More than 80 years ago, Himmelsbach proposed the theory that the physiological ability to maintain homeostasis in response to addictive substances may underlie drug tolerance [14]. This theory predicts that the brain undergoes adaptation to maintain homeostasis in response to such substances. This neuroadaptation is not apparently visible because altered homeostasis represents a “new normal” or allostasis in the presence of the substance. Interestingly, allostatic changes are unveiled and emerge when the substance in question is absent, and this disinhibition effect in the absence of addictive substances results in withdrawal syndromes [14]. In line with this view, it has been proposed that alcohol-induced brain adaptation and allostasis underlie alcohol dependence and withdrawal [15, 16]. Allostasis is defined as the process of maintaining stability by changing the body physiology to a new state in response to environmental changes; a chronic deviation of the regulatory systems from the normal state of operation establishes a new set point, or allostatic state [17–20]. This model predicts the development and progression of AWS as follows. At the beginning of alcohol consumption, alcohol exerts stimulatory effects in the brain, and alcohol’s positive effects are a major driving force towards alcohol consumption. However, alcohol eventually functions as a depressant for the central nervous system (CNS). In order to counterbalance alcohol’s effect as a depressant, the brain undergoes neuroadaptive changes in a direction that makes the brain hyperactive; however, this new normal state (i.e. hyperactive) or allostatic state is masked due to the continued action of alcohol’s depressive functions. The consequence of such allostatic changes is that the brain becomes tolerant to the previous amount of alcohol, and thus increased amounts of alcohol are needed to produce the same level of positive reinforcement. Despite the increased alcohol consumption, the brain becomes tolerant again and reaches the next level of allostatic state, which will motivate more alcohol consumption to retain alcohol’s positive effects. This repeated cycle of alcohol exposure and tolerance is expected to ultimately drive towards an alcohol dependence state. In this alcohol-dependent state, the equilibrium has been shifted significantly to make the brain more excitable in order to counterbalance alcohol’s depressive functions, but these neuroadaptations are veiled by the continued presence and action of alcohol [15, 16, 21].

The consequences of allostatic state in alcohol dependence emerge and are expressed when alcohol consumption is suddenly terminated [22]. Upon AW, alcohol’s function as a CNS depressant is removed, and the compensatory changes that have made the brain more excitable during an alcohol dependence state are expressed. This resultant hyperexcitability of the brain is thought to be manifested in the form of AWS, raising the hypothesis that alterations to the E/I balance in the brain may result in AWS. In this context, increased excitatory and/or decreased inhibitory neurotransmission will be discussed as potential mechanisms that can increase the E/I and shift the brain equilibrium to a hyperexcitable state. In addition, we will review how hippocampal neurogenesis, a process that provides a significant level of plasticity to the brain, contributes to the maintenance of E/I balance, and how disrupted neurogenesis may underlie AWS.

Increased excitatory neurotransmitter, enhanced expression and function of glutamate receptors, and altered receptor subtypes and subunits can contri-bute to the hyper-glutamatergic state during AW [23–28]. Here we use selected examples to discuss how increased glutamatergic neurotransmission can be attributed to alterations to the E/I balance.

Extracellular glutamate concentrations in various brain regions have been measured by microdialysis after AW. In one study, the concentration of extracellular glutamate was significantly increased in the striatum within 12 hours of AW compared to that of sucrose-treated controls [29]. Glutamate levels remained elevated for the subsequent 12 hours and returned to control levels within 36 hours of AW. Other microdialysis studies found increased glutamate concentrations in the nucleus accumbens and the hippocampus at 12 hours of AW [30, 31]. Interestingly, repeated cycles of ethanol exposure and withdrawal significantly increased glutamate concentration, i.e., glutamate concentration is much higher in the hippocampus after a third cycle of ethanol exposure and withdrawal compared to that of the first cycle [31].

Glutamate receptors include both the ionotropic (iGluRs) and metabotropic (mGluRs) families [32]. iGluRs include three major receptor subtypes: N-methyl-d-aspartate (NMDA) receptors (NMDARs), α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (AMPARs), and kainic acid receptors. Eight members of the G-protein-coupled mGluRs have been identified (mGluR1–8) and they are categorized into group I (mGluR1 and 5), group II (mGluR2 and 3), and group III (mGluR4, 6, 7, and 8). NMDARs have a tetrameric structure that consists of NR1, NR2A–D or NR3A-B subunits. Typically, NMDARs are composed of two NR1 subunits and any combination of NR2 and NR3 subunits. The AMPAR family is heteromeric and consists of four major subunits (GluR1-4). In particular, GluR2 pl-ays an important role in the control of neuronal activity because AMPARs containing GluR2 reduce Ca^2 + permeability [33, 34]. The number and subunit compositions of both NMDARs and AMPARs vary among different brain regions and/or neuronal cell types, and changes in composition are associated with development, synaptic plasticity, and various pathological states [33, 35, 36]. AW can alter glu-tamate receptor-mediated synaptic transmission. Acute exposure and withdrawal induced long-term potentiation (LTP) of NMDAR-mediated postsynaptic currents in the dorsal striatum (DS) [37]. Six-teen hours after a 7-day ethanol exposure to rats, NMDAR currents were significantly increased in dorsomedial striatal (DMS) neurons. Moreover, an extended ethanol administration period from 7 to 14 days upregulated NMDAR activity, which lasted for up to 40 hours after the last administration [38].

A chronic intermittent ethanol (CIE) exposure is characterized by repeated cycles of alcohol exposure and withdrawal that result in a more persistent and severe form of AW syndrome. This effect is often referred as to a kindling effect [39, 40]. Because each cycle of alcohol exposure and withdrawal aggravates withdrawal symptoms as the cycle proceeds, this model closely recapitulates key aspects of human alcoholism [41]. For example, 48 hours after CIE exposure, enhanced glutamatergic transmission was reported in layer II/III pyramidal neurons of the medial prefrontal cortex (mPFC) [42]. Withdrawal from CIE has been shown to increase the NMDA/AMPA current ratio of layer V pyramidal neurons in the mPFC [43]. The NMDA/AMPA current ratio is often used to measure differential expression or synaptic transmission of NMDAR or AMPAR dependent currents. Thus, the NMDA/AMPA current ratio has been linked to synaptic plasticity, as well as synapse formation, gr-owth, and stability [44]. This increase in NMDA/ AMPA current ratio was associated with increased NR1 and NR2B subunit expression [43]. Prolonged withdrawal after CIE also induced a redistribution of AMPAR subunit composition by increasing GluR2-lacking AMPARs in the postsynaptic site. GluR2-lacking AMPARs are calcium permeable, have a higher single-channel conductance, and is associated with increased glutamatergic synaptic strength [45]. Another study has shown that CIE caused an increase in cell surface expression of GluR2/3, whereas withdrawal resulted in an increase in both GluR1 and GluR2/3 in the basolateral amygdala(BLA) [46]. Together, these cases support the concept that AW induces differential changes in glutamate release, glutamate receptor subtypes, and glutamate receptor subunits in response to alcohol exposure or withdrawal.

Alcohol is a CNS depressant and modulates the function of gamma-aminobutyric acid (GABA), a major inhibitory neurotransmitter in the mammalian brain. Acute and chronic alcohol exposure modulates GABA receptors and GABA release in many different areas. The GABA receptor (GABAR) family consists of GABA_AR_, GABA_BR, and GABA_CR. [47]. GABA_ARs are ionotropic receptors that gate chloride ion channels, whereas GABA_BRs are met-abotropic G-protein-coupled receptors. Although different alcohol exposure paradigms differently modulate the activity of GABARs, it has been generally accepted that alcohol exposure upregulates GABA_ARs and downregulates GABA_BRs [48]. Different combinations of 19 GABA_AR subunits (α1–6, β1–3, γ1–3, δ, ɛ, θ, π, and ρ1–3) can constitute different forms of GABA_ARs [48, 49]. GABA_ARs are comprised of two α and two β subunits, along with a single γ2 or δ subunit. The composition of GABA_AR subunits determines their functional properties and localization at synaptic or extrasynaptic sites. For example, the γ subunit mediates phasic inhibition and is mostly localized within the synapse, whereas receptors containing α4, α5, α6, or δ subunits are largely located at extra- or non-synaptic sites [48, 50]. Tonic GABA_AR-mediated currents are produced by extrasynaptic GABA_ARs containing the δ receptor subunit [51, 52]. GABA_B1a, GABA_B1b, and GABA_B2 are the three main GABA_BR subunits, and a combination of GABA_B2 subunit with a GABA_B1a or GABA_B1b subunit makes functional GABA_BRs [53]. In the following paragraph we will discuss some examples where AW affects GABAergic transmission.

In the hippocampus, kindling-like effect after CIE exposure was associated with hypofunction of GABA_ARs [54, 55]. At 2 days of AW after CIE ex-posure, mRNA expression of the α4 subunit of GABA_ARs was significantly increased in the DG, CA3, and CA1 of the hippocampus. As this upregulation of α4 subunit transcription has been observed in several animal models of temporal lobe epilepsy [56–58], it is plausible that elevated levels of GABA_AR α4 subunits may be associated with AWS risk [59]. Similar alterations in GABA subunit composition were observed in the amygdala during the withdrawal phase of CIE [60]. While decreased surface expression of GABA_AR subunits α1 and δ was observed in the basolateral amygdala (BLA) at 40 days after CIE treatment, surface expressions of α4 and γ2 subunits of GABA_AR were increased. The majority of the GABAergic postsynaptic currents in BLA are mediated by α1- and α2-containing GABA_ARs. Thus, this decrease in α1-containing GABA_ARs may contribute to GABAergic transmission during AW [61].

It was popularly thought that the adult brain is hard-wired and that no new neurons are regenerated. However, it has been established that new neurons are continuously generated and integrated into neural circuits in two structures of most mammalian brains, referred to as neurogenic niches [62–65]. One neurogenic niche is the subventricular zone (SVZ) of the lateral ventricle [66–68]. Neural stem cells (NSCs) that have the potential to self-renew and differentiate into multiple neural cells reside and produce immature neurons in the SVZ. Immature neurons migrate following the rostral migratory stream, mature and become interneurons, and integrate into the olfactory bulb. The second neurogenic niche is the subgranular zone (SGZ) of the dentate gyrus (DG) of the hippocampus [69–75]. NSCs are populated in the SGZ and differentiate into excitatory neurons or dentate granule cells (DGCs). Unlike the production of interneurons that are born in the SVZ and migrate to the olfactory bulb, adult-born DGCs are born in and are also integrated into the DG. Newborn neurons in both the olfactory bulb and DG continuously remodel local neural circuits. This life-long production and circuitry integration of newborn neurons provides a great deal of plasticity at the level of cell genesis, and this plasticity is required for cognition and emotion [76–80]. In physiology, integration of newborn DGCs plays a critical role in the maintenance of the E/I signal balance in the hippocampus. Because neurogenesis in the hippocampus plays a key role in cognitive and emotion as well as in maintaining the balance of the E/I signals, hippocampal neurogenesis has been implicated in characteristic features of AUD and AWS.

NSCs in the SGZ represent a heterogeneous cell population because they differ in morphology, marker expression, proliferation kinetics, and differentiation potential [74, 81, 82] (Fig. 2). Radial glial (RG) cells or Type-1 cells have radial morphology and represent slowly-dividing NSCs that can give rise to identical RG cells and Type-2 NSCs that subsequently produce neurons and astrocytes in the hippocampus [65, 81–83]. Type-2 NSCs do not have the distinct radial morphology, but they represent rapidly-dividing NSCs that have the potential to generate Type-3 cells and downstream neural cells. Type-3 cells represent neuroblast cells that transiently proliferate and ultimately differentiate into DGCs in the hippocampus. Different signaling mechanisms along with various neurotransmitters released at local synaptic circuits regulate the continued proliferation and activation into more differentiated cells [84]. The different stages of neurogenesis are shown in Fig. 2.

Hippocampal neurogenesis is tightly regulated by intrinsic as well as extrinsic factors. Among such regulators, brain activity that is translated by the expression of neurotransmitters plays a critical role in distinct stages of neurogenesis. In particular, GABA plays critical roles in the regulation of multiple steps of neurogenesis. First, it has been shown that GABA regulates proliferation of RG cells though GABA_ARs [85]. Specifically, GABA action on GABA_ARs containing γ2-subunit restricts proliferation of Type-1 RG cells, leading to decreased production of newborn DGCs. GABA also acts on GABA_ARs expressed in Type-2 cells and it is required to transform Type-2 cells into mature neurons [86, 87]. Interestingly, GABA depolarizes immature DGCs. This is due to higher intracellular chloride concentrations compared to the extracellular space. As a result, the opening of GABA receptors upon GABA binding onto its receptors allows chloride ions to move out, making immature DGCs depolarized [88–90]. This GABA-mediated depolarization induces the extension of apical dendrites of immature neurons into the molecular layer, where immature DGCs make synaptic connections with entorhinal cortical neurons and the axon projections to the CA3, forming the mossy fiber track [90]. Thus, GABA-mediated depolarization is essential for the integration of newborn neurons into hippocampal neural circuits.

Glutamate also plays a key role in the regulation of hippocampal neurogenesis. It has been shown that NMDA administration rapidly decreases cell proliferation, while NMDAR antagonists induce the opposite effect [91, 92]. Moreover, induction of LTP invoked by perforant path axon fibers of entorhinal neurons that project to DGCs promotes the proliferation of NSCs, further supporting the idea that glutamate has additional roles during neurogenesis [93, 94]. Specific deletion of an essential subunit of the NMDA glutamate receptor, NR1, in newborn neurons showed that glutamate-mediated excitation of newborn neurons is required for the survival of newborn DGCs that are integrated into neural circuits [95–97]. Synaptic connections from the perforant path can be detected within three to five weeks after their birth [98], whereas functional connections formed with efferent CA3 neurons can be observed four to six weeks after their birth [99]. Within a time window of four to six weeks, adult-born DGCs mature and form the characteristic tri-synaptic circuits connecting the entorhinal cortex to the DG, and the DG to the CA3. These four- to six-week-old neurons show reduced induction threshold and increased LTP amplitude in response to a physiological stimulation [100]. Thus, adult neurogenesis represents an ongoing developmental process during which the nervous system is continuously remodeled. This provides an expanded capacity of plasticity in response to experience [100].

Despite the fact that DGCs receive a significant amount of excitatory inputs, mainly from the entorhinal cortex, they are sparsely activated [101]. This unique function of hippocampal newborn DGCs stems from their distinct intrinsic properties and neuronal connectivity [100, 102–104]. The sparse activation of DGCs is achieved by feedback inhibition: DGCs project to different types of interneurons located mainly in the hilus, and these interneurons in turn project back to DGCs. This strong feedback inhibition endows hippocampal DGCs with a “gate” function that limits the excessive excitatory signals from entering the DG [12, 105, 106].

In addition to the contribution of DGCs to the for-mation of a feedback inhibitory loop, DGCs have unique intrinsic properties that facilitate sparse act-ivation. These include hyperpolarized resting membrane potential, low input resistance, and relatively high threshold for firing [107–109]. During DGC development, the action of GABA is critical to dec-rease the E/I ratio, suggesting that this inhibition is crucial for reduced spiking of maturing DGCs [103]. Immature DGCs are highly responsive due to low excitatory innervation, high intrinsic excitability, and high E/I ratio, whereas mature DGCs are minimally responsive to a broad array of cortical activity patterns due to feedback inhibition [13, 103]. Indeed, blocking neurogenesis in mice can elevate overall hippocampal activity, suggesting that the gate function of DGCs is critical for the functional physiology of the hippocampus [110].

Disrupted hippocampal neurogenesis and circuitry integration of hippocampal newborn DGCs alter the E/I ratio, which may underlie hippocampal hyperactivity manifested by the expression of seizures in some neurological conditions. Uncontrolled and synchronized bursts of neurons lead to electrical and behavioral seizures, and chronic conditions of seizure occurrence are key features of epilepsy. Dramatic changes in hippocampal neurogenesis have been reported in epilepsy. In rodent epilepsy models, seizure activity dramatically increased cell proliferation in the DG, leading to increased neurogenesis [111–113]. A number of mechanisms have been proposed to explain increased proliferation. It is possible that activation of GABA receptors expressed by NSCs [84] or differential epigenetic modification induced by seizures on the NSCs can also play important roles in their proliferation [114, 115]. Moreover, seizure-induced expression of trophic factors, such as brain-derived neurotrophic factor (BDNF), vascular endothelial growth factor (VEGF), and others can induce NSC proliferation [116–118). Of note, other studies have suggested that newborn neurons induced by seizures are prone to degenerate eventually 119, 120].

In addition to increased neurogenesis in rodent models of temporal lobe epilepsy (TLE), seizures also induced abnormal development of hippocampal newborn DGCs. While mature neurons are resistant to seizure-induced morphological changes, DGCs that are born immediately before and after SE develop abnormally [111, 121–124]. They display mossy fiber sprouting, abnormal migration and dispersion into the hilus and molecular layer, and ectopic basal dendrite formation. All of these structural abnormalities have been implicated in aberrant circuitry integration that leads to hyperactivity of the hippocampus [105, 123, 125, 126]. In particular, aberrant circuitry integration of hippocampal newborn DGCs is thought to disrupt the gate function of the DG. Zhou et al. showed that hippocampal newborn DGCs received increased excitatory inputs from the entorhinal cortex, as well as ectopic innervations from the CA3 and neighboring DGCs, while the connectivity ratio of newborn DGCs with hilar interneurons did not change [127]. This altered neuronal connectivity is expected to increase the flow of excessive excitatory signals from the entorhinal cortex to the DG and to amplify excitatory signals via de novo formation of recurrent excitatory loops between the DG and CA3 in epileptic brains [127, 128]. In addition, both normotopic and ectopic DGCs showed higher excitability and action potentials compared to control rats after SE [129]. These observations clearly suggest that abnormal development, integration, and neurophysiology of hippocampal newborn DGCs anatomically facilitate the formation of pro-epileptic neural circuits and physiologically disrupt the gate function of the DG, all of which contribute to hyperactivity of the hippocampus.

Generalized tonic–clonic seizures are one of the most prominent features of AW syndromes. Previous studies strongly suggested that neuroadaptations during alcohol exposure play a critical role in the expression of seizures when alcohol is not present. In addition to mechanisms discussed previously, abnormal hippocampal neurogenesis has been proposed as a neural substrate that transmits the effects of alcohol exposure and withdrawal into brain activity [142, 145].

In the study by Lee et al., animals had voluntary access to a nutritionally-adequate liquid diet containing ethanol for 4 weeks [145]. Upon sudden AW, mice displayed two distinct waves of seizures. The first wave consisted of a surge of multiple seizures and epileptic spikes that occurred immediately after AW and lasted for only a few days during abstinence. In the second wave, the frequency of the epileptic spikes was reduced, but they occurred for a protracted period of abstinence in a time-dependent manner. Starting from week 1 of abstinence, the frequency of spikes progressively increased, reaching a peak at week 4 and then gradually decreased and was abolished by week 8 of abstinence. AW also increased the spine density of hippocampal newborn DGCs, which correlated with an increase in epileptic spikes (i.e. 2 and 4 weeks of abstinence); however, dendritic spine density returned to the level of control mice when AWS disappeared. Changes in the density of mushroom spines were responsible for overall changes in spine densities during AW. Most importantly, DREADD (designer receptor activated by designer drug)-mediated functional studies revealed that hippocampal newborn neurons are necessary and sufficient for the expression of AWS [145]. DREADDs are genetically-engineered receptors that can be activated by administration of the exogenous ligand CNO (Clozapine N-oxide) [94, 146, 147]. The interaction of CNO with hM3Dq or hM4Di DREADDs results in the activation or inhibition of neuronal activity, respectively. This DREADD-mediated specific activation and inhibition of hippocampal newborn DGCs increased and decreased the expression of seizures during the second wave of AWS, respectively [145]. This functional study showed that hippocampal neural circuits, which are continuously remodeled by the integration of newborn DGCs, underlie the expression of AWS, suggesting that hippocampal neurogenesis and hippocampal newborn DGCs are involved in neuroadaptation during alcohol exposure and withdrawal.