The GABA system, a new target for medications against cognitive impairment—Associated with neuroactive steroids

The prevalence of cognitive dysfunction, dementia, and neurodegenerative disorders such as Alzheimer's disease (AD) is increasing in parallel with an aging population. Distinct types of chronic stress are thought to be instrumental in the development of cognitive impairment in central nervous system (CNS) disorders where cognitive impairment is a major unmet medical need. Increased GABAergic tone is a mediator of stress effects but is also a result of other factors in CNS disorders. Positive GABA‐A receptor modulating stress and sex steroids (steroid‐PAMs) such as allopregnanolone (ALLO) and medroxyprogesterone acetate can provoke impaired cognition. As such, ALLO impairs memory and learning in both animals and humans. In transgenic AD animal studies, continuous exposure to ALLO at physiological levels impairs cognition and increases degenerative AD pathology, whereas intermittent ALLO injections enhance cognition, indicating pleiotropic functions of ALLO. We have shown that GABA‐A receptor modulating steroid antagonists (GAMSAs) can block the acute negative cognitive impairment of ALLO on memory in animal studies and in patients with cognitive impairment due to hepatic encephalopathy. Here we describe disorders affected by steroid‐PAMs and opportunities to treat these adverse effects of steroid‐PAMs with novel GAMSAs.

The prevalence of cognitive dysfunction, dementia, and neurodegenerative disorders such as Alzheimer's disease (AD) is increasing in parallel with an aging population. Distinct types of chronic stress are thought to be instrumental in the development of cognitive impairment in central nervous system (CNS) disorders where cognitive impairment is a major unmet medical need. Increased GABAergic tone is a mediator of stress effects but is also a result of other factors in CNS disorders. Positive GABA-A receptor modulating stress and sex steroids (steroid-PAMs) such as allopregnanolone (ALLO) and medroxyprogesterone acetate can provoke impaired cognition. As such, ALLO impairs memory and learning in both animals and humans. In transgenic AD animal studies, continuous exposure to ALLO at physiological levels impairs cognition and increases degenerative AD pathology, whereas intermittent ALLO injections enhance cognition, indicating pleiotropic functions of ALLO. We have shown that GABA-A receptor modulating steroid antagonists (GAMSAs) can

Introduction
The population of the industrialized countries is increasing with age. At the same time, the prevalence of memory, learning disorders, and dementia is increasing [1]. Distinct types of chronic stress are believed to be instrumental in the development of cognitive impairment. This review will look at the effects of stress on cognitive processes as well as the involvement of neurosteroids and GABA-A receptors in that process.

Stress, memory, and steroid-PAM in acute stress
Acute stress can in certain situations both increase and decrease the strength of memories, but in chronic stress, the ability to remember decreases [2]. Acute stress disorder and post-traumatic stress disorder (PTSD) can occur in severe acute stress with large activation of the adrenal cortex, which often leads to memory impairment [3,4]. In acute stress, there is a pronounced activation of the adrenal gland's production of steroids. The increase in cortisol production in humans and corticosterone in rodents is well known and described in acute stress, whereas the parallel production of positive GABA-A receptor modulators (steroid-PAMs) has been less described. One of the most potent steroid-PAMs is allopregnanolone (ALLO), a metabolite of progesterone that is formed in the adrenal cortex. ALLO production increases during stress at the same time as other stress steroids, such as cortisol in humans and corticosterone in rodents [5][6][7][8]. Increases in steroid-PAM correlate with changes in amygdala volume and anxiety symptoms in women [9,10]. Many repeated acute stress events can eventually turn into chronic stress and thus damage cognitive functions [11]. However, even at acute stress under chronic stress conditions, steroid-PAM production, the acute stress steroid levels, increase in central nervous system (CNS) [5,12,13].

Memory and cognitive function in chronic stress
There are several types of chronic stress, like PTSD, psychosocial work stress, burnout syndrome, and/or chronic fatigue disorder often have cognitive problems [3,11,[14][15][16]. Rats with chronic stress show downregulated ALLO production, and the plasma concentration is low. Female and male rats under chronic stress show a larger acute stress response with higher ALLO concentrations than during non-chronic stress conditions when exposed to an acute stressor [6]. ALLO acute stress response in humans under chronic stress conditions is comparable to that seen in rats, but in women, the response is smaller compared to men [17]. In burnout syndrome, memory is often disturbed. With repeated stress-work-related, psychosocial, and/or emotional of various kinds-the risk of dementia increases [18][19][20]. Stress in animal models of Alzheimer's disease (AD) impairs memory and learning [21,22]. This means that chronic stress must be considered a risk for developing dementia and AD [23]. The biological relationship between memory disorders and chronic stress is not known, but steroid-PAM, elevated chronic GABAergic tone, and change in steroid-PAM sensitivity in the GABA-A system may be factors of interest to investigate further.

Steroid-PAM in chronic stress and disorders with cognitive impairment
In the case of chronic stress, steroid production decreases to low concentrations in the blood, so also for ALLO [24]. However, this does not occur in all individuals, and some continue with high steroid production even under chronic stress conditions. In these individuals, poorer cognitive function is related to increased cortisol concentrations in the blood [14,25]. In addition, GABA-A receptor sensitivity to steroid-PAM appears to increase due to the altered composition of GABA-A receptor subunits and thus with different sensitivity to steroid-PAM [26]. GABA-A receptor sensitivity to ALLO and other PAMs can be determined using GABA-A receptor-controlled functions-for example, saccadic eye velocity (SEV) and sedation [27,28]. In this regard, women with PTSD showed reduced sensitivity to a positive allosteric modulator, namely, diazepam (a benzodiazepine) both in terms of sedation and SEV [29]. The PTSD group showed no difference in the early initial ALLO response compared with controls but showed a faster recovery in SEV after ALLO injection. This suggests a GABA-A receptor subtype change in the PTSD group to a receptor subtype unresponsive to benzodiazepines-namely, α4,βx,δ, the subunit in the region responsible for the SEV effect. In patients with burnout syndrome, a similar response is noted with a reduced sensitivity to benzodiazepines but an increased effect of ALLO. We know that the α4,βx,δ subunit is highly sensitive to ALLO but insensitive to benzodiazepines, which would indicate an upregulation of the α4,βx,δ subunit GABA-A receptor in these disorders [30].

GABA-A receptors involved in memory
In adults, an activation of the GABA-A receptor leads to an inward flow of chloride ions. This leads to a hyperpolarization-that is, inhibition of the neuron [31]. In fetuses, in neuronal stem cells and certain neurological disorders, the chloride flow is reversed when the GABA-A receptor opens, and GABA thus becomes excitatory [32][33][34]. Postsynaptic GABA-A receptors can be localized to the synaptic cleft (intra-synaptic) but also outside the synaptic cleft itself and are then called extra-synaptic receptors [35]. Intra-synaptic GABA-A receptors generally contain α1, α2, or α3 subtypes. These receptors are less sensitive to GABA and require  [39,26,35].
high GABA concentrations to open (typical physiological concentrations 100 μM), whereas the extrasynaptic receptors generally contain α4, α5, or α6 and require lower GABA levels for activation (typical physiological concentrations 0.3 μM) [26,36,37]. The GABA-A receptor has about 20 different subtype sets in the brain, and these sets have specific localization in different areas of the brain (Fig. 1). Thus, certain subtypes will be able to influence the function of the brain area where the receptor is located [38]. More information about receptor subtype area function is described elsewhere [31,39]. The receptor type with the α5 subunits is well known to be related to cognition and is found in the hippocampus. Subtypes containing α4,βx,δ are abundant in the thalamus and are highly sensitive to steroid-PAM, which also controls their expression [26,40].

Acute and chronic treatment with positive GABA-A receptor modulators on memory
Many drugs affect GABA-A receptors-for example, benzodiazepines [41,42], barbiturates [43], ethanol [44], zinc [45], and steroid-PAMs [13,[46][47][48][49]. Several steroid-PAMs-such as ALLO, 3α-OH,5β-pregnanolone (3α5β-P), tetrahydrodeoxycorticosterone (3α5β-THDOC), and 3α-OH,5αandrostanediol-have CNS depressant effects that are even able to induce anesthesia. Such agents are all positive GABA-A receptor-modulating steroids-that is, steroid-PAMs [50][51][52]. Several steroid-PAMs can impair memory and learning at low concentrations. For example, ALLO reduces hippocampal neuronal activity in rats [53,54]. Rats given ALLOs have impaired learning and poor memory when evaluated in the Morris water maze (MWM) memory model [55,56]. In women, episodic memory is blocked after an ALLO injection [57], and that is interesting because episodic memory is disrupted early in AD [58]. Long-term potentiation is a phenomenon linked to memory development in the hippocampus; steroid-PAM inhibits that development [59], and ALLO inhibits the cholinergic effect in the hippocampus [60]. In humans, we have seen that progesterone and/or its metabolite ALLO impair the ability to retrieve memories from distinct parts of the brain [61]. Positive GABA-A receptor modulators, benzodiazepines, barbiturates, and ethanol cause in humans longterm acute and permanent cognitive impairments [62][63][64] and rats [65]. Positive GABA-A receptor modulators inhibit neurotransmission and thus cognition by activating GABAergic mechanisms [66,67]. This suggests that exposure to positive modulation of GABA-A receptors during prolonged periods increases the risk of cognitive impairment.

Medroxyprogesterone MPA treatment and dementia development
Hormone replacement therapy (HRT) in postmenopausal women with the steroids estrogen and progestin has for a long time been discussed for memory improvement and protection against dementia [68,69]. HRT with estrogen plus medroxyprogesterone acetate (MPA) is a common variant of HRT. In a large study started to investigate whether HRT could improve memory function and prevent the development of dementia-the Women's Health Initiative Memory Study-postmenopausal women were treated with estrogen plus MPA for 4-7 years in a placebo-controlled randomized trial. The results of the study were contrary to expectations in that the number of women with suspected dementia was twice as many in the estrogen plus MPA group compared to the placebo group [70]. In the group treated with estrogen alone, there was no difference compared to the placebo group [71]. The increase in suspected dementia was also not dependent on ischemic lesions or stroke, and increased dementia rate was due to a biological factor that was likely to be MPA [72,73]. MPA can also prevent and attenuate the positive effects of estrogen [74]. Studies in animal models have shown impaired ability to learn and remember with estradiol plus MPA treatment, or with MPA treatment alone compared to estrogen treatment [75][76][77]. MPA has the same properties as progesterone and ALLO-that is, it causes sedation and anesthesia [51,78,79]. This suggests that MPA acts via the GABA-A receptor. With the results of MPA treatments described above, especially on memory and learning, it was of interest to investigate whether MPA has a modulatory effect on GABA-A receptors.

MPA is a positive GABA-A modulator (PAM)
There is a lack of knowledge about MPA effects on subtypes of GABA-A receptors involved in cognition and mood. We have therefore performed studies of MPA on recombinant human GABA-A receptors [80]. Our studies in human embryonic kidney 293 cell lines permanently expressing the α1β2γ 2L, α5β3γ 2L, or α2β3γ 2S subtypes-using electrophysiological patch-clamp technology-examined MPA compared to well-known steroid-PAMs such as THDOC and ALLO. We chose the α1 subtype because it is the most common in the CNS and has been the receptor subtype most often investigated in previous studies [35,81,82]. The α5 subtype is found in the hippocampus and is related to memory, learning, and dementia development [31,83,84]. Depression is a common side effect of MPA, and the α2 subtype has been discussed in association with depressive symptoms [39,85]. The results show that MPA has large and differentiated effects as a positive modulator of GABA's effect and has a direct activating effect on α5β3γ 2L and α2β3γ 2S GABA-A receptors. However, MPA did not affect the α1β2γ 2L GABA-A receptor evaluated up to a concentration of 10 μM. Patch-clamp studies were also performed in cells from the preoptic area of the rat hypothalamus. MPA also showed its direct effect without adding GABA in rat hypothalamic cells. These results demonstrate that MPA is a positive allosteric modulator of GABA-A receptor subtypes α5β3γ 2L and α2β3γ 2S, and influences cells taken from the hypothalamic preoptic area [80].

Effect of PAM on AD TG mice
As mentioned above, progesterone has anticonvulsant and anesthetic effects [79,86]. These effects are due to the metabolite ALLO, which is a positive GABA-A receptor modulator more potent than MPA [87]. The GABA-A receptor is in adult mature neurons the major inhibitory receptor, and ALLO can induce anesthesia via its action on it [48,51]. In the rat hypothalamus, a dosage of only 2 nM is needed to activate the GABA-A receptor [88]. This concentration is reached in the CNS during acute stress, for example, and in women during the menstrual cycle [5,12]. Hippocampus is a CNS region intricately linked to cognition [89,90]. Changes in the hippocampus are seen early in AD development, resulting in a poorer ability to consolidate new experiences (episodic memory). Disorientation and reduced spatial perception are also early symptoms of AD. A frequently used memoryrelated technique in animal experiments is the MWM because it is particularly suitable for investigating spatial hippocampus-related memory [91]. ALLO concentrations as in mild stress continuously elevated or given as repeated injections of ALLO several times a week lead to reduced mem-ory and poorer learning ability as well as the faster progression of AD disease in transgenic (TG) AD mice [92][93][94]. In our studies, a low-stress concentration of subcutaneous ALLO was given continuously with Alzet mini pumps systemically for 1 month to TG-APPSwe/Arc mice, and for 3 months to TG-APPSwe/PS1 mice (Fig. 2). Both mouse types have double mutations in the amyloid-beta precursor protein (APP) gene (Swedish/Artic or Swedish/PSEN1δE9), giving an increase in Aβprotein production and plaque formation. In both experiments, the TG mice had impaired learning and memory compared to placebo-treated mice and wild-type mice [92,93]. The experiments were done 1 month after the Alzet pumps, and thus, the ALLO exposure was removed. At 1 month after the removal of the Alzet pumps, no external ALLO remained in the body at the time of the memory and learning test. This indicates that the memory and learning difficulties were not due to the presence of the drug in the brain, but rather that the ALLO treatment had caused permanent damage to memory and learning ability. The results show that chronically mildly elevated levels of ALLO permanently impair memory. One month after the medication had stopped, and the memory impairment was still present in the MWM test (Fig. 2). ALLO had a greater effect on male APPSwe/PS1 mice compared to females. Why there was a gender difference is not known. TG-APPSwe/Arc mice showed impaired learning already after 1 month of ALLO treatment in low-stress concentration. In the TG-APPSwe/Arc mice, the effect was as great in females as in males, especially in the so-called probe test where the mice must find a hidden platform under the water surface in the pool. Comparable results have been obtained with frequent injections of ALLO (3 injections/week) to 3xTgAD mice, where impairment of memory and learning was shown. However, it has been shown that ALLO injections at longer intervals-that is, weekly injections in AD mice with a pharmacological dose of ALLO-induce growth in neural stem cells that form new neurons in the hippocampus [94,95]. It is well known that immature progenitor cells are excited by GABA and ALLO and not inhibited as in mature neurons. We also treated wild-type mice with ALLO for 5 months. The wild-type mice also had a residual memory impairment and a smaller hippocampal volume [96]. Previously, however, we have shown that ALLO has a tolerance development toward high concentrations, but despite this, we saw memory impairment in the wild-type mice [96,97]. However, the concentration obtained continuously in the wild-type mice was lower than in the rat MWM experiments, and the development of tolerance is dependent on dosage and frequency of administration [96,97]. A summary is given in Table 1.

Human diseases with high steroid-PAM concentrations, and impaired memory and learning
A disease with high ALLO levels in the CNS and blood is hepatic encephalopathy (HE). High GABAergic tone has also been shown to be associated with impaired memory and learning [98][99][100][101]. CNS concentrations of ALLO have been investigated in autopsy material, which are increased compared to patients without HE [102]. In animal models of HE, levels of ALLO and THDOC are also high [98,103]. The most important pathological factor in HE is an increase in ammonia production. Ammonia stimulates the production of steroid-PAM by increasing the transport of cholesterol into mitochondria through the upregulation of translocator protein (TSPO) [104]. TSPO is also a marker of neuroinflammation that is increased in HE [105][106][107]. ALLO formation increases due to the increase in the intramitochondrial production of pregnenolone and by the cytoplasmic enzymes further down the pathway to progesterone and ALLO [104,108,109]. Primary biliary cholangitis is another liver disease with high ALLO concentrations and cognitive impairment [110]. Obesity and overweight are known to have high ALLO levels, even though cognitive impairment is not a prominent symptom [111]. Menstrual cycle-related mood changes are known to develop in parallel with increasing ALLO levels during the luteal phase of the menstrual cycle [112]. Conditions such as burnout syndrome, PTSD, and premenstrual dysphoric disorder show a different GABA-A receptor sensitivity to ALLO compared to controls [27,29,30].

GABA-A modulating steroid antagonists (GAMSAs)
We know that ALLO enhances the effect of GABA, and that ALLO binds to its own sites on the GABA-A receptor [113]. A minor change in ALLO's chemical structure at the hydroxy group in the three positions with the hydroxy group in the 3β position instead of the 3α position makes a marked difference in the effect on the GABA-A receptor. ALLO's 3β-isomer (3β-hydroxy-5αpregnan-20-one [iso-allo]) has no intrinsic effect on GABA-A receptors [114]. On the other hand, we have shown that iso-allo can counteract   [94]. AD, Alzheimer's disease. Source: "Reprinted from Ref. [92], with permission from IOS Press."
Steroid-PAM Continuous stress levels of ALLO give reduced memory and learning and faster AD progression in transgenic AD mice [92][93][94].
Steroid-PAM Chronically mildly elevated ALLO level permanently impair memory in wild-type mice [96].
GAMSA Iso-allo antagonizes anesthetic effect of ALLO in rats [115]. GAMSA UC1011 antagonized ALLOs impairment of learning and memory in rat Morris Water Maize 118 GAMSA GR3027 inhibits the effect of PAMs such as ALLO and THDOC [28]. GAMSA GR3027 improves motor skills, coordination memory learning in rat models of hepatic encephalopathy (HE) [28].
GAMSA GR3027 antagonizes ALLO induced sedation and slowing of saccadic eye velocity in humans [122]. GAMSA GR3027 shows in HE patients' effect on an objective electroencephalogram (EEG) measure and on memory, sleeping score, and reaction time [134].
ALLO/THDOC's effects-for example, in hippocampal slices, concentration-dependent. In the experiment, we obtained a complete block of the ALLO effect with the same concentration of iso-allo as ALLO [115,116]. Using a different technique, we studied the GABA-mediated flux of chloride ions [Cl − ] through the GABA-A receptor. In cortex homogenates from adult rats, iso-allo markedly inhibited ALLO-induced Cl-uptake. This result indicates that iso-allo or a similar substance could be used as a drug against diseases caused by steroid-PAMs [117]. We obtained comparable results when we studied 3β-steroids in individual rat hypothalamic neurons using the patch-clamp technique-that is, ALLO's effect was reduced by the 3β-steroids [13]. The disadvantage of iso-allo is that it must be given parenterally. However, we have developed a new orally active 3β-steroid-GR3027 (GAMSA)-that can selectively antagonize ALLO-and THDOC-enhanced activation of GABA-A receptors. A summary is given in Table 1.

Novel GABA-A-modulating steroid antagonists (GAMSA) effects on memory
To investigate whether it is possible to counteract ALLO's negative effect on memory, we gave rats a combination of ALLO plus a GAMSA (3β-20βdihydroxy-5α-pregnane) and compared the combination against the same dose of ALLO alone. GAMSA (3β-20β-dihydroxy-5α-pregnane) significantly shortened the time until the rats found the platform during a 5-day test period compared to the ALLO-injected group, and memory performance was on par with the placebo-treated group [118]. Pregnenolone sulfate (PS) is a GABA antagonist, and it also has antagonistic effects against ALLO [119]. However, PS works differently than 3βhydroxysteroids because it is a GABA antagonist [117,120] and can induce epileptogenic seizures like, for example, bicuculline [121]. We have developed an oral 3β-OH compound GR3027 because oral preparation is easier to administer as a drug than a parenteral preparation [122]. With GR3027, we can inhibit the effect of positive GABA-A receptor modulators such as ALLO and THDOC. We have identified HE as a group for which treatment with GR3027 against cognitive and motor impairment could be appropriate. We have investigated GR3027 in two rodent models of HE. The rat models both have increased ammonia production as a basis. One model induced hyperammonemia by having ammonia in the food. In the second model, a portacaval shunt was surgically performed. In both models, the rats had impaired motor skills and impaired memory. GR3027 significantly counteracted the HE symptoms in motor coordination, memory, and learning [122].

GAMSA in relation to neuroinflammation
In the CNS, microglia and astrocytes respond to various brain insults through activation. This activation occurs with neuroinflammation-for example, with hyperammonemia. An enhancement of the GABAergic effect occurred with the activation of microglia and astrocytes. The intracellular [Cl − ] concentration in the neuron decreases through an increase in the effect of the chloride pump (KCC2) out of the cell [123][124][125]. TSPO increases during neuroinflammation, which increases cholesterol intake to the mitochondria and thus pregnenolone synthesis. Via cytoplasmic enzymes, ALLO is formed, which leads to an increase in the concentration of ALLO in the brain. We know that ALLO is elevated in the CNS of rats with hyperammonemia [126,127] and in autopsy material from brains of patients with cirrhosis [102]. Studies in rats with hyperammonemia show that treatment with GR3027 reduces systemic inflammation [106]. GR3027 lowers, for example, the TNFα and IL-10 levels, and turns off the activation of microglia and astrocytes in the hippocampus and cerebellum. At the same time, cognition and motor skills are improved, and coordination is normalized. However, the mechanisms behind GR3027's effects are still unclear [106]. However, there are conflicting pro-and anti-inflammatory data on GABA's effect, so the area is not thoroughly investigated yet [128][129][130][131]. ALLO has also been reported to have antiinflammatory effects, but ALLO's results do not appear to be via GABA-A receptors and can also be activated by pregnenolone, which does not bind to the GABA-A receptor [132].

Clinical trials with GR3027 in HE
In awake humans, the effect of steroid-PAM can be examined with measurements of SEV [133]. GABA-A receptors minutely control SEV, an objective measure that is not affected by volition. SEV has often been used to determine the effect of sedative PAM on the individual and has, among other things, been used as evidence of driving under the drug influence. SEV can also be used to check the efficacy and sensitivity of steroid-PAMs such as ALLO. Using SEV and sedation estimates, we have shown that ALLO has a dose-dependent CNS effect that also shows a strong relationship with plasma concentrations [133]. We have used SEV and sedation to investigate different GABAergic antagonists-including new drugs against the overproduction or effect of endogenous steroid-PAMs, that is, GAMSA. For this purpose, we have synthesized oral substances, and one-GR3027, also called golexanolone-has a GABA-A receptor subunit profile that shows selectivity for the α5 receptor subtype, and thereby the treatment of cognitive and alertness impairment. Studies we have done in rodents have shown that GR3027 can inhibit ALLO's cognitive-disrupting effects very effectively. We have therefore tested GR3027 through a toxicological procedure, which has shown that GR3027 is safe and does not show any serious side effects, and can be used as a drug against excessive GABAergic tone in the CNS. To demonstrate a target engagement of GR3027, we have used ALLO-induced reduction in SEV and an increase in sedation. We gave men GR3027 orally but ALLO intravenously at a previously tested dose in a placebo-controlled randomized crossover study. GR3027 treatment significantly reduced the ALLO effect in both SEV and sedation [28]. As mentioned above, HE is a disease that has a high GABAergic tone and high ALLO levels in the blood. We have conducted a clinical trial with GR3027 in adult patients with HE and cognitive impairment [134]. We conducted a double-blind randomized placebo-controlled trial with GR3027 (golexanolone) for 21 days of treatment. Outcome parameters were an objective measure measured with an electroencephalogram (EEG, theta frequency; delta + theta/alpha + beta DT/AB ratio) that can be used for diagnosis and measuring treatment effect in HE subjects. In addition, patients were asked to complete various tests for encephalopathy score, animal naming, continuous reaction time, and Epworth Sleepiness Scale. The tests were administered at baseline, and after 10 and 21 days of treatment. Compared to baseline measurements, the group receiving golexanolone treatment showed a significant improvement, relative to placebo, in EEG performance (by normalizing the theta frequency and the ratio of DT/AB frequencies), Epworth Sleepiness Scale, continuous reaction time, psychometric liver encephalopathy score, and animal naming [134]. Further studies are ongoing in patients with primary cholestasis.