Despair-associated memory requires a slow-onset CA1 long-term potentiation with unique underlying mechanisms

The emotion of despair that occurs with uncontrollable stressful event is probably retained by memory, termed despair-associated memory, although little is known about the underlying mechanisms. Here, we report that forced swimming (FS) with no hope to escape, but not hopefully escapable swimming (ES), enhances hippocampal α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR)-dependent GluA1 Ser831 phosphorylation (S831-P), induces a slow-onset CA1 long-term potentiation (LTP) in freely moving rats and leads to increased test immobility 24-h later. Before FS application of the antagonists to block S831-P or N-methyl-D-aspartic acid receptor (NMDAR) or glucocorticoid receptor (GR) disrupts LTP and reduces test immobility, to levels similar to those of the ES group. Because these mechanisms are specifically linked with the hopeless of escape from FS, we suggest that despair-associated memory occurs with an endogenous CA1 LTP that is intriguingly mediated by a unique combination of rapid S831-P with NMDAR and GR activation to shape subsequent behavioral despair.

immobility was suggested to score a learned behavioral despair 33 , although this viewpoint has been debated [34][35][36] . Notably, we established a hopefully escapable swimming (ES) paradigm by placing a floating platform into the water (Fig. 1A), such that the rats could attempt to climb onto that unstable platform even though they were difficult to stay on it. This way, ES rats were exposed to stress similar to FS group, but with hopefully potential escape.

Despair-associated memory depended on the NMDAR-mediated endogenous LTP in hippocampal CA1.
We directly addressed the possibility that FS could induce an endogenous CA1 LTP to underlie the FS-induced despair-associated memory 26,27 . We recorded the field excitatory postsynaptic potentials (fEPSP) from the Schaffer/commissural-CA1 pathway in freely moving rats 26 before and after FS/ES, and measured test immobility in these animals. FS induced a slow-onset synaptic potentiation in CA1, with an initial and transient synaptic depression and fully reversed to LTP within about 40-60 min and lasted for about 4 h (thus an endogenous slow-onset CA1 LTP) ( , indicating also that other potential influences of FS/ES such as body temperature changes may be unrelated to the synaptic potentiation. Furthermore, this slow-onset synaptic potentiation was found to be a form of NMDAR-dependent CA1 LTP, because the NMDAR antagonists KET (i.p. 15 mg/ kg) or AP5 (i.c.v., 10 mM) blocked it without affecting the transient synaptic depression (Fig. 4C, n = 5, AP5 = 97.3 ± 2.2%; n = 5, KET = 87.9 ± 8.5%; AP5 or KET vs. baseline, p = 0.315 or 0.102). Thus, controllable stress had no significant effect on synaptic efficacy and test immobility but uncontrollable stress induced CA1 LTP and increased test immobility (Fig. 4H), both of which were blocked by NMDAR Scientific RepoRts | 5:15000 | DOi: 10.1038/srep15000 antagonists, landing further support to the new concept that the CA1 LTP may underlie despair-associated memory that leads to despair-like behavior.
Blockade of the endogenous CA1 LTP blocked despair-associated memory. To further examine whether S831-P is a prerequisite to induce the CA1 LTP and increase of test immobility in the same rats, we administered CaMKII inhibitor KN62 (25 μ g/5 μ l, i.c.v.) after 40-min baseline recordings. KN62  significantly reduced basal fEPSP, resulting in a synaptic depression which lasted for several hours (Fig. 4D, n = 6, KN62 = 59.6 ± 4.5%, vs. baseline, p = 0.002), during which KN62 prevented FS to induce the CA1 LTP (Fig. 4E, n = 6, pre-FS = 73.1 ± 10.5% vs. KN62 = 74.5 ± 11.1%, p = 0.855) and reduced test immobility (Fig. 4H). As a yet additional test to the association between the CA1 LTP and despair-associated memory we examined the effect of deep brain stimulation (DBS). DBS is found to disrupt memory and is used in recent years to treat treatment-resistant MDD in patients 37 . DBS applied to the hippocampal CA1 area via the stimulating electrode blocked the CA1 LTP (Fig. 4F, n = 4, DBS = 80.6 ± 5.2%, vs. baseline, p = 0.091) and reduced test immobility (Fig. 4H); sham stimulation had no significant effect on test immobility (n = 6, 115.9 ± 20.8% of VEH + FS). We summarized the electrophysiological and behavioral data, because the accumulating evidence strongly supports a possible causal relation between the CA1 LTP (Fig. 4G,  The theory of metaplasticity provides the prediction of the consequences of the FS-induced CA1 LTP, such as occlusion of the subsequent LTP induction by using high-frequency stimulation (HFS) 26,27 . Just as the prediction, we found that HFS failed to induce CA1 LTP in FS rats, that showed increase of test immobility, but CA1 LTP could be induced in ES rats that did not exhibit increase of test immobility (Supplementary Figure 6). These findings are compatible with the notion that uncontrollable but not controllable stress may induce an endogenous form of CA1 LTP which is critical for the formation of despair-associated memory, mediating not only the increase of test immobility but also the subsequent influences on synaptic plasticity.

Discussion
The FS/ES protocol we employed here was effective for studying the underlying mechanisms of despair-associated memory that might shape cognition and behavior in the future. We found that animals in FS/ES groups shared similarities such as behavioral signs of stress (such as defecation), corticosterone levels (Fig. 2), an increase of S845-P and a transient synaptic depression. Despite that, were only in the FS group an increase in S831-P and a slow-onset CA1 LTP found. Most importantly, was only in the FS group a subsequent elevated immobility observed during the test, in agreement with the concept of the controllable and uncontrollable stress 26,39,40 . These findings are consistent with a recent report Scientific RepoRts | 5:15000 | DOi: 10.1038/srep15000 indicating that several types of inescapable stress including FS trial for 20-min increased NMDARand AMPAR-mediated excitatory postsynaptic currents in rat prefrontal cortex when GR was activated about 1 h after the stress exposure 41 . Thus under the uncontrollable situation, the AMPAR and CaMKII activity-induced rapid GluA1 phosphorylation, NMDAR and GR activation may work together to induce the slow-onset CA1 LTP that underlies despair-associated memory.
The different outcomes of FS and ES pinpoint the necessity of the uncontrollable stress nature in the formation of despair-associated memory. GluA1 S831-P may serve as a synaptic tagging for the subsequent molecular events in the selected synapses, because the increase of S831-P in FS group is independent from NMDAR and GR activation (Figs 3D and 5C). This disassociation between the synaptic tagging and the CA1 LTP is highly consistent with the synaptic tagging and capture hypothesis for the cellular consolidation of memory 42 . In particular, S831-P, NMDAR and GR were required for both the CA1 LTP and increased test immobility, suggesting that the molecular events could be bridged by the synaptic tagging mechanisms. It has been reported that GluA1 S831-P is involved in encoding emotional significance, facilitating LTP induction and enhancing emotional memory under stress situation 22,43 . So despair situation that have induced the synaptic tagging and the following LTP may be a major contributor to subsequent synaptic dysfunctions and behavioral abnormalities [25][26][27]44,45 .
Many of the hypotheses for MDD have proposed that chronic and mild stressful experience must have certain contributions to the development of the disorder 13,14,35,[46][47][48] . The powerful memory mechanisms for emotionally charged information would be highly conserved in evolution because they are critical in shaping survival cognition and behavior 5 . Here, we find that the uncontrollable type of acute stress induces a slow-onset CA1 LTP that underlies despair-associated memory. Given the unique mechanisms of the despair memory we demonstrated here, it is possible that the effects of chronic mild stresses on emotion could be accumulated by the memory mechanisms and thus likely led to psychopathological changes. In addition, the retrieval of despair-associated memory was prevented by NMDAR antagonists when treated 30 min before test trial (Fig. 6, n = 12, VEH = 100 ± 5.45%; n = 12, KET = 49.9 ± 4.4%, vs. VEH, p < 0.001; n = 12, MK801 = 24.9 ± 6.4%, vs. VEH, p < 0.001). Such a possible relation of despair memory mechanisms with MDD is also implicated by the antidepressant effects of ketamine, because a single treatment of ketamine reduced test immobility and the effect was maintained for about 7 d 49 , similar to the clinical features of ketamine in treating MDD patients 16 . Furthermore, the DBS treatment, which is known to disrupt memory and produces antidepressant efficacy in clinic, disrupted the uncontrollable stress-induced CA1 LTP and despair-associated memory. Thus, as the schematic diagram shown in Fig. 7, our evidencefor the hippocampal mechanisms of the 'despair memory' formation and retrieval should thus provide a new insight into not only the development of the stress-induced MDD but also the prevention and treatment of such psychopathologies.

Materials and Methods
Animals. Male Sprague Dawley rats (Animal Housing Center, Kunming Medical University, Kunming, China) weighing 250-300 g were used. Animals were group housed, but single housed after surgery, in ventilated cages with free access to water and food in a 12-h light/dark cycle and temperature-regu-

The forced swimming (FS) test and hopefully escapable swimming (ES). The experimental
protocols were modified from those described previously 33,50 . Briefly, rats were allowed to adapt to test room for 1 h before FS or test trial. All experiments were conducted between 8:30 and 18:00. The apparatuses were glass cylinders, 50 cm in height and 26 cm in diameter, and filled with 35 cm in height of tap water at 25 ± 1 °C that was the same as the regulated-room temperature. A hopefully escapable swimming (ES) paradigm was established by placing a floating platform into the water (Fig. 1A). This floating platform was a waterproof plastic box, 12 cm length, 10 cm width, 9 cm height, and stones were fixed to the inside bottom of the box to reach a final weight around 950 g. The floating platform was placed into the glass cylinders filled with water during ES trial for 15-min, but it was removed during test trial 24 h later. Notably, this floating platform was unstable, as ES rats could attempt to climb onto it but they were actually difficult to stay on it. After swimming, rats were dried very carefully by using a towel before returned to their homecages or the recording chamber for electrophysiological studies. Immobility was scored by skillful technicians to measure the accumulative time of the rats spent in immobile (slight movements of limbs that are necessary for keeping nose above the water are also account as immobility). Since there was no behavior equivalent to immobile posture in the ES trial on day 1, the immobility could be measured only during the test trial on day 2 in the ES group when the platform was removed. Each individual behavioral data were normalized (%) to the mean value of control group and then the mean and SEM were further calculated, except for the behavioral data in Fig. 1B,C which were normalized (%) to 5-min.
Surgery and drugs injection. Under pentobarbital sodium anesthesia (i.p., 60 mg/kg), rats were implanted with stainless steel guide cannulas (26 gauge, 11 mm, from RWD Life Science Co. Ltd., Shenzhen, Guangdong, China) that were affixed to the skull with dental cement by using techniques similar to those described 26,[51][52][53] . The cannulas were located into the lateral cerebroventricle (0.1 mm caudal of the Bregma, 1.6 mm lateral from the midline and 4.0 mm depth from the brain surface) or the bilateral hippocampal CA1 areas (3.8 mm caudal of the Bregma, 2.8 mm lateral from the midline and 2.2 mm depth from the brain surface Electrophysiological studies. Experimental protocols for the recordings of the field excitatory postsynaptic potentials (fEPSP) in the Schaffer/commissural-CA1 pathways in vivo are similar as those described previously 26,[53][54][55] . Briefly, Implantation of the electrodes was performed in the rats under pentobarbital sodium (60 mg/kg, i.p.) anesthesia and body temperature was maintained at 37 ± 0.5 °C by a heating blanket. Three stainless-steel screws (one serving as a reference electrode, the second acting as a ground electrode and all of three serving as anchors) were inserted into the skull through drilled holes without piercing the dura. Electrodes were made by gluing together a pair of twisted Teflon-coated 90% platinum/10% iridium wires (50-μ m bare diameter, 100-μ m coated diameter, World Precision Instruments, Sarasota, Florida, USA) and were placed into the CA1 area at the following coordinates: FS induced a rapid increase of GluA1 S831P①, followed by the conjunctive activation of NMDAR② and GR③. These molecular events led to an endogenous CA1 LTP④, and contributed to despair-associated memory (increase of test immobility). According to this model, antidepressants that target on AMPAR, NMDAR (such as ketamine) and GR may exert antidepressant effect by preventing the formation and retrieval of despair-associated memory.
Scientific RepoRts | 5:15000 | DOi: 10.1038/srep15000 3.8 mm caudal of the Bregma, 2.8 mm lateral from the midline and 2.4 mm depth from the brain surface for the recording electrode; 4.8 mm caudal of the Bregma, 3.8 mm lateral from the midline and 2.4 mm depth for the stimulating electrode. The optimal depth of the electrodes was determined by electrophysiological criteria and was verified by post-mortem examination. For the fEPSP recordings in anaesthetized rats, after 40 min stable baseline recordings, a high-frequency stimulation (HFS, 200 Hz, 10 trains × 20 pulses/train, intertrain interval 2 s) at the baseline stimulation intensity was delivered. For the fEPSP recordings in freely moving rats and the deep brain stimulation (DBS), the entire assembly of the electrodes was sealed and fixed to the skull using dental acrylic. The rats were then placed individually into their homecages for at least 2 weeks for recovery. DBS was applied through the stimulating electrode using a clinically used protocol (100 μ s pulse widths, 150 μ A intensity, 130 Hz for 1.5 h) described previously 37,56 . Immunoblot assays. The synaptic proteins were made from filtered synaptoneurosomes that prepared from hippocampal tissue as those described previously 22,57 . The hippocampal samples were obtained in experimental or naïve rats. Dissections of the hippocampus were performed using ice-cold phosphate buffered saline (PBS) rapidly, and the samples were homogenized in ice-cold homogenization buffer (10 mM Hepes, 1.0 mM EDTA, 2.0 mM EGTA, 0.5 mM DTT, 1 mM PMSF, 10 mg/liter leupeptin, 1 mg/ liter pepstatin A, 10 mM NaF, 1 mM Na 3 VO 4 , 0.2 mM β -glycerophosphate). The homogenization was performed by glass/glass tissue homogenizers, and the homogenate was passed through two 100-μ m-pore nylon mesh filters (Millipore, NY1H02500), and then through a 5-μ m-pore filter (Millipore, SLSV025NB). The filtered homogenates were centrifuged at 1000 g for 15 min at 4 °C. Resultant pellets (filtered synaptoneurosomes) were resuspended in 350 μ L boiling 1% SDS, boiled for 5 min. After 20 μ L of each sample being saved for BCA assay, the rest was mixed with 4x Laemmli Buffer in a 3:1 ratio, heated at 80 °C for 15 min, then stored at − 80 °C 58 .
Immunoblot analysis was performed as previously described 59 . Synaptic protein samples (40 μ g/lane) were size-separated by electrophoresis in SDS-PAGE (10% acrylamide) and transferred to PVDF membranes (Immobilon TM -P PVDF membrane, from Millipore Co, Billerica, Massachusetts, USA). Membranes were blocked at room temperature for 120 min with TBS-T (0.9% NaCl, 10 mM tris, 0.1% tween-20, PH7.4) containing 1% BSA on an orbital shaker. After blocking, the membrane was reacted overnight at 4 °C with the primary antibody diluted in TBS-T containing 1% BSA (Rabbit × GluA1, 1: 5000, from Millipore Co, Billerica, Massachusetts, USA; Rabbit × GluA1 S831p, 1:15000, Rabbit × GluA1 S845p, 1:15000, from Epitomics Co., Burlingame, California, USA; Mouse × β -tubulin, 1:30,000, from CWbiotech Co. Ltd., Beijing, China). After four washes of 10 min each with TBS-T, the membranes were subsequently incubated for 2 h at room temperature with HRP-linked secondary antibody diluted in TBS-T (Goat × Rabbit/Mouse, HRP-linked, 1:20000, KangChen Bio-tech Inc., Shanghai, China). The membrane was washed for another four times before exposure to the chemiluminescent HRP substrate (Luminata TM HRP substrates, from Millipore Co, Billerica, Massachusetts, USA). The light emitted by ECL reagent is subsequently captured on X-ray film (FUJIFILM Super RX, from Fuji Photo Film Co Ltd, Tokyo, Japan). Films were scanned by an Epson scanner. Then the images were transferred to grayscale and the bands intensity were analyzed by NIH ImageJ software. Elevated platform stress. A mild stress was evoked by elevated platform (EP) as described previously 26,52 . Briefly, animals were placed on an elevated platform (10 cm × 10 cm × 1.6 m) in the middle of a bright room for 30 min, and signs of stress such as freezing, defecation and urination were observed. Immediately after the EP or homecage exposure, the rats were subjected to FS. Serum preparation and corticosterone immunoassay. Serum preparation and corticosterone determination was performed as described previously 44 . Briefly, blood was collected by cardiac puncture under pentobarbital sodium (60 mg/kg, i.p.) anesthesia, 15 minutes after 15′FS, ES, 5′FS, 30 minutes after corticosterone injection (5 mg/kg, i.p.), or immediately after 30′ exposed to elevated platform stress. Samples were left undisturbed for at least 1 hour, then centrifuged at 1000 g twice to remove the clot. Approximately 0.5 ml of serum from each rat was collected into 1.5 ml Eppendorf tubes and stored at − 80 °C. Corticosterone concentration was assessed using ENZO ADI-900-097 ELISA kits (from ENZO life science, New York, USA). The sensitivity of the corticosterone assay was 27 pg/mL. Statistical Analysis. Paired or unpaired Student's t test or a one-way ANOVA followed by post hoc analysis with least significant difference (LSD) was used for comparisons. The significance level was set at p < 0.05. All values were reported as mean ± SEM.