H 2 O 2 , the electroactive reporter molecule. The other Pt recording site (self-referencing or sentinel site) was coated with the BSA/glutaraldehyde solution, which is unable to enzymatically generate H 2 O 2 from l-glutamate. A potential of + 0.7 V vs a Ag/AgCl reference electrode was applied to the Pt recording surfaces resulting in oxidation of H 2 O 2 . While + 0.7 V is capable of oxidizing potential interferants, such as AA and DA, lower potentials are unable to adequately oxidize H 2 O 2 and subsequently detect glutamate 62  Microelectrode array/micropipette assembly. Glass micropipettes (1.0 mm outer diameter, 0.58 mm internal diameter; World Precision Instruments, Inc., Sarasota, FL) were pulled using a vertical micropipette puller (Sutter Instrument Co., Novato, CA). The tip was "bumped" to create an internal diameter of 12-15 µm. The micropipette tip was positioned between the pair of recording sites and mounted 100 µm above the MEA surface ( Fig 6D) www.nature.com/scientificreports/ soldered to a gold-plated connector (Newark element14 Chicago, IL). The other stripped end was placed (cathode) into a 1 M HCl bath saturated with NaCl that also contained a stainless steel counter wire (anode). Passing a + 9 V DC to the cathode versus the anode for 15 min deposits Ag/Cl onto the stripped wire.
In vivo anesthetized recordings. One week after MWM, mice were anesthetized using 1.5-2.0% isoflurane (Abbott Lab, North Chicago, IL) in a calibrated vaporizer (Vaporizer Sales & Service, Inc., Rockmart, GA) 63 . The mouse was placed in a stereotaxic frame fitted with an anesthesia mask ( Fig. 6C; David Kopf Instru-   64 . Recordings were conducted using a two electrode system whereby a Ag/AgCl reference wire was positioned beneath the skull and rostral to the craniotomy and a working electrode was positioned in one of the hippocampal subregions (Fig. 6D). Constant voltage amperometry (4 Hz) was performed with a potential of + 0.7 V vs the Ag/AgCl reference electrode applied by the FAST16mkIII. MEAs reached a stable baseline for 60 min before a 10 s basal glutamate determination and pressure ejection studies commenced. Once five reproducible signals were evoked, the MEA was repositioned into a new hippocampal subfield, which was randomized for each mouse. The FAST software saved amperometric data, time, and pressure ejection events. Calibration data, in conjunction with a MATLAB (MathWorks, Natick, MA; RRID:SCR_001622) graphic user interface program was used to calculate basal, stimulus-evoked, and clearance of extracellular glutamate. The evoked glutamate signals in each hippocampal subfield were averaged into a representative signal for comparison.
Amyloid plaque staining and semi-quantification. Sections were prepared, stained and quantified as previously described 22,23 . After electrochemistry, brains were removed and post-fixed in 4% paraformaldehyde for 48 h and then transferred into 30% sucrose in 0.1 M phosphate buffer for at least 24 h prior to sectioning. Twenty µm coronal sections through the hippocampus were obtained using a cryostat (Model HM525 NX, Thermo Fisher Scientific). Mounted sections were treated with 10% H 2 O 2 in 20% methanol for 10 min, transferred to 70% ethanol solution for 5 min, and then washed with PBS for 2 min. Sections were incubated for 10 min in Amylo-Glo RTD (1:100; Biosensis, Temecula, CA), submerged in physiological saline for 5 min, and rinsed three times in separate PBS solutions for 2 min. Sections were coverslipped using Fluoromount-G (South-ernBiotech; Birmingham, AL). Staining intensity was controlled for by imaging all sections the next day. Images were captured with an Olympus 1 × 71 microscope equipped with an Olympus-DP73 video camera system, and a Dell Optiplex 7020 computer. National Institutes of Health Image J Software (v. 1.48; RRID:SCR_003070) was used to measure relative staining density by using a 0-256 Gy scale. Staining density was obtained when background staining was subtracted from mean staining intensities on every sixth section through the hippocampus. Individual templates for the DG, CA3, and CA1 were created and used on all brains similarly. Measurements were performed blinded, and approximately four sections were averaged to obtain one value per subject. Amyloid plaques were identified by a dense spherical core of intense staining that were often surrounded by a less compact spherical halo.
Data analysis. Prism (GraphPad Prism 8 Software, Inc., La Jolla, CA; RRID:SCR_002798) software was used for statistical analyses. For glutamate measurements and amyloid plaque staining, hippocampal subregions were examined independently. For statistical analysis, genotypes were compared within age groups and all tests are listed in the figure legends. Outliers were identified with a single Grubbs' test (alpha = 0.05) per group. Data are represented as mean ± SEM and statistical significance was defined as p < 0.05.