Highlights

•

Grik4 overexpression in mice induces anxiety, social deficits, and amygdala output imbalance

•

Grik4 normalization in BLA restored activity in regular but not late-firing cells.

•

Grik4 dose normalization in BLA abolished anxiety, depression, and social deficits

•

Regular firing neurons are key regulators of affective disorder-related behaviors

Summary

Anxiety and depression are highly prevalent psychiatric disorders with poorly understood neural mechanisms. The amygdala, particularly its hyperactivity, is strongly implicated in anxiety. Mice overexpressing the Grik4 gene display anxiety, depression, social deficits, and disrupted amygdala excitability, inducing output circuit imbalance. To dissect the role of specific amygdala neuron populations, we created mice with extra copies of Grik4 and floxed native alleles. We normalized Grik4 dosage selectively in the basolateral amygdala (BLA) pyramidal cells via stereotaxic injection of AAV-CRE-GFP, using AAV-GFP as a control. Electrophysiological recordings from centrolateral amygdala (CeL) revealed that the normalization of Grik4 restored synaptic strength in regular but not late firing neurons. Behaviorally, this intervention reversed anxiety, depression, and social deficits, but not object recognition memory impairments. These results highlight the critical role of regular firing CeL neurons in affective disorders and suggest that targeting their activity may offer new strategies for treating anxiety and depression.

Graphical abstract

Subject areas

1. Behavioral neuroscience
2. Cellular neuroscience

Introduction

Although psychiatric disorders such as anxiety and depression are highly prevalent, the neural circuitry underlying these abnormal behaviors remains incompletely understood. Like autism, anxiety disorders are also characterized by deficits in social behavior1,2,3 and anxiety and depression are often comorbidities of autism.4,5 The absence of targeted pharmacological remedies for these neuropsychiatric conditions,6 among others, underscores the necessity for a deeper comprehension of the neural mechanisms governing social behaviors and their susceptibility to anxiety-related disorders.7,8

Pervasive brain pathologies can emerge because of variations in the dosage of certain genes.9,10,11 Perhaps a well-defined example is the Down syndrome, where the chromosome 21 is fully triplicated. De novo insertions, deletions, inversions, and duplications result in loss or gain of gene function, and when affected genes are active at synapses, these alterations modify normal circuits and their performance.12,13 Enhanced doses of Grik4, a gene encoding a high-affinity subunit of the kainate receptor (KAR) family (GluK4) produces in mice unanticipated effects on the synaptic efficiency of hippocampus and amygdala pathways.14,15 In addition, the behavioral effects of GluK4 overexpression recapitulate abnormal human behavior and may be significant for understanding the etiopathology of human disorders, likely establishing new ways to approach their experimental understanding.

The amygdala consists of functionally and morphologically diverse subnuclei interconnected in a complex manner.16,17,18,19 Most neurons in the basolateral nucleus (BLA) are pyramidal cells20 that project to the centrolateral (CeL) amygdala, where they activate GABA neurons.21 Several studies have indicated the relationship of the BLA to the central amygdala (CeA) pathway with anxiety processes.8,22 However, most of these valuable indications are based on the non-physiological stimulation of CeA afferents, often using the non-specific activation of the BLA areas or optogenetic methods and, in fact, differ from the expected neuronal activity that might be basally altered during pathological situations.

Kainate receptors play relatively minor roles in mediating postsynaptic currents but strongly influence neuronal network dynamics by modulating neurotransmitter release.23,24 The kainate receptor gene Grik4 is widely expressed in the amygdala,15 and its overexpression in BLA pyramidal neurons produces a bidirectional action on CeL amygdala neurons. On one hand, the input to the so-called regularly spiking cells is enhanced, while the excitation of the late spiking cells, the other large population of GABA neurons found in CeL amygdala,15 is reduced. This is essentially achieved by enriching presynaptic terminals with high-affinity KARs, which make them sensitive to environmental glutamate.14,15 Therefore, modest increases in GluK4 protein result in unbalanced circuit outputs from the CeL amygdala. Mice overexpressing GluK4 display signs of depression, anxiety, and social impairment,14,15 and closely reflect the human endophenotypes associated with autism and schizophrenia.2,3

The availability of the transgenic mouse overexpressing GluK4, where basal synaptic transmission is altered in several brain structures, including intra-amygdala circuits, may allow scrutiny of the role played in anxiety disorders by identified neurons. This is a natural situation, given the frequency of pathological cases in which synaptic genes appear down- or up-regulated. In these circumstances, subtle alterations of basal synaptic transmission are to be expected in patients with these symptoms. GRIK4 has been found duplicated de novo in cases of autism,25 and there is cytogenetic and genetic evidence supporting a link of GRIK4 overexpression with schizophrenia and bipolar disorders.26,27,28,29

In the mouse model overexpressing Grik4, we interrogated the role of altered inputs from BLA to CeL amygdala neurons in anxiety and social behavior by normalizing Grik4 dosage exclusively in BLA pyramidal neurons. Our data show that GluK4 normalization in BLA neurons restored the synaptic input onto regular firing cells but not on late firing cells of the CeL amygdala. This synaptic normalization was sufficient to reinstate anxiety and depression as well as social behavior to normal, indicating a critical role for the specific activity of these neurons in affective behaviors.

Results

Contrary to Grik4 knock out mice, which express clear anxiolytic and anti-depressant phenotypes,30 mice overexpressing GluK4 show a profound anxiety and depression-like behaviors and altered synaptic transmission in several parts of the brain14 including a disequilibrium in amygdala outputs.15 In this mouse model, the expression of the Grik4 transgene is under the control of calcium/calmodulin-dependent protein kinase II α (CaMKII) promoter. Several studies have demonstrated that CaMKII is a useful marker for pyramidal neurons. CaMKII is found in all BLA pyramidal neurons but is not found in GABA neurons.31 We wondered what of the observed behavioral anomalies in GluK4 overexpressing mice could be due to the altered inputs to the CeL amygdala provided by principal neurons from BLA and adopted a strategy of gene dose normalization in this specific population of neurons.

Strategy for Grik4 normalization on basolateral amygdala neurons

To investigate the influence of specific neuron populations on behavioral abnormalities induced by the overexpression of GluK4 subunits, we generated a mouse line with extra copies of Grik4, in which the native gene alleles were floxed (Figure 1A) by crossing TgGrik4 with Grik4**fl/fl. We don’t know exactly how many copies of Grik4 were inserted in our TgGrik4, but previous quantification of GluK4 protein suggested that this gene dosage was just duplicated.14 The Grik4**fl/fl line was initially originated in 129/S6 line, and backcrossed with C57BL/6J mice, until getting an almost uniform C57/6J background. It has been found that the 129/S6 strains carry a natural mutation in the Disrupted in Schizophrenia 1 (Disc1) gene, which encodes DISC1, a structural protein at the excitatory postsynaptic density.32 This mutation consists of a 25-bp deletion in exon 6 of the Disc1 allele, causing a truncated form of DISC1 protein. The presence of this mutation leads to behavioral phenotypes of hyperlocomotion, behavioral despair, anxiety, and so forth, resembling symptoms of human schizophrenia and depression.33,34 Although the resulting Grik4fl/fl:TgGrik4 strain had a C57BL/6J background, it could still harbor the DISC deletion as part of chromosome 8 of the 129/S6 donor strain.33 Therefore, we genotyped our animals for the presence of this mutation and determined that ca. 30% of mice exhibited Disc1 mutation either in hetero- or homozygosis (Figure 1A). Since the random presence of Disc1 mutations could affect the interpretation of our results, the colony was reconstituted from individuals that lacked Disc1 deletions, thus eliminating this mutation from all our mice, as confirmed through PCR (Figure S1A).

Figure 1 Mouse model for normalizing Grik4 gene dosage

(A) We obtained Grik4fl/fl:TgGrik4 mouse line, having the native alleles of Grik4 gene floxed as well as extra copies of this gene (right), by crossing TgGrik4, a mouse line with extra copies of Grik414 and Grik4**fl/fl (left; kindly provided by Dr. Anis Contractor). As mouse lines originated from 129sv background could carry mutations in the Disc1 gene (center panel), the line used in further experiments was reconstituted from animals lacking this mutation.

(B–D) Examples of activity tracking records of activities in the open field arena (B), elevated plus maze (C), and forced swimming test (D) of both types of animals, showing that the new line retained anxiety and depression-like behaviors observed in the mice overexpressing GluK4.14 Boxplots represent data collected from 7 (Grik4**fl/fl) and 9 (Grik4fl/fl:TgGrik4) animals of each genotype.

(E) Whole-cell patch clamp recordings of a typical regular firing cell from CeL amygdala (top, left) and mEPSC activity (lower left) in both types of genotypes. The mEPSC frequency was increased in Grik4fl/fl:TgGrik4 mice*,* overexpressing GluK4. Boxplots represent data from 15 cells/8 mice (Grik4**fl/fl) and 17 cells/7 animals (Grik4fl/fl:TgGrik4). In this and the following figures, the horizontal line represents the median, while the mean is indicated by a cross inside the box. Single data points are superimposed. ∗p ≤ 0.05, ∗∗p ≤ 0.005; unpaired t-test. Data are expressed as box and whiskers plots, where data are represented as the median and mean (black horizontal line and cross inside the box, respectively), the first and third quartiles (bottom and top of the box, respectively), the 10th and 90th percentiles (whiskers above and below the box, respectively). Single data points are superimposed on the boxplots.

Then, we wanted to determine whether behavioral abnormalities, previously described in mice overexpressing GluK4 (i.e. TgGrik4), such as anxiety and depression-like behaviors, were retained in the new mouse line. Therefore, animals were subjected to a battery of behavioral tests. The new mouse line exhibited anxiety and depression-like behaviors (Figures 1B–1D), like those previously described in TgGrik4 mice.14 One of the most striking effects of GluK4 overexpression in the intra-amygdala circuits was the synaptic transmission from the BLA to CeL amygdala, which was clearly altered. Such functional alterations from BLA to CeL amygdala inputs were also observed in Grik4fl/fl:TgGrik4, fully reproducing the phenotype found in mice overexpressing GluK4 (i.e., TgGrik4) (Figure 1E).

The overexpressing GluK4 in BLA neurons lead to an unanticipated effect on the BLA synaptic transmission to CeL amygdala neurons, which caused an unbalance in the amygdala output.15 This phenomenon agrees well with the observation that most forms of human anxiety disorders are associated with BLA hyperactivity.35 Since this circuit disequilibrium could be ascribed exclusively to the enrichment of GluK4-containing presynaptic KARs in BLA axon terminals contacting GABAergic CeL amygdala neurons, we decided to normalize GluK4 expression in BLA principal cells. For this reason, we injected Adeno Associated Virus (AAV) engineered to express the recombinase Cre under the control of CaMKII promoter (AAV-CaMKII-Cre-GFP) to drive expression exclusively in BLA pyramidal cells. As a control, we used the same AAV from which the DNA coding for Cre had been removed (AAV-CaMKII-GFP). Localized injections into the BLA resulted in most BLA pyramidal neurons infected (Figures 2A and 2B), sparing neurons in CeL amygdala. The extension of infections was histologically verified after carrying out behavioral analysis, and those animals in which infections were not mostly restricted to the BLA were discarded. The Figure 2 shows examples of typical injections of either AAV-CamKII-GFP (Figures 2B and 2C) in which neurons expressing GFP were mostly restricted to BLA, and that GFP positive axons can be seen in the CeL amygdala neuropil (Figure 2C). We also found that neurons expressing GFP also expressed Cre recombinase by Cre immunostaining (Figure 2D). To further verify that Cre expression in BLA neurons induced the removal of native GluK4 and in the absence of a specific antibody for GluK4 (see Figure S3), we performed two different types of measurements. First, we quantified GluK4 mRNA in cells collected from the amygdala that were cell-sorted according to the expression of GFP (GFP+, GFP− respectively) in the assumption that Cre expression paralleled GFP expression (Figure 2D). GluK4 mRNA was more than double in GFP− cells than in GFP+ cells from Grik4fl/fl:TgGrik4 mice, indicating the effectiveness of Cre to remove endogenous Grik4 alleles (Figure 2E) (0.378 ± 0.110 GFP- cells versus 0.120 ± 0.040 GFP+ of relative values; Mean ± SEM). In the same way genomic DNA analysis from GFP+ cells showed the effective elimination of the Grik4 exon 14 when we performed a PCR with primer flanking outside of the loxP sides fragments (Figures S2A and S2B). To further verify that such a decrease in mRNA content led to the reduction of functional GluK4 protein, we made electrophysiological recordings from BLA neurons in brain slices from Grik4fl/fl:TgGrik4 mice. We applied kainate (3 μM) in the presence of LY303070 to selectively activate KARs present in the cell membrane. The results showed that the kainate-induced responses were reduced in GFP+, Cre-positive cells to approximately the levels observed in animals not overexpressing GluK4 (Grik4**fl/fl 190.4 ± 21.4 pA, n = 16 cells/3 mice; Grik4fl/fl:TgGrik4 245.1 ± 47.6 pA, n = 9 cells/2 mice), which was approximately half of the amplitude observed in TgGrik4**fl/fl neurons (420 ± 43.4 pA, n = 11 cells/3 mice) (Figure 2F).

Figure 2 Normalization of Grik4 gene dosage specifically in the basolateral amygdala of Grik4fl/fl:TgGrik4 mice

(A) A cartoon shows the bilateral infusion of indicated viral vectors in the basolateral amygdala (BLA).

(B) Representative confocal image showing that the expression of GFP was mostly restricted to BLA, and that centro-lateral amygdala (CeLA) lacked labeling.

(C) Enlargement of the insets shown in B. Most BLA principal neurons expressed GFP (green) and their terminals could be seen in CeLA region (center and right panels) (green, GFP; blue, DAPI).

(D) Confocal images of BLA neurons of a mouse injected with AAV-CaMKIIα-Cre-GFP showing that CRE protein is detected by immunocytochemistry (center panel, red fluorescence) in GFP expressing cells (left panel). The right panel is the merge of the two previous panels (green, GFP; blue, DAPI; red, CRE; colocalization appears yellow).

(E) Quantification of Grik4 mRNA in BLA neurons expressing GFP or not after infection with AAV-CaMKIIα-Cre-GFP. GFP-positive and GFP-negative neurons were isolated by FACS and subjected to qPCR. Boxplots represent data obtained from 10 different animals. ∗p < 0.05, unpaired t-test.

(F) Kainate (3 μM)-induced currents in BLA neurons from brain slices prepared from animals with the indicated genotype. Note that in Grik4fl/fl:TgGrik4, injection of AAV-CaMKIIα-Cre-GFP reduced the membrane levels of GluK4 to levels of the WT mice. Data are from 16 cells/3 mice, 9 cells/2 mice, and 11 cells/3 mice, respectively. ∗∗∗p ≤ 0.001, one-way ANOVA. Data are expressed as the mean ± S.E.M.

Altogether, these data support the strategy chosen to reduce the expression of GluK4 to normal values in a selective population of BLA neurons.

Assessment of Grik4 gene dose normalization in basolateral amygdala neurons on behavior

To evaluate if normalizing GluK4 in BLA principal cells restores the synaptic and behavioral anomalies observed in mice that overexpress this subunit, we performed a series of behavioral and electrophysiological analyses of Grik4fl/fl:TgGrik4 mice injected with AAV-CaMKII-Cre-GFP, which should normalize the GluK4 levels in pyramidal neurons of BLA. Injections of AAV-CaMKII-GFP in Grik4fl/fl:TgGrik4 or in Grik4**fl/fl mice served as the control of GluK4 overexpressing mice and WT mice, respectively. Behavioral tests and electrophysiological analysis were conducted after four weeks of injection (Figure S4). The open field (Figure 3) and elevated plus maze (Figure 4) tests were chosen to gather information on locomotor activity and anxiety levels. As we expected, the total distance traveled by Grik4fl/fl:TgGrik4 mice injected with AAV-CaMKII-GFP was much less than that traveled by WT Grik4**fl/fl mice. Although there was a tendency for the overexpressing mice to travel less through the center of the arena and to spend less time in this part, in this series of experiments we did not observe statistical differences between all groups when we measured the total distance traveled in the center or the total time spent in the center area (Figures 3C and 3D). However, when we analyzed the session by sorting it into 5-min bins (Figures 3E and 3F), Grik4fl/fl:TgGrik4 mice injected with AAV-CaMKII-GFP traveled less distance in the center area than Grik4**fl/fl since the very beginning, indicating significant levels of stress. Normalization of GluK4 levels in Grik4fl/fl:TgGrik4 mice by injecting AAV-CaMKII-Cre-GFP recovered normal behavior in terms of total distance traveled, as well as the distance traveled in the center of the arena, indicating recovery of normal behavior in the open field test (Figures 3A, 3E, and 3F).

Figure 3 Normalizing Grik4 in basolateral amygdala neurons restores locomotor activity and anxiety levels

(A) Illustrative schema of the open field and examples of tracking records of all groups of mice throughout the session.

(B) Total distance traveled by the three mouse groups. ∗p = 0.027, ∗∗p = 0.008, one-way ANOVA Holm-Šidák’s multiple comparisons test.

(C and D) Boxplots showing differences in the distance traveled or time spent in the center between all groups. nsp = 0.07 (C), nsp = 0.086 ad 0.170, respectively, and nsp = 0.930 in both cases (D), one-way ANOVA Holm-Šidák’s multiple comparisons test.

(E and F) When total distances traveled were sorted into 5-min bins, differences were found between Grik4**fl/fl and Grik4fl/fl:TgGrik4 mice, which were abolished when Grik4fl/fl:TgGrik4 was injected with AAV driving Cre expression in BLA, indicating a significant level of stress/anxiety. Data collected from the indicated animals. ∗p ≤ 0.05, ∗∗p ≤ 0.01 One-way ANOVA Holm-Šidák’s multiple comparisons test. Data are expressed as box and whiskers plots (A–D), where data are represented as the median and mean (black horizontal line and cross inside the box, respectively), the first and third quartiles (bottom and top of the box, respectively), the 10th and 90th percentiles (whiskers above and below the box, respectively). Single data points are superimposed on the boxplots. In E and F data are expressed as the mean ± S.E.M.

Figure 4 Restoring the Grik4 dosage in the basolateral amygdala normalizes anxiety and depression levels

(A) Examples of tracking records recorded from all groups of mice in the elevated plus maze.

(B and C) Total distance (B) and distance traveled in closed arms (C) of the plus maze were similar in the three groups, indicating normal locomotor activity and exploration of the closed arms. nsp>0.16 (B), nsp>0.76 (C) one-way ANOVA.

(D) Boxplot illustrating the distance traveled in open arms by the three groups of mice.

(E) Time of exploration of anxiogenic open arms. ∗p < 0.05, ∗∗p < 0.01; ∗∗∗p < 0.001 one-way ANOVA Holm-Šidák’s multiple comparisons test.

(F) Measurement of immobility during the forced swimming test of mice with different genotypes. ∗p < 0.05 one-way ANOVA Holm-Šidák’s multiple comparisons test.

(G) Immobility times during the swimming test of the three groups of mice displayed in 1-min intervals across the experimentation time. ∗p < 0.05 one-way ANOVA Holm-Šidák’s multiple comparisons test. Data collected from the indicated animals. Data are expressed as box and whiskers plots (A–F), where data are represented as the median and mean (black horizontal line and cross inside the box, respectively), the first and third quartiles (bottom and top of the box, respectively), the 10th and 90th percentiles (whiskers above and below the box, respectively). Single data points are superimposed on the boxplots. In G data are expressed as the mean ± S.E.M. Data are expressed as the mean ± S.E.M.

Signs of anxiety of Grik4fl/fl·TgGrik4 mice were even clearer in the elevated plus maze test. Interestingly, in this test there was no evidence of less distance traveled by GluK4 overexpressing mice (Figures 4A and 4B), but these mice entered open arms with more difficulty, traveled less, and spent substantially less time in anxiogenic open than in closed arms (Figures 4D and 4E). As it was the case for the open field test, normalizing GluK4 levels by injecting AAV-CaMKII-Cre-GFP in the amygdala of these mice restored anxiety values to normality (Figures 4D and 4E). Something similar occurred when the animals were subjected to the forced swim test (FST) (Figure 4F), a test that measures behavioral despair, which is normally taken as a proxy indicator of depression. The overexpressing mice (Grik4fl/fl:TgGrik4) presented higher immobility as compared to WT animals, an abnormal behavior that was not observed in animals with the doses of GluK4 restored to normality (Figures 4F and 4G).

In addition to displaying behavioral phenotypes such as anxiety, depression, and reduced locomotor activity, mice overexpressing GluK4 presented deficits in social behavior. To delineate the behavior of mice with the GluK4 doses restored in BLA principal neurons, we subjected these animals to the three-chamber test and the Figures 5A–5D shows the summary of the results. We first allowed some time for habituation to the three-chamber cage (Figure 5A), and all groups showed similar mobility, suggesting no difference in the exploratory activity. Then, we introduced a non-familiar mouse inside a wire cup in one chamber and an object in the other chamber. The social preference was measured as the time spent in each room and the time interacting with the mouse versus the time interacting with the object. All groups preferred the chamber containing the mouse and the calculated discrimination index (Figure 5B), indicated that sociability was completely normal in all three groups of mice in the sense that they preferred to interact with a partner rather than with an inanimate object. However, in contrast to Grik4**fl/fl wild type mice, Grik4fl/fl:TgGrik4 failed the social novelty test in that they did not discriminate between novel and familiar animals, similarly interacting with both (Figure 5C). This indicates that mice overexpressing GluK4 in BLA neurons exhibit deficits in social behavior. The Figure 5D represents the graphs quantifying the time spent by each type of mouse in each of the three chambers, where such a deficit in social behavior can be clearly seen. Interestingly, the normalization of GluK4 doses in BLA neurons by Cre injections restored the social behavior of these mice. We then tested these animals for social memory. A never-before-met juvenile mice was placed into the cage together to a test animal, and the duration of the social investigatory behavior within 5 min was scored, and then the juvenile animal was removed. After 1 h, the same animal was returned to the cage for the same period, and the time the animals spent interacting was measured. (Figure 5E). Wild type Grik4**fl/fl mice interacted less time with the now familiar mice, which contrasted with the transgenic Grik4fl/fl:TgGrik4 mice which again showed symptoms or social interaction deficits, since the interaction time was about the same in both trials (Figure 5F). The normalization of Grik4 expression after Cre injection into BLA restored social memory to normality (Figures 5F and 5G).

Figure 5 Restoring the dosage of the Grik4 gene in the basolateral amygdala rescues the social deficits

(A) Schematic representation of the habituation phase (top) and boxplots showing total distance traveled (bottom) for all groups, which demonstrated similar locomotor activity. ns**p>0.47, one-way ANOVA Holm-Šidák’s multiple comparisons test.

(B) Illustrative schema of the sociability phase (top) and boxplots representing the discrimination index for the mice treated or not with AAV expressing Cre. nsp>0.98, one-way ANOVA Holm-Šidák’s multiple comparisons test.

(C) Representative schema of the social novelty phase (top) and a boxplot representing quantifications of novelty preference. Grik4**fl/fl:TgGrik4 mice did not prefer novel over familiar mice revealing a deficit in social interaction, which was restored when animals were injected with AAV expressing Cre. ∗p < 0.05 one-way ANOVA Holm-Šidák’s multiple comparisons test.

(D) Heat maps (top) and quantification of time spent in each chamber of Grik4fl/fl, Grik4fl/fl·TgGrik4 and Grik4fl/fl·TgGrik4 injected with AAV expressing Cre*,* showing how when GluK4 levels are restored, mice preferred to spend more time in the chamber housing the novel mouse, behaving as wild type (Grik4**fl/fl). ∗p < 0.05, nsp = 0.252, ANOVA Holm-Šidák’s multiple comparisons test. Data collected from the indicated animals.

(E) Scheme for social memory test. A never-met juvenile female was introduced for 5 min in the subject animal’s arena, and the interaction time was scored (trial 1). After 1 h, the test was repeated (trial 2).

(F) Interaction time during trail 1 and trial 2 of the three groups of mice. Dots are the mean ± SEM and the single score values are linked by a gray line for each mouse. ∗p = 0.042, ∗∗p = 0.008, nsp = 0.091, paired t-test.

(G) boxplots representing the discrimination index for the mice treated or not with AAV expressing Cre. ∗p < 0.05, ANOVA Holm-Šidák’s multiple comparisons test. Data collected from the indicated number of animals. Data are expressed as box and whiskers plots (A–D and G), where data are represented as the median and mean (black horizontal line and cross inside the box, respectively), the first and third quartiles (bottom and top of the box, respectively), the 10th and 90th percentiles (whiskers above and below the box, respectively). Single data points are superimposed on the boxplots. In F data are expressed as the mean ± S.E.M.

Mice overexpressing GluK4 also showed some deficits in object recognition memory. We used the novel object recognition (NOR) task to evaluate this particular type of novelty preference in GluK4 overexpressing mice and determine whether this behavior was also recovered by normalizing GluK4 levels in amygdala neurons. In the training session (Figure 6A), the different groups of mice showed no preference for any of the two identical objects (Figure 6B). One hour after the training session, we replaced one of the objects with a novel one, and in the test session, the animals were allowed to explore the objects. Normal mice (Grik4**fl/fl) explored more the novel object reflecting normal recognition memory, while Grik4fl/fl:TgGrik4 showed no preference for the novel object. In contrast to other types of behaviors, the normalization of Grik4 gene dosage in the BLA was insufficient to rescue the altered novelty recognition memory (Figure 6C), indicating that this behavior does not depend on the amygdala but probably on the hippocampus, perirhinal cortex and other brain areas, where GluK4 doses were not restored to normal in these experiments.

Figure 6 Normalizing the Grik4 gene dosage in the basolateral amygdala did not rescue object recognition memory deficits

(A) Illustrative schema of the training and test sessions.

(B) Discrimination index values during the training session, showing similar preference for both objects of groups of animals (nsp>0.90, ANOVA Holm-Šidák’s multiple comparisons test.).

(C) Values of discrimination index during the novel object recognition test (left) and heat maps reporting example of trajectories of each mouse type (right). *Grik4fl/fl·*TgGrik4 mice presented a deficit of object recognition that was not abolished by injecting AAV expressing Cre (Grik4**fl/fl 25.25 ± 4.23%; Grik4**fl/fl:TgGrik4 0.70 ± 5.63%, Grik4**fl/fl:TgGrik4 with Cre injected 6.26 ± 4.02%). ∗∗p = 0.0013; ∗p = 0.017; nsp = 0.420). one-way ANOVA Holm-Šidák’s multiple comparisons test. Data collected from the indicated animals. Data are expressed as box and whiskers plots, where data are represented as the median and mean (black horizontal line and cross inside the box, respectively), the first and third quartiles (bottom and top of the box, respectively), the 10th and 90th percentiles (whiskers above and below the box, respectively). Single data points are superimposed on the boxplots.

Assessing the role of basolateral amygdala in intrinsic anxiety

In our cohort of C57BL/6J transgenic mice, we observed substantial inter-individual variability in anxiety-like behavior when assessed in the open-field arena. Similarly, the elevated plus maze test revealed a high degree of variability (Figure 7), indicating that some mice naturally exhibit greater anxiety levels than others. The origin of this spontaneous anxiety remains unclear but likely reflects inherent individual differences and emphasizes the importance of considering individual variability and the relationship between multiple behavioral measures.36 We then hypothesized that examining the phenotypic extremes of high- and low-anxiety behavior could offer valuable insights into natural anxiety states.37 This approach may provide a more accurate characterization of anxiety compared to states induced by specific gene mutations.38,39,40 Indeed, previous studies in mice and rats have used the performance in the elevated plus maze to establish animal models with high- and low-anxiety behaviors.41,42,43 Building on this, we investigated whether modulating excitability along the BLA to CeL amygdala axis by manipulating GluK4 levels could influence these spontaneous or idiosyncratic anxiety traits. We first analyzed mice whose anxiety levels had been measured using two different methods: the open-field test and the elevated plus maze. Cluster analysis36 of open-field data revealed two distinct groups based on distance covered in the center of the arena (Figure S5B). These groups also showed distinct behavior in the elevated plus maze, supporting the presence of animals with naturally higher anxiety levels, hereafter referred to as highly anxious mice, and supporting that open-field test parameters reliably predict stable anxiety levels when assessing them from elevated plus maze performance (Figure S5E). Next, we selected a cohort of 21 Grik4-floxed (Grik4**fl/fl) mice. As shown in Figure 7, cluster analysis of open-field behavior again identified two distinct groups with different levels of anxiety. These two groups were injected in the BLA with AAV-CaMKII-GFP (control) or AAV-CaMKII-Cre-GFP. Low anxiety animals did not show any alterations after knocking down Grik4 compared to controls (Figure 7D), while anxiety levels in highly anxious mice were partially relieved after reducing GluK4 doses in BLA neurons. (Figure 7E). These data further support the idea that BLA input strength to CeL amygdala is highly dependent on GluK4 levels in BLA neurons15 (see later in discussion) and plays an important role in controlling intrinsic anxiety.

Figure 7 Assessing the effect of GluK4 knock-down in BLA neurons on intrinsic anxiety

(A) A cohort of 21 Grik4fl/fl mice underwent cluster analysis based on the distance covered in the center of an open-field arena.

(B) This analysis identified two distinct populations with radically different behaviors, classified as non-anxious and highly anxious mice.

(C) These two groups were injected in the BLA with an AAV expressing either GFP (Control) or Cre recombinase and subsequently subjected to the elevated plus maze test.

(D) Knockdown of GluK4 had no effect on anxiety levels in non-anxious mice.

(E) In highly anxious mice, GluK4 knockdown mildly attenuated anxiety. For statistical analysis, non-anxious mice (injected with either GFP or Cre) were pooled and compared to highly anxious mice (injected with either GFP or Cre). Examples of mice performance in the elevated plus maze are shown on the right. ∗p < 0.05, ns p = 0.5, one-way ANOVA Holm-Šidák’s multiple comparisons test. Data are expressed as box and whiskers plots (A, D, and E), where data are represented as the median and mean (black horizontal line and cross inside the box, respectively), the first and third quartiles (bottom and top of the box, respectively), the 10th and 90th percentiles (whiskers above and below the box, respectively). Single data points are superimposed on the boxplots. In C data are expressed as the mean ± S.E.M.

Effects of Grik4 dose normalization on functional connectivity of basolateral amygdala and centrolateral amygdala neurons

In the TgGrik4 mouse model, overexpressing GluK4, the synaptic transmission from BLA to CeL amygdala neurons is specifically altered. This alteration consisted of enhanced synaptic input onto discrete CeL amygdala regular firing neurons, and depressed input onto the so-called late firing cells. We reasoned that this imbalance could explain some of the abnormal behaviors observed in TgGrik4 mice. To study if altered inputs to CeL amygdala neurons were affected by normalizing the GluK4 doses expressed by BLA neurons, we recorded AMPAR-mediated miniature EPSC (mEPSC), under the configuration of whole-cell, from these two main populations of CeL amygdala neurons in brain slices, since the frequency of mEPSC reflects activity occurring at the presynaptic compartment, i.e., BLA neuron synaptic terminals. The two types of CeL amygdala neurons could be easily identified by depolarizing pulses before applying TTX to measure mEPSCs. Regular firing cells from Grik4fl/fl:TgGrik4 mice presented an increased frequency of mEPSC (Figures 8A and 8B) as compared to Grik4**fl/fl, while the amplitude of mEPSCs was slightly increased. In contrast, the synaptic input to the late firing cells was largely depressed (Figures 8D–8F) in these animals in that both mEPSC frequency and amplitude were decreased. Interestingly, the synaptic attenuation observed in these neurons persisted after normalizing GluK4 expression in BLA neurons. In contrast, the synaptic input to regular firing cells was totally restored upon GluK4 normalization in BLA neurons (Figure 8A).

Figure 8 Grik4 dose normalization in BLA rescued synaptic activity in regular firing but not in late firing cells from the CeL amygdala

(A) Schematic representation of the recording area and an example of a CeL amygdala (CeLA) regular firing cell. On the right, AMPAR-mediated mEPSC were recorded at −65 mV of membrane potential from this type of cell from each group of animals.

(B) Cumulative distribution (left) and boxplot representation (right) of mEPSC frequency in regular firing cells from each mouse type (18 cells/8 mice from Grik4**fl/fl; 17 cells/5 mice from Grik4fl/fl:TgGrik4 and 17 cells/5 mice from Grik4fl/fl·TgGrik4 injected with AAV expressing Cre). ∗p = 0.043, ∗∗p = 0.0046, one-way ANOVA Holm-Šidák’s multiple comparisons test.

(C) Amplitude values of AMPA receptor-mediated mEPSC recorded from regular firing cells. nsp = 0.350, ∗∗p = 0.009, one-way ANOVA Holm-Šidák’s multiple comparisons test.

(D) As in (A), but for the case of late firing cells.

(E) Cumulative distribution (left) and boxplot representation (right) of mEPSC frequency in late-firing cells of Grik4**fl/fl, Grik4fl/fl:TgGrik4 and Grik4**fl/fl:TgGrik4 mice injected with AAV expressing Cre. (E) Cumulative distribution (left) and boxplot representation (right) of mEPSC frequency in late firing cells of these mice (data collected from 23 cells/6 mice; 18 cells/4 mice and 19 cells/5 mice for each type, respectively).

(F) Cumulative distribution (left) and boxplot representation (right) of amplitudes of mEPSC recorded from late firing cells. ∗∗∗p < 0.0003; ∗p < 0.05; nsp = 0.224 one-way ANOVA Holm-Šidák’s multiple comparisons test. Colored bands along cumulative distribution curves represent the Standard Deviation point to point. Data are expressed as box and whiskers plots, where data are represented as the median and mean (black horizontal line and cross inside the box, respectively), the first and third quartiles (bottom and top of the box, respectively), the 10th and 90th percentiles (whiskers above and below the box, respectively). Single data points are superimposed on the boxplots.

Miniature EPSC frequency and amplitude are reliable parameters for measuring neuronal excitation. However, since they are generated by all synaptic inputs to a neuron regardless of origin, we aimed to directly investigate the specific effects of selectively manipulating GluK4 in BLA neurons. To achieve this, we conducted a series of experiments in which we infected BLA neurons of Grik4fl/fl:TgGrik4 mice with AAV expressing Channelrhodopsin-2 (AAV-CaMKII-hChR2_eYFP) or AAV expressing Cre recombinase (AAV-CaMKII-Cre-GFP) plus AAV with floxed Channelrhodopsin-2 under hEF1 promotor (AAV-hEF1α-lox-HChR2_mcherry-lox), to normalize the GluK4 levels and make GluK4-normalized neurons light-sensitive. Our results revealed a 2.3-fold increment in the input to the regular firing cells in the CeL amygdala from BLA neurons overexpressing GluK4 compared to neurons with normal GluK4 levels (107.9 ± 39.6 pA in Grik4**fl/fl vs. 357.5 ± 79.7 pA in Grik4fl/fl:TgGrik4) (Figures 9A and 9B). Notably, the normalization of GluK4 expression completely reversed this enhancement (Figure 9B), confirming previous findings where excitatory input to the regular firing cells was estimated using mEPSC analysis. Indeed, the increase in synaptic strength appeared even more pronounced when analyzed with this targeted approach (331% vs. 182.6% from mEPSC analysis). Interestingly, BLA-specific input to the late firing cells did not show any alterations (Figure 9C), in contrast to the results obtained when mEPSCs were measured in these cells. This likely reflects the multiple origins of mEPSCs, of which BLA neurons constitute only a small fraction.

Figure 9 Optogenetic activation of BLA neurons in GluK4-overexpressing mice

Dose normalization rescued synaptic strength in regular-firing but not in late-firing cells of the CeL amygdala.

(A) Schematic representation of the light stimulation and recording arrangement, along with representative light-evoked responses of CeL regular-firing cells stimulated at different intensities at −65 mV membrane potential in each group of animals.

(B) Average response amplitudes for each group in regular firing cells. Notably, GluK4 normalization in BLA neurons fully restored synaptic strength.

(C) Average response amplitudes for each group in late firing cells. ∗p < 0.05. one-way ANOVA Holm-Šidák’s multiple comparisons test. Data are expressed as the mean ± S.E.M.

Taken together, these results indicated that GluK4 normalization in BLA neurons exclusively reinstates synaptic input to CeL amygdala regular firing cells from BLA, pointing to this neuron class as a critical player in the behavioral abnormalities observed in GluK4-overexpressing transgenic mice.

Discussion

This study shows that in a mouse model of anxiety and depression, induced by overexpressing the GluK4 high affinity subunit of the KAR in forebrain principal cells,14 the normalization of synaptic activity in BLA principal cells is enough to remove signs of anxiety, depression, and the associated alterations of social behaviors. These altered behaviors are then the result of an imbalance in the main output of the amygdala, likely due to the bidirectional modulation of the connection between BLA and the two subpopulations of GABA neurons that mostly form the CeL amygdala. In the transgenic mouse overexpressing GluK4, the input to regular firing cells was enhanced, while input to late firing cells was depressed, although at a lesser extent. The two populations of CeL amygdala neurons were functionally identified by their response to depolarizing pulses. Our data indicate that the attenuation of anxiety is at least achieved by the input normalization of BLA pyramidal cells onto regular firing cells, a type of neuron present at the CeL amygdala, which are at the crossroad of amygdala output. On one hand, regular firing cells directly project to the periventricular nucleus of the thalamus and periaqueductal gray matter and, on the other, they strongly inhibit the population of late firing cells that exerts a strong control of CeM amygdala projecting neurons, which form part of the anxiolytic pathway.19 From a mechanistical point of view, our results lead to hypothesize that lowering regular firing cells activity conveys a disinhibition of late firing cells resulting in a relief of the imposed anxiety. In support of this, it is known that hyperactivity or increased excitatory transmission through the stimulation of amygdala leads to the perception of anxiety in humans and rodents.44 Hyperexcitability of amygdala, especially BLA, is a common finding in patients with generalized anxiety, social anxiety, and panic disorders.45 In our model, these phenotypes were recapitulated by the enhanced basal synaptic input from BLA to regular firing cells of the CeL amygdala. Previous work has shown that when the BLA to CeA pathway is selectively activated using optogenetics, mice show lower anxiety.22 However, what types of cells are preferentially activated within CeA was undetermined. Our data further support the hypothesis that the BLA to CeL to CeM amygdala circuit could be causally involved in anxiety, as suggested by Tye and collaborators22 and delineate the critical role played by regular firing cells in this circuit. Alternatively, a recent work identified that some CeL amygdala neurons biochemically classified as somatostatin expressing neurons, of which a significant proportion are regular firing cells, inhibit GABA neurons of a region comprising central sublenticular extended amygdala to induce anxiety.46 Normalization of regular firing cells activity is anticipated to exert an anxiolytic effect through this pathway.

We cannot discard that the normalization of other BLA outputs may also contribute to recovering normal phenotypes, since BLA connects directly to several brain structures (the striatum, thalamus and cerebral cortex).47 However, in a broad sense, the CeA receives most of the projections from the BLA and supplies most of the output from the amygdala. Indeed, most of the CeA output is channeled from the CeM amygdala to brainstem effector structures such as the BNST, PAG, periventricular thalamus (PVT), and so forth, and these projections, depending on the degree of activity of CeM amygdala cells, can trigger fear, anxiety, and reward behaviors.17,48,49

We have classified CeL amygdala neurons into two subgroups from a functional point of view. Although it has been described at least three different types of neurons in the CeL amygdala,50 these two groups seem to be the vast majority, accounting for 90% of all neurons.18,51 With some reservations, regular and late firing cells have been equated to CeLON and CeLOFF cells recorded during fear conditioning17,52 and, from a biochemical point of view to cells expressing (LFC) or not (RFC) somatostatin.18 A classification based on the expression of PKCδ is less accurate as half of PKCδ+ or PKCδ− cells are LFC and about half of PKCδ+ neurons are RFCs.18,48 This difficulty in classifying these cells has posed challenges in the study of RFCs compared to other amygdala neuron types, contributing to the poor understanding of their role in anxiety and other behaviors. What seems to be clear from anatomical studies is that CeLON cells projections onto CeLOFF cells are 3 times more frequently than vice-versa,48 suggesting that, together with the faster responding capabilities, RFCs may exert profound inhibitory control over LFCs, which are the main source of innervation of CeM amygdala projecting neurons.

Surprisingly, the normalization of Grik4 expression in BLA did not restore the synaptic alteration observed on LFC, i.e., depressed synaptic inputs as measured by mEPSC frequency. In contrast, when BLA neurons were specifically activated using optogenetics, GluK4 overexpression did not produce clear alterations in the BLA input to LFC. As expected, the normalization of GluK4 levels in BLA neurons also had no further effect on the functional connectivity between these BLA neurons and LFC. Although this result was beneficial for the purpose of this study, the discrepancy remains difficult to explain. The most parsimonious interpretation is that many of the connectivity to LFC originate from neuronal populations distinct from BLA neurons. Therefore, the change in excitatory input to LFC, which did not reverse upon GluK4 normalization specifically in BLA neurons, may reverse when normalization is global.15 Unfortunately, the microcircuitry of the BLA to CeL amygdala is not entirely delineated, but BLA neurons innervating LFC appear to be ventrally displaced in the complex53 and it is possible that our injection did not routinely reach them. Whatever the reason, this result made it possible to underscore a role of RFC activity in anxiety. Similarly, another phenotype, such as object recognition memory, was not recovered either after Grik4 normalization. This result can be considered as a positive control for the specificity of our AAV injections to the BLA, since rather than amygdala, brain structures implicated in object recognition memory include the hippocampus and neighboring cortical regions, the medial prefrontal cortex, the subiculum, and the thalamus,54 where Grik4 dose remained increased.

The amygdala is considered a composite of parallel circuits that affect multiple aspects of emotional behavior (see19;44 for reviews). Despite extensive neuropharmacological research to find drugs for the treatment of pathol