Introduction

The v-AKTs are essential regulators of cell metabolism, growth, proliferation, and survival [1, 2]. The AKT family consists of three homologous kinases, AKT1 (PKBα), AKT2 (PKBβ) and AKT3 (PKBγ), encoded in humans by three separate genes on chromosomes 14q.32.33, 19q13.1-q13.2 and 1q44, respectively [3, 4]. AKT1-3 share a high level of sequence homology, including identical phosphorylation sites at amino acid residues Serine(Ser)-473 and Threonine(Thr)-308 [5]. Through activation of these sites the AKT kinases mediate their cellular functions through phosphorylation of downstream substrates, including mTOR and GSK3 [6].

Although the AKTs share a high degree of sequence similarity, and exhibit overlapping functions, emerging evidence suggests physiological divergence between the isotypes [7, 8]. Importantly, each of the Akt isoforms display differential tissue-specificity, with Akt1 and Akt2 being ubiquitously and highly expressed in brain, Akt2 also being expressed abundantly in skeletal muscle, heart, adipose tissue and testes, and Akt3 being restricted to the brain and testes [9,10,11]. Consistent with non-overlapping tissue expression profiles, each Akt isoform is implicated in different human diseases [12]. Akt1 has been extensively implicated in cancer [12], Akt2 with insulin resistance and diabetes mellitus [13] and Akt3 to brain growth anomalies and schizophrenia [14,15,16]. Convergently, in rodent, targeted genetic modification results in unique physiological profiles, with Akt1 mice displaying reduced overall organismal size and disrupted placental development, Akt2 mice, mild fasting hyperinsulinemia, and Akt3 mice showing reduced brain size and corpus callosum disorganization [17,18,19,20,21,22].

Notably, it is important to recognize that all AKT isoforms are highly expressed in the human [11] and rodent brain [9, 10], where Akt signaling is crucial for neuronal development and synaptic plasticity [9]. In vitro RNA interference studies show that Akt2 and Akt3, but not Akt1, are important for axonal growth and neuronal viability [23] and altered electrophysiological and neuromorphological properties are observed in Akt1 mutant mice [24, 25]. Neurobehavioral characterization of Akt1 or Akt3 deficiency in mice reveals that Akt1 loss results in deficits of working memory, fear conditioning, sociability, sensorimotor gating and dopamine signaling [24,25,26,27,28], while Akt3 deficiency results in impairments in social cognition, sensorimotor gating and aspects of learning and memory [22]. Together rodent studies of Akt1 and Akt3 support genetic association to schizophrenia [16, 26] and pharmacological targeting of AKT is posited to be a key molecular mechanism of action of existing [29] and novel preclinical antipsychotic agents [30, 31].

To date, little characterization of the neurobiological role of Akt2 has been performed, with most of the research being limited to studying its role in glucose homeostasis [9, 17, 18] and studies of CNS-related phenotypes being solely limited to examination of KO animals [32, 33]. Notably, with regards to neuropsychiatric phenotypes, polymorphisms in the AKT2 gene have been associated with personality traits pertaining to anxiety and depression in humans [34], and the segment of chromosome 7 in mice, which harbors Akt2, contains quantitative trait loci that predict murine anxiety-related phenotypes [35] and reduced brain AKT2 expression is observed in bipolar disorder [36]. Together these studies suggest a novel and unappreciated neurological role for Akt2, that may be physiologically distinct from that of Akt1 and Akt3.

Here, we sought to comprehensively characterize the role of Akt2 in multiple behavioral domains, as well as investigation of the neuromechanistic and electrophysiological correlates of reduced Akt2 signaling in both heterozygous and homozygous Akt2 null mice. We highlight the novelty and importance of studying heterozygosity (in comparison to the null condition) for reverse translation to inform human populations and clinical findings in disorders where Akt signaling is reduced but not absent, e.g., schizophrenia, autism and intellectual disability. Through a comprehensive battery of neurobehavioral tasks, we reveal that reduced Akt2 expression negatively impacts a range of cognitive domains including learning and memory, social behaviors, anxiety-related phenotypes, and brain dopaminergic tone. Moreover, we show that reduced Akt2 levels impair long-term potentiation (LTP) and electrophysiological properties of neurons in the cortex and hippocampus and produces a molecular signature involving mTOR and GSK3β, that does not overlap with phenotypes observed in Akt1 or Akt3 genetically modified mice. Together these studies support a unique physiological role of the Akt2 isotype in brain and identify impaired mTOR signaling as a potential biochemical mechanism underlying observed learning and memory impairments.

Materials/subjects and methods

Akt2 genetically modified mice

Akt2 HET mice (strain B6.Cg-Akt2tm1.1Mbb/J, http://jaxmice.jax.org/strain/006966.html), originally donated by the laboratory of Dr. Birnbaum [17] were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). Mice were maintained on a 12:12 h light: dark cycle in standard plastic cages with free access to food and water. A HET harem breeding trio scheme was used (2 females HET-Akt2tm1.1Mbb x 1 male HET-Akt2tm1.1Mbb) to obtain WT, HET and KO littermates. HET mice are fertile and breed normally. Pups were weaned at postnatal day 21 and housed up to 5 mice/cage with same sex littermates. All animal procedures were in accordance with and approved by the NIMH Animal Care and Use Committee and the University of Colorado Denver Institutional Animal Care and Use Committee. Sample size estimates were based on extensive prior studies in our lab [22, 30, 31].

Genotyping

Genotyping was performed on tail snip using standard polymerase chain reaction (PCR) (http://jaxmice.jax.org/protocolsdb/f?p=116:2:2226059766934507::NO:2:P2_MASTER_PROTOCOL_ID,P2_JRS_CODE:4745,006966). For DNA extraction, tails were digested for 2 h in lysis buffer (1 M Tris HCl pH 7.4, 0.5 M EDTA, 5 M NaCl, 10% SDS and 0.5 mg/ml Proteinase K) at 65 °C. Proteinase K was subsequently deactivated at 95 °C for 15 min, and samples were diluted to a concentration of 50–100 ng/μl. The following primers were used for genotyping: oIMR6745 TAC ACT TCA TTC TCA GTA TTG TTT TGC (mutant reverse), oIMR6746 ACC AAC CCC CTT TCA GCA CTT G (WT reverse), oIMR6747 TGC ACA ATC TGT CTT CAT GCC AC (forward) to generate a 277 bp band for the mutant gene and a 110 bp band for the endogenous gene.

Behavioral testing

3-month-old male Akt2 HET, KO and WT littermate mice were used for behavioral analyses and subsequently euthanized at 4–5 months of age for collection of brain tissue for biochemical analyses. In preparation for behavioral testing, mice were handled for three days in alternate days of the week preceding the test and allowed to habituate for 1 h before testing in a room adjacent to the behavioral testing room. Mice were divided into three cohorts and tested in the following behavioral tasks from the least to the most aversive test: Cohort 1: general health, basal locomotor activity, temporal order object recognition test, (11 WT, 12 HET, 6 KO); Cohort 2: sociability, novel object exploration, fear extinction (10 WT, 11 HET, 6 KO); Cohort 3: object location recognition, context fear, amphetamine-induced locomotor (10 WT, 10 HET, 5 KO). Investigators were blind to genotype until behavioral analyses were complete.

Amphetamine-induced locomotor activity

Mice were habituated in an open field arena (Accuscan 42x42x30, Columbus, OH, USA) for 1 h the day preceding the test. Locomotor activity was recorded by photo beam sensors installed on the edges of the arena. 24 h later the mouse was introduced into the same arena for 10 min to measure the baseline locomotor activity. Subsequently, the mouse was then removed from the arena, and randomly selected for either I.p injection with 3 mg/kg amphetamine or vehicle (saline) and returned into the test space for an additional 75 min. Locomotor activity was measured as distance traveled in the arena; values of distance traveled were graphed every 5 min.

Fear conditioning

The conditioned fear response is an established Pavlovian paradigm used to study the neurological basis of learning and memory. For fear conditioning, experimental mice are exposed to a neutral conditioning stimulus (tone) and to a tone paired to an aversive unconditioned stimulus (foot shock). The conditioning chamber (Med Associates Inc, Georgia, VT, USA) was equipped with a foot shock floor, speaker, and a digital video camera. Freeze Monitor software automatically scored the number of seconds the mouse spends freezing, defined as no movement except only those needed for respiration, given that the innate response to fear in mice is freezing behavior. Freezing behavior was calculated as the average time the mouse spends motionless (time frozen/allotted time x 100). Mice were tested for contextual fear memory and fear extinction learning as described in the Supplementary Material. Between experiments, mice were returned to their home cage for 24 h. The apparatus was cleaned with a 50% ethanol solution between experimental animals.

Recognition memory for objects, location, and temporal order tasks

All three cognitive tasks consist of an acquisition or sample phase and a retention or test phase [37, 38] wherein mice are presented with small heavy glass objects of different shape and color (black cuboids and white cones) which are randomly counterbalanced between subjects during testing; between phases mice are returned to their home cage. Mice were habituated to the testing apparatus, an open field arena (Plexiglas, Accuscan 42 × 42 × 30 cm), for 1 h the day before testing. During this habituation period spontaneous locomotor activity was recorded by photo beam sensors installed on the edges of the arena. Photo beams allowed calculation of both the distance traveled and time spent in the center and borders of the arena. We quantified the overall distance traveled to evaluate baseline locomotion, and the time spent in the center as a measure of anxiety.

Temporal order recognition test, novel object preference and object location recognition tasks

The temporal order recognition, novel object and object location tasks were performed as described previously [31]. Tasks are also described in detail in the Supplementary Material.

Social behavior

Social function was tested using methods previously described for assessment of sociability and preference for social novelty [31, 39]. Social approach was tested in an automated three chambered apparatus [39]. Mice used as the novel stimulus target were C57BL/6 matched to the subject mice by sex and age.

Electrophysiology

8–10-week-old male mice were euthanized by decapitation following isoflurane anesthesia. Brains were quickly removed and sliced with a vibrating blade microtome (Leica VT1000S, Leica Systems, Wetzlar, Germany) in oxygenated (95% O2 and 5% CO2) Ice-cold Na+-free sucrose solution containing 2.5 mM KCl, 1.25 mM NaH2PO4, 26 mM NaHCO3, 0.5 mM CaCl2, 4.0 mM MgCl2, 10 mM glucose and 250 mM sucrose. 3–4 mice (8–14 brain slices)/genotype were used for the investigation of hippocampal LTP, current clamp and voltage clamp assessment of mPFC pyramidal neurons. Details are provided in the Supplementary Material.

Western blotting

Proteins were extracted from fresh frozen mPFC using a lysis buffer containing 1 M DDT (Dithiothreitol, Sigma, St. Louis, MO, USA), protease inhibitor cocktail and T-PER Tissue protein Extraction buffer (Thermo Scientific, Rockford, IL, USA). Tissue was homogenized with a sonicator (Ultrasonic Processor model GE50 Sonics, Newtown, CT, USA). After 30 min on ice, samples were centrifuged at 13000 rpm for 3 min at 4 °C, and the supernatant was collected, and protein concentration quantified using a BCA tm protein assay kit (Thermo Scientific). 25 µl of protein/sample were loaded into a 15 well gel and after running, transferred onto PVDF membranes (Invitrogen Life Technology, Grand Island, NY, USA). Membranes were incubated overnight at 4 °C in the appropriate primary antibody (see Supplementary Material for full list) diluted in 5% milk in TBST (Tris-Buffered Saline 1% Tween 20. The following day after a series of washes, the membranes were incubated in the species appropriate HRP-conjugated secondary antibodies (Santa Cruz, CA, USA) at a 1:5000 dilution. Chemiluminescence signal was obtained using a Syngene image analyzer G:BOX (Frederick, MD, USA). The optical density of the protein bands was calculated using Image J software (NIH).

IC87114 testing in the temporal order object recognition task

To determine the preclinical potential of IC87114 to reverse recency discrimination impairments in Akt2 HET and KO mice a separate naïve cohort of mice were examined at 3–5 months of age. Testing and IC87114 administration was performed as previously described [31]. In brief, each genotype group was randomly split into two treatment groups, one group received a single intraperitoneal injection of 0.25% DMSO in saline, the other group received 0.1 mg/kg IC87114 dissolved in 0.25% DMSO in saline. Injections were performed 30 min before the test phase of the temporal order recognition task as described in the Supplementary Material.

Mouse ex vivo hippocampal neuronal dissociated cultures

Primary neuronal cultures were prepared from hippocampi of individual Akt2 mouse embryos on embryonic day 18 (E18) as described previously [31, 40]. Briefly, hippocampi were removed and dissociated from individual embryos and resuspended in supplemented DMEM. Cells were plated at a density of 100,000 cells/well on poly-D-lysine/laminin coated glass coverslips (BD Biosciences). Culture media was changed to Neurobasal (+2% B27, +1% Pen/Strep, +1%Glutamax) 24-h after initial plating and half of the media was changed every 7 days. Cells were transfected with a DNA plasmid containing pVenus-YFP [41] using Lipofectamine2000 (Invitrogen) per the manufacturer’s protocol apart from modifying the incubation period to 1 h for optimal cell viability. Two to three days following transfection, neurons were fixed using 4% paraformaldehyde/ 4% sucrose in PBS and subsequently immunostained with an antibody against GFP (Santa Cruz Biotechnology). Images were taken on a Zeiss inverted LSM700 confocal microscope with 10 and 20× objectives collected through a series of 5–8 z-sections max projected to create a 2D composite image for analysis in ImageJ (NIH). Measurements from at least 3 embryos/genotype were performed at DIV 3 (axon length and soma size) and DIV 10 (dendritic arborizations). Axons, somas, and dendrites were traced manually, and the Sholl analysis plugin on ImageJ, with a radius step size of 5 microns, was used as an additional measure of dendritic complexity. Viability assessments were performed by co-incubating neurons at DIV 4 with Calcein-AM (CAM) and Propidium Iodide (PI) (both Invitrogen) diluted in DMEM media for 30 min at 37 °C. Individual images of PI and CAM positive neurons were taken with a Zeiss epifluorescent inverted microscope at four quadrants of each coverslip. Images were co-localized in ImageJ (NIH) and the ratio of C-AM:PI neurons was calculated. Averages of ratios in all four quadrants of each coverslip was calculated to ensure reliability. All imaging and quantification were performed by a researcher blinded to genotype.

Data analysis

Data are expressed as the mean ± SEM throughout. All analysis was performed using IBM SPSS Statistics 21 software (Armonk, NY, USA). T-tests, one-way or two-way ANOVA, with repeated measurement (RM ANOVA) or multivariate (MANOVA) were used when appropriate, as described in the text. Poisson regression was used to model count variables. Logistic binomial regression was used to predict categorical variables from sets of predictor variables. Post-hoc analyses were Bonferroni corrected for multiple testing. P values ≤ 0.05 were considered statistically significant.

Results

Akt2 mutant mice display anxiety-like phenotypes and hyposensitivity to amphetamine

To determine the in vivo neurological consequences of full or partial genetic deletion of Akt2, we first performed a comprehensive screen of general health and motor reflex parameters in male mice genetically engineered with heterozygous (HET) or homozygous (KO) germ-line deletion of the Akt2 gene, compared to their wild type (WT) littermates. Akt2 HET and KO mice showed no physical abnormalities of body weight, fur condition, whisker presence, reflexes (righting, eye blink, ear twitch, and whisker twitch) or body and limb tone (Table 1). Additionally, no differences were observed in open cage behaviors such as transfer freezing, digging events and grooming events. Conversely, a significantly increased number of rearing events was observed in both Akt2 HET and KO mice (Poisson regression: Wald Chi-square = 47.840, df = 2, p ≤ 0.0001; sequential Sidak post-hoc: p ≤ 0.0001 WT vs HET, p ≤ 0.0001 WT vs KO). While there were no significant differences in the wire hang performance between genotypes, significantly decreased forepaw reaching (Logistic binomial regression: Wald Chi-square = 7.798, df = 2, p = 0.02; sequential Sidak post-hoc: p ≤ 0.0001 WT vs HET, p = 0.013 WT vs KO) and increased positional passivity (42% HET, 33% KO compared to 0% of WT, one-way ANOVA: F(2,29) = 3.24, p = 0.05) were also identified in both Akt2 HET and KO mice. These convergent findings are indicative of anxiety-like behaviors, rather than a muscle weakness phenotype.

Table 1 General health profile of AKT-2 HET and KO mice.
Full size table

In a separate cohort of animals we identified that heterozygous or homozygous deletion of the Akt2 gene did not impact brain weight (g) (WT = 0.438 ± 0.004, HET = 0.441 ± 0.005, KO = 0.424 ± 0.006, one-way ANOVA: F(2,45) = 2.341, p ≥ 0.05), body weight (g) (WT = 26.967 ± 0.703, HET = 25.860 ± 1.147, KO = 26.667 ± 0.804, one-way ANOVA: F(2,45) = 4.325, p ≥ 0.05), nor the ratio of brain:body weight (WT = 0.0165 ± 0.0052, HET = 0.0173 ± 0.00082, KO = 0.0161 ± 0.00054, one-way ANOVA: F(2,45) = 0.723, p ≥ 0.05). These observations confirm that gross basal metabolic abnormalities are not observed in the context of Akt2 deficiency.

Next, to further examine the anxiety-like phenotype we tested locomotor performance for Akt2 HET and KO mice in the open field arena compared to WT mice. Consistently, Akt2 HET traveled significantly less distance than WT littermates during the 60 min test (RM ANOVA: effect of genotype F(2,26) = 3.516, p = 0.045; Bonferroni post-hoc p = 0.014 WT vs HET; Fig. 1a, b). Furthermore, both HET and KO mice spent significantly less time in the center of the arena than WT littermates (RM ANOVA: effect of genotype F(2,26) = 3.369, p = 0.05, Bonferroni post-hoc p = 0.04 WT vs HET, p = 0.035 WT vs KO; Fig. 1c).

Fig. 1: Akt2 HET and KO mice show altered locomotor activity, response to amphetamine, associative learning of environmental context, and reassessment of a learned aversive cue.
figure 1

Distance (cm) traveled during 1-h locomotor activity (a) every 5 min and (b) overall traveled distance of Akt2 HET and Akt2 KO mice and their WT littermates. HET mice travel less distance than WT littermates (p = 0.014). c Overall time (s) spent in center of the arena during the 1-h open field task of Akt2 HET and Akt2 KO mice and their WT littermates. HET and KO mice spend less time in the center of the arena than WT littermates (p = 0.04 and 0.035, respectively). N = 11 WT, 12 HET, 6 KO. Bars represent mean ± SEM. d 10 min locomotor activity in naïve AKT2 HET and KO mice and their WT littermates, followed by injection of amphetamine 3 mg/kg or vehicle and subsequent 75 min recording of locomotor activity every 5 min. HET and KO mice injected with amphetamine travel less distance than WT mice (p ≤ 0.0001). N = 5/genotype/treatment. e Contextual fear memory of Akt2 HET and Akt2 KO mice and their WT littermates. The duration of immobility (% freezing) during baseline conditions (no-stimuli), post-training and context performed after 24 h. % freezing is significantly lower in Akt2 mutant mice (p = 0.016) at post-training (F(2,22) = 4.912, p = 0.017; post hoc HET vs WT p = 0.025, KO vs WT p = 0.099) and at context24h (F(2,22) = 8.639, p = 0.002; post hoc HET vs WT p = 0.002, KO vs WT p = 0.057). N = 10 WT, N = 10 HET and 5 KO. f Fear extinction learning of Akt2 HET and Akt2 KO mice and their WT littermates. Single freezing response to 5 tones and to 5 tone+shock pairings during day 1 (conditioning phase) and to 5 tones-only during day 2 and 3 (extinction phase, 24 and 48 h after conditioning, respectively). No extinction was observed in Akt2 HET and KO mice (day 2 tone×genotype F(10,120) = 3.434, p = 0.001; day 3 tone×genotype F(10,120) = 2.276, p = 0.018). N = 10 WT, 11 HET, 6 KO. Data represent mean ± SEM. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

Full size image

Next, in a separate naive cohort of Akt2 WT, HET and KO mice we sought to determine if Akt2 expression impacts brain dopaminergic tone through sensitivity to amphetamine challenge. Intraperitoneal injection of amphetamine strongly induces locomotor activity through the inhibition of dopamine reuptake [42]. Thus, we monitored locomotor response to amphetamine (3 mg/kg i.p.) in Akt2 KO and Akt2 HET mice compared to their WT littermates (Fig. 1d). 3 mg/kg amphetamine injection significantly increased locomotor activity in the open field (RM ANOVA: treatment F(1,24) = 99.24, p < 0.0001). However, a significant genotype x treatment interaction was observed (genotype x treatment; F(2,24) = 32.72 p < 0.00001), whereby Akt2 HET and KO mice exhibited a significantly blunted locomotor response to amphetamine compared to WT mice (Bonferroni post-hoc WT vs HET p ≤ 0.0001, WT vs KO p = 0.001; Fig. 1d). These data demonstrate that dopaminergic tone is altered in the context of reduced Akt2 expression.

AKT2 mutant mice exhibit impaired associative learning and memory

Neurons encoding fear memories are located in the lateral nucleus of the amygdala [43] and send information to the hippocampus and medial prefrontal cortex (mPFC), where extinction of fear and contextual memory are elaborated [44, 45]. Akt2 WT, HET and KO mice were tested for contextual (Fig. 1e) and extinction fear memory (Fig. 1f) using classic Pavlovian paradigms. During contextual fear conditioning, there was a main effect of genotype (MANOVA: p = 0.016), which was significant in both the post-training (F(2,22) = 4.912, p = 0.017) and in the context condition (F(2,22) = 8.639, p = 0.002). At baseline freezing was equal to zero in all groups while at post-training a robust freezing response was observed in WT mice but a much lower extent of freezing was seen in mutant mice; post-hoc analyses showed a significant reduction of freezing in Akt2 HET mice, in Akt2 KO mice the same trend of reduction in freezing was observed (Bonferroni post-hoc: HET vs WT p = 0.025, KO vs WT p = 0.09). Re-exposure to the same context after 24 h, without cue, elicited a similar pattern of % freezing to that observed during the post-training with a significant reduction of freezing in Akt2 HET mice and a marginal significance in Akt2 KO mice compared to WT (Bonferroni post-hoc: HET vs WT p = 0.002, KO vs WT p = 0.057). Our data indicate that Akt2 HET and KO mice fail to associate the auditory cue and the context with the aversive event (foot shock), suggesting that Akt2 is essential for encoding of associative learning and acquisition memory.

In the fear extinction paradigm (Fig. 1f) all three genotypes responded with substantial freezing to tone+shock parings during training (RM ANOVA: effect of stimulus F(9,216) = 46.387, p ≤ 0.001; tone×genotype F(18,216) = 0.336, p = 0.995) with no effect of genotype (F(2,24) = 0.355, p = 0.705). However, during the extinction phases (days 2 and 3), Akt2 HET and KO mice showed a stable % freezing without any extinction trend as indicated by a significant tone by genotype statistical interaction (day 2 RM ANOVA: within-subject effect of tone F(5,120) = 3.674, p = 0.004, tone×genotype F(10,120) = 3.434, p = 0.001; day 3 RM ANOVA: within-subject effect of tone F(5,120) = 1.761, p = 0.126, tone×genotype F(10,120) = 2.276, p = 0.018). Together, these data demonstrate that Akt2 is critical for associative learning, memory, and extinction, demonstrating that even a partial loss of Akt2 expression leads to a malfunctioning of the neural circuity between the amygdala, hippocampus and mPFC.

AKT2 mutant mice demonstrate altered PFC- and hippocampus-dependent memory tasks

To further probe the role of Akt2 signaling in the neural circuitry involving the mPFC and hippocampus we tested separate naïve cohorts of WT, HET and KO Akt2 mice in a series of three recognition memory tasks differentially associated with function of the mPFC, perirhinal cortex and hippocampus [31, 37]. Recognition memory requires accurate identification of the previous occurrence of stimuli made on the basis of the relative familiarity of objects, or by integrating information concerning object location, or by using recency discrimination [31, 37, 38]. In the temporal order discrimination test, which requires intact functioning of the mPFC, perirhinal cortex and hippocampus, Akt2 HET and Akt2 KO mice failed to discriminate between the least recent and the most recent object (one-way ANOVA: F(2,26) = 11.667, p ≤ 0.0001; Bonferroni post-hoc p = 0.002 HET vs WT, p = 0.001 KO vs WT; Fig. 2a) indicating that Akt2 is critical for recency discrimination memory and development of the mPFC, perirhinal cortex and/or hippocampus. The novel object preference task, which requires intact function of the perirhinal cortex alone, was unimpaired in both Akt2 HET and KO mice with no effect of genotype being observed in the performance of this task (one-way ANOVA: F(2,24) = 1.095, p = 0.351; Fig. 2c). Conversely, Akt2 genotype significantly impacted performance in the object location recognition test (one-way ANOVA: F(2,22) = 3.481, p = 0.049; Fig. 2e), post-hoc analysis revealed that Akt2 KO (but not Akt2 HET) displayed impaired task performance dependent upon intact hippocampal functioning (Bonferroni post-hoc p = 0.049 KO vs WT). Importantly, the overall extent of exploration (object exploration time) did not differ between genotypes in any of the tasks denoting normal exploration (Fig. 2b, d, f). Together, these data demonstrate that cognitive function dependent on intact PFC-hippocampal neural circuitry requires intact Akt2 signaling. Our data also support that the extent of perturbed circuit function is dependent upon the magnitude of Akt2 impairment, with HET mice displaying selective impairment in mPFC function, whereas complete knockout of Akt2 leads to impairments of both mPFC and hippocampal function.

Fig. 2: Akt2 genetically modified mice display impaired recency discrimination memory and object location memory, but intact novel object memory.
figure 2

a, b Temporal order recognition test of Akt2 HET and KO mice and their WT littermates. a Recency discrimination ratio of time spent exploring the more recent object compared to a less recent. The ratio is lower in HET (p = 0.002) and KO mice (p = 0.001). b Overall exploration time of both objects during samples 1, 2 and 3. N = 11 WT, 12 HET, 6 KO. c, d Novel object recognition test of Akt2 HET and KO mice and their WT littermates. c Preference index for time spent exploring the novel object compared to a familiar one. d Overall exploration time of the objects during samples 1 and 2. N = 10 WT, 11 HET, 6 KO. e, f Object location recognition of Akt2 HET and KO mice and their WT littermates. e Preference index for the time spent exploring the misplaced object compared to object in its original location. The index is significantly lower in KO mice (p = 0.049 KO vs WT). f Overall exploration time of the objects during samples 1 and 2. N = 10 WT, 10 HET, 5 KO. Data represent mean ± SEM. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

Full size image

Akt2 HET and KO mice display impaired social interaction

To assess the impact of Akt2 deficiency on social cognitive domains we tested Akt2 WT, HET and KO mice in rodent tasks of social behavior [39, 46]. Using a three-chamber apparatus, murine social preference is measured as the predisposition to spend more time in the chamber containing a stranger conspecific (i.e., novel mouse 1) than the chamber containing an inanimate object (i.e., novel object). In the sociability test a significant main effect of chamber was observed for sniff time (RM ANOVA F(1,24) = 174.044, p ≤ 0.0001) and time in chambers (RM ANOVA F(1,24) = 8.919, p = 0.006). An interaction of genotype by chamber was observed for time in chamber (RM ANOVA: genotype×chamber F(2,24) = 4.359, p = 0.024), whereby Akt2 HET (p = 0.005) and WT (p = 0.003) mice, but not AKT2 KO mice, demonstrated a preference for the chamber containing the novel mouse 1 over the chamber containing the novel object (Fig. 3a). These findings demonstrate that Akt2 KO mice show impaired social preference and fail to discriminate between a novel mouse and novel object. In all three genotypes, sniff time of the novel object was lower than that of novel mouse 1 (Akt2 WT p ≤ 0.0001, HET p ≤ 0.0001, and KO p = 0.006) (Fig. 3b).

Fig. 3: Abnormal sociability and social novelty preference in Akt2 HET and KO mice.
figure 3

a, b 10 min social preference test of Akt2 HET and KO mice and their WT littermates. a Time spent in the chambers containing the novel mouse 1 or the novel object, and in the empty center chamber. WT (**p = 0.003) and HET (**p = 0.005) mice spend more time in the novel mouse 1 chamber compared to novel object chamber. b Time spent sniffing the novel mouse 1 and the novel object. Sniff time of the novel object is lower than that of novel mouse 1 in all genotypes (WT and HET ***p ≤ 0.0001, KO**p = 0.006). c, d 10 min social novelty preference test of Akt2 HET and KO mice and their WT littermates. c Time spent in the chambers containing the novel mouse 1 or the novel mouse 2, and in the empty center chamber. WT mice spend more time in the novel mouse 2 chamber compared to the novel mouse 1 chamber (***p ≤ 0.001). d Time spent sniffing the novel mouse 1 and the novel mouse 2. Sniff time of novel mouse 2 was higher in WT mice (**p ≤ 0.01). e Number of entries in left and right chambers during the initial 10-min habituation period. N = 10 WT, 11 HET, 6 KO. Data represent mean ± SEM.

Full size image

Preference for social novelty is defined as the propensity to spend more time in the chamber containing a stranger conspecific (novel mouse 2) compared to the chamber containing a familiar conspecific (novel mouse 1). A significant effect of chamber was observed on the time spent in chamber (RM ANOVA: F(1,24) = 11.794, p = 0.002). Moreover, an interaction of genotype by chamber was found (RM ANOVA: genotype×chamber F(2,24) = 6.671, p = 0.005), whereby WT (p ≤ 0.001), but not Akt2 HET or KO mice, demonstrated a preference for the chamber containing the novel mouse 2 rather than the chamber containing the novel mouse 1 (Fig. 3c). These data demonstrate deficits of social novelty in both Akt2 Het and KO mice.

A significant effect of chamber was observed on sniff time in the preference for social novelty test (RM ANOVA: F(1,24) = 5.510, p = 0.027). No genotype x chamber interaction (p = 0.332) or genotype effect (p = 0.341) was observed. However, Akt2 WT mice, but not Akt2 HET and KO mice, spent more time sniffing novel mouse 2 than novel mouse 1 (p = 0.005).

Together, these results indicate a selective dose-dependent effect of Akt2 deletion on sociability and social preference, with KO mice exhibiting a pronounced deficit in sociability and social novelty, while HET mice show a selective deficit in social preference. Importantly, we did not observe any effects of genotype on entries into the left or right chambers during the habituation phase of the task (Fig. 3e).

Hippocampal long-term potentiation (LTP) is impaired in AKT2 KO mice

As a neural substrate underlying the observed memory impairments dependent on hippocampal function, we investigated whether reduced Akt2 expression alters synaptic plasticity using field recordings of excitatory postsynaptic potentials (fEPSPs) and measurements of long-term potentiation (LTP). Extracellular field recording to measure LTP was conducted in the CA3-CA1 synapses of acute hippocampal slices from 8–10-week-old mice (Fig. 4a, b). A main effect of genotype was observed on LTP (ANOVA, F(2,30) = 3.479, p = 0.044), whereby Akt2 KO mice displayed significantly reduced LTP compared to WT mice (Bonferroni post-hoc, p = 0.041). Specifically, LTP was only induced at 118 ± 7% potentiation at 50–60 min in Akt2 KO mice, whereas LTP was induced at 142 ± 11% potentiation in Akt2 HET and 153 ± 8% in WT mice. Paired pulse facilitations (PPFs; Fig. 4c) were virtually identical over a wide range of interpulse intervals in all genotypes and basal transmission in hippocampal CA3-CA1 synapses (Fig. 4d) was not influenced by genotype. These data demonstrate that Akt2 KO mice exhibit impaired LTP in the absence of changes in paired pulse facilitations, indicating a postsynaptic, rather than presynaptic mechanism [47] of hippocampal circuit dysfunction.

Fig. 4: Significant impairment of hippocampal LTP and decreased mPFC pyramidal neuron excitability in Akt2 KO mice.
figure 4

a Representative fEPSP recordings before (black lines) and 60 min after LTP induction (red lines) in hippocampal slices of Akt2 HET and KO mice and their WT littermates. b Complete time courses of fEPSP in hippocampus: the black bar represents the injection time point of 4 TBS at 100 Hz to induce LTP. fEPSP is reduced in KO mice compared to WT (KO vs WT: p = 0.044). N = 4 WT (8 slices), 4 HET (9 slices), 4 KO (8 slices). c PPF ratio in hippocampus. The ratios of the second and first EPSP slopes were calculated, and mean values are plotted against different inter-pulse intervals (IPI, 12.5–1600 ms). N = 3 WT (8 slices), 3 HET (8 slices), 3 KO (8 slices). d Basal synaptic transmission in hippocampus. Input-output curves were generated by plotting the postsynaptic response (initial slope of fEPSP amplitude, mV) as a function of the stimulation intensity. N = 4 WT (12 slices), 4 HET (10 slices), 4 KO (14 slices). e Representative traces of pyramidal neuron excitability in mPFC lamina II/III of Akt2 HET and KO and their WT littermates mice. f Representative traces of pyramidal neuron excitability in lamina V of mPFC of Akt2 HET and KO mice. g Neuronal firing frequency (spikes) of mPFC lamina II/III. Firing frequency is significantly lower in Akt2 KO mice compared mice when injection current intensities were from 350 pA to 450 pA (p value, from < 0.0001 to 0.0025). N = 3 WT (7 slices, 11cells), 3 HET (6 slices, 9 cells), 3 KO (9 slices, 21 cells). h Neuronal firing frequency (spikes) of mPFC lamina V. Firing frequency is significantly lower in Akt2 KO mice when injection current intensities were from 350 pA to 450 pA (p value, from < 0.0006 to 0.05), respectively. No difference was observed in mPFC lamina V pyramidal neuron excitability between HET and WT mice. N = 3 WT (7 slices, 12 cells), 3 HET (8 slices, 15 cells), 3 KO (9 slices, 21 cells). Data represent mean ± SEM. *p ≤ 0.05, **p ≤ 0.01.

Full size image

Reduced intrinsic excitability of mPFC pyramidal neurons of lamina II/III and V in Akt2 KO mice

To study the physiological ramifications of Akt2 deficiency in the prefrontal cortex, we assayed intrinsic excitability in the form of repetitive firing patterns using whole-cell current clamp recording in pyramidal neurons of lamina II/III and V of the mPFC slices from adult mice (Fig. 4e–h). A two-way ANOVA analysis revealed a significant interaction between genotype and stimulation intensity on the excitability of lamina II/III pyramidal neurons (F(16,342) = 3.792, p < 0.0001). Neuronal firing frequencies were significantly lower in Akt2 KO mice when injection current intensities ranged from 300 pA to 450 pA (p value range, from < 0.0001 to 0.0025, Fig. 4g). No difference was observed in mPFC lamina II/III pyramidal neuron excitability in Akt2 HET mice (p = 0.409, Fig. 4g). Next, we measured firing frequencies of mPFC lamina V pyramidal neurons. Consistently, a significant interaction between genotype and stimulation intensity on neuronal firing frequency was observed (two-way ANOVA analysis F(16,405) = 3.436, p < 0.0001). Again, neuronal firing frequencies in Akt2 KO mice were significantly lower when injection current intensities were from 350 pA to 450 pA (p value, from 0.024 to 0.001; Fig. 4h), respectively. No difference was observed in mPFC lamina V pyramidal neuron in Akt2 HET mice (p = 0.343, Fig. 4h). The resting membrane potential did not differ between genotypes in either lamina II/III or lamina V neurons (Fig. 4e, f). Together, these data indicate that Akt2 is critical for normal membrane potential, mediated through potential alterations in the biophysical properties of voltage gated ion channels, leading to reduced firing rates and reduced excitability of pyramidal cells within layers II/III and V of the mPFC.

Akt2 deficiency causes abnormal excitatory and inhibitory transmission in mPFC pyramidal neurons

To investigate excitatory and inhibitory pre- and post-synaptic transmission in prefrontal cortical pyramidal neurons, whole-cell voltage-clamp recording was performed in mPFC slice preparations of 8–10-week-old mice (Supplementary Fig. 1). In Akt2 KO mice, lamina V pyramidal neurons exhibited significantly higher mEPSCs amplitude compared to WT (KO −7.21 ± 0.2 pA vs. WT −5.71 ± 0.17 pA; one-way ANOVA: F(2,45) = 5.890, p = 0.005; Bonferroni post-hoc: p = 0.006 KO vs WT, Supplementary Fig. 1D) and faster mEPSCs rise time (KO 1.85 ± 0.06 ms vs. WT 2.25 ± 0.05 ms; one-way ANOVA: F(2,45) = 11.863, p ≤ 0.001; Bonferroni post-hoc: p ≤ 0.001 KO vs WT; Supplementary Fig. 1E). No deficits in mEPSC frequency (Supplementary Fig. 1C) or decay time (Supplementary Fig. 1F), were observed. Together, these data demonstrate that Akt2 is critical for normative cortical lamina V pyramidal excitatory transmission and indicate altered density, conductance, or kinetics of postsynaptic AMPA-type ionotropic glutamate receptors [48, 49] rather than presynaptic release probability or excitatory synaptic number. Interestingly, whole-cell voltage-clamp recording in lamina II/III mPFC slices revealed no abnormalities of mEPSC frequency, amplitude or kinetics in Akt2 KO mice, suggesting that at least in the context of complete loss of function of Akt2, glutamatergic neurotransmission abnormalities appear laminar-specific.

In lamina V, increased frequency of miniature inhibitory postsynaptic currents (mIPSC) was observed in Akt2 KO mice (KO 7.01 ± 0.31 Hz vs. WT 5.60 ± 0.31; one-way ANOVA: F(2,41) = 21.637 p ≤ 0.001; Bonferroni post hoc: p = 0.001 KO vs WT; Supplementary Fig. 1I). No differences were observed in mIPSC amplitude (Supplementary Fig. 1J), indicating that Akt2 directly or indirectly modulates presynaptic GABAergic terminal release machinery to enhance inhibitory synaptic strength onto excitatory pyramidal neurons. Interestingly, mIPSCs event kinetics in lamina V pyramidal neurons of Akt2 KO mice also exhibited faster rise time (KO 2.82 ± 0.06 ms vs. WT 3.42 ± 1.92 ms; one-way ANOVA: F(2,41) = 4.147, p = 0.023; Bonferroni post-hoc: *p ≤ 0.02 KO vs WT; Supplementary Fig. 1K), additionally implicating altered gating properties of postsynaptic GABAA receptor channels. Overall, these data suggest that Akt2 KO mice show altered inhibitory and excitatory synaptic transmission in pyramidal neurons within lamina V in mPFC, with complex pre- and post-synaptic mechanisms.

Interestingly, a more moderate and somewhat contrasting pattern of altered neurotransmission was observed in the context of Akt2 heterozygosity, with increased mEPSCs frequency in lamina II/III pyramidal neurons (HET 5.77 ± 0.60 Hz vs. 2.87 ± 0.22 Hz WT); one-way ANOVA: F(2,43) = 19.540, p ≤ 0.001; Bonferroni post-hoc: p ≤ 0.001 HET vs WT; Supplementary Fig. 1C) and decreased frequency of mIPSCs in lamina V pyramidal neurons (HET 4.32 ± 0.33 Hz vs. WT 5.60 ± 0.31; oneway ANOVA: F(2,41) = 21.637 p ≤ 0.001; Bonferroni post hoc: p = 0.005 HET vs WT; Supplementary Fig. 1I) being observed.

Pharmacological inhibition of p110δ improves recency discrimination deficits in Akt2 HET and KO mice

To determine whether modulation of AKT activity, in the context of Akt2 deficiency could rescue cognitive impairments in Akt2 HET and KO mice, we evaluated the generality of systemic effects of pharmacological inhibition of p110δ with the small molecule inhibitor, IC87114 [30, 31], in a separate naïve cohort of mice. Preclinical administration of IC87114 increases cortical Akt phosphorylation and has shown antipsychotic and pro-cognitive efficacy in several rodent models of psychiatric relevance [30, 31]. Here we chose to assess the efficacy of IC87114 in the temporal order recognition task as both Akt2 HET and KO animals display prominent impairments in this task, and prior work demonstrated preclinical efficacy of IC87114 to rescue such deficits in other rodent models associated with altered Akt signaling [31]. Firstly, as reported in Fig. 2 in a separate, replication cohort of mice, Akt2 HET and KO mice consistently showed an allele-dependent deficit in recency discrimination memory (main effect of genotype, F(5,49) = 6.524, p < 0.01). IC87114 treatment rescued deficits seen in Akt2 HET and KO mice with discrimination ratios comparable to IC87114 treated WT mice (treatment x genotype interaction F(2,49) = 3.418, p < 0.05; Fig. 5a). Although discrimination ratios in treated WT mice appeared to be lower than untreated WT mice, no main effect of treatment was observed. These data suggest that cognitive deficits observed in the context of Akt2 loss-of-function are amenable to pharmacological intervention, via novel a therapy aimed at modulating PI3K/AKT signaling [30, 31].

Fig. 5: Preclinical relevance of PIK3CD signaling and pharmacological inhibition in Akt2 mice and evidence that Akt2 regulates mTOR signaling.
figure 5

a Discrimination ratio in Akt2 WT, HET and KO mice in the temporal order object recognition task. Mice received a single intraperitoneal injection of 0.1 mg/kg IC87114 or saline (vehicle) 30 min prior to the test phase. (WT/VEH n = 7, WT/IC87114 n = 9, HET/VEH n = 13, HET/IC87114 n = 14, KO/VEH n = 6, KO/IC87114 n = 6). b–f Relative expression of select signaling proteins in the AKT pathway. N = 4/genotype. Data represent mean ± SEM. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.0001.

Full size image

Proteomic changes in mPFC of Akt2 HET and KO mice

We next assessed a panel of select synaptic and intracellular proteins associated with PI3K/AKT/mTOR signaling in mPFC lysates of Akt2 mice. Notably, mTOR total protein was dramatically reduced in Akt2 HET (Students T test t = 4.77; p = 0.003) and KO mice (Students T test t = 3.15; p = 0.02; Fig. 5d). Akt2 loss (KO) was also associated with increases in the activity of PDK1, as assessed by levels of phosphorylation of Ser-241, (Students T test t = −3.28; p = 0.017; Fig. 5b), implicating loss of feedback inhibition from mTORC1. Consistently, we observed significant increases in the total phospho-AKT-Thr308 in Akt2 KO mice (Students T test t = −2.968; p = 0.025; Fig. 5c). Since PDK1 is responsible for the phosphorylation of AKT at Thr308, these data demonstrate that Akt2 deficiency results in reduced mTOR levels and compensatory overactivation of remaining AKT signaling, potentially through loss of feedback inhibition to PDK1. The observation that phosphorylation levels mTOR were overall significantly increased in the context of Akt2 deficiency (WT vs HET, Students T test t = −4.37; p = 0.005; WT vs KO, Students T test t = 4.05; p = 0.007; Fig. 5d) further supports this hypothesis. Consistently, mTOR activity is a key regulator of translation and biosynthesis of the ribosomal subunit protein S6 whose total levels were increased (WT vs HET, Students T test t = −2.85; p = 0.029; WT vs KO; Students T test t = 3.71; p = 0.01; Fig. 5d). Activated Akt phosphorylates numerous downstream substrates, including glycogen synthase kinase-3 (GSK3β), which regulates glycogen and glucose metabolism. Notably, levels of total GSK3β were upregulated in Akt2 HET (Students T test t = −4.88; p = 0.003) and KO mice (Students T test t = −3.54; p = 0.012; Fig. 5e).

Total protein levels of Akt1 and Akt3 were not changed in the context of Akt2 reduction, suggesting that haplo- or null- insufficiency of Akt2 does not result in compensatory expression changes in other Akt isoforms in the brain (Fig. 5c). As expected, Akt2 protein was not detected in Akt2 KO mice (Students T test t = 10.1; p < 0.0001; Fig. 5c) and its level was approximately 50% in Akt2 HET mice (Students T test t = 3.50; p = 0.013; Fig. 5c). Consistently phospho-AKT2-Ser474, the specific phosphorylation site at the Serine residue for Akt2, was absent in Akt2 KO mice, however was significantly elevated in Akt2 HET mice (WT vs HET, Students T test t = −10.548; p < 0.001 Fig. 5c).

Akt2 loss-of-function impacts hippocampal neuronal morphology and dendritic anatomy

Finally, to assess the impact of genetic insufficiency of the Akt2 gene on neuronal development at the morphological level, we examined ex vivo primary hippocampal neuronal cultures from Embryonic day 18 (E18) WT, Akt2 HET and Akt2 KO embryos. At day in vitro (DIV) 3 hippocampal neuron axon length was significantly affected by Akt2 genotype (one-way ANOVA: F(2,387) = 3.946, p < 0.05), with both Akt2 HET and KO neurons having significantly shorter axons (Bonferroni post-hoc: WT vs HET, p < 0.05, WT vs KO, p < 0.01, Supplementary Fig. 2a). Conversely, soma size was increased in the context of Akt2 insufficiency, this was selectively seen in neurons derived from Akt2 KO mice (one-way ANOVA: F(2,387) = 3.95, p < 0.05), Bonferroni post hoc: WT vs KO, p < 0.01; Supplementary Fig. 2b). Hippocampal neuron viability was unaffected by Akt2 genotype, as assessed by the percentage of calcein-AM to propidium iodide positive neurons reflective of the ratio of viable/apoptic neurons at DIV4 (Supplementary Fig. 2c). Upon longer culturing conditions to assess dendritic maturation, at DIV10 hippocampal neurons derived from Akt2 HET and KO embryos had significantly altered dendritic patterning as assessed by Sholl analysis. While there was the expected significant reduction in dendritic intersections as distance from soma increased (RM ANOVA: Distance F(37) = 202.191, p < 0.001), both Akt2 HET and KO neurons displayed a downwards shift in their slope resulting in a significant distance by genotype interaction (F(74) = 1.381, p = 0.017), and a main effect of genotype (F(2,238) = 4.525, p < 0.01). This simplification in dendritic maturation was observed in both heterozygous and Akt2 KO neurons (Bonferroni post hoc: WT vs KO, p < 0.01; WT vs HET, p < 0.05). This simplification was mainly evident at distances between 75–165 µm from the soma (Supplementary Fig. 2d). Since the structure of a neuron’s dendritic tree and axonal morphology defines how the cell receives and transmits synaptic inputs, these data provide anatomical evidence that hippocampal neuron impairment following Akt2 expression reduction may mechanistically underlie learning and memory dysfunction in Akt2 mice.

Discussion

The Akt signaling network is essential for normative brain development and alterations in the pathway are implicated in several neuropsychiatric and neurocognitive disorders, with Akt drawing attention as a potential novel target for antipsychotic, antidepressant and pro-cognitive drug development [30, 31, 50]. Given that three Akt isoforms exist, all of which are highly expressed in the brain, a comprehensive understanding of the distinct physiological roles of each isoform in the context of brain development and function is necessary. While Akt1 and Akt3 have been more extensively studied, little is known about the neurological functions of Akt2, with prior studies [32, 33] limited to Akt2 KO mice and examination of limited phenotypes. Together our findings demonstrate that Akt2 mice exhibit only a partial overlap in phenotypes with Akt1 and Akt3 mutant mice, providing further evidence for the non-redundant isoform-specific roles of Akt isoforms in the brain.

Consistent with a previous preliminary study [32], we demonstrate that Akt2 expression regulates anxiety-like behavioral phenotypes. We expand previous findings to demonstrate that heterozygosity of the Akt2 allele is sufficient to exert an anxiogenic phenotype, which is likely more clinically relevant to human studies of the involvement of the AKT2 locus in neurological disorders [34, 36]. Anxiety-like features have not been observed in Akt3 [22], nor in Akt1 deficient mice [24, 28], suggesting that this affective domain may be selectively associated with Akt2.

Like Akt1 deficient mice, we found that reductions in Akt2 attenuated response to amphetamine. A reciprocal regulation of AKT by dopamine type-2 receptors and serotonin has been proposed as a molecular mechanism of therapeutic efficacy of neuroleptic drugs [29]. Our findings in combination with previous studies [25, 26] suggest that Akt1 and Akt2 regulate dopaminergic tone. We posit however, that Akt1 and Akt2 regulate dopaminergic neurotransmission via alternative mechanisms. Akt2 deficient mice develop hyper-insulinemia [17], and insulin growth factor (IGF) signaling in the brain has been shown to directly regulate the actions of dopamine via altering the expression of surface dopamine receptors [51]. Furthermore, dopamine transporter expression is altered selectively by pharmacological Akt2 inhibition, but not by inhibition of Akt1 [52].

Heterozygous and homozygous deletion of Akt2 dramatically impaired recency discrimination memory but had no impact on novel object preference, suggestive of intact perirhinal- but impaired mPFC and hippocampal function [37]. Additionally, Akt2 KO mice selectively demonstrated impaired object location recognition, an index of hippocampal dysfunction that appears to be unique to complete loss of Akt2 function. As described by rat micro-lesion studies, the combination of the three tests allows for the identification of specific brain region malfunctions [37, 38]. These findings suggest that heterozygous deletion of Akt2 disrupts mPFC function, whereas complete knockout of Akt2 additionally impacts hippocampal physiology (consistent with the deficits in hippocampal LTP observed). Interestingly, prior data show that Akt1 homozygous deletion does not affect the execution of a PFC-dependent working memory task [24]. We have previously shown that Akt3 heterozygous and knockout mice also display recency discrimination and spatial location memory deficits [22], suggesting that Akt2 and Akt3 have overlapping physiological functions in these cognitive domains. In contrast, Akt3 genetic deletion mice do not display impaired associative learning and memory in Pavlovian fear-conditioning tasks [22]. Given that contextual fear conditioning is commonly used to measure cortical/hippocampal communication with the amygdala during emotion-associated learning [53], and that the processing of fear extinction involves the activation of the infralimbic PFC [54], these findings provide further evidence that Akt2 is critical for cortico-hippocampal circuit development and function.

Consistent with our neurobehavioral findings, observation that Akt2 heterozygous mice harbor a more selective prefrontal cortical deficit, whereas hippocampal function is additionally impacted in Akt2 KO; electrophysiological analyses revealed that both Akt2 KO and HET mice show alteration of inhibitory and excitatory neurotransmission in pyramidal neurons of mPFC, that serves to alter excitatory/inhibitory (E/I) balance. Additionally, Akt2 KO mice display reduced pyramidal neuron intrinsic cell excitability and hippocampal LTP. Since the pair-pulse facilitation ratio was normal, the LTP deficit is typically attributed to postsynaptic mechanisms [47]. Postsynaptic currents are indicative of kinetics in AMPA or GABAA receptors, which mean the open and close speed of the receptors. Since some variation was observed between Akt2 HET and KO mice, and lack of changes in total protein levels of AMPA receptors were observed, this suggests that there may be a change in either 3-dimensional receptor structure or functionality. Similar to Akt2 KO mice, Akt1 KO mice showed reduced hippocampal LTP [27], whereas Akt3 KO mice display normal hippocampal LTP [55].

In addition to altered electrophysiological properties, we also found that hippocampal neuronal morphology and dendritic patterning was altered in Akt2 HET and KO mice. In keeping with a previous study [23] we report that reduced Akt2 expression leads to decreased axiogenesis, and expand on these findings to show that soma size is increased in Akt2 KO neurons, likely as a compensatory mechanism. We also found that neurons derived from Akt2 HET and KO mice display immature dendritic patterning, which may provide the cellular mechanism underlying impaired hippocampus-dependent cognitive behaviors. Impaired hippocampal and striatal neuronal dendritic complexity has also been observed in Akt1 KO mice [24, 25]. While it has been previously suggested that immunohistochemical localization of Akt2 is confined to astrocytes of the hippocampus [33], single cell RNA-seq transcriptomic studies in both human and mouse brain, while confirming higher expression in astrocytes, also confirm expression in neurons [56, 57]. The cellular dynamics and cell-type specific effects of how Akt2 reduction impacts neuronal development requires further investigation.

At the biochemical level, we demonstrate that expression of key members of the PI3K/AKT/mTOR pathway are altered in the context of Akt2 deficiency. Notably, both Akt2 HET and KO mice display dramatic (>50%) reductions in total mTOR. mTOR activity is a critical regulator of neurodevelopment and is disrupted in several neurological diseases, including autism, schizophrenia, and intellectual disability [58,59,60]. Studies of pharmacological blockade of mTOR signaling (via Rapamycin) consistently show that attenuation of mTORC1 impairs learning and memory, through inhibition of translation control, synaptic plasticity and blockade of LTP [61,62,63,64], suggesting that decreased expression of mTOR, mechanistically underlies observed learning and memory deficits in Akt2 mutant mice. In addition, the known roles of mTOR in regulating neuronal physiology and morphology [65,66,67] are mechanistically consistent with our observations of reduced axiogenesis and dendritic patterning. Furthermore, decreased intrinsic excitability of cortical pyramidal neurons in Akt2 mutant mice is biologically consistent with loss of mTOR function, and consistent with the converse knowledge that hyperactivation or mTOR pathogenically underlies epileptogenic activity [66].

Signaling pathways are not linear but are tightly regulated by positive and negative feedback mechanisms [68]. Studies of compensatory-activation of PI3K/AKT signaling in preclinical cancer models, mouse models and human cells, have demonstrated that a prominent consequence of mTORC1 inhibition, is a dramatic increase in Akt phosphorylation at Thr308 by PDK1, due to the loss of feedback inhibition of the AKT pathway via failure of inhibition and degradation of the IRS-1 receptor [68,69,70]. In this context our findings are consistent. Our observations of attenuated mTOR levels in the context of increased total pAKT308 and increased phosphorylation and activity of mTOR demonstrate that loss of Akt2 specifically results in downregulation of mTOR protein which results in loss of normal feedback inhibition and subsequent overactivation of remaining AKT isotypes and deregulated AKT signaling. The mechanisms by which Akt2 activity influences proteostasis of mTOR require further investigation, and studies of IRS signaling are currently underway. Finally, another key downstream target of AKT signaling altered in Akt2 deficient mice is GSK3β, which was increased in both genotypes. GSK3 signaling plays critical roles in neurodevelopment via regulation of glucose metabolism, insulin signaling and inflammatory mechanisms [71, 72] and is the target for the mood stabilizer lithium, the first line therapy for bipolar disorder. Lithium inhibits GSK3β activity, directly- and indirectly through modulation of AKT activity with consistent antidepressant and anxiolytic effects observed in rodent models of depression [72, 73]. Our data suggest that increased GSK3β may represent a molecular mechanism underlying the anxiety and depressive [32] phenotypes observed in the context of Akt2 deficiency. Furthermore, emerging evidence suggests that impaired inhibitory regulation of GSK3β may contribute to social dysfunction [74], consistent with those observed in our model.

In summary, our convergent data demonstrate that Akt2 in the murine brain regulates learning and memory and mPFC function, in addition to depression and anxiety-like phenotypes. Akt2 loss-of-function produces cognitive deficits dependent upon cortico-hippocampal circuit function, which have partial overlap with previous observations in Akt1 and Akt3 genetic deletion mice and can be rescued by pharmacological modification of PI3K signaling. Moreover, we demonstrate that Akt2 is critical for physiological, morphological, and biochemical properties of neurons that serve to alter the basic parameters of synaptic neurotransmission and circuit function. Together these data provide novel biological insight into the importance of Akt2 signaling in normative brain development, and in concert with ongoing studies of Akt isoform expression in human postmortem brain, across neocortical development and aging, and in psychiatric diagnoses, will serve to shed further light on the involvement of specific AKT isotypes in neurological syndromes such as autism, schizophrenia, and depression.