Spine Formation Pattern of Adult-Born Neurons Is Differentially Modulated by the Induction Timing and Location of Hippocampal Plasticity

In the adult hippocampus dentate gyrus (DG), newly born neurons are functionally integrated into existing circuits and play important roles in hippocampus-dependent memory. However, it remains unclear how neural plasticity regulates the integration pattern of new neurons into preexisting circuits. Because dendritic spines are major postsynaptic sites for excitatory inputs, spines of new neurons were visualized by retrovirus-mediated labeling to evaluate integration. Long-term potentiation (LTP) was induced at 12, 16, or 21 days postinfection (dpi), at which time new neurons have no, few, or many spines, respectively. The spine expression patterns were investigated at one or two weeks after LTP induction. Induction at 12 dpi increased later spinogenesis, although the new neurons at 12 dpi didn’t respond to the stimulus for LTP induction. Induction at 21 dpi transiently mediated spine enlargement. Surprisingly, LTP induction at 16 dpi reduced the spine density of new neurons. All LTP-mediated changes specifically appeared within the LTP–induced layer. Therefore, neural plasticity differentially regulates the integration of new neurons into the activated circuit, dependent on their developmental stage. Consequently, new neurons at different developmental stages may play distinct roles in processing the acquired information by modulating the connectivity of activated circuits via their integration.


Introduction
Many works have clarified that new neurons are continually being generated in the adult hippocampus (HP) of different mammalian species including monkey and human [1], although most results have been proposed from model animals, especially rodents. The dentate gyrus (DG) of the HP integrates newly born neurons throughout adult life. Adult-born neurons are required for various types of brain function including HP-dependent memory [2][3][4][5]. Numerous factors associated with various behavioral and cognitive states of animals regulate adult neurogenesis in the DG, and HP-dependent learning is one of the major regulators of this neurogenesis [5]. Learning of HP-dependent tasks, but not HPindependent tasks, enhances neurogenesis in DG [5,6]. In addition, HP-dependent learning has been suggested to selectively add and remove new neurons according to their maturity and functional relevance [7,8].
In the central nervous system, postsynaptic spines on neuronal dendrites interact with presynaptic axonal terminals, and the majority of glutamatergic excitatory inputs are received by the dendritic spines of postsynaptic neurons [9]. The dendritic spine is a critical site for synaptic plasticity. Therefore, spine expression pattern is one indicator of the integration of neurons into excitatory synapses. Retrovirus (RV)-based gene transfer is a useful methodology to mark progenitor cells and their progeny in the DG because the viral genome is only integrated into proliferating cells [10]. Therefore, green fluorescence protein (GFP)-RV has been used to visualize the morphology of newly born neurons derived from DG progenitor cells [11]. GFP-RVmediated labeling has revealed distinct morphological stages of newly born neurons during development [11]. Their spine expression pattern is closely correlated with the excitatory postsynaptic response of developing newly born neurons in adult DG [11][12][13][14].
Spatial learning is dependent on HP and influences neurogenesis in adult DG [6][7][8]. Notably, training in spatial learning by water maze during new neurons' second week of age enhances the complexity of their dendritic arbor and spine formation rate [15]. Thus, new neurons pass through several developmental stages, during which each neuron could be differentially integrated into circuits in an experience-dependent manner. However, it remains unclear whether the later synaptic integration pattern of new neurons is differentially modulated by the timing of experiences, and whether the integration is regulated locally in a synaptic inputspecific manner or as a cell-wide event.
Long-term potentiation (LTP) is a designated model of activitydependent synaptic plasticity. One type of learning has been reported to induce LTP in HP and produce the same change in mechanisms underlying the LTP induction of excitatory synapses [16]. Therefore, LTP is now widely accepted as a critical component of neural mechanisms underlying learning and memory [17,18]. DG LTP induction in freely moving animals has the advantage of allowing the long-lasting effects of synaptic plasticity induction on various phenomena to be observed. In DG, each neuron consisting of a granule cell (GC) receives two inputs from the entorhinal cortex, via the medial and lateral perforant path (MPP and LPP, respectively). MPP and LPP share the lamination domain of the molecular layer (ML), comprising its middle third (MML) and its outer third (OML), respectively [19][20][21] (Figure 1A). Newly born neurons also form functional synapses with the MPP and LPP [22,23].
High-frequency stimulation (HFS) of MPP or LPP fibers specifically results in LTP at the MML or OML synapses, respectively [21]. Establishment of layer (input)-specific long lasting late LTP (L-LTP) is mediated by a synaptic tagging mechanism underlying the input-specific function of proteins at activated synapses that are newly synthesized at soma [24,25].
In this paper, we address the questions of whether the LTP induction at new neurons' different ages affects later spine expression patterns in an induction-timing-dependent manner, and whether these effects are restricted to the area where LTP is induced or if it is spread cell-wide. Our results indicate timingdependent and area-specific regulation of spine expression patterns.

Animals
All procedures involving the use of animals complied with the guidelines of the National Institute of Health and were approved by the Animal Care and Use Committees of Mitsubishi Kagaku Institute of Life Sciences and University of Toyama. Male Wistar ST rats (Japan SLC Inc., Shizuoka, Japan) approximately 20 weeks of age were used for retroviral labeling and LTP experiments.

Retrovirus Production
The pMXs-SIN-CAG-GFP was developed from RV vector pMXs [26]. First, a PvuI-SacI fragment was deleted from pMXs, followed by blunting the SacI site using a Blunting Kit (Takara, Japan), and ligating PvuI and the blunted SacI. The resulting vector pMXs-SIN (self-inactivating) was linearized by PacI in the 59-multi-cloning site and blunted as above, where the CAG promoter was inserted. Finally, a GFP cDNA was inserted downstream of the CAG promoter using the EcoRI and NotI sites.
The pMXs-SIN-CAG-GFP-b-actin was prepared as follows. Rat b-actin cDNA prepared from rat PC-12 cell mRNAs was kindly donated by Prof. M. Takahashi, Kitasato University, Japan. The b-actin open reading frame cDNA was subcloned into EcoRI and BamHI sites of pEGFP-C2 (Clontech, Mountain View, CA). The EGFP-b-actin fragment was generated by Eco47III and BamHI digestion, and then replaced with the GFP fragment of the pMXs-SIN-CAG-GFP vector. The pMXs-SIN-CAG-GFP or -GFP-b-actin was co-transfected with pVSV-G (Clontech) into PLAT-gp cells, counterparts of PLAT-E cells lacking the transgene for envelope [27], by FuGENE HD (Roche Applied Science, Indianapolis, IN). RV-containing culture supernatants were collected after 2 days, and the RV was concentrated by ultracentrifuge at 5610 4 6g for 90 min at 4uC. The RV pellet was resuspended to 0.5% of original volume in phosphate-buffered saline (PBS).

Labeling of New Neurons
For high accuracy injection of RV solution into DG, the needle tip of a Hamilton syringe was shaped as shown in Figure 1B using an electrical grinder. Rats were anesthetized with pentobarbital (55 mg/kg, i.p. injection), and 3 ml per hemisphere of RV solution was infused bilaterally into the DG positioned 3.6 mm posterior; 61.9 mm lateral, and 4.2 mm ventral to bregma (0.5 ml/min).
Procedure to label newly born cell with bromodeoxyuridine (BrdU) was described previously [4]. Briefly, rats were injected with BrdU in 0.9% NaCl solution (100 mg/kg, i.p., Sigma) twice a day (6 h interval) for three consecutive days, and were perfused 27 days later the last injection.

DG LTP in Unanesthetized Freely Moving Animals
The surgical procedure was as described previously [21]. After the RV injection procedure, the electrode stimulating the MPP fibers was positioned 8.7 mm posterior, 5.3 mm lateral, and 5.3 mm ventral to bregma. A recording electrode was implanted ipsilaterally 4.0 mm posterior, 2.5 mm lateral, and 3.8 mm ventral to bregma. Electrode placement was only performed on the right ( = ipsilateral) hemisphere.
LTP experiments on freely moving animals were performed as described previously [21,28]. LTP was induced by tetanic stimuli with biphasic square waveform, 200-ms pulse width. Maximal population spike (PS) amplitude was determined, and the intensity of the stimulus current was set to elicit 60% of the maximal PS amplitude. This intensity was used for baseline recording and HFS(500) experiments. The animals were transferred to the recording chamber, and the baseline response was monitored by delivering test pulses (0.05 Hz) for 15 min. After the baseline monitoring, LTP was induced by high frequency tetanic stimulation, HFS(500), consisting of 10 trains with 1-min intertrain intervals. Each train consisted of five bursts of 10 pulses at 400 Hz, delivered at 1-s interburst intervals. Synaptic transmission was monitored for 15 min immediately, 1 day, 7 days, 12 days, and 16 days after HFS(500) stimulation. The local induction at MML in each animal was confirmed by F-actin staining.
In CPP experiments, the baseline monitoring was performed 1 day before HFS(500). CPP was dissolved in saline (0.9% saline), and the rats were injected i.p. with 10 mg/kg CPP 2 h before the beginning of HFS(500) stimulation. Synaptic transmission was monitored for 15 min at 1 day and 7 days after HFS(500) stimulation.

Histochemistry
Rats were deeply anesthetized with an overdose of pentobarbital solution and perfused transcardially with PBS, pH 7.4, followed by 4% paraformaldehyde (PFA) in PBS. The brains were removed and further post-fixed by immersion in 4% PFA in PBS for 2 h at 4uC. Each brain was equilibrated in 25% sucrose in PBS, then frozen in dry-ice powder. For F-actin staining, coronal sections (14-mm thickness) were cut on a cryostat and washed with PBS. The sections were incubated with phalloidin-TRITC (0.1 ng/ml) at 4uC for 24 h. The sections were treated with DAPI (1 mg/ml) or DRAQ5 (1:5000) and then washed with PBS three times, 10 min/ wash. For GFP staining, coronal sections were cut on a cryostat at 50-mm thickness, 3.1 mm to 4.9 mm from bregma (36 sections total) and transferred to 12-well cell culture plates (Corning, Corning, NY) containing PBS. After washing with PBS, the floating sections were treated with PBST (PBS supplemented with 0.5% Triton X-100) at room temperature (RT) for 20 min, followed by two 10-min washes with PBS. The sections were then treated with blocking buffer (3% bovine serum albumin in PBS) at RT for 1 h. Reactions with primary antibodies were performed in blocking buffer containing rabbit anti-GFP (1:1000) and mouse anti-synaptophysin (1:5000) antibodies at 4uC overnight. After three 10-min washes with PBS, the sections were incubated with FITC-conjugated anti-rabbit and AlexaFluor 546-conjugated antimouse IgG secondary antibodies at RT for 3 h. Sections were treated with DAPI (1 mg/ml) or DRAQ5 (1:5000) and then washed with PBS three times, 10 min/wash. For BrdU staining, coronal sections were cut on a cryostat, and every other 20 mm coronal section was collected from 3.0 mm to 5.0 mm from bregma (50 sections total). The sections were boiled with 0.01 M sodium citrate buffer (pH 6.0) for 10 min, treated with 2 M HCl for 30 min, and rinsed in 0.1 M boric acid (pH 8.5) for 10 min, as described previously [29]. Sections were blocked with 5% donkey serum in PBS containing 0.1% Triton X-100 at room temperature for 1 h. After blocking, sections were incubated Mounting of sections on slide glasses was performed with ProLong Gold antifade reagents (Invitrogen). The fluorescent signals were examined with a laser-scanning confocal microscope (LSM5 PASCAL or LSM700 ZEN, Carl Zeiss, Jena, Germany).

Immunoelectron Microscopy
Rats were deeply anesthetized with an overdose of pentobarbital solution and transcardially perfused with 4% PFA and 0.05% glutaraldehyde in PBS. The brains were postfixed for 48 h in 4% PFA, equilibrated in 25% sucrose in PBS, then 50-mm coronal vibratome sections were cut. The sections were cryoprotected in 30% sucrose in 0.1 M phosphate buffer (PB) for 2 h and freezethawed two times in liquid nitrogen. Immunostaining was performed with an avidin-biotin immunoperoxidase using a VECTASTEIN Elite ABC kit (Vector Laboratories, Burlingame, CA). The sections were then treated with blocking buffer (3 drops of normal goat serum solution of VECTASTEIN Elite ABC kit in 10 ml PBS) at RT for 1 h. Reactions with primary antibodies were performed in blocking buffer containing rabbit anti-GFP (1:1000) antibody at 4uC overnight. After washing in PBS, the sections were incubated for 3 h at RT in biotinylated secondary antibody (1 drop of goat antibody to rabbit IgG of VECTASTEIN Elite ABC kit in 10 ml blocking solution). To reveal this labeling, we incubated sections at 4uC overnight in avidin-biotin peroxidase complex (VECTASTEIN Elite ABC kit). After washing in 0.1 M PB, the sections were visualized with the 3,39-diaminobenzidine tetrachloride with nickel (DAB substrate kit, Vector Laboratories) for 5 min. The sections were then postfixed with 1% glutaraldehyde in 0.1 M PB for 10 min, washed in 0.1 M PB, postfixed with osmium tetroxide in 0.1 M PB for 1 h, washed in water, and then treated with 5% uranyl acetate in 50% ethanol for 30 min. The sections were dehydrated by passing through a graded series of ethanol and propylene oxide, and embedded in Epon 812. Ultra thin sections were cut (about 150 nm thick) and observed by tilting

Data Analysis
Quantitative measurements of F-actin levels and synaptophysin expression were defined by average signal intensity using Metamorph Software (Molecular Devices, Downingtown, PA).
For quantitative measurements of spine density and crosssectional area, analyses were performed with neurons that were relatively isolated from neighboring GFP+ neurons and had untruncated dendrites from MML through OML as Figure 1C. Images of GFP-actin signal were acquired from a clearly extending dendritic fragment of individual cell without intersecting with other GFP+ dendrites and at the center of middle third and outer third of ML under the condition of 0.7-mm intervals using a Plan-Apochromat 636/1.4 oil lens and digital zooms of 610 and 67 with LSM510 (for Figures 1, 2, 3, 4) and LSM700 (for Figure 5) laser-scanning confocal microscopes, respectively (Carl Zeiss, Jena, Germany). Z-stack images were created by LSM image browser or ZEN 2009 software with maximum intensity projections of confocal z-series (Carl Zeiss, Jena, Germany). To define spine density, the number of spines was counted manually, and the length of dendritic segments was determined by tracing the center of the dendritic shaft with the ''Distance'' function of Metamorph software (Molecular Devices, Downingtown, PA). To define the cross-sectional area of spines, each z-stack image was binarized by appropriate thresholding to bring out the strong edges while minimizing noises and false edges [30]. The threshold for each images were set to be almost same average background pixel number. The edge of each spine was manually traced in Metamorph software (Molecular Devices, Downingtown, PA), and then the area was determined in pixel numbers with the Metamorph ''Area'' function. The cross-sectional area of each spine was calculated based on the obtained number of pixels and the reduced scale.
For quantitative measurements of BrdU + and BrdU + Zif268 + cells in the dorsal hippocampus, all BrdU + cells, regardless of size or shape, were counted using a 406 objective (BX41, OLYM-PUS), as described previously [29]. A cell was counted if it adjoined the subgranular zone (SGZ) or was positioned in the SGZ or GC layer (GCL) (excluding the hilus region), as described previously [29,31,32].
All statistical analyses were performed using StatView Software (Abacus Concepts, Berkeley, CA). Comparisons between two-group data were analyzed by Student's t-test (two tailed). If the data did not meet the assumptions of the t-test, the data were analyzed using the Mann-Whitney U-test (two tailed). Multiple-group comparisons were assessed using a one-way analysis of variance (ANOVA), followed by the post-hoc Scheffe's test when significant main effects were detected. If the data met the assumptions, they were analyzed using the post-hoc Fisher's test. The null hypothesis was rejected at the P,0.05 level. Quantitative data in this study are shown as mean 6 SEM.

Spinogenesis of New Neurons for the First 4 Weeks after Birth in the Rat Adult DG
Because the actin filament (F-actin) is the major cytoskeletal structure in dendritic spines [33,34], GFP-b-actin-RV provides a clear image of the spines of newly born neurons. We constructed a compatible system of spine labeling of newly born neurons in DG of adult rats by GFP-b-actin-RV infection and in vivo LTP induction. To observe spine development, newly born neurons were retrovirally labeled with GFP-b-actin ( Figure 1C). GFPactin allows visualization of spines without altering their morphology [34,35] and enters all spines [36]. At 12 days postinfection (dpi), no spines were observed, and the distal tips of dendrites did not expand beyond the middle part of the ML in all cells observed (n = 17 neurons from 2 animals) ( Figure 1C, panels a and d, and Figure 1F). Few spines were found at 16 dpi (spine density: 3.9360.44/10 mm from 2 animals, n = 26 dendritic fragments) ( Figure 1C, panels b, e, and f, and Figure 1F), similar to previous reports in mice [11] and rats [37]. Spinogenesis dramatically progressed from 16 to 18 dpi (spine density at 18 dpi: 14.7460.95/10 mm, n = 29 dendritic fragments from 2 animals; 16 vs. 18 dpi, P,0.001, Mann-Whitney U-test) ( Figure 1C, panels c, g, and h, and Figure 1F). Moreover, the first dramatic progression of spinogenesis nearly reached plateau by 3-4 weeks post-infection ( Figure 1C, panels i and j, and Figure 1F), showing interaction with presynaptic terminals ( Figure 1D). These observations were similar to previous reports in mice [11,22]. Moreover, immunoelectron microscopy revealed that about 80% of GFP-actin-positive spines on 28 dpi neurons contained postsynaptic density and contacted with synaptic vesicles containing structure (78.161.9%, n = 3 neurons, total 90 GFP+ protrusions) ( Figure 1E, panels a and b). Other protrusions did not satisfy the criteria of functional synapses ( Figure 1E, panel c).
Next, to determine whether spine development reflects responsiveness to excitatory input, LTP-inducible stimulation, HFS(500), was delivered in ipsilateral MPP-GC synapses ( Figure 1A) at 12 and 28 dpi, and brains were then dissected 1 h after the stimulation ( Figure 1G). The HFS(500) in the MML increased the expression rate of Zif268, a product of an immediate early gene, in the GC layer (GCL) ( Figure 1H) where GC somata localize ( Figure 1A, C). Zif268 expression was significantly increased by HFS in GFP-positive neurons at 28 dpi (control hemisphere, 15.55% 64.73%; ipsilateral ( = LTP) hemisphere, 52.25% 66.26; P,0.01, Student's t-test) ( Figure 1I, J). By contrast, no detectable signals in GFP-positive neurons were observed in either control or ipsilateral hemispheres at 12 dpi ( Figure 1I, J). These results indicate that 4-week-old neurons have features of excitatory neurons. In this study, we therefore mainly observed the spine formation pattern of new neurons at 28 dpi to be an indicator of integration pattern, at which time new neurons passed through the first phase of dramatic development.

Induction of LTP at the No-spine Stage of New Neurons Specifically Enhances their Later Spinogenesis in the MML
New neurons have no, few, or many spines at 12, 16, or 21 dpi, respectively ( Figure 1C, F). LTP was induced at MPP-GC synapses at each representative stage of new neurons' spinogenesis, and, at first, the spine expression pattern was observed at 28 dpi ( Figure 2A). First, LTP was induced in the ipsilateral MML at 12 dpi by delivering 500 pulses of HFS (HFS(500)) to the MPP. Potentiation of population spike (PS) amplitude persisted for 16 days (Figure 2B), field excitatory postsynaptic potential (fEPSP) slope was maintained at least for 7 days ( Figure S1). LTP induction at the DG in vivo is associated with actin cytoskeleton reorganization, including the long-lasting increase of F-actin content within spines [21]. Thus, the layer specificity of the HFS(500) was monitored by the significant increase of F-actin signal in the MML of the HFS(500)-treated hemisphere. F-actin was quantified by histochemical analysis at 28 dpi using phalloidin, a F-actin-specific probe (in the same animals used for spine investigations below, control side, F (2, 6) = 1.22, P.0.360, ANOVA; LTP side, P,0.002, ANOVA; MML vs. IML, P,0.002; vs. OML, P,0.016; post-hoc Scheffe's test) (Figure 2C1, C2, G). Heterosynaptic LTD is formed in LPP synapses by LTP induction in MPP synapses [38]. We observed significant decrease of F-actin signal at IML in LTP side compared with control side (P,0.05, Student's t-test), and it may be mediated by the heterosynaptic depression which is induced by MML LTP, similarly with a case of OML. By contrast, presynaptic content revealed by synaptophysin signal in the MML and OML at 28 dpi was unchanged by MML LTP induction ( Figure 2D1, D2 Figure 2E, F, K). These results indicate that LTP induction at 12 dpi locally enhances later spinogenesis in new neurons within the LTP-induced layer by 4 weeks of age, although no spines are present at the time of LTP induction ( Figure 1C, F).
We also observed the spine expression pattern at 19 dpi (with LSM700, control hemisphere, n = 15 dendritic fragments, HFS hemisphere, n = 15 dendritic fragments from 3 animals). In this case, no significant difference in spine density were appeared between the LTP-induced layer and other layers (control MML, 13

Induction of MML LTP after the Drastic Spinogenesis Stage of New Neurons Results in MML-specific Spine Enlargement
Next, MML LTP was induced at 21 dpi by delivery of HFS(500) ( Figure 4A). Potentiation of PS amplitudes and fEPSP slopes were maintained at 28 dpi ( Figure 4B, Figure S1), and the increased Factin content in the MML of HFS(500)-treated DG was confirmed (in the same animals used for spine investigations below, control hemisphere, F (2,6) Figure 4E, F, K). These results indicate that LTP induction at 21 dpi induces spine enlargement specifically in the LTP-induced layer. This observation is similar to the responses of spines of mature CA1 neurons following LTP induction [39,40].
We also observed the spine expression pattern at 35 dpi, timing at 1 week later from the first observation (with LSM700, control hemisphere, n = 13 dendritic fragments, HFS hemisphere, n = 14 dendritic fragments from 2 animals). Although significant increase of the spine size in the LTP-induced layer was observed at 28 dpi, the difference of average spine size of new neurons disappeared at

NMDAR Activity during the Delivery of HFS to MPPs is Required for LTP Induction and all Changes in the Spine Expression Pattern of New Neurons
Because NMDA receptors (NMDARs) play a crucial role in memory formation and long-lasting neural plasticity including LTP [17], we finally investigated whether the changes in spine expression pattern of new neurons are mediated by NMDAR activity during LTP induction. The selective NMDAR antagonist 3-(2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP) was used to block NMDAR. CPP (10 mg/kg) was intraperitoneally (i.p.) injected 2 h prior to HFS(500) ( Figure 5A) at 12, 16, or 21 dpi, and the spine expression pattern of new neurons was then observed at 28 dpi. CPP application with this protocol clearly blocked LTP induction in DG ( Figure 5B), as described previously [29,41]. Similarly, F-actin signal intensity at 28 dpi within the ipsilateral DG MML showed no change compared with other layers after HFS delivery at 12, 16, or 21 dpi (HFS at 12 dpi, control hemisphere, F (2,6) Figure 5I, J, K). Therefore, the differential regulation of later spine expression pattern observed in this study was mediated by the activity of NMDAR at the time of HFS delivery.

Induction of LTP in New Neurons Approximately 12 Days in Age Enhances their Later Functional Integration into Preexisting Circuits
To evaluate functional integration of the 4-week-old new neurons that underwent LTP induction 16 days prior, we investigated the rate of circuit activity-dependent Zif268 expression in BrdU + cells. We established three experimental groups that were injected with BrdU for 3 consecutive days: ''HFS 12d'' and ''HFS 28d'', in which LTP was induced 12 or 28 days after the second day of BrdU injection, respectively ( Figure 6A, B, C), and ''HFS/HFS'', in which LTP was induced at 12 and 28 days after the second day of BrdU injection ( Figure 6A, D).
BrdU is incorporated into the newly synthesized DNA of dividing cells. The majority of 4-week-old BrdU + cells that were localized in or attached to the GCL express NeuN, a marker of mature neurons [29,31,32]. All BrdU + Zif268 + cells were NeuN + ( Figure 6E). We mainly focused on differences in the Zif268 induction ratio between the ipsilateral and control hemispheres of each animal. In ''HFS 12d'', no significant difference in the Zif268 expression rate of 4-week-old BrdU + cells was observed between control and ipsilateral DGs (control side, 11.3% 63.3; ipsilateral side, 15.2% 62.9, P.0.42, Student's t-test). By contrast, in the ''HFS 28d'' and ''HFS/HFS'' groups, LTP induction at day 28 significantly enhanced Zif268 expression in 4-week-old cells (HFS  Figure 6F). HFS(500) induced at any time elicits no change in the ratio of neuronal differentiation in 4-week-old BrdU + cells [29,31]. In addition to the spine expression pattern results shown in Figure 2, these previous findings and our data strongly suggest that when LTP is induced in neurons at approximately 12 days of age, their later functional integration into activated circuits is enhanced.

Developmental Stages of Newly Born Neurons
In the hippocampal DG, new neurons are continually generated and integrated into preexisting circuits throughout adulthood. The spine expression pattern is closely correlated with the excitatory postsynaptic response of developing newly born neurons in adult DG [11][12][13][14]. Our observations of the dendritic morphogenesis of newly born neurons labeled by RV-mediated expression of GFPactin identified several developmental stages of spine formation in adult rats. The initial timing of spinogenesis occurred in new neurons aged at ,16 days ( Figure 1C, F), which is consistent with previous reports both in mice [11] and rats [37]. In the DG of adult mice, most dendritic protrusions of new neurons labeled by RV are formed between 21 and 30 dpi [11,22]. However, our results indicated that spinogenesis increased sharply at around 18 dpi ( Figure 1C, F). Thus, the dramatic progression of spine growth occurs about 1 week earlier in rat DG than mouse DG. New neurons in adult rats have been reported to show a mature neuronal marker profile by 4 weeks after birth, which is 1-2 weeks earlier than mice [32]. Immunoelectron microscopy revealed that about 80% of spines on 28 dpi neurons meet structural requirements for excitatory synapse ( Figure 1E). Accordingly, the date of our observation, 28 dpi, is appropriate for investigating the effects of earlier experiences during developmental on the spine patterns of mature new neurons in adult rat. In adult mice, gradual development of spines and lower LTP induction threshold are still observed in new neurons up to 180 days and 6-8 weeks of age, respectively [11,42]. Thus, new neurons pass through several developmental stages, during which each neuron could be differentially integrated into preexisting circuits in an experiencedependent manner.

Effects of LTP at 12 dpi on Later Spine Formation Pattern
Interestingly, we observed that LTP induction at 12 dpi specifically increased spine expression rate and spine size within the LTP-induced layer by 4 weeks but not 3 weeks of neuronal age (Figures 2, 7A), although no spines were present at the time of LTP induction ( Figure 1C, F). This observation is quite similar to the effect of HP-dependent learning in new neurons at the second week of age [15]. Thus, these new neurons seem to be preferentially integrated later into neural circuits in which synaptic plasticity had been induced ( Figure 7A).
How does LTP induction at 12 dpi influence later spine expression pattern when 12-dpi neurons have no spines? Spatial learning facilitates the long-term survival (.1 month) of 2-weekold new neurons in mice [43]. The surviving new neurons have also been suggested to preferentially incorporate into the neural circuits that encode the learned behavioral experiences [43]. Therefore, LTP induction and experience of spatial learning may similarly tag some molecular traces that are required for changes in synaptic morphology and physiology and would later enhance the spine formation of new neurons. Some candidate molecular traces are found in extracellular matrix. For example, hippocampal matrix metalloproteinase-3 (MMP-3) and -9 (MMP-9) are transiently increased during hippocampal-dependent learning in an NMDAR activity-dependent manner, and inhibition of their activity prevents LTP and hippocampal-dependent learning [44]. Neural cell adhesion molecule (NCAM) has been implicated in synaptic plasticity including LTP and learning and memory [45]. DG LTP induction in the perforant path (PP) synapses increases NCAM expression at the spine synapses of GCs in the LTPinduced layer [45]. In contrast, HP progenitors and new neurons initially receive depolarizing GABAergic input before glutamatergic synapses are established. Therefore, mechanisms for activitydependent regulation of adult neurogenesis may sense neuronal network activity through local ambient GABA levels before forming glutamatergic synapses. However, inhibition of GABA A receptor from 14-27 dpi does not exert significant effects on the spine density or dendritic length of 28-day-old new neurons in mice [46]. This finding suggests a minor role of GABAergic signaling in the maturation process of new neurons after their second week. Alternatively, although 12-dpi neurons have no spines, about 30% of 12-day old neurons in adult rats receive glutamatergic inputs in an in vitro slice preparation [47], which may lead to subsequent spine growth. Future studies are needed to identify and clarify the mechanisms underlying later enhancement of spinogenesis in new neurons.
What is the functional relevance of the enhanced later integration of new neurons into LTP-induced circuits? Spine density possibly correlates with the available postsynaptic capacity. Repeated delivery of HFS to PP synapses results in impaired spatial learning in rats [48], suggesting strongly that the HP capacity for information processing is saturated by repeated HFS. Therefore, HFS(500) applied to the MPP in this study may decrease postsynaptic capacity of preexisting circuits in the LTPinduced layer. Given this situation, continuous and increased later integration of new neurons after LTP induction possibly compensates for the decreased processing ability for new information within the LTP-induced layer. Our present data suggest that later functional integration is specifically enhanced in activated circuits when LTP is induced in new neurons at approximately 12 day of age (Figures 2, 6). In addition to the increase in spine density and enhancement of functional integration, it is predicted that new neurons willingly process the new input based on their features of increased excitability and lower LTP induction threshold [42,49]. Indeed, spine size had been already enlarged in LTP induced layer 1 week after induction whereas the spine density was not changed between control and LTP-induced layer ( Figure 7A). Enlarged spines are implicated as sites of acquisition of information, because of correlation between spine volume and number of functional AMPA receptors [39,40].
Alternatively, the enhanced integration of new neurons after LTP induction may promote the gradual decay of DG LTP. DG LTP is reversed by exposure of rats to an enriched environment after LTP induction [50]. Blockade of NMDAR activity after LTP induction leads to slower decay of DG LTP [51]. Both enriched environment and NMDAR activity positively regulate adult neurogenesis [52][53][54]. In addition, the inhibition of neurogenesis in DG sustains LTP of the fEPSP slope [4]. Integration of new neurons has been suggested to have an impact on the wiring pattern of preexisting circuits [55] due to the likelihood of a competitive situation between new neurons and preexisting GCs [22]. We did not observe an increase in presynaptic content, as measured by synaptophysin labeling, within the LTP-induced layer ( Figure 2D, H). Finally, LTP of the fEPSP slope returns to basal levels 2 weeks after induction ( Figure S1) [4]. Taken together, a likely scenario can been imagined in which new protrusions from new neurons actively invade into LTP-induced preexisting synapses more frequently than normal (non-LTP) synapses at the timing of initial spinogenesis at around 16-18 dpi, several days after the LTP induction. These protrusions would then synapse with presynaptic boutons that have already made synaptic contact with spines of mature neurons, gradually enlarge, and finally win against the older spines, leading to the gradual LTP decay. Thus, time-lagged spinogenesis of new neurons may play an active role in specific and positive renewal of the DG circuits that have acquired and stored old information. The time courses of the maturation and functional integration of new neurons and the progressive decay of the hippocampal dependency of hippocampus-dependent memory are quite similar (,1 month) in rodents. The decay of the memory process in hippocampus is positively regulated by adult neurogenesis in mice [4]. In monkeys, the maturation time of new neurons is approximately 6 times longer than that in rat [56], and the retention time of memories in the hippocampus is approximately 3 months [57]. The correlation of time windows between the processes of adult neurogenesis and the hippocampal dependency of acquired memories in rodents and monkeys may suggest that adult neurogenesis in primates selectively participates in information clearance from preexisting circuits of the dentate gyrus, even if the rate of adult neurogenesis is lower than rodents [1].

Effects of LTP at 16 dpi on Later Spine Formation Pattern
LTP induction at 16 dpi inhibited later spine expression specifically within the LTP-induced layer (Figures 3, 7B). To our knowledge, this result is the first report of the negative regulation  Figure 1F) and black circles, respectively. Right, Graph shows the mean values of ipsilateral (Ipsi)/contralateral (Contra) ratio of spine cross-sectional area in MML and OML. Abbreviations: GCL, granule cell layer; IML, MML, and OML, inner, middle, and outer molecular layer, respectively. doi:10.1371/journal.pone.0045270.g007 of spine expression by neural activity. Thus, the question arises as to how LTP induction at 12 dpi and 16 dpi direct opposite patterns in spine expression at 28 dpi. Differences in developmental stage may underlie this effect. A sharp increase in spine expression occurs at 16 dpi ( Figure 1C, F). LTP induction should increase the spine volume of mature neurons [58], and the enlarged state may persist for at least 24 h in the DG [41]. In this situation, synaptic contact at the LTP-induced synapses of mature neurons should be tight immediately following LTP induction, and there may be no space for filopodia budding by new neurons. Indeed, electron microscopy analysis suggests that new neurons in the adult DG compete with existing GCs at excitatory synapses [22]. Thus, competitive bias from the mature neurons at activated synapses inhibits the later spine expression of new neurons. The 12-dpi neurons begin to protrude spines several days after LTP induction. DG LTP of the fEPSP slope elicited by HFS(500) decays gradually, and the magnitude of LTP at 1 week is about half the initial magnitude [4]. Therefore, this competitive bias does not inhibit the spinogenesis of the 12-dpi neurons.

Effects of LTP at 21 dpi on Later Spine Formation Pattern
The neurons in which LTP was induced at 21 dpi passed the period of sharply increased spinogenesis, and these neurons had a number of spines on their dendritic shafts ( Figure 1C, F). In rats, the expression of neural activity marker Zif268 is detectable in new neurons from day 16 of age in a LTP induction-dependent manner [31]. Thus, the 21-dpi neurons were already functionally integrated into preexisting DG circuits at PP synapses. LTP induction at 21 dpi induced only spine enlargement of new neurons in the LTP-induced layer (Figures 4, 7C). DG LTP is associated with an increase in spine volume without any change in spine density on GCs [58]. Thus, the 21-dpi and preexisting mature neurons show a quite similar response to LTP induction. Actually, the spine enlargement returned to control level by 2 weeks after LTP induction ( Figure 7C), and the time course is quite similar with decay of LTP of the fEPSP slope ( Figure S1) [4].
By contrast, spine size in the OML ipsilateral to the LTPinduced DG showed tendency to decrease compared with both layers of the contralateral side ( Figure 4C, D, and I). Long-term depression (LTD) is another type of neural plasticity, and LTD of hippocampal neurons accompanies shrinkage of dendritic spines [59]. Heterosynaptic LTD is formed in LPP synapses by LTP induction in MPP synapses, and this plastic change persists for 2 weeks [38]. Thus, heterosynaptic OML LTD may be coincidently induced in 21-dpi neurons by the MML LTP induction in a similar manner to mature GCs.

Timing-and Layer-specific Effects
Morris water maze, a HP-dependent spatial learning task, promotes the survival of new neurons approximately 10-days-old and enhances the cell death of younger neurons about 7-days-old [8]. Similarly, LTP induction specifically enhances the survival of new neurons at around 10 days of age, but not at 4 or 20 days of age [29]. We observed LTP timing-dependent bidirectional regulation of the spine formation pattern of new neurons, in which LTP differentially enhanced or suppressed later spine formation, dependent on new neurons' developmental stages (Figures 2, 3, 4, 7). LTP and HP-dependent learning may selectively add or remove new neurons in the DG according to their maturity by regulating their spine formation pattern. Consequently, the different stages of new neurons' development could possibly play various roles in information processing by modulating the connectivity of preexisting circuits.
Morris water maze training promotes the later spine expression of new neurons in their second week after birth [15]. However, whether the effect is cell-wide or restricted to certain spines had remained unclear. In the present study, we have shown that MML LTP directs various patterns of later spine expression by new neurons only in the MML. Therefore, LTP does not affect later cell-wide spine expression; instead, it locally influences synapse formation at the restricted area that has received synaptic inputs.

NMDAR Dependence
NMDAR activity is required for spatial learning and induction of long-term synaptic plasticity at various HP synapses [60] including LTP of PP-GC synapses ( Figure 5B) [29,61]. Blockade of NMDAR activity at the time of LTP induction inhibited all changes in the spine expression of new neurons observed in this study ( Figure 6). These results strongly suggest that activitydependent neural plasticity, but not a simple transient increase in neural activity, in DG circuits is required to alter the later spine expression pattern of new neurons.
In this paper, we addressed the questions of whether the hippocampal LTP induction affects later spine expression patterns of new neurons in an induction-timing-dependent manner, and whether these effects are restricted to the area where LTP is induced. Our results indicate that LTP differentially regulates the integration of new neurons into the activated circuit, dependent on their developmental stage. Consequently, new neurons at different developmental stages may play distinct roles in processing the acquired information by modulating the connectivity of activated circuits via their integration.