Information processing in the hippocampus depends on a fined tune interaction of excitatory and inhibitory systems (Ego-Stengel and Wilson, 2007). Inhibitory neurotransmission is defined by the action of γ-aminobutyric acid (GABA) that is synthesized via two isoenzymes, glutamic acid decarboxylase (GAD)65 and GAD67, in so-called interneurons. While pharmacological studies using classical modulators of GABA receptors indicate a rather memory-impairing role of GABAergic neurotransmission in hippocampus-dependent tasks (D'Hooge and De Deyn, 2001; Myhrer, 2003), more recently developed drugs, e.g., the GABA analog Gabapentin (Celikyurt et al., 2011), show the reverse effect. The involvement of GABAergic neurotransmission in hippocampus-dependent tasks therefore appears complex and requires a careful evaluation to exclude potentially confounding side effects of drugs. Moreover, GABA is expressed in diverse populations of interneurons with different functions across hippocampal subregions (Freund and Buzsaki, 1996). These subpopulations are characterized by specific expression of calcium-binding proteins and neuropeptides like neuropeptide Y (NPY), somatostatin (SST) and also cholecystokinin (CCK). These neuropeptides themselves modulate inhibition and excitation in local circuits (Sperk et al., 2007; Tallent, 2007; Lee and Soltesz, 2011) and affect thereby also hippocampal tasks as spatial learning in a Morris water maze (MWM) (Thorsell et al., 2000; Dyer and Cain, 2007; Lo et al., 2008).

We previously demonstrated that spatial learning in a MWM under different levels of stress alters synaptic plasticity in the hippocampus in a region specific-manner (Kavushansky et al., 2006). Such dissociated responses within the hippocampal subregions cornu ammonis field 1 (CA1) and the dentate gyrus (DG) have been observed particularly after activation of the basolateral amygdala (BLA) complex (Vouimba et al., 2004; Vouimba and Richter-Levin, 2005). Amygdala inputs appear to regulate the expression of different interneuron subpopulation markers in distinct subregions of the hippocampus (Berretta et al., 2001, 2004) and exposure to stress alters GABAergic functioning at the molecular and physiological levels in the hippocampus (Jacobson-Pick et al., 2008; Yarom et al., 2008; Jacobson-Pick and Richter-Levin, 2010). GABAergic signaling within the BLA is also modulated by stress (e.g., Bergado-Acosta et al., 2008; Heldt et al., 2012) and such changes have the potential to modulate synaptic plasticity within the hippocampus as well as hippocampus-dependent spatial memory (Kim et al., 2005). These findings suggest an involvement of the GABAergic system in both the amygdala and hippocampus during spatial learning and a modulation of their interaction under conditions of elevated stress.

In this study we began to address this question by investigating gene expression of GABAergic factors, including the key enzymes for GABA synthesis, GAD65 and GAD67, and GABA-associated neuropeptides in subregions of the hippocampus and in the BLA after a spatial learning task in the MWM. MessengerRNA (mRNA) levels were compared to a group of animals that experienced comparable physical stress exposure to the water without an invisible platform to locate. The increased stress response due to the water maze experience was confirmed in both groups by assessing corticosterone (CORT) plasma levels. Through measurement of ERK 1/2 phosphorylation, we demonstrated a differential activation of the amygdala in rats that have acquired the spatial learning task compared to animals that have been water exposed only. The differential regulation of mRNA expression levels of selected target genes under these conditions provide evidence for a differential, stress- and memory related regulation of GABA interneuron function in different hippocampal subregions and within the amygdalo-hippocampal system.

The MWM task is commonly used for the evaluation of learning and memory in rodents, however the complex involvement of the GABAergic system in these processes has not yet been completely solved (D'Hooge and De Deyn, 2001). Here we demonstrate changes in molecular markers of GABAergic function in subregions of the hippocampus and the BLA after spatial learning in this task. In addition, we show a differential expression of the selected target genes in the BLA after spatial learning compared to exposure to the water maze itself. Moreover, a selective increase in the phosphorylation within the BLA of ERK 1/2 was observed in animals that have been exposed to the water maze without the possibility to learn the location of the escape platform, whereas CORT plasma levels were increased in both compared to handled control animals. These data suggest a differential involvement of hippocampal local GABAergic circuit neurons in spatial learning.

Spatial learning of the platform location induced a distinct pattern of changes in expression of the GABA-synthetizing enzymes GAD65 and GAD67 as well as in the expression of selected neuropeptides NPY, SST, and CCK that modulate GABAergic neurotransmission and mark different subsets of GABAergic interneurons. Intriguingly, the pattern of changes was subregion-specific, with increased levels of NPY in the DG, increased levels of SST in the CA3 and decreased levels of CCK in the CA1 region of the dorsal hippocampus. These changes may highlight the activation of distinct subsets of interneurons that are preferential located in a specific subregions of the hippocampus to modulate local network activity (Maccaferri and McBain, 1995; Houser, 2007) thereby shaping the information flow throughout the hippocampal formation.

The DG as the main input station to the hippocampal formation is strongly modulated by its local inhibitory network (Acsady and Kali, 2007); GABAergic interneurons thus contribute to the DG function in pattern separation, with implications for spatial memory formation (Richter-Levin et al., 1995; Acsady and Kali, 2007; Kesner, 2007). An important interneuron subpopulation that exerts feedback inhibition on DG inputs are the so-called HIPP cells (= hilus perforanth path associated cells),that have their cell bodies located in the hilus and send axons to the distal dendrites of DG granule cells (Houser, 2007; Sperk et al., 2007). The increased NPY mRNA expression after spatial learning in the MWM, without additional significant changes for GAD65 and GAD67 or the other neuropeptides, therefore most likely reflect a training-specific activation of these NPY-positive HIPP cells subpopulation. The contribution of NPY appears to be region-specific, since a general increase in NPY by constitutive overexpression, in mice, impairs spatial memory in the MWM (Thorsell et al., 2000). Indeed, in our experiments we observed no modulation of NPY expression in the SO of dorsal hippocampus areas CA1 and CA3 after spatial learning.

However, in the CA3 region we found a learning-induced increase in SST expression. SST in interplay with serotonergic signaling in the hippocampus is required for spatial memory formation in the MWM (Dyer and Cain, 2007) and depletion of SST impairs spatial learning (Matsuoka et al., 1995). The CA3 region of the dorsal hippocampus is one of the key regions for acquisition of spatial memory in the water maze (Florian and Roullet, 2004; Teather et al., 2005) and SST can facilitate LTP of mossy fiber inputs into this region, but not into CA1 (Matsuoka et al., 1991). In our study the SO of the CA3/CA1 was isolated via LCM and in this region SST is confined to the oriens/lacunosum-moleculare (OLM) cells. Comparable to the HIPP cells in the DG, these cells exert feedback inhibition on principal cells and shape the information flow in the hippocampal formation (Katona et al., 1999). OLM cells further elicit and control rhythmical activity in the hippocampus (Somogyi and Klausberger, 2005) that is essential for spatial and temporal integration during navigation and memory formation (Buzsaki and Moser, 2013).

In contrast to these selective increases of neuropeptide expression in hilus and CA3, changes in the CA1 region were characterized by reduced expression of the GABA synthesizing enzymes GAD65 and GAD67. This indicates a more general activation of local interneurons in CA1 and may potentially support spatial memory formation by reducing inhibition in CA1 pyramidal cells, which are critical for this form of learning (Teather et al., 2005; Goodrich-Hunsaker et al., 2008). In fact, drugs that enhance GABAergic signaling by allosteric modulation of GABA_A receptors impair spatial learning in the MWM (White et al., 1997). Finally, the reduction in CCK expression observed under spatial learning conditions in the CA1 also supports spatial memory formation. Within the CA1 region, CCK is mostly expressed in basket cells, with cell bodies located in the stratum pyramidale. These cells are involved in stress coping and emotional memory formation via inputs from the serotonergic and cannabinoid system (Touma, 2011; Haring et al., 2012; Keimpema et al., 2012) and control rhythmic activity of principal cells. CCK-positive interneurons are also observed in the SO (Schiffmann and Vanderhaeghen, 1991; Tsou et al., 1999), but the function of this subpopulation is not well understood.

Together, these data demonstrate that spatial learning in the water maze induces a distinct subregion-specific expression pattern of GAD65 and GAD67 as well as of the selected neuropeptides. Exposure to the water maze alone induced no alterations in hippocampal mRNA expression levels for any of the factors examined.

However, in the BLA we observed a differential expression between groups being only exposed to the water maze or gaining spatial memory during the task. Here, GAD65 and GAD67 as well as the neuropeptides NPY, SST, and CCK show a rather reduced expression after spatial learning compared to the water exposed animals. For GAD65 and GAD67 reduced expression in the BLA was observed in the spatial learning group compared to handled controls, indicating effects of spatial learning on the amygdala expression of these factors. A learning-specific reduction in GAD65 mRNA in the BLA has also been reported after fear conditioning (Bergado-Acosta et al., 2008). Such a reduction appears to be linked to transiently reduced inhibition in the BLA (Szinyei et al., 2007) and a reduction in extracellular GABA (Stork et al., 2002) that may contribute to memory consolidation. Indeed, GABAergic signaling in the BLA can modulate spatial memory formation and LTP in the hippocampus (Kim et al., 2005) and both regions, BLA and hippocampus, are activated during retrieval of spatial memory (Vanelzakker et al., 2011).

The differential expression of neuropeptides in the BLA may reflect, as in the hippocampus, the activation of specific interneuron subpopulations that could contribute to altered LTP and inhibitory control observed after stress (Vouimba et al., 2004; Rodriguez Manzanares et al., 2005). On the other hand, NPY and SST themselves are powerful modulators of anxiety-related behavior and have anxiolytic properties when administered to the amygdala (Heilig, 2004; Yeung and Treit, 2012). Increased levels of NPY in the amygdala have been linked to coping behavior after repeated exposure to stressors (Thorsell et al., 1999) and in an animal model of posttraumatic stress disorder, increased levels of NPY in amygdala and hippocampus have been associated with resilience to traumatic stress (Cohen et al., 2012). SST microinfusions in the amygdala produce anxiolytic- and antidepressant-like effects (Yeung and Treit, 2012), likely through modulation of GABA_A receptor signaling (Engin et al., 2008), whereas administration of CCK elicits anxiety-like behavioral responses (Mathew et al., 2008; Sherrin et al., 2009). Thus the transcriptional regulation of these factors in the BLA likely is part of an adaptive response to a stressful experience.

Analysis of CORT plasma levels revealed increases in both, the Spatial Learning and the Water Exposure groups, compared to Control. Elevated CORT reflects the activation of the hypothalamus-pituitary-adrenocortical axis (Armario et al., 2012) and is an important marker for the hormonal stress response. Stress itself thereby consists of different components and includes in part stress by physical activity during swimming (Zheng et al., 2006), but also psychological components due to exposure to a new, challenging environment (the water at a certain temperature; Akirav et al., 2001).

Stress further increases the expression of plasticity markers in the BLA (Zoladz et al., 2012) and activates the signaling and transcription factor ERK 1/2 in this region through phosphorylation (Maldonado et al., 2013). Strikingly, only animals that had been exposed to the water maze without an invisible platform displayed an increase in p-ERK 1/2 in the BLA. However, the increase in p-ERK 1/2 levels in the water exposed animals can neither be attributed to spatial learning per se nor to stress per se, since similar CORT levels were observed in both experimental groups. Consequently, the observed increase in ERK 1/2 phosphorylation may be related to another main difference between the experimental groups: the level of gained controllability over a stressful situation.

While animals that were allowed to learn the location of an invisible platform can be said to have gained controllability over the stressful exposure to the water, animals that have been exposed to the water for the same duration (exposure time was matched to average of learning group per trial) but without an escape platform were left in an uncontrollable situation. Controllability thereby describes the perceived ability to alter the onset, duration, intensity or pattern of an aversive experience (Maier and Seligman, 1976) and contributes strongly as a psychological factor to the emotional impact of a stressful event. The lack of controllability can interfere with the performance in cognitive tasks (Overmier and Seligman, 1967; Seligman and Maier, 1967). In turn, by acquiring escape strategies subjects may effectively reduce the stress levels associated with a given task. We have previously shown that gaining controllability over a stressful situation, i.e., by avoidance learning in a two-way shuttle box, attenuates phosphorylation of ERK 1/2 in the amygdala (Ilin and Richter-Levin, 2009).

Therefore, it is interesting to compare the observed effects between the experimental groups in the light of controllability as a factor. Although their physical stress experience, i.e., the exposure to the water and the swimming itself, was matched, differential expression profiles and transcriptional activation of the BLA were observed between the Water Exposed and the Spatial Learning groups. Animals in both groups may share the motivation to end the challenging and stressful situation of the water exposure, but only the Spatial Learning group can actively reach this goal by learning the location of the hidden platform. Thus, spatial learning in the MWM may be described as leading to gaining controllability over a stressful situation. In line with that, the differential changes in the BLA observed in this study might be, at least in part, related to the difference in the level of gained controllability. Since the BLA critically affects information processing in the hippocampus and hippocampus-dependent memory (Kim et al., 2005) in a subregion-specific manner (Vouimba and Richter-Levin, 2005), the observed changes could contribute also to the differential effects of water exposure on LTP in the DG versus the CA1 reported previously (Kavushansky et al., 2006).

In summary, in this study we could demonstrate a highly specific expression regulation of the interneuron markers NPY, SST, and CCK in distinct subregions of the dorsal hippocampus by spatial learning in the MWM. In the BLA, while spatial learning itself led to reduced expression of GAD65/67, a differential expression of all selected markers was observed in water exposed animals compared to animals that learned the location of a hidden escape platform. Together with the increased ERK 1/2 phosphorylation after water exposure only, the observed changes in the BLA appear to be related to the stress experience and may related also to aspects of lack of controllability. Future studies will address how these molecular changes affect information processing in the amygdalo-hippocampal system and the adaptation to learning under stress.

Osnat Hadad-Ophir, Anne Albrecht, Oliver Stork and Gal Richter-Levin conceived and designed the experiments. Osnat Hadad-Ophir performed the experiments. Osnat Hadad-Ophir, Anne Albrecht, Oliver Stork and Gal Richter-Levin analyzed and discussed the data. Osnat Hadad-Ophir, Anne Albrecht, Oliver Stork and Gal Richter-Levin wrote the manuscript.