Development of Depotentiation in Adult-Born Dentate Granule Cells

Activity-dependent synaptic plasticity, i.e., long-term potentiation (LTP), long-term depression (LTD) and LTP reversal, is generally thought to make up the cellular mechanism underlying learning and memory in the mature brain, in which N-methyl-D-aspartate subtype of glutamate (NMDA) receptors and neurogenesis play important roles. LTP reversal may be the mechanism of forgetting and may mediate many psychiatric disorders, such as schizophrenia, but the specific mechanisms underlying these disorders remain unclear. In addition, LTP reversal during the development of adult-born dentate granule cells (DGCs) remains unknown. We found that the expression of the NMDA receptor subunits NR2A and NR2B displayed dynamic changes during the development of postnatal individuals and the maturation of adult-born neurons and was coupled with the change in LTP reversal. The susceptibility of LTP reversal progressively increases with the rise in the expression of NR2A during the development of postnatal individual and adult-born neurons. In addition, NMDA receptor subunits NR2A, but not NR2B, mediated LTP reversal in the DGCs of the mouse hippocampus.

INTRODUCTION inputs in the developing Xenopus retinotectal system, which depend on activation of NMDARs (Zhou et al., 2003). Similarly, high frequency stimulation induces LTP in the rat hippocampus in familiar environments, which can be reversed by exploration of a novel environment (Xu et al., 1998).
NMDA receptors (NMDARs) are glutamate-gated ion channels, which are crucial for synapse structure and function in the CNS (Paoletti et al., 2013). NMDARs are heteromers (Dingledine et al., 1999) composed of two essential subunits (NR1) and two or three regulatory subunits (NR2). In the hippocampus, the regulatory subunit is usually composed of a combination of NR2A and NR2B. The NMDAR expression patterns (region, composition and level) change dynamically during postnatal development, and NR2B is gradually replaced by NR2A (Sheng et al., 1994;Loftis and Janowsky, 2003). Development-related changes may alter the functional properties of NMDARs, especially the kinetic properties of the channels (Monyer et al., 1992;Sheng et al., 1994;Takahashi et al., 1996;Flint et al., 1997;Rumbaugh and Vicini, 1999). For instance, it has been confirmed that the NR2A-containing NMDARs deactivate faster than NR2B-containing NMDARs in HEK 293 cells (Monyer et al., 1994;Vicini et al., 1998). The direction of synaptic plasticity (potentiation or depression) is controlled largely by the NMDAR deactivation kinetics and the amount of Ca 2+ influx through NMDARs, and therefore, the composition of NR2A and NR2B in NMDARs plays a significant role in the brain (Philpot et al., 2007;Wang et al., 2011). In addition, some reports suggested that the NMDAR subunit NR2B was involved in LTD, whereas NR2A was involved in LTP reversal (Liu L. et al., 2004;Massey et al., 2004;Zhu et al., 2005).
Adult neurogenesis has been observed in different brain regions of multiple species, such as in humans and mice (van Praag et al., 2002;Boldrini et al., 2018;Pilz et al., 2018), yet its physiological significance remains essentially unknown (Schinder and Gage, 2004). Increasing evidence suggests that newborn dentate granule cells (DGCs) contribute to hippocampus-specific forms of learning and memory, which might be different from the function of mature DGCs in the adult brain (Clelland et al., 2009;Aimone et al., 2014). Adult neurogenesis in the dentate gyrus (DG) is regulated by NMDAR activation. Application of the NMDAR antagonist resulted in an increase in the number of newborn DGCs in the adult hippocampus (Cameron et al., 1995;Gould et al., 1997). In addition, it has been reported that conditional reduction in adult neurogenesis impairs LTP and LTD in the hippocampus, and the deficits in bidirectional synaptic plasticity are completely rescued when dentate gyrus neurogenesis is recovered (Massa et al., 2011). Adult-born DGCs increase the LTP amplitude and decrease the LTP induction threshold compared with mature DGCs in the adult brain (Ge et al., 2007;Dieni et al., 2016). Moreover, there are many evidences that schizophrenia animal models show impaired generation of newborn DGCs in the adult brain (Reif et al., 2006;Mao et al., 2009;Ouchi et al., 2013). A recent report indicated that Erbb4 knock-out mice show a deficit in LTP reversal, which is coupled with behavioral deficits that are thought to be the positive symptoms of schizophrenia (Shamir et al., 2012). A recent report indicated that the developmental reinforcement of LTP reversal susceptibility is associated with increased forgetting (Ge et al., 2019). However, no report has studied the reversal of LTP in adult-born DGCs, which may contribute to a better understanding of the physiological significance of adult neurogenesis and the pathologic mechanism of related schizophrenia.
We examined LTP reversal in mature and adult-born DGCs with whole-cell patch-clamp. The adult-born DGCs were labeled with retrovirus to test synaptic plasticity during their maturation. We found that LTP reversal occurred only in mature and adult-born 6-week DGCs, not in adult-born 3-week DGCs. This difference was related to the NMDAR subunit NR2A expression level elevation in adult mouse mature DGCs. Similarly, NR2A expression increased coupled with LTP reversal in adult mice during postnatal development. In conclusion, the NMDAR subunit NR2AR mediated LTP reversal in hippocampal DGCs, and NMDAR subunit NR2A expression increased with individual postnatal development and adult-born neuron maturation.

Retrovirus Injection
Highly concentrated murine moloney leukemia virus-based retroviral (RV) stock (6.74 × 10 8 TU/ml) carrying the green fluorescent protein (GFP) was injected into the DG of the mouse hippocampus through stereotaxic surgery to label dividing newborn neurons. The injection site for DG was determined using spatial coordinates relative to the bregma in the mouse brain (details shown in Table 1). The mice used for RV injection were 6-7-week-old male C57BL/6 mice, who were housed in standard cages for further experimentation after surgery. All animal experimental procedures were approved by the Committee of Animal Care of Huazhong University of Science and Technology.
The bregma was used as a zero point to measure the anterior to posterior (A-P) and medial to lateral (M-L) coordinates. Dorsal to ventral (D-V) positions were measured from the skull at the bregma. The bregma and the lambda were on the same horizontal plane to obtain the correct coordinates. The injection quantity of the RV for each position was 0.1 µl. Images of newborn DGCs in the hippocampus with GFP + were acquired on an inverted single-photon laser scanning confocal microscope (Zeiss, LSM780).

Electrophysiology
The mice were housed in standard cages for different periods of time after RV injection that expressed GFP were anesthetized. The mouse brains were quickly removed and placed in 4 • C artificial cerebrospinal fluid (ACSF) composed of (in mM) choline chloride 110, KCl 2.5, NaH 2 PO 4 1.25, NaHCO 3 26.0, CaCl 2 0.5, MgCl 2 7, D-glucose 10, Na-ascorbate 11.6, Napyruvate 3.1, and atropine sulfate 0.01. Horizontal slices (200 µm thick) were cut using a Leica VT1000S vibratome and then incubated in standard ACSF containing (in mM) NaCl 119, KCl 2.5, NaH 2 PO4 1.25, NaHCO 3 26, CaCl 2 2.5, MgCl 2 1.3, D-glucose 10 saturated with 95% O 2 and 5% CO 2 . Hippocampus slices from the mice were kept at 35 • C approximately 40 min, followed by ACSF incubation at room temperature for at least 60 min before recording. Electrophysiological recordings were performed as previously described (Mu et al., 2015). Hippocampus DGCs were visualized by an upright microscope (Olympus, BX51WI) with infrared differential interference contrast (DIC) optics. The newborn DGCs from RV-injected mice were visually identified by their GFP and neuronal morphology, usually located in the inner DGC layer. The excitatory postsynaptic potentials (EPSPs) of mature DGCs (GFP − ) located in the outer DGC layer (Kuhn et al., 1996;Schmidt-Hieber et al., 2004;Esposito et al., 2005) in the brain slice were recorded. The slice was perfused with oxygen-saturated ACSF containing 50 µM picrotoxin to block GABAergic synaptic transmission during recording at room temperature in the chamber. The micropipettes (3-5 M) were tip-filled with internal solution composed of (in mM) Kgluconate 128, KCl 17.5, NaCl 9, MgCl 2 1, EGTA 0.2, and HEPES 10 (pH 7.4) and back-filled with the same internal solution containing amphotericin B (200 µg/ml) to perform the whole-cell perforated patch recordings for recording the EPSPs. A bipolar electrode (World Precision Instruments) was placed in the stratum molecular to stimulate the medial perforant pathway input to the dentate gyrus in the hippocampus slice. DGCs were held at −70 mV to record the EPSPs in currentclamp mode. For characterization of excitatory postsynaptic currents (EPSCs), potassium salt was substituted with cesium salt in the intracellular solution, which was composed of (in mM) Cs-methanesulfonate 128, CsCl 17.5, NaCl 9, MgCl 2 1, EGTA 0.2, and HEPES 10 (pH 7.4). α-Amino-3-hydroxy-5methyl-4-isoxazole-propionic acid receptor (AMPAR)-mediated EPSC recorded at −70 mV in whole-cell voltage-clamp with picrotoxin (PTX, 50 µM), NMDAR-mediated EPSC recorded in the presence of CNQX (10 µM) and PTX at +40 mV in currentclamp mode. Deactivation decay times of averaged EPSCs from 10 continuous EPSC sweeps were derived from fitting to double exponential equations, , where A1 and A2 are the amplitudes of the fast and slow decay components, and tau1 and tau2 are their respective decay time constants used to fit the data. A weighted mean decay time constant was used to compare decay times: tau = A1 A1+A2 × tau1 + A2 A1+A2 × tau2 (Rumbaugh and Vicini, 1999). The antagonists for NMDAR subunits were added to the ACSF. All drugs were purchased from Sigma. Data were collected using a Multiclamp 700B computercontrolled current and voltage clamp (Molecular Devices) and acquired with a Digidata 1550B low-noise data acquisition system (Molecular Devices) at 5 kHz. The input resistance was obtained from hyperpolarizing current injections of 5 pA into the DGCs, which were monitored continuously during recordings. The series resistance was controlled below 20 M . Data were accepted for analysis only if both series and input resistances remained relatively constant (<20% change) throughout the experiment. LTP was induced with five episodes of theta-burst stimulation (TBS) applied at 0.1 Hz. Each episode of TBS consisted of 10 trains of stimuli delivered every 200 ms, with ten pulses at 100 Hz in each train (Supplementary Figure S1). Reversal LTP was induced with a combination of presynaptic stimulation and postsynaptic spikes, in which the presynaptic stimulation was 70 ms anterior to the postsynaptic spikes. The presynaptic stimulation and the postsynaptic spikes both consisted of 120 trains at 0.2 Hz, and each train contained 5 pulses at 20 Hz (Supplementary Figure S1). The extent of LTP was quantified by averaging the amplitude of EPSPs during the last 10 min of experiments and normalizing the result to the mean baseline value. Data were compared with either Student's t-test (paired or unpaired), and results associated with p < 0.05 were considered statistically significant.

DGC-Specific and DG-Specific Real-Time PCR
Bilateral DGs were separated from the hippocampus under a stereomicroscope to extract the tissue RNA of DGs. The RNA was converted to cDNA through reverse transcription (Vazyme, R223) according to the operation manual for realtime PCR. Entire DGCs were harvested from acute mouse brain slices by patch-clamp capillaries with a tip diameter of approximately 5 µm (Song et al., 2016). A total of 20 cells were collected into one pipette to increase the quantity of RNA, and the pipette contents were then moved into a sterile 0.5 ml PCR tube containing TRK buffer (TakaRa, RR047Q) to extract the neurons RNA of DGCs. The RNA was converted to cDNA through reverse transcription (OMEGA, R6831) according to the product instructions. The DG and DGC cDNA samples were subsequently analyzed by forty rounds of PCR amplification (Vazyme, Q311) using specific primers for Grin2A (NR2A), Grin2B (NR2B) and the housekeeping gene GAPDH ( Table 2). GAPDH could be detected reliably from all samples tested. Real-time PCR experiments were performed in triplicate, and each sample was normalized to the control housekeeping gene GAPDH. Electrophoresis images were obtained using the gel imaging analysis system to verify the product of NR2A, NR2B and GAPDH after PCR amplification.

Statistical Analysis
Statistical analyses were performed using two-tailed Student's t tests (paired or unpaired). Data are expressed as the mean ± SEM. A p-value less than 0.05 was considered statistically significant.

Characteristics of LTP and LTP Reversal in the DGCs of the Adult Mouse Hippocampus
To evaluate the LTP and LTP reversal, EPSPs of DGCs were recorded in current-clamp mode with low-frequency stimulation (20 s per interval) of the medial perforant pathway ( Figure 1A).
To examine the synaptic plasticity of DGCs, we used a TBS paradigm (Supplementary Figure S1) that mimics the physiological firing pattern commonly observed in hippocampal DGCs in vivo (Skaggs et al., 1996) to induce LTP (Larson and Lynch, 1986). Figures 1B,D illustrates a typical LTP of an EPSP recorded from DGCs with TBS stimulation. Here, we calculate the normalized EPSP amplitude by averaging the baseline as the standardized reference value, and we set the time point after applying the TBS as 0 min for all LTP and LTP reversal cartograms unless otherwise stated. During the recordings, LTP was maintained at a stable level for more than 60 min ( Figure 1E, paired t-test, compared with the baseline, 0-10 min: 1.45 ± 0.1715, P = 0.0468; 30-40 min: 1.498 ± 0.062, P = 0.0005; 50-60 min: 1.545 ± 0.07196, P = 0.0006; n = 6 cells). After 10 min of LTP induction, an associated 10-min stimulus of presynaptic stimulation and postsynaptic spikes (associated stimulation, AS, Supplementary Figure S1) that simulates the spontaneous activity and firing pattern of Xenopus tectal neurons exposed to a random visual input which can reverse the LTP in vitro (Zhou et al., 2003) was applied to reverse the LTP (Figures 1C,D). LTP (open circles, n = 6 cells) and LTP reversal (solid circles, n = 7 cells) were successfully induced after applying the relevant stimuli to the hippocampus ( Figure 1D). The LTP amplitude was efficiently reversed almost to the baseline level after delivering the AS ( Figure 1F, paired t-test, compared with baseline, 0-10 min: 1.475 ± 0.09572, P = 0.0025; 30-40 min: 1.026 ± 0.0616, P = 0.6891; 50-60 min: 1.015 ± 0.08825, P = 0.8691; n = 7 cells). The extent of LTP reversal was 93.95 ± 16.86% ( Figure 1I, n = 7 cells). Poo's group reported that random flashes were effective in reversing LTP only when they were presented within the first 20 min after the induction of LTP in tectal neurons of Xenopus (Zhou et al., 2003). Therefore, we examined the time window of the LTP reversal during which the AS was delivered after 30 min of LTP induction. We found that the extent of LTP reversal was 7.211 ± 28.15% after 30 min of LTP induction (Figures 1G,I; unpaired t-test, compared with the extent of LTP reversal when AS was delivered after 10 min of LTP induction, P = 0.0185, n = 5 cells). AS could obviously reverse LTP within 10 min after the induction of LTP in adult mature DGCs. Taken together, these results showed that LTP could be effectively reversed in a time-dependent pattern.
We sought to reverse LTP with the single stimuli model of presynaptic stimulation or postsynaptic spikes to test whether presynaptic and postsynaptic stimuli are both required for inducing LTP reversal. The presynaptic stimulation could transiently reverse LTP within 10 min of LTP induction, but it slowly recovered to the original LTP level after 20 min ( Figure 1J, solid circles, n = 6 cells). Postsynaptic spikes made no difference in LTP during the first 10-min period after the induction of LTP (open circle, n = 6 cells). Therefore, both presynaptic and postsynaptic mechanisms are involved in LTP reversal in the DG. Usually, LTP reversal can be induced by low-frequency stimulation (LFS) both in vitro and in vivo (HzFroc et al., 2000;Sugita et al., 2016), in which LFS is ineffective at inducing LTD (Huang et al., 1999). This phenomenon of LTP reversal might be the distinct nature of LTD. Therefore, we first tested the effect of AS on basal synaptic plasticity. We recorded the EPSPs before and after giving the AS to test the influence on the basal synaptic transmission. As we could see, AS failed to induce stabilized LTD, although the EPSP amplitude decreased during the first 10min period (open circles, n = 8 cells). We found that 1 Hz LFS (900 pulses) failed to induce LTP reversal (open circle, n = 5 cells), but it could induce stable LTD ( Figure 1K, solid circles, n = 7 cells).

LTP and LTP Reversal Varies With Developmental Stage of the Adult-Born DGCs in the Hippocampus
It has been reported that synaptic plasticity dynamically changes during the development of adult-born neurons, especially LTP (Ge et al., 2007), although the changes in LTP reversal during maturation of adult-born DGCs in the hippocampus remain unknown. To test LTP and LTP reversal at different stages during the development of adult newborn DGCs, we labeled newborn neurons in the adult hippocampus with stereotaxically injected retroviruses (RV) expressing green fluorescent protein (GFP) into the DG of adult mice (for details see Experimental Procedures). GFP + newborn DGCs in fresh brain slices prepared from retrovirus-infected adult mice were recorded with wholecell patch-clamp (Figure 2A). Obvious LTP of EPSPs was induced with TBS in GFP + DGCs after 3 weeks of RV infection ( Figure 2B, paired t-test, compared with baseline, 0-10 min: 2.022 ± 0.3148, P = 0.0228; 30-40 min: 2.196 ± 0.3054, P = 0.0112; gray-solid circles, n = 6 cells), when most newborn DGCs are already fully integrated into the existing circuitry (Esposito et al., 2005;Ge et al., 2006). We recorded EPSPs from GFP + DGCs at 6 weeks after RV infection ( Figure 2B, blacksolid circles, n = 6 cells) when newborn neurons are nearly entirely morphologically  and physiologically matured (Laplagne et al., 2006). Interestingly, LTP was reversed successfully in newborn 6-week-old DGCs after delivering the AS post-10 min of LTP induction (paired t-test, compared with baseline, 0-10 min: 1.895 ± 0.275, P = 0.0226; 30-40 min: 1.087 ± 0.121, P = 0.5036; n = 6 cells), but newborn 3-week-old DGCs failed to reverse the LTP when the same AS paradigm was transmitted (paired t-test, compared with baseline, 0-10 min: 1.817 ± 0.1083, P = 0.0006; 30-40 min: 1.87 ± 0.1638, P = 0.0032; n = 6 cells). These results demonstrated that LTP reversal gradually emerged during the maturation of adultborn DGCs. It has been reported that adult-born DGCs present different electrophysiological properties, including higher input resistance and lower thresholds for the induction of LTP than mature DGCs (Mu et al., 2011;Marin-Burgin et al., 2012;Dieni et al., 2016). Thus, we quantified the input resistance of DGCs. Consistent with previous reports, the mean input resistance of adult-born DGCs was significantly larger than adult mature DGCs (Figure 2C, unpaired t-test, compared with adult mature DGCs, adult mature DGCs: 0.2003 ± 0.006771 G , n = 8 cells; adult-born 3 W: 0.3514 ± 0.03797 G , P = 0.0007, n = 6 cells; adult-born 6 W: 0.3469 ± 0.05898 G , P = 0.0140, n = 6 cells). A previous study reported that newborn 2-3-week-old DGCs begin to receive excitatory input from entorhinal cortical perforant pathway synapses, and their physiology and anatomy begin to approach maturation at 4-8 weeks, coupled with slowly developing dendritic arborizations and axonal projections in adult brain (Aimone et al., 2014). We first observed morphological changes during newborn DGC maturation in the adult hippocampus. Here, we can see that newborn 3-week-old DGCs extend their apical dendrites into the molecular layer, but newborn 6-week-old neurons displayed abundant dendrite arborizations in the molecular layer in the DG ( Figure 2D).

Developmental Alterations in Adult Mouse 3-Week-Old and 6-Week-Old Newborn DGC NMDAR Subunit Composition
NMDARs are important for adult newborn DGC maturation in the hippocampus (Ge et al., 2007;Mu et al., 2015). Acute treatment with NMDA rapidly decreases the number of newborn neurons, whereas NMDAR antagonist injection increases the birth of DGCs in the rat hippocampus (Cameron et al., 1995). We sought to determine the NMDAR subunit distribution and contribution alterations during newborn DGC development in the adult hippocampus. It was reported that NR2A mRNA expression increases in association with downregulation of NR2B mRNA expression during neuron maturation in cultured cerebellar granule cells (Vallano et al., 1996). We detected NR2AR and NR2BR mRNA in adult mouse 3-week-old and 6-week-old newborn DGCs by single-cell real-time PCR. The NR2AR mRNA expression level of newborn 3-week-old DGCs was standardized as a reference. The NR2AR mRNA relative expression of newborn 6-week-old DGCs was distinctly higher than that of newborn 3-week-old DGCs (Figure 3A, unpaired t-test, P = 0.0303, adultborn 3 W: 1 ± 0.121, n = 16 mice; adult-born 6 W: 1.619 ± 0.2337, n = 18 mice) in adult mice. There was no difference in NR2BR mRNA relative expression between the two groups of adultborn DGCs in the hippocampus (unpaired t-test, P = 0.2699, 3-week-old adult-born: 1.341 ± 0.1634, n = 16 mice; 6-week-old adult-born: 1.682 ± 0.2469, n = 18 mice).

LTP Reversal and the Expression of NMDAR Subunits in the DG Change With Postnatal Individual Development
To test whether synaptic plasticity related to LTP and LTP reversal during the development of postnatal mice presents similar changes with the maturation of adult-born neurons in the hippocampus, we recorded the EPSPs in juvenile mouse (postnatal 19-21 days) mature DGCs and adult mouse (postnatal 60-70 days) mature DGCs. A previous study suggested the reversal of LTP changes with development in postnatal individuals (Kramár and Lynch, 2003). Here, we showed that the same AS protocol that reversed the LTP in adult mice mature neurons could not reverse the LTP in juvenile mice mature DGCs under the same conditions ( Figure 4A, paired t-test, compared with the baseline, 0-10 min: 2.047 ± 0.2534, P = 0.0061; 30-40 min: 1.887 ± 0.2063, P = 0.0051; blacksolid circles, n = 7 cells). The normalized EPSP amplitude of juvenile mouse mature DGCs was maintained at a stable level after delivering the TBS (Figure 4A, paired t-test, comparing the normalized EPSP amplitude at 0-10 min with 30-40 min, P = 0.1440, 0-10 min: 1.864 ± 0.2663; 30-40 min: 2.197 ± 0.2209; open circle, n = 7 cells). We also found the mean LTP amplitude of juvenile mouse mature DGCs during the whole LTP (0-60 min) was significantly larger than that of adult mouse mature DGCs [ Figure 4B, unpaired t-test, P = 0.0269, juvenile mice: 2.133 ± 0.2207, n = 7 cells ( Figure 4A, open circle); adult mice: 1.503 ± 0.06319, n = 6 cells ( Figure 1D, open circle)]. NMDARs are important for synaptic plasticity during postnatal individual development (Sheng et al., 1994). It has been reported that the NMDAR subunits NR2AR and NR2BR change with postnatal individual development of the brain (Wenzel et al., 1997;Liu X.B. et al., 2004). To explore whether the NMDAR subunits NR2A and NR2B changed during postnatal individual development, we detected the NR2A and NR2B mRNA expression levels in hippocampal DG tissue using realtime PCR. We found that the relative expression of NR2A in the adult DG was significantly higher than that in the juvenile DG ( Figure 4C, unpaired t test, P = 0.0225, juvenile mice: 1 ± 0.09187, n = 11 mice; adult mice: 1.494 ± 0.1888, n = 9 mice). The NR2BR mRNA relative expression level showed no difference between the juvenile DG and the adult DG (unpaired t-test, P = 0.1715, juvenile mice: 1.326 ± 0.1672, n = 11 mice; adult mice: 1.696 ± 0.2026, n = 9 mice). The reliability of the real-time PCR was confirmed by running a gel imaging analysis (Supplementary Figure S2). In addition, evoked EPSCs were recorded with whole-cell voltage-clamp from both juvenile and adult hippocampus DGCs. As reported previously, NMDA-EPSCs from adult Sprague Dawley rat hippocampi exhibit much faster decay times than juvenile rats (Rumbaugh and Vicini, 1999). Here, we observed that NMDA-EPSC decay kinetics are significantly faster in adult hippocampus DGCs than in juvenile hippocampus DGCs (Figure 4D, unpaired t-test, P = 0.0066, juvenile mice: 157.5 ± 13.37 ms, n = 9 cells; adult mice: 114.9 ± 2.895 ms, n = 9 cells), which further demonstrated that NR2AR subunit expression was apparently higher in juvenile DG NMDARs than in adult DG NMDARs. There were no obvious differences in AMPA-EPSC decay times between juvenile and adult DGCs (unpaired t-test, P = 0.1189, juvenile mice: 15.94 ± 1.212 ms, n = 9 cells; adult mice t: 12.87 ± 1.417 ms, n = 9 cells). Similar to the changes during maturation of adult newborn DGCs in the hippocampus, the percentage increase in NMDA-EPSC decay time in the adult mouse mature DGCs was significantly higher than that in juvenile mouse mature DGCs after perfusion with PEAQX ( Figure 4E, unpaired t-test, P = 0.0190, adult mice: 154 ± 11.34% of the decay time before perfusion, n = 7 cells; juvenile mice: 117.9 ± 4.921% of the decay time before perfusion, n = 6 cells). In contrast, the decreased NMDA-EPSC decay time percent in the adult mouse mature DGCs was not significantly different between juvenile mouse mature DGCs after perfusion with ifenprodil (unpaired t-test, P = 0.8407, adult mice: 81.58 ± 7.064% of the decay time before perfusion, n = 7 cells; juvenile mice: 83.28 ± 4.748% of the decay time before perfusion, n = 8 cells). These results indicated that the proportion of NR2AR in NMDAR subunits was greater in juvenile mouse mature DGCs than in juvenile mouse mature DGCs in the adult brain.

DISCUSSION
Long-term potentiation reversal leads to prior acquired memory deficiency (Kim and Cho, 2017), which is likely to be the mechanism by which postlearning interfering stimuli induce forgetting (Wixted, 2004). Our finding demonstrated that the susceptibility to LTP reversal progressively increases during the development of postnatal individuals and maturation of adultborn neurons, which was coupled with NMDA receptor subunit NR2A expression augmentation. In addition, the reason for the increase in LTP reversal susceptibility is the increase in NR2A relative expression. LTP reversal can be achieved within 10 min of induction (Huang et al., 1999;Staubli and Scafidi, 1999). Figure 1 shows the induction of LTP and LTP reversal in the DG of the adult mouse hippocampus by TBS and AS, which is distinct from a protein synthesis-dependent LTP consolidation that requires several hours to begin (Nguyen et al., 1994). However, it has been reported that new dendritic structures emerge approximately 15-30 min after LTP induction in the hippocampus CA1 (Engert and Bonhoeffer, 1999;Maletic-Savatic et al., 1999). It is believed that LFS (2 Hz) or theta-frequency stimulation (5 Hz) can effectively reverse LTP in the brain (Zhou and Poo, 2004). LTP reversal occurred at the input that received the LTP reversal stimuli but not at another potentiated input to the same postsynaptic cell (Muller et al., 1995). Our experiment adopted a modified AS stimuli pattern to simulate the spontaneous activity of tectal neurons during LTP reversal in the visual system (Zhou et al., 2003) to investigate the physiological significance of LTP reversal, which effectively reversed the LTP in the adult hippocampus DG region (Figure 1). In addition, this AS pattern revealed that both pre-and postsynaptic mechanisms are needed to induce LTP reversal. The same AS paradigm for reversing LTP could not induce LTD in adult mature DGCs ( Figure 1K). However, LTD could be induced by LFS (1 Hz) in the presence of PTX, which could not reverse LTP in the adult brain ( Figure 1L). These findings further illuminated that LTP reversal is a nervous system phenomenon involving persistent changes in synaptic responsiveness, which differs from LTD (Wagner and Alger, 1996).
Long-term potentiation amplitude and the extent of LTP reversal change during development of postnatal individual and maturation of newborn neurons (Wagner and Alger, 1995;Kramár and Lynch, 2003;Ge et al., 2007). The dynamic changes in the mRNA expression of NMDAR subunits NR2A and NR2B during the maturation of adult-born DGCs, in which LTP reversal presented dramatic corresponding changes, have not been previously reported. We used single-cell real-time PCR to test the NMDAR subunit NR2A mRNA expression and whole-cell patch-clamp to recording the NMDA-mediated EPSCs in adult-born 3-week-old and adult-born 6-week-old DGCs. We can see that NR2A subunit expression increased and the decay time decreased in adult-born 6-week-old DGCs (Figure 3), which was coupled with successfully induced LTP reversal compared with the failure to induce LTP reversal in adult-born 3-week-old DGCs (Figure 2). Similar to changes in the maturation of adult-born DGCs, synaptic plasticity presented relevant changes during development of postnatal individuals (Kramár and Lynch, 2003). However, the differences in stimulation protocols and brain regions may account for the discrepancy between the experiments regarding the finding that LTP reversal susceptibility increased with increasing age ( Figure 4B). During the development of postnatal individuals and the maturation of adult-born neurons, NR2A increased coupled with LTP reversal under the same stimulus pattern (Figures 2-4). Whole-cell patch-clamp to record the NMDAmediated EPSCs showed that the decay time of NMDA-EPSCs decreased with the augmentation of NMDAR subunit NR2A mRNA expression during individual postnatal development (Figure 4). We found that NR2A subunit expression was increased and activity-dependent NMDAR responses were decreased, as previously reported in the visual cortex during individual postnatal development (Carmignoto and Vicini, 1992;Liu X.B. et al., 2004).
NR2A-containing NMDARs produce faster (Monyer et al., 1994;Flint et al., 1997) and smaller EPSCs (Barria and Malinow, 2002), which may influence the amount of Ca 2+ entering the intracellular space. It has been hypothesized that these properties could be responsible for the reduced synaptic plasticity observed in the brain during individual postnatal development (Carmignoto and Vicini, 1992;Crair and Malenka, 1995) or after learning (Quinlan et al., 2004). It is known that the patterns of intracellular Ca 2+ rise dictate the direction of NMDARdependent synaptic plasticity (Yang et al., 1999). Higher levels of NMDAR activation produce LTP, while low levels of activation induce LTD (Artola and Singer, 1993). The reversal of LTP immediately after LTP induction depends on a small amount of Ca 2+ entering the intracellular space through NMDARs during stimuli, which in turn activates protein phosphatase 1 (PP-1) to change the cytoskeleton and gene transcription to mediate synaptic plasticity Zhou and Poo, 2004). The amount of Ca 2+ entering the intracellular space through NMDARs may be decided by a subunit substitution in the NMDARs, which influences the decay time of NMDARs. A previous study demonstrated that the decay time is significantly slower for NR1/NR2B NMDARs than for NR1/NR2A NMDARs (Vicini et al., 1998), in which the shorter NMDAR decay time may decrease the amount of Ca 2+ entering the intracellular space. NMDAR activation has been shown to play a role in learning and memory (Paoletti et al., 2013) as well as in adult neurogenesis (Deisseroth et al., 2004;Mu et al., 2015). Adult neurogenesis and NMDARs mediate bidirectional hippocampal synaptic plasticity (Liu L. et al., 2004;Massa et al., 2011). LTP reversal can be mimicked by a brief application of NMDA during the initial susceptible phase of LTP in the CA1 region of hippocampus slices (Lee et al., 1998). It has been shown that NMDARs mediate LTP reversal (Kramár and Lynch, 2003;Zhou et al., 2003), similar to our findings, and NMDAR-specific antagonist APV significantly inhibited the reversal of LTP in DGCs (Figures 5A,B). NR2Aand NR2B-containing NMDARs are the major subtypes of NMDARs expressed in the hippocampus (Monyer et al., 1994), each playing distinct roles in different forms of synaptic plasticity; NR2A-containing NMDARs are required for LTP, whereas NR2B-containing NMDARs are required for LTD in rat brains (Liu L. et al., 2004;Massey et al., 2004). We found that the NMDAR subunit NR2A antagonist PEAQX obviously depressed the reversal of LTP in adult hippocampal mature DGCs and in adult-born 6-week-old DGCs (Figures 5C-F).
Beside NMDA receptors, mechanisms underlying LTP reversal are presumably related to the reversal of cellular changes associated with LTP induction. AMPA receptors are phosphorylated at Ser831 on GluR1 subunit by CaMKII. Such phosphorylation is responsible for the increase in the conductance of AMPA receptors after LTP (Benke et al., 1998). The LTP-induced phosphorylation of Ser831 is reversed during LTP reversal (Lee et al., 2000;. Incubation of the tectal neuron with a selective PP-1/2A blocker okadaic acid eliminated the reversal of LTP in a developing visual system (Zhou et al., 2003). The similar phenomena are shown in the hippocampus (O'Dell and Kandel, 1994;Huang et al., 1999;Kang-Park et al., 2003). Taken together, the interaction between kinases and phosphatases can determine whether LTP is induced and how long it will last. If the phosphatase activity overwhelms that of kinases, LTP can be reversed. The possibility remains that additional mechanisms may also contribute to LTP reversal. Application of metabotropic glutamate (mGlu) receptors with mGlu5 subtype-selective antagonists prevented the LTP reversal in the hippocampus (Qi et al., 2013). Increasing extracellular adenosine may underlie the LTP reversal by inhibiting the cAMP-dependent signaling cascade induced LTP via the A1 receptor subtype coupled to Gi/o proteins to prevent the PKA signaling (Huang et al., 1999).
A previous study reported that reversal of LTP by spontaneous activity may serve as a protective mechanism against persistent synaptic changes triggered by an incidental insignificant event in the developing visual system (Zhou et al., 2003). Therefore, we speculated that LTP reversal may function as an eraser of LTP to modify synaptic plasticity, which may mediate forgetting. NR2Amediated reversal of LTP was impaired in the hippocampal adult-born DGCs, which might cause adult-born DGCs not eliminate the subsequent transient memory; namely, adult-born DGCs could resist the incidental episodes of uncorrelated activity to form stabilized LTP. Similar to LTP reversal deficiency in juvenile mice, juvenile animals learn faster than adult animals. A recent report showed by using hippocampus-dependent object-location memory and contextual fear conditioning tasks confirmed that juvenile mice displayed deficits in long-term memory retention compared with adult mice (Tsai et al., 2018), which was probably because juvenile mice have insufficient memory encoding (Ge et al., 2019).
Synaptic plasticity dynamically changes during adult neurogenesis (Ge et al., 2007). Adult-born DGCs mediate pattern separation, whereas mature DGCs facilitate pattern completion (Clelland et al., 2009;Nakashiba et al., 2012). Our work indicates that LTP reversal is impaired in hippocampal adult-born 3-week-old DGCs, which may make it easier for newborn DGCs to build reliable connections between adult-born cells to participate in memory consolidation (Kitamura et al., 2009). It was reported that LTP reversal disappeared in a schizophrenia mouse model (Shamir et al., 2012). Schizophrenia may be caused by subtle differences in neural development leading to erroneous neuronal connections (Lewis and Levitt, 2002). In addition, a recent report claimed that hippocampal adult-born DGCs were reduced and working memory was impaired in a schizophrenia model (Ouchi et al., 2013). We wondered whether deficiency of LTP reversal in adult-born DGCs of adult mice contributes to cognitive and behavioral impairments in schizophrenia. It remains unknown whether impaired adult hippocampal neurogenesis is a causal factor underlying relevant pathology, including schizophrenia. Nevertheless, targeting adult-born neurons could be a potential therapeutic strategy for schizophrenia (Christian et al., 2014). Additional studies are needed to elucidate the molecular mechanisms of LTP reversal in adult-born DGCs and their relationship with forgetting and schizophrenia.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

ETHICS STATEMENT
The animal study was reviewed and approved by the Committee of Animal Care of Huazhong University of Science and Technology.