GABAergic Inhibitory Interneuron Deficits in Alzheimer’s Disease: Implications for Treatment

1Alzheimer’s disease (AD) is the most common neurodegenerative disease. AD is characterized clinically by severe cognitive deficits and pathologically by amyloid plaques, neuronal loss, and neurofibrillary tangles. AD patients often have learning deficits, memory loss, language difficulties, confusion in place and time, and emotional and other behavioral changes. These changes are not only a burden to the individual but also to families, communities, and healthcare systems. However, the molecular mechanisms underlying the cognitive deficits seen in AD are unclear, and there remains no effective treatment to slow or halt progression of AD.

2Many factors are known to contribute to AD pathophysiology, with abnormal aggregation of amyloid β-protein (Aβ) in the brain a likely dominant initiating factor. Aβ is formed from cleavage of amyloid precursor protein (APP) by beta (β)-(BACE1) and gamma (γ)-secretase, and Aβ exists in different forms including monomers, oligomers, fibrils, and senile plaques. Excessive Aβ can affect the expression of synapse-related proteins, decrease dendritic spine density, inhibit excitatory synaptic transmission, affect synaptic plasticity, and cause cognitive impairments both in AD patients and AD mouse models including Tg2576, 5xFAD, 3xTg-AD, APP/PS1, hAPP-J20, and hAPPJ9/FYN mice. A novel AppNL–F/NL–F knock-in AD model mouse shows age-dependent accumulation of Aβ, neuroinflammation, and neurodegeneration similar to AD patients. Recent studies performed in this model indicate that, in the early stages of AD, synaptic dysfunction originates in the lateral entorhinal cortex (LEC) and then spreads to other brain areas including the hippocampus in combination with synaptic hyperexcitation, severely disrupting excitatory–inhibitory inputs and resulting in synaptic imbalance and dysfunctional synaptic homeostasis. Dysfunctional synaptic homeostasis induces a decrease in long-term potentiation (LTP), and it has been shown that high Aβ concentrations inhibit LTP, which is essential for the normal development of cognitive functions such as learning and memory, whilst enhancing long-term depression (LTD). LTP inhibition and LTD enhancement disrupt normal neural network activity and may induce cognitive impairment in AD. Network activities supporting cognition, including activation and deactivation deficits, abnormal oscillatory rhythmic activity, and network hypersynchrony, are altered in AD, schizophrenia, epilepsy, and other neurological and psychiatric diseases. It has been shown that network hypersynchrony and altered oscillatory rhythmic activity may lead to cognitive abnormalities in AD. During normal brain activity, excitatory and inhibitory neurons are co-activated and maintain normal network activities. The balance between neuronal excitation and inhibition is disrupted in neurodegenerative and psychiatric disorders including epilepsy, schizophrenia, Parkinson’s disease (PD), and AD. γ-aminobutyric acid (GABA) inhibitory interneurons (GABAergic neurons) regulate excitatory neurons in different brain regions through GABA release. Imbalanced GABAergic and glutamatergic transmission has been shown to impair adult neurogenesis in animal models of AD. Thus, abnormally increased neural network activity in AD may be due to excitatory-inhibitory imbalance.

3A recent clinical imaging study combining positron emission tomography with a novel analytical framework showed that distant Aβ induces regional metabolic vulnerability, and the interaction between local Aβ and a vulnerable environment drives the clinical progression of dementia. Similar results were obtained in transgenic Aβ rats that do not form neurofibrillary tangles, suggesting a novel mechanism of cognitive deterioration. Therefore, the causal link between Aβ and AD remains controversial, and further work is still required to fully establish how Aβ participates in the definition, etiology, and diagnosis of AD. The genetic risk factors for AD may act through both Aβ-dependent and Aβ−independent mechanisms to influence disease onset and progression. Despite this, the presence of extracellular Aβ deposition as neuritic plaques and intracellular accumulation of hyperphosphorylated Tau (pTau) as neurofibrillary tangles remain the primary neuropathological diagnostic criteria for AD. Moreover, plasma Aβ is a reliable biomarker for AD in the clinical setting, along with other biomarkers. Aβ still plays an important role in AD pathology, and here we explore both Aβ−dependent and Aβ−independent mechanisms of neurocognitive decline and how they might be exploited for clinical benefit.

4Initial studies on the pathological changes and cognitive impairment induced by Aβ focused on Aβ’s effect on excitatory neurons. In excitatory neurons, Aβ can reduce synaptic transmission by activating the glutamate N-methyl-D-aspartate receptor (NMDAR). Aβ inhibits excitatory synaptic transmission, reducing neuronal network excitability. However, many studies have shown that Aβ can also induce aberrant neural network activity. Aβ leads to rapid excitatory amino acid transporter (EAATs) mislocalization and internalization and alters the clearance of glutamate, resulting in excitatory neurotoxicity during AD development. Synaptic damage can be significantly inhibited by treatment with the vitamin E derivative, Trolox. An intrasynaptic vesicle APP domain also promoted the release of the excitatory neurotransmitter glutamate. Recent studies in APP/PS1 mice have shown that, in the early stages of AD, microRNA-34a expression increases before Aβ production and cognitive impairment. MicroRNA-34a caused synaptic damage and cognitive impairment by inhibiting NMDAR and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor expression. Moreover, increased microRNA-34a expression inhibited synaptogenesis by suppressing the expression of two synaptic proteins, synaptotagmin I and syntaxin 1A, in AD patients. microRNA-34a knockdown in vitro significantly increased NMDA and AMPA receptor expression, thereby inhibiting synaptic damage and cognitive deficits.

5AD pathogenesis is also associated with significant cholinergic and glutamatergic neurotransmitter system dysfunction, including changes in the levels of these neurotransmitters and neuronal network dysfunction in AD patients. Furthermore, the basal forebrain cholinergic system (BFCS) undergoes severe atrophy in AD patients as well as in AD animals. Based on these studies, US Food and Drug Administration (FDA)-approved drugs for AD can be divided into two categories: the acetylcholine inhibitors including tacrine, donepezil, galanthamine, and rivastigmine, which enhance residual cholinergic activity to treat mild AD, and memantine, an NMDAR antagonist. However, these therapeutic drugs do not fundamentally delay disease progression and are required at large doses that induce adverse reactions; they are also too expensive for many patients. Therefore, there is a clinical imperative to develop different treatment schemes for different stages of AD development. Regulating the expression of NMDA and AMPA receptors using NMDAR antagonists to inhibit excitatory neurons and EAAT expression may improve synaptic damage and cognitive deficits, but there is clearly a need for further studies on regulating excitatory neurons and the cholinergic system to identify new therapeutic modalities.

6Studies have demonstrated that GABA is an important inhibitory neurotransmitter that balances neural excitability and inhibition and is ubiquitously expressed in the brains of mammals including humans. GABA’s inhibitory function is mainly mediated by three different GABA receptors: GABAA, GABAB, and GABAC. GABA inhibitory activities can be divided into two main types: phasic inhibition and tonic inhibition. In phasic inhibition, during the action potential, the membrane depolarizes and GABA is released from presynaptic vesicles to rapidly increase GABA concentrations in the synaptic cleft. GABA activates GABA receptors in the postsynaptic membrane, decreasing postsynaptic neuron excitability. Phasic inhibition is induced by the transient or phasic activation of GABA receptors by GABA from presynaptic vesicles. The postsynaptic γ2 subunits of GABAA receptors have been shown to be the primary mediators of phasic inhibition. Extrasynaptic GABAA receptors containing the pi (π) subunit mediate tonic inhibition in most brain regions, and the alpha (α)-5 and delta (δ) subunits are the major GABAA receptors mediating inhibition in the hippocampus. Hippocampal neurons receive inhibitory charges from tonic inhibition that account for ~75% of the total inhibitory charge received by hippocampal neurons. In this way, GABA released from the synaptic cleft can activate extrasynaptic GABA receptors to persistently inhibit neurons. Furthermore, α5 GABAA receptor-mediated tonic inhibitory conductance in hippocampal pyramidal neurons may regulate memory and neuroexcitatory disorders. Moreover, α5 GABAA receptor activity predominates over synaptic inhibition in modifying the strength of both synaptic plasticity in vitro and certain forms of memory in vivo under specific conditions. Tonic inhibition could be deemed a continuous “brake on the system,” regulating excitation due to long-lasting hyperpolarization. Therefore, tonic currents play an important role in mediating neuronal excitability, network oscillations, synaptic plasticity, neurogenesis, neuronal development, information processing, and cognition.

7Cortical microcircuits containing the hippocampus are critical for normal cognitive functions in the mammalian brain. Microcircuits comprise two major neuronal classes: excitatory principal cells and inhibitory interneurons, which release the neurotransmitters glutamate and GABA, respectively. GABA inhibitory interneurons account for 10–20% of cortical neurons. They can regulate the activity of excitatory neurons to maintain normal neural circuitry and neural network activities. According to differential neuronal molecular expression, cortical interneurons are divided into at least five types including parvalbumin (PV) neurons, neuropeptide somatostatin (SST/SOM) neurons, neuropeptide Y (NPY) neurons, vasoactive intestinal peptide (VIP) neurons, and cholecystokinin (CCK) neurons. Of these, SST neurons account for 20–30% and PV inhibitory interneurons 40–50% of GABA inhibitory interneurons. Wamsley et al. classified cortical interneurons into four types that do not overlap according to their expressed molecular markers, namely PV, SST, VIP, and RELN (Reelin, non-SST) neurons. Tremblinglay et al. divided cortical interneurons into PV, SST, and ionotropic serotonin receptor 5HT3α (5HT3αR) neurons, which were further divided into VIP and non-VIP neurons. Rudy et al. found that PV, SST, and 5HT3αR interneurons account for nearly 100% of neocortical GABAergic neurons, representing 40, 30, and 30%, respectively. Neurogliaform cells, also called spiderweb cells, belong to VIP-negative/SST-negative 5HT3αR neurons. In rat and human cerebral cortex, neurogliaform cells are unique in that they elicit slow, long-lasting, inhibitory postsynaptic potentials (IPSPs) on pyramidal cells and other interneurons through the combined activation of slow GABAA and GABAB receptors. Neurogliaform cells produce hyperpolarizing responses in a large fraction of nearby neurons via axonal varicosities containing synaptic vesicles without requiring synapses to produce inhibitory responses, suggesting that the cells influence target neurons by volume release of GABA. Moreover, the inhibitory effect of neurogliaform cells is dependent on GABAAδ receptors, which preferentially localize to neurogliaform cells of all the cortical interneurons.

8Early studies showed that GABA inhibitory interneurons are not vulnerable to Aβ attack. Therefore, research efforts have mostly focused on the effect of Aβ on excitatory neurons or excitatory synaptic transmission. Recent studies, however, have shown that AD patients suffer from memory and cognitive impairment that is obviously different to normal age-related decline and partly due to hippocampal neuron over-activity caused by GABA inhibitory interneuron dysfunction (;;;; Table 1). These neural circuits are overactive in different transgenic AD mouse models including hAPP-J20, Tg2576, 5xFAD, 3xTg-AD, and APP/PS1 mice (;;;;;; Table 2). The brain wave activity recorded by electroencephalography (EEG) in APP transgenic mice contains epileptiform discharges, suggesting that neural network excitability and neuronal discharge synchronization are enhanced in AD. Since neural network activity is regulated by both excitatory and inhibitory neurons, these seemingly contradictory results suggest that Aβ not only affects the function of excitatory neurons but also inhibitory neurons, breaking down the balance between neuronal excitability and inhibition. Therefore, the role played by GABA inhibitory interneurons in AD development is attracting increasing attention.

9Further studies have indicated that abnormal Aβ production and aggregation interfere with GABA inhibitory interneuron function, aberrantly activating hippocampal neurons and promoting cognitive impairment in AD mice. There is abnormal hippocampal overactivity in patients with moderate AD and young people carrying the apolipoprotein E (APOE4) gene. Imbalanced excitatory and inhibitory neurons in the hippocampus may be important in the cognitive impairment seen in aging and AD patients. Aβ can reduce the number and activity of GABA inhibitory interneurons, resulting in abnormal synaptic transmission and aberrant neural network activity, ultimately causing cognitive impairment in both AD patients and mice. Moreover, the altered patterns of synchronous activity might be caused by the loss of GABAergic septohippocampal pathway (SHP) axons in AD patients, which might also modulate hippocampal network activities. Therefore, targeting GABA inhibitory interneurons to rescue cognitive impairment might be a valuable therapeutic option, perhaps by increasing the number or activity of GABA inhibitory interneurons or the number of GABA inhibitory interneuron synapses or regulating GABA neurotransmitters (Figure 1 and Table 3).

10Previous studies have shown that neuronal loss is common in AD (; Figures 1, 2 and Tables 1, 2). Excessive Aβ may downregulate GABA inhibitory interneuron activity, cause GABA inhibitory interneuron loss, and thereby induce functional deficiencies in GABA inhibitory interneurons. A recent study showed that pTau accumulates in the hippocampal subgranular cell zone and hilus in AD patients. Most of these pTau-positive cells have been identified as GABAergic interneurons by co-labeling with glutamate decarboxylase 67 (GAD67) in PV and SST neurons, further suggesting GABA inhibitory interneuron function is disturbed in AD. Compared with controls, GABA inhibitory interneurons and SST neurons in the brains of AD patients are abnormally reduced, along with the inhibitory neurotransmitters (;;;;; Table 1). There are significantly fewer SST neurons in the frontal and temporal lobes of AD patients as measured by high-performance liquid chromatography analysis or immunohistochemical staining. PV neurons in the temporal neocortex have been shown to be relatively resistant to degeneration in AD, with only one in seven patients with AD having significantly reduced PV neuron numbers. However, PV neurons are decreased in dentate gyrus (DG)-CA4 and CA1–CA2 but not in CA3 of the hippocampus in AD patients. Furthermore, there is a clear decrease in PV neurons in parts of the entorhinal cortex in AD patients. Other studies have shown that SST neurons, calretinin (CR) neurons, and PV neurons are decreased in the olfactory tubercle, piriform cortex, amygdala, and EC of AD patients. Takahashi et al. observed a dramatic reduction in the mean densities of CR neurons in the hippocampal DG in AD patients compared to controls by immunofluorescent detection. However, compared to controls, there were no differences in PV neuron densities in the same region.

11Later studies indicated that the numbers of GABA inhibitory interneurons in the hippocampus were significantly decreased in transgenic mice such as TgCRND8, Tg2576, 5xFAD, and TauPS2APP (;;;;;;;;;; Table 2). Loss of PV and CR interneurons in the hippocampal CA1 of aged 3xTg-AD mice might be in part due to the global network excitability defects associated with AD. In APP/PS1 mice, there was a loss of PV neurons (40–50%) in CA1-2 and a loss of CR neurons (37–52%) in the DG and hilus, as well as a loss of SST neurons (50–60%) in the hippocampus. Another immunohistochemical study in this model showed a substantial decrease (35–45%) in CR-positive interneurons in CA1 and CA2/3 hippocampal subfields in very young mice (4 months) compared to age-matched controls. At 4 months, Cheng et al. found that there was a significant reduction (20%) in PV neurons in the frontal cortex and also some decreases in CR neurons in APP/PS1 mice. SIRT3 haploinsufficiency (Sirt3+/–AppPs1) aggravated the loss of both PV and CR neurons by >50% compared to control mice and Sirt3 haploinsufficiency alone mice. Feeding with a ketone ester-rich diet increased SIRT3 expression and prevented PV and CR neuron degeneration in Sirt3+/–AppPs1 AD mice, indicating that the aggravated GABAergic neuronal loss seen in these mice may be caused by reductions in SIRT3 that could be rescued by increased SIRT3 expression. Further results obtained from APP/PS1 AD mice showed that there was also a decrease in PV, SST, and CR neurons in the olfactory cortex. Moreover, CR neurons decreased during the early stages of the pathology, while PV neurons decreased at the later stages of the disease in the olfactory bulbs of APP/PS1 mice. However, there have been contradictory analyses of these neurons in AD mice, with other studies revealing no changes in either the hippocampus or the entorhinal cortex. A recent study showed that SOM (SST)-positive neurons were significantly less frequent in the perirhinal cortex of 6-month-old APP/PS1 mice and in AD patients (91% in Braak V, VI cases). PV interneurons in the perirhinal cortex of APP/PS1 mice were unaffected, but a pronounced and significant loss (69%) has been detected in the corresponding area in AD patients. Therefore, further studies are needed to ascertain the changes in the numbers of these neurons in AD models and their clinicopathological correlation with human disease (Table 2).

12The synaptic imbalance seen in AppNL–F/NL–F AD mice results from a reduction in the number of PV interneurons and a reduction in the somatic inhibitory axon terminals in the lateral entorhinal cortex. Another study indicated that CCK and SST interneurons decrease in an age-dependent manner in AppNL–F/NL–F AD mice. A novel mechanism of memory loss in AD suggests that the synaptic terminals of pyramidal neurons in EC layer II directly innervate CA1 PV neurons and are selectively degenerated in AD mice. Synaptic loss between the pyramidal and PV neurons disturbs the excitatory and inhibitory balance in the CA1 circuit and ultimately impairs learning and memory. APOE4 is an AD susceptibility factor and high-risk gene that can significantly lower the age of onset of AD; about 60–75% of AD patients are APOE4 gene carriers. There are GABA and SST decreases in APOE4 carriers, resulting in higher brain activity during rest and in the face of work requiring memory. In APOE knock-in mice, the hilar GABAergic interneurons, especially SST neurons, significantly decline with age. Enhancing GABA signaling in APOE4 knock-in mice prevented age-dependent GABAergic interneuron decline and learning and memory deficits during middle adulthood. The APPE693Δ (Osaka) mutation is associated with familial AD, reduces the number of GABA inhibitory interneurons, and inhibits their functional activity, thereby causing a series of downstream AD symptoms. The studies presented above suggest that the number of GABA inhibitory interneurons is significantly reduced during the process of cognitive deficits induced by excessive Aβ and may be accompanied by reduced neural activity. Thus, increasing numbers of GABA inhibitory interneurons might be potential therapeutic option in AD.

13It has been reported that the treatment of transgenic CRND8 mice with alpha-melanocyte stimulating hormone (MSH) can prevent cognitive deficits by inhibiting the loss of hippocampal SST inhibitory interneurons. Enriched environment rescues PV cell numbers and improves cognitive deficits. Citalopram, a selective serotonin reuptake inhibitor (SSRI), can reduce Aβ formation in vitro and reduce Aβ plaques in APP/PS1 mice. Furthermore, citalopram administration decreases the quantity of newly generated Aβ in young healthy humans. Clinical studies indicate that citalopram has therapeutic effects in AD patients, including rescuing non-cognitive neuropsychiatric behaviors such as depression, anxiety, irritability, and apathy, although the benefits on agitation and cognition remain controversial. Results from socially isolated rats suggested that citalopram improves learning and memory by promoting synaptic plasticity and hippocampal neurogenesis. Furthermore, chronic citalopram administration in APP/PS1 mice rescued impaired short-term memory and ameliorated non-cognitive behavioral deficits such as sociability dysfunction, depressive-like behaviors, and repetition-like behaviors. After treatment with citalopram, PV neurons increased in the cortex but not in the hippocampus of APP/PS1 mice. The improvements in behavior might be due to an increase in PV neurons, since PV neurons are essential for social interactions, the progression of depression, and PV neurons in the prefrontal cortex can affect short-term memory in mice. Cortical PV neurons are indispensable for short-term memory and social interactions via cortical circuit plasticity. PV neurons were found to be decreased in the cortex of 6-month-old APP/PS1 mice, and citalopram treatment increased the numbers of cortical PV neurons, thus rescuing behavioral performance.

14Transplantation of GABA progenitor neurons into the hippocampus also improved the learning and memory function of Aβ-overexpressing mice. Recent studies have shown that medial ganglionic eminence (MGE) progenitor cells derived from day 13.5 (E13.5) mouse embryos and transplanted into the DG hilum of 14-month-old AD model mice had the same electrophysiological characteristics as endogenous normal GABA inhibitory interneurons after 90 days. These neurons regulated neuronal activity in the DG region, significantly inhibiting memory impairment in AD mice and reducing anxiety-related behaviors to normal levels. APOE4-overexpressing mice showed GABA inhibitory interneuron loss and cognitive impairment, while APOE4 knockout significantly increased the number of GABA inhibitory interneurons and rescued cognitive deficits. In addition, cells derived from human embryonic stem cells (hESCs) or human pluripotent stem cells (hPSCs) can be differentiated into MGE progenitor cells after induction and then transplanted into the hippocampus of AD mice. In these experiments, transplanted MGE progenitor cells differentiated into GABA inhibitory interneurons with normal function and significantly improved neuronal circuit dysfunction, thereby improving the cognitive ability of AD mice. Therefore, GABA inhibitory interneurons are dysfunctional during AD development, and transplantation of GABA progenitor cells may improve cognition in AD mice (Table 3). In the early stages of AD, it might be possible to improve the cognitive ability of AD patients by increasing the number of normal GABA inhibitory interneurons by transplantation.

15There are now several studies suggesting that abnormal neuronal activity is associated with the increase in Aβ seen in AD patients. Synaptic inhibition and abnormally active neural network activity may coexist in both AD mice and AD patients, although further studies are needed to confirm this. Clusters of hyperactive neurons have been detected near amyloid plaques, and the hyperactivity is presumably due to a relative decrease in synaptic inhibition. Previous studies have shown that glutamate decarboxylase 65 (GAD65) protein levels are significantly reduced in AD patients and that, in these patients, the GABAergic system is severely affected. GAD65 deficits may contribute to AD pathogenesis through a loss of GABAergic inhibitory activity. In hAPP-J20 mice, without inhibiting the action potential, the frequency of spontaneous inhibitory postsynaptic currents (sIPSCs) significantly decreases, as recorded in the granular cells in the hippocampal DG and pyramidal neurons in cortical layers II/III, suggesting that Aβ overexpression affects the activity of GABA inhibitory interneurons. Subsequent studies have shown that these GABA inhibitory interneurons are dysfunctional, and GABA release is impaired in the hippocampus of hAPP-J20 AD mice, leading to the dysfunction of neural circuits and cognitive impairment during AD development. A relative decrease in synaptic inhibition contributed to neuronal hyperactivity near amyloid plaques in the cerebral cortex of APP23 × PS45 mice, indicating a reduction in activity of GABA inhibitory interneurons. Oriens lacunosum-moleculare (O-LM) interneurons are one type of SST-positive interneuron in the hippocampus. In APP/PS1 mice, progressive axonal loss and increased turnover of dendritic spines suggested defective O-LM interneuron connectivity, resulting in O-LM interneuron dysfunction associated with memory deficits. Similarly, significant GABA inhibitory interneuron functional impairment and GABA release inhibition were found in the hippocampi of APOE4 transgenic AD mice. Therefore, GABA inhibitory interneurons not only decrease in number but also in activity during excessive Aβ-induced cognitive deficits, resulting in GABA inhibitory interneuron functional impairment. However, further studies are necessary to clarify the mechanisms of functional impairment mediated by GABA inhibitory interneurons.

16With respect to the molecular mechanism underpinning GABA inhibitory interneuron functional impairment, it has been shown that the amplitude of the action potential induced in PV inhibitory interneurons in the cortex of hAPP-J20 mice is smaller than that of control mice. This further confirms that Aβ overexpression affects GABA inhibitory interneuron activity. Recent studies have shown that Aβ may inhibit PV inhibitory interneuron activity by interacting with ErbB4, subsequently causing pathological abnormalities and cognitive impairment. Further studies have shown that APP can regulate GABA inhibitory interneuron function by changing L-type Ca2+ channels. The number of L-type Ca2+ channels and Ca2+ current in GABA inhibitory interneurons were obviously increased after knocking out APP by enhancing the function and plasticity of GABA inhibitory interneurons, a phenomenon reversed by re-expressing APP. Aβ may also affect GABA inhibitory interneuron function by regulating other ion channels. For example, the deleterious voltage-gated sodium channel Nav1.1 mutation impairs telencephalic inhibitory neurons and results in Dravet Syndrome, an intractable form of childhood epilepsy. Nav1.1 expression is decreased in PV inhibitory interneurons of hAPP-J20 AD mice, affecting their intrinsic excitability and inhibiting γ oscillations, thereby triggering the synchronization of excitatory pyramidal neurons and increasing abnormal network activity with characteristics similar to epilepsy. The γ oscillation is crucial for maintaining normal learning and memory. Upregulation of Nav1.1 in PV inhibitory interneurons enhanced PV inhibitory interneuron activity, significantly increased PV inhibitory interneuron-dependent γ oscillation, inhibited the synchronization of neural network activity, and improved the cognitive ability of AD mice. A further study indicated that Nav1.1-overexpressing interneuron transplants derived from the embryonic MGE enhanced PV inhibitory interneuron-dependent γ oscillatory activity, reduced network hypersynchrony, and improved cognitive functions in hAPP-J20 AD mice. Since γ oscillations play a vital role in cognitive activity and neural network regulation, many studies have shown that Aβ deposition in the brain can be reduced by enhancing γ oscillations in PV inhibitory interneurons, which may improve cognition in AD mice. A non-invasive light flicker stimulus obviously enhanced PV inhibitory interneuron activity, increased γ oscillations, reduced Aβ plaque deposition and pTau levels, and ultimately improved the cognition of AD mice. 40 Hz visual stimulation entrained γ oscillations in the visual cortex (V1), CA1, and pre-frontal cortex, reduced neuronal and synaptic loss, modified synaptic signaling and synaptic plasticity-related proteins, and rescued spatial learning and memory in both Tau P301S and CK-p25 mice. The Aβ levels in the cortex and hippocampus as well as the number of plaques in 5xFAD mice were significantly reduced after continuous treatment with a 40 Hz auditory stimulation, thereby inhibiting the cognitive deficits. Combined 40 Hz auditory stimulation with 40 Hz light flicker further induced γ oscillations in the hippocampal CA1 and auditory cortex, reducing amyloid levels and improving memory in AD models. Hippocampal in vivo multi-electrode recordings revealed that optogenetic activation of channelrhodopsin-2 (ChR2)-expressing SST and PV interneurons in Aβ-injected mice selectively restored Aβ induced reduction of the peak power of theta and gamma oscillations, respectively, and resynchronized CA1 pyramidal cell spikes. Similar studies by revealed that Aβ-induced impairments of gamma oscillogenesis and oscillation-induced timing-dependent LTP were fully restored by optogenetic activation of PV and SST interneurons, respectively. Hence, these results suggest that SST neurons and PV neurons are potential therapeutic targets for restoring hippocampal network oscillations in early AD (Table 3 and Figures 1, 2).

17It has been shown that Tau reduction prevents Aβ-induced defects in axonal transport. Tau knockout in hAPP-J20 mice can inhibit epileptic-like EEG activity and improve learning and memory, perhaps due to the increased activity of GABA inhibitory interneurons. Expressing mutant Tau in the entorhinal cortex also induces excitatory neuron loss, grid cell dysfunction, and spatial memory deficits. Further studies have shown that Tau accumulates predominantly in excitatory neurons rather than inhibitory neurons, not only early in the entorhinal cortex but also in areas later affected by AD. These studies suggest that both Aβ and Tau can affect excitatory neurons and GABA inhibitory interneurons during AD development, probably via different mechanisms, but ultimately disturbing the balance between neuronal excitation and inhibition. High levels of Tau phosphorylation induce excessive metabolism of excitatory glutamatergic neurons and GABA inhibitory interneurons, disrupting the excitation-inhibition balance and resulting in abnormal neural network activity in TauP301L mice.

18Recent studies demonstrated that ErbB4 protein knockdown in PV inhibitory interneurons significantly increased LTP and improved memory ability in hAPP-J20 mice without changing Aβ plaque deposition. One plausible mechanism for this is that ErbB4 ablation increases PV inhibitory interneuron activity, although this requires further confirmation. By regulating GAD67 expression in GABA inhibitory interneurons, half-quantity expression of GAD67 significantly reduced Aβ deposition and improved cognitive performance in 5xFAD mice due to elevated GABA inhibitory interneuron activity. Moreover, nicotine enhanced synaptic plasticity in the CA3–CA1 synapses, and GABA receptor antagonists inhibited this enhancement, suggesting that GABAergic interneurons are required for nicotine-treated adult anti-NGF mice (AD11), a comprehensive animal model of AD. However, recent studies demonstrated that at the early stage of AD (16 weeks), Aβ induces hyperexcitability of hippocampal PV interneurons and contributes to neural network dysfunction and memory impairment in APP/PS1 mice. Suppressing PV interneuron hyperexcitability restored PV interneuron properties to wild-type levels, thereby reducing inhibitory inputs into pyramidal cells and rescuing memory deficits. One possible reason for the discrepancy may be due to the different ages and models of mice used in different studies. While further studies are needed to clarify this complex issue, these results suggest that targeting GABAergic interneurons might be a potential therapeutic target for AD patients.

19Both GABA and glutamate levels are significantly decreased in the temporal cortex of AD patients, suggesting deficient synaptic function and neuronal transmission in AD. Using magnetic resonance spectroscopy (1H-MRS), a significant decrease in GABA levels was detected in the parietal region of AD patients. Further studies showed that the decreased GABA neurotransmitter levels in the CSF were associated with age and AD. SST expression was reduced by 50% in AD, which was related to the formation of Aβ oligomers. SST might inhibit Aβ metabolism by upregulating neprilysin activity, with elevated neprilysin activity reducing the accumulation of both soluble and fibrillar Aβ in APP transgenic mice. Recent studies revealed that SST interfered with Aβ fibrillization and promoted the formation of Aβ assemblies characterized by a 50–60 kDa core, and these findings may signify a new role for SST in AD. A liquid chromatography-mass spectrometry study suggested that GABA levels are significantly reduced in cultured APP/PS1 mouse hippocampal neurons and the CSF collected from the hippocampal regions in 0-, 2-, 6-, and 8-month-old APP/PS1 mice compared to age-matched controls. Direct administration of GABA into APP/PS1 AD mice at 2 months but not 6 or 8 months improved cognitive function, suggesting that GABA administration during early life may have potential as a treatment for AD. 5xFAD transgenic AD mice treated with fingolimod had restored hippocampal GABA, decreased brain Aβ levels, and inhibited activation of microglia and astrocytes, ultimately improving memory. Therefore, it is possible to improve the cognitive impairment seen in AD model mice and AD patients by regulating the levels and release of GABA from GABAergic neurons during AD pathogenesis (; Table 4).

20Previous in vitro and in vivo studies have indicated that Aβ suppresses synaptic inhibition via downregulation of GABAA receptors. Therefore, upregulation of GABAA receptors may be a useful AD treatment. GABAA receptor agonists have been tested in vitro and in vivo and shown to have a positive effect on AD pathological changes. In the foraging/homing task, etazolate, a selective GABAA receptor modulator, significantly improved the cognitive deficits in aged rats at 28–31 months. A further study indicated that etazolate can protect cortical neurons against Aβ-induced toxicity due to etazolate stimulating sAPPα production in rat cortical neurons and in guinea pig brains, enhancing GABAA receptor signaling and rescuing the cognitive deficits. In a later clinical Phase IIA study, EHT0202 (etazolate hydrochloride) was shown to be safe and generally well tolerated in a 3 month, randomized, placebo-controlled, double-blind study in AD patients. Similar results were obtained using another GABAA receptor modulator baicalein. Zhang et al. found that activating GABAA receptors with baicalein significantly reduced Aβ production by increasing sAPPα in wild-type APP-overexpressing CHO cells and also had an effect in 6-month-old Tg2576 AD mice treated for 8 weeks, decreasing AD-like pathology and improving cognitive performance. Chronic stimulation of GABAA receptors with their agonist muscimol reduced amyloid Aβ25–35-induced neurotoxicity in cultured rat cortical cells, and the effect was completely reversed by the GABAA receptor antagonist bicuculine. A further study showed that AD mice exhibit increased spontaneous EEG delta (2–4 Hz) and decreased spontaneous EEG alpha (8–12 Hz) activity in the hippocampus and decreased Y-maze EEG theta (4–8 Hz) activity in the pre-frontal cortex. Low-dose muscimol and the GABAA receptor antagonist bicuculine reduced EEG delta activity and increased EEG theta activity in the pre-frontal cortex and improved spatial recognition memory during Y-maze testing in APP/PS1 mice. Another GABAA receptor agonist, propofol, is a commonly used anesthetic that has been demonstrated to reduce Aβ40 and Aβ42 levels in the brain tissues of aged (18 month) mice by decreasing brain levels of BACE1 to decrease Aβ generation and increasing brain neprilysin levels to increase Aβ degradation. Moreover, propofol treatment improved cognitive function by attenuating Aβ-induced mitochondrial dysfunction and caspase activation in both aged (18 month) wide-type mice and 19 month APP/PS1 mice.

21The clinical sedative and hypnotic chlomethiazole (CMZ) is neuroprotective and has anti-inflammatory properties in animal models due to GABAA-potentiating actions in vitro and sedative activity in vivo. CMZ protects against neurotoxic oligomeric Aβ1–42 in primary neurons. A CMZ analogue, 4-methyl-5-(2-(nitrooxy)ethyl) thiazol-3-ium chloride (NMZ), retains the GABAA potentiating actions of CMZ in vitro and sedative activity in vivo. NMZ, but not CMZ, restored the LTP in hippocampal slices from APP/PS1 AD mice and also restored memory consolidation in these mice. These effects of NMZ can be antagonized by the GABAA receptor antagonist bicuculline, an allosteric inhibitor of GABAA receptor channel opening, indicating that NMZ’s neuroprotective effect is mediated by the GABAA receptor. A further study revealed that the effect of NMZ is mediated by the α1β2γ2 GABAA receptor without having direct actions on the ion current gated by this receptor. However, the mechanisms of action of NMZ as an anti-AD agent remains unclear. Previous studies have indicated that the NO/sGC/cGMP/CREB pathway is vital for cognition, learning and memory, and neuroprotective effects. NMZ is one of a series of NO-releasing hybrid agents that can be used as potential therapeutic compound to treat AD. A recent study showed that NMZ is a multifunctional drug that does not directly target Aβ and Tau pathology. NMZ successfully attenuated the hallmarks of AD pathology and rescued cognitive deficits in different mouse models of AD including three familial AD models (APP/PS1, 3xTg, APOE4) and a novel model of sporadic Aldh2–/– AD mice in different ways. In APP/PS1 mice, NMZ restored cognition and reduced Aβ levels. In 3xTg mice, NMZ restored LTP via NO/cGMP, enhanced CREB activity, reversed cognitive deficits, and reduced Aβ and pTau levels. In APOE4 transgenic AD mice, NMZ lowered Aβ and elevated CREB phosphorylation and PSD-95. In Aldh2–/– mice, NMZ restored synaptic plasticity and attenuated Aβ and pTau levels. These multifunctional properties of NMZ make it a potential drug to treat the mixed pathology seen in AD patients. NMZM is a maleate salt of NMZ with higher efficacy due to its improved solubility and absorption. A recent study showed that NMZM may improve learning and memory by alleviating LTP suppression induced by scopolamine in rat DGs. Moreover, the protective effects of NMZM against scopolamine-induced depression of LTP was partly dependent on the potentiation of GABAA receptors.

22Alpha-asarone (ASA) is an essential oil isolated from the traditional Chinese medicinal herb Acorus gramineus that has been used to treat respiratory diseases and neural disorders for centuries in traditional Chinese and Indian herbal medicine. ASA has positive neuroprotective effects and improves cognition in rodent models. Chen et al. found that ASA had an anti-AD effect in aged rats with cognitive deficits by inhibiting neuronal injury and decreasing Aβ1–42 levels in the hippocampus, ultimately rescuing the cognitive deficits. ASA also had neuroprotective effects on primary hippocampal neurons impaired by glutamate. Further computer modeling and whole-cell patch-clamp recording studies suggested that ASA, as a GABAAR-positive allosteric modulator, can bind to GABAAR and improve cognitive function by reducing neuronal overexcitation. Therefore, these studies suggest that treatment with GABAA receptor agonists or positive allosteric regulators could be a potential strategy for improving cognitive function in the elderly and in AD patients.

23GABAA receptors are ligand-gated pentameric structures around a central chloride channel assembled from different combinations of 19 subunits including α1–6, β1–3,γ1–3, ρ1–3, δ, θ, ε, and π. Different subunits may have specific functions. For example, the postsynaptic γ2 subunits mainly mediate phasic inhibition, while extrasynaptic GABAA receptors contain π subunits that mediate tonic inhibition in most brain regions, with the α5 and δ subunits the major GABAA receptors for mediating inhibition in the hippocampus. α5-containing GABAA receptors play an important role in cognitive processes by controlling a component of synaptic transmission in the hippocampus. Ablation or reduction of the a5 subunit of GABAA receptors alters GABAergic synaptic transmission and enhances hippocampus-dependent memory and spatial learning in mice. Many studies have shown that inverse agonists of the GABAA receptor α5 subunit negatively regulate receptor activity by binding to the α5 subunit to improve learning and memory. MRK-016, α5IA, and α5IA-II have shown positive effects in improving cognition in animal models. In a preclinical and clinical study, α5IA was well tolerated in young and elderly subjects, and the efficacy of α5IA with respect to cognitive deficits requires further clarification. 3-iodo-8-(pyridin-4-ylmethoxy)pyrazolo[5,1-c][1,2,4]benzotriazine 5-oxide 1 selectively, safely, and significantly improved mouse memory processes in strict chemical relationships with α5IA, α5IA-II, and MRK-016. Several other compounds with pyrazolo[5,1-c][1,2,4]benzotriazine cores acting at GABAA receptors have been confirmed to selectively display anti-amnesic and pro-cognitive activities in vitro and in vivo. α7-nicotinic acetylcholine receptor (α7-nAChR) agonists have entered clinical trials as pro-cognitive agents for treating schizophrenia and AD. FRM-17848, an α7-nAChR agonist, enhanced LTP in rat septo-hippocampal slices, at least in part dependent on increased GABAergic neurotransmission mediated by GABAA α5-receptors. Therefore, targeting α5-containing GABAA receptors is an attractive strategy for treating disorders associated with cognitive defects such as AD.

24A recent study found that astrocytes are also involved in tonic inhibition. In the DG of 5xFAD AD mice (6–8 months) and AD patients, compared with normal controls, the GABA content in astrocytes was significantly increased and the GABA release mediated by GABA transporter GAT3/4 obviously enhanced the tonic inhibition of hippocampal DG cells. Moreover, suppressing tonic inhibition either with SNAP-5114 to block astrocytic GABA release or with L-655,708 to block α5 GABAA receptors rescued the LTP and memory deficits in 5xFAD AD mice. Consistent with this study, Jo et al. found that in APP/PS1 AD mice, activated astrocytes can induce excessive tonic gliotransmitter GABA secretion through B-type monoamine oxidase-B (Maob) and promote GABA release through the bestrophin 1 channel in DG. The released GABA decreased the spike probability of granule cells by acting on presynaptic GABA receptors. More importantly, astrocytic GABA and Maob were significantly upregulated in AD patients. However, the impaired spike probability, synaptic plasticity, and learning and memory could be fully restored by inhibiting GABA production or release from reactive astrocytes with selegiline, a selective and irreversible Maob inhibitor, which has been FDA-approved for patients with PD. Therefore, selective inhibition of astrocytic GABA synthesis, release, or GABAA receptors may be a potential effective therapeutic strategy for treating memory impairment in AD.

25GABAB receptors are located pre-synaptically, post-synaptically, and on extrasynaptic membranes in the hippocampus. Synaptic transmission in the brain is tightly regulated by GABAB receptors in two ways: presynaptic GABAB autoreceptors inhibit the release of a variety of neurotransmitters, and postsynaptic GABAB receptors generate inhibitory K+ currents that hyperpolarize the membrane and inhibit neuronal activity. Extrasynaptic GABAB receptors can probably be activated by ‘spill-over’ of GABA from neighboring synapses. The presynaptic GABAB receptors decline in response to neural activity and in AD. In AD patients, the numbers of GABAB receptor R1 protein (GABABR1)-positive neurons were found to be significantly reduced in the CA1 field of the hippocampus, and GABABR1 immunoreactivity but not neuron number was increased in the CA4 and CA3/2 fields. The changes in hippocampal GABABR1 may reflect the balance between excitatory and inhibitory neurotransmitter systems and result in dysfunction of the hippocampal circuitry in AD. Recent studies have indicated that secreted Aβ precursor proteins such as sAPPα can act as GABABR1α ligands to modulate synaptic transmission, thereby providing a potential target for the development of GABABR signaling–specific therapeutics in AD. Previous reports have suggested that astrocytes are abnormally activated in AD mouse models and human AD patients. Reactive astrocytes can induce excessive tonic gliotransmitter GABA secretion, and the released GABA binds to neuronal GABAB receptors at extrasynaptic sites, inhibiting synaptic release in APP/PS1 AD mice. Thus, suppressing GABAB receptor function may rescue synaptic release and improve the cognitive deficits seen in AD. The first GABAB receptor antagonist in clinical trials, SGS742 (CGP36742), displays pronounced cognition-enhancing effects in mice, young and old rats, and in Rhesus monkeys according to several different cognitive and learning-related behavioral tests. A further study has shown that SGS742 blocks the late IPSP and the paired-pulse inhibition (PPI) of population spikes recorded from rat CA1 pyramidal neurons in vitro and in vivo. SGS742 significantly enhanced the release of glutamate, aspartate, glycine, and somatostatin in vivo. Moreover, SGS742 induced significant enhancements in mRNA and protein levels of nerve growth factor (NGF) and brain-derived natriuretic factor (BDNF) in the cortex and hippocampus of rats. Acute in vivo administration of SGS742 in rats improved memory, at least in part due to reduced total hippocampal CREB2 activity. Oral administration of SGS742 for 8 weeks in a Phase II clinical trial significantly improved attention and working memory in patients with mild cognitive impairment and mild-moderate AD. Another GABAB receptor antagonist, CGP55845, showed a complete reversal of olfactory discrimination learning deficits in cognitively impaired aged Fischer 344 rats. Further results revealed that the intra-hippocampal microinjection of the GABAB receptor antagonist, CGP35348, ameliorated the learning, memory, and cognitive impairments induced by microinjection of Aβ in rats, suggesting that GABAB receptor antagonism may be a therapeutic agent against the progression of acute Aβ toxicity-induced memory impairment. Therefore, it is possible to improve the cognitive impairment seen in AD model mice and AD patients by regulating the level and release of GABA from GABAergic neurons and astrocytes during AD pathogenesis (Figure 3). However, the mechanisms underpinning GABA content and release remain unclear, with more research needed in this area.

26During AD development, Aβ leads to the loss and dysfunction of GABA inhibitory interneurons, with abnormal activity resulting in structural and functional impairment of nerve circuits and ultimately the cognitive deficits seen in AD patients and mice. Targeting GABA inhibitory interneurons by transplantation or functional enhancement may rescue the cognitive impairment seen in AD animals. Further studies of AD animal models transplanted with GABA inhibitory interneurons to stimulate GABA release or GABA receptor expression are needed. It will also be necessary to clarify the molecular mechanisms underpinning Aβ-induced dysfunction of GABA inhibitory interneurons, the aberrant activity of neuronal networks, and the cognitive deficits seen in AD. Nevertheless, GABA inhibitory interneurons represent a promising therapeutic target for the treatment of AD.

27HZ and YX conceived and designed the study. YX, MZ, and YH summarized the related literature. HZ and MZ conceived the images. MZ drew the images. HZ, YX, and MZ wrote the manuscript. All authors read and approved the final manuscript.

28The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.