ReportsDevelopment

A disynaptic feedback network activated by experience promotes the integration of new granule cells

See allHide authors and affiliations

Science  28 Oct 2016:
Vol. 354, Issue 6311, pp. 459-465
DOI: 10.1126/science.aaf2156

Integration of adult-born brain cells

Physical exercise or exploration of a novel environment greatly influences the production, maturation, and connectivity of adult-born neurons. Alvarez et al. investigated how experience affects the incorporation of adult-born neurons into the hippocampal network. A brief period of sensory enrichment when new neurons were 9 to 10 days old led to neurons having larger dendrites and more functional spine synapses. A disynaptic preexisting feedback circuit promoted the growth and integration of the new cells.

Science, this issue p. 459

Abstract

Experience shapes the development and connectivity of adult-born granule cells (GCs) through mechanisms that are poorly understood. We examined the remodeling of dentate gyrus microcircuits in mice in an enriched environment (EE). Short exposure to EE during early development of new GCs accelerated their functional integration. This effect was mimicked by in vivo chemogenetic activation of a limited population of mature GCs. Slice recordings showed that mature GCs recruit parvalbumin γ-aminobutyric acid–releasing interneurons (PV-INs) that feed back onto developing GCs. Accordingly, chemogenetic stimulation of PV-INs or direct depolarization of developing GCs accelerated GC integration, whereas inactivation of PV-INs prevented the effects of EE. Our results reveal a mechanism for dynamic remodeling in which experience activates dentate networks that “prime” young GCs through a disynaptic feedback loop mediated by PV-INs.

Neural stem cells (NSCs) of the adult hippocampus follow a multifaceted developmental program that culminates in the generation of dentate granule cells (GCs) capable of information processing (14). The pathway from NSC to GC offers multiple checkpoints that are controlled by physiological and pathological conditions, which ultimately modulate the efficacy and quality of the neurogenic process (59). Thus, simple experiences such as physical exercise or exploration of novel environments can influence the production, maturation, survival, and connectivity of adult-born GCs (1014). Yet the early transition from NSC to neuron is much better understood than the processes that control the subsequent steps resulting in the functional integration of new neurons into the preexisting network. We investigated how experience is translated into local signals that can shape the developmental profile of new GCs.

Adult-born GCs were studied during their initial 3 weeks in the temporal hippocampus, where development occurs at a slow pace and is sensitive to behavior (12, 15, 16). We first asked whether an experience that activates the dentate gyrus might work as a signal to shape neuronal development. Because spatial exploration actively involves hippocampal processing (1719), an enriched environment (EE) was selected as a stimulus to promote hippocampal activity. A brief exposure to EE reliably activated a substantial fraction of principal cells in the GC layer (fig. S1). We subsequently tested whether exposure to EE during a restricted period modulates GC development and whether such modulation is dependent on the developmental stage of new GCs. Mice received a retroviral injection in the temporal dentate gyrus to express red fluorescent protein (RFP) in a restricted cohort of new GCs (RFP-GCs) and were exposed for 48 hours to EE at different days post-injection (dpi). The morphology of RFP-GCs was examined at 21 dpi (Fig. 1A). RFP-GCs displayed longer dendrites when exposure to EE occurred within a restricted time window beginning at 9 days of age (EE 9), but not before or after that period (Fig. 1, B and C). A more detailed analysis revealed morphological features of enhanced connectivity, such as higher dendritic complexity and spine density, in GCs exposed to EE at 9 days versus nonexposed controls (Fig. 1, D to G, and fig. S2, A to C). We also noted more RFP-GCs in sections obtained from EE mice, which suggested increased survival (fig. S2D) (8).

Fig. 1 EE accelerates growth and integration of new GCs during a critical period.

(A to G) Brief experience in EE within a critical period accelerates morphological maturation. (A) Experimental design. RFP-expressing retrovirus was delivered to the temporal dentate gyrus to label adult-born GCs. Mice were allowed to freely explore in EE for 48 hours at different times after labeling (blue segments) or were left in a regular cage (control). Morphological parameters were analyzed in RFP-GCs at 21 dpi by immunofluorescence and confocal microscopy. (B) Representative confocal images of 21-dpi RFP-GCs for the different periods. Scale bar, 50 μm. (C) Dendritic lengths of neurons exposed to EE at different ages. **P < 0.01, ***P < 0.001 for analysis of variance (ANOVA; P < 0.0001) followed by Bonferroni post hoc tests, with n = 31, 27, 24, and 27 neurons (from 3 or 4 mice) for EE onset at 5, 9, 14, and 17 days, respectively. Error bars denote SEM. [(D) to (G)] Separate set of experiments to compare GCs at 21 dpi from control versus EE 9 mice. (D) Representative confocal images, with 4′,6-diamidino-2-phenylindole (DAPI; blue) labeling the GC layer. Scale bar, 20 μm. (E) Dendritic lengths measured for the two conditions. ***P < 0.0001, t test with Welch’s correction, with n = 15 and 44 neurons (from 3 to 5 mice) for control and EE 9 mice, respectively. Orange circles correspond to example neurons shown in (D). (F) Higher-resolution images to highlight dendritic spines at 21 dpi. Scale bar, 5 μm. (G) Spine densities for control and EE 9 mice. *P < 0.05 after t test with Welch’s correction with n = 17 and 23 segments (from 3 to 5 mice) for control and EE 9 mice, respectively. (H to O) Functional properties of new GCs from control and EE 9 mice. (H) Experimental design. Ascl1CreERT2;CAGfloxStopTom mice received TAM to label new GCs and were exposed to EE 9. Whole-cell recordings were carried out at 20 dpi in Tom-GCs. (I) Representative images of Tom-GCs obtained from control or EE 9 mice, with immunofluorescence for the neuronal marker NeuN (cyan) labeling the GC layer. Scale bar, 100 μm. (J) Top: Representative voltage traces depicting spiking in response to depolarizing current steps (60 pA, 500 ms) for control and EE 9 GCs. Scale bars, 20 mV, 50 ms. Bottom: Representative current traces (2 cells each) depicting sEPSCs obtained from control or EE 9 neurons held at –70 mV. Scale bars, 4 pA, 2 s. [(K) and (L)] Passive membrane properties. *P < 0.05 after Mann-Whitney test with n = 17 neurons for each condition. (M) Maximum number of spikes elicited by depolarizing current steps. Orange circles correspond to traces shown in (J). [(N) and (O)] sEPSC frequency and amplitude. ***P < 0.001 after Mann-Whitney test with n = 17 and 16 neurons for control and EE 9 mice, respectively. In (E), (G), and (K) to (O), central horizontal bars denote means and error bars denote SEM.

To determine whether the morphological changes observed in stimulated mice are linked to functional integration, we made electrophysiological recordings in 3-week-old GCs after EE 9. Adult-born GCs were labeled using Ascl1CreERT2;CAGfloxStoptdTomato mice and tested 3 weeks after activation of the recombinase by tamoxifen (TAM; Fig. 1, H and I) (2022). In general, GCs expressing tdTomato (Tom-GCs) showed electrical properties typical of immature neurons, such as high input resistance, low membrane capacitance, and limited capacity to spike repetitively (Fig. 1, J to M). Tom-GCs recorded from mice exposed to EE 9 showed membrane resistance and firing properties similar to those from control mice, indicating similar degrees of functional maturation. Their slightly higher values of membrane capacitance were consistent with the more elaborated dendrites. However, Tom-GCs from stimulated mice displayed a factor of >3 increase in the frequency (but not the amplitude) of spontaneous excitatory postsynaptic currents (sEPSCs), reflecting enhanced integration into excitatory networks through an increased number of input connections (Fig. 1, N and O).

We next investigated how exploratory experience may be transduced into local signals that act on functional integration of new individual cells. Because exploration strongly activates the GC layer (fig. S1), we asked whether activation of the local network would be sufficient to influence developing GCs. In recent work, we used the synthetic receptor hM3D to activate adult-born GCs upon binding of the synthetic ligand clozapine-N-oxide (CNO) (23, 24). Here, the hM3D-expressing retrovirus was injected in the dentate gyrus to transduce a cohort of new GCs (hM3D-GCs) and, 7 weeks later, a RFP-expressing retrovirus was used to label a new cohort of developing GCs in the same region (Fig. 2A). This resulted in simultaneous expression of hM3D in fully mature GCs (hM3D-GCs) and RFP in developing GCs (RFP-GCs). Mice received CNO for 2 days to stimulate hM3D-GCs at 8 weeks, at which time neighboring RFP-GCs were undergoing the critical period of high sensitivity to experience that was expressed at about 9 days of age. Dendritic growth was then assessed in RFP-GCs at 21 days. Activation of a limited group of mature adult-born GCs for 2 days increased the complexity of the dendritic tree and the density of dendritic spines of RFP-GCs to an extent similar to that elicited by EE (Fig. 2, B to F).

Fig. 2 Chemogenetic activation of mature GCs promotes the integration of new GCs undergoing the critical period.

(A) Experimental design. A retrovirus expressing hM3D-EGFP was delivered to the dentate gyrus to transduce new GCs (hM3D-GCs). After 7 weeks, RFP-expressing retrovirus was injected to label a second cohort of new GCs (RFP-GCs). Nine days later, mice received CNO for 48 hours to stimulate mature hM3D-GCs neighboring 9-dpi RFP-GCs. Morphological analysis was done on RFP-GCs at 21 dpi. (B) Left: Representative dentate gyrus images depicting RFP-GCs at 21 dpi (red) and mature (70-dpi) hM3D-GCs from mice that received vehicle or CNO. Right: Representative examples of RFP-GCs. Scale bars, 50 μm. (C and D) Dendritic length (C) and Sholl analysis (D) demonstrate substantial dendritic growth in RFP-GCs from mice that received CNO. ***P < 0.001, t test with Welch’s correction, with n = 21 and 23 neurons (from 5 mice) for control and activated circuits, respectively. (E and F) Neuronal segments shown to highlight dendritic spines. Scale bar, 5 μm. Spine density is higher after hM3D-GC activation. ***P < 0.001 after t test with Welch’s correction, with n = 22 and 20 segments (from 5 mice) for control and activated circuits, respectively. In (D), triangles connected by straight lines denote means and error bars denote SEM. In (C), inset of (D), and (F), central horizontal bars denote means and error bars denote SEM.

One possible pathway to convey activity-mediated signaling would be direct or indirect synaptic transmission from mature GCs onto developing GCs undergoing the critical period. At this early developmental stage, new GCs receive depolarizing γ-aminobutyric acid–releasing (GABAergic) inputs but lack glutamatergic afferents (2528). A transient direct connection between mature and immature GCs was recently proposed (29). We therefore selectively expressed the light-activated channel Channelrhodopsin-2 (ChR2) in mature GCs by means of Ascl1CreERT2;CAGfloxStopChR2EYFP mice induced with TAM at a young adult stage (30). Seven weeks after TAM administration, a cohort of new GCs was labeled by retroviral transduction of RFP (Fig. 3, A and B). Acute slices were obtained 10 days later to monitor responses in 10-day-old RFP-GCs (fig. S3) elicited by light-mediated stimulation of mature ChR2-GCs. No glutamatergic responses were evoked from mature to immature GCs (0 of 9 cells; Fig. 3C). However, mature ChR2-GCs reliably recruited inhibitory postsynaptic currents (IPSCs) onto RFP-GCs (9 of 10 cells) that were blocked by the GABAA receptor antagonist picrotoxin or the ionotropic glutamate receptor blocker kynurenic acid, indicating a polysynaptic GABAergic feedback connection (Fig. 3D). IPSCs recorded from RFP-GCs displayed small amplitude and slow kinetics, in contrast to those recorded from mature GCs, consistent with a period of incipient GABAergic synaptogenesis (Fig. 3, E and F).

Fig. 3 Mature GCs signal onto developing GCs through PV interneurons.

(A to F) Mature GCs connect to developing GCs. (A) Experimental design. Ascl1CreERT2;CAGfloxStopChR2EYFP mice received TAM to label a pool of adult-born GCs (mature ChR2-GCs). After 7 weeks, an RFP-expressing retrovirus was injected to label a different cohort of new GCs (RFP-GCs). Whole-cell recordings were carried out 10 days later in RFP-GCs. (B) Confocal image showing TAM-induced mature ChR2-GCs and a 10-dpi RFP-GC. Scale bar, 20 μm. (C) Representative current traces depicting IPSCs and EPSCs obtained from 10-dpi RFP-GCs in response to laser-evoked activation of mature ChR2-GCs (1 ms; blue line). Average (red) and individual traces (gray) are shown. Scale bars, 5 pA, 20 ms. (D) IPSC currents were abolished (>90% block) by 100 μM picrotoxin (PTX; n = 3) and 4 mM kynurenic acid (KYN; n = 2). Scale bars, 5 pA, 20 ms. (E) Comparison of light-evoked IPSCs from unlabeled mature GCs and 10-dpi GCs. Scale bars, 50 pA, 20 ms. Inset shows normalized traces. (F) Characterization of disynaptic IPSCs from 10-dpi RFP-GCs (n = 9 cells, 4 mice) and mature GCs (n = 3 cells, 2 mice). (G and H) Mature GCs activate PV-interneurons. (G) PVCre;CAGfloxStopTom mice received ChR2-expressing retrovirus to transduce adult-born GCs (mature ChR2-GCs). Whole-cell recordings were carried out 8 weeks later in Tom-PV-INs. (H) Example voltage traces depicting spiking of a PV-IN upon light-activation of mature ChR2-GCs (1 ms; blue line). Scale bars, 20 mV, 2 ms. (I to N) PV-INs synapse onto developing GCs. (I) PVCre;CAGfloxStopChR2EYFP mice received RFP-expressing retrovirus to label new GCs. Whole-cell recordings were carried out in 10-dpi RFP-GCs. (J) Confocal image of a 10-dpi RFP-GC and mature GCs surrounded by axons from ChR2-PV-INs. Scale bar, 20 μm. (K) Representative laser-evoked (200 μs; blue line) IPSC obtained from a 10-dpi RFP-GC. Scale bars, 10 pA, 50 ms. (L) IPSCs were blocked >99% by 20 μM bicuculline (BIC; n = 5). Scale bars, 10 pA, 50 ms. (M) Light-evoked IPSCs from unlabeled mature GCs and 10-dpi GCs. Scale bars, 50 pA, 20 ms. Inset shows normalized traces. (N) Monosynaptic IPSC parameters from 10-dpi RFP-GCs (n = 9 cells, 3 mice) and mature GCs (n = 3 cells, 2 mice). In (F) and (N), central horizontal bars denote mean and error bars denote SEM.

We next investigated the nature of the feedback pathway. Parvalbumin-expressing GABAergic interneurons (PV-INs) are main targets of GCs and also contact developing GCs at early stages of development (8, 31). Activation of mature GCs can reliably recruit spiking of PV-INs (24) (Fig. 3, G and H). To determine whether PV-INs can signal onto GCs at early developmental stages relevant to the critical period observed here, we delivered RFP-expressing retrovirus to PVCre;CAGfloxStopChR2EYFP young adult mice to monitor connectivity from ChR2-expressing PV-INs to 10-day-old RFP-GCs (Fig. 3, I and J). Laser-evoked stimulation of ChR2-PV-INs reliably elicited GABAergic IPSCs in RFP-GCs (9 of 9 cells) that were blocked by bicuculline (Fig. 3, K and L). In contrast to IPSCs recorded from mature GCs, RFP-GCs displayed IPSCs with small amplitude and slow kinetics, indicative of immature monosynaptic contacts (Fig. 3, M and N). GABAergic synaptic responses in 10-dpi RFP-GCs displayed similar properties when evoked by direct stimulation of ChR2-PV-INs and indirectly by stimulation of mature ChR2-GCs (fig. S4). Because of the high intracellular chloride concentration, GABAergic transmission is depolarizing in immature GCs and it becomes inhibitory at 3 to 4 weeks of age (2628). Current-clamp recordings were performed to determine whether 10-dpi GCs undergo noticeable depolarization despite their weak GABAergic postsynaptic currents. RFP-GCs were depolarized (~20 mV) by laser-mediated stimulation of ChR2-PV-INs in brief trains that mimic PV-IN activity, but not by single pulses (fig. S5). Therefore, mature GCs recruit a disynaptic GABAergic feedback mediated by PV-INs that depolarizes developing GCs.

To determine whether this feedback circuit may link EE exploration to local dentate signaling, we investigated the possibility that activation of dentate PV-INs in vivo modulates the integration of GCs. We used an adeno-associated virus (AAV) to express the synthetic ligand-gated cationic channel PSAM-5HT3-GFP under control of a flex switch in PVCre mice to allow specific expression in PV interneurons and, consequently, restricted activation of PV-INs upon delivery of the synthetic agonist PSEM308 (32). Mice received both AAV-PSAM-5HT3 and RFP retrovirus; PSEM308 or vehicle were administered from 9 to 11 dpi, and morphology of RFP-GCs was analyzed at 21 dpi (Fig. 4, A and B). Stimulation of PV-INs produced robust dendritic growth in new GCs, mimicking the effects of EE (Fig. 4, C to E). We then investigated the consequence of simply depolarizing young GCs. A retrovirus expressing hM3D-EGFP (hM3D fused to enhanced green fluorescent protein) was used to transduce a cohort of new GCs in young adult mice, which received CNO for 48 hours starting at 9 or 14 days of age. Morphology of hM3D-EGFP–expressing GCs analyzed at 21 days showed that brief intrinsic stimulation within the critical period (9 dpi), but not outside (14 dpi), was sufficient to accelerate dendritic growth (Fig. 4, F to K). These findings suggest that the signal conveyed by intrinsic depolarization is sufficient to promote the integration of GCs by 9 dpi, but not later. Accordingly, chronic CNO stimulation beyond this time rendered effects on hM3D-GCs similar to the effects of a 2-day CNO stimulus during the critical period (fig. S6). Enhanced growth was detected rapidly upon stimulation at early stages of GC development, suggesting a process that triggers the initial preparation for input integration (fig. S7).

Fig. 4 PV interneurons and intrinsic activity during the critical period promote integration of new GCs.

(A to E) PV-INs accelerate GC integration. (A) Experimental design. PVCre mice received AAV-PSAM-5HT3 and RFP retrovirus to transduce PV-INs and new GCs. PSEM308 (10 mg/kg) or vehicle were delivered at 9 dpi (8 intraperitoneal injections in 48 hours) to stimulate PSAM-5HT3-PV-INs. Morphological analysis was done on RFP-GCs at 21 dpi. (B) Representative confocal images of RFP-GCs surrounded by labeled PV-INs and their axons, obtained from mice treated with vehicle (left) or PSEM308 (right). Scale bar, 20 μm. (C) Dendritic lengths measured for the different treatments. The orange symbol corresponds to the example neuron shown in (B). [(D) and (E)] Sholl analysis for dendrites of new GCs under different treatments. ***P < 0.001 after Kruskal-Wallis ANOVA followed by Dunn post hoc test, with n = 35, 37, and 32 GCs (from 3 or 4 mice) for RFP + PSEM, PSAM-5HT3 + vehicle, and PSAM-5HT3 + PSEM, respectively [(C) and (E)]. (F to K) Intrinsic depolarization accelerates GC integration. (F) Retroviruses expressing hM3D-EGFP or RFP were delivered to the dentate gyrus to transduce new GCs. Mice received CNO for 48 hours to stimulate hM3D-GCs at 9 or 14 dpi, and morphological analysis was done on RFP-GCs and hM3D-GCs at 21 dpi. (G) Representative confocal images of 21-dpi neurons in the temporal dentate gyrus obtained from mice that received RFP + CNO (left), hM3D-EGFP and RFP + vehicle (center), or hM3D-EGFP and RFP + CNO (right). Insets show colocalization of RFP and EGFP in the soma of co-infected neurons. Scale bar, 20 μm. (H) Dendritic lengths measured for the different conditions. ***P < 0.001 after ANOVA followed by Bonferroni post hoc test, with n = 22, 29, 16, and 28 GCs (from 4 or 5 mice) for RFP, hM3D + vehicle, hM3D + CNO 9, and hM3D + CNO 14, respectively. Orange symbols correspond to example neurons shown in (G). (I) Sholl analysis demonstrates substantial dendritic growth induced by CNO 9 treatment in hM3D-GCs. **P < 0.01, ***P < 0.001 after Kruskal-Wallis ANOVA followed by Dunn post hoc test. (J) Neuronal segments shown to highlight dendritic spines. Scale bar, 5 μm. (K) Spine density is higher after hM3D-GC activation. *P < 0.05, ***P < 0.001 after ANOVA followed by Bonferroni post hoc test, with n = 16, 19, and 13 GCs (from 4 or 5 mice) for RFP, hM3D + vehicle, and hM3D + CNO 9, respectively. (L to P) PV interneurons mediate the effects of EE. (L) PVCre mice received AAV-hM4Di and RFP-expressing retrovirus at the indicated times and were exposed to EE 9. CNO was delivered from day 8 to day 12 to block GABA release from hM4Di-PV-INs. Morphology of RFP-GCs was analyzed at 21 dpi. (M) Representative confocal image of RFP-GCs surrounded by labeled PV-INs expressing HA-HM4Di. Scale bar, 20 μm. Right panels show colocalization of anti-PV and anti-HA-hM4Di immunofluorescence from neurons in the dashed area. [(N) to (P)] Dendritic lengths and Sholl analysis for dendrites of new GCs after different treatments. *P < 0.05, **P < 0.01 after ANOVA followed by Bonferroni post hoc test, with n = 24, 36, and 28 GCs (from 4 or 5 mice) for CNO, EE, and CNO + EE, respectively. All error bars denote SEM.

To determine whether PV-INs mediate the signaling from EE to new GCs, we used an AAV to express the synthetic ligand-gated G protein–coupled receptor hM4Di with a hemagglutinin tag (AAV-flex-HA-hM4Di) in PVCre mice to block GABA release from PV-INs of the dentate gyrus upon delivery of the synthetic agonist CNO (33). Mice received AAV at postnatal day 21 and RFP retrovirus at 6 to 7 weeks, and were then exposed to EE 9 in the presence or absence of CNO (Fig. 4, L to P). Silencing dentate gyrus PV-INs during exposure to EE abolished the accelerated integration of developing GCs triggered by experience.

Our results reveal a positive feedback loop of circuit remodeling in the GC layer with direct influence from behavior, whereby mature GCs activate PV-INs that accelerate the functional integration of incoming cohorts of very immature neurons—a process that we call “priming.” Under basal conditions, new GCs become potentially relevant for information processing after about 4 weeks of age. At this time, they receive sufficient cortical excitation to become activated by experience and acquire the capacity to refine their cortical inputs in an activity-dependent manner, a process that might be relevant for learning (4, 34, 35). Our results show that a brief experience in EE acts upon developing GCs long before they acquire cortical excitatory inputs, and accelerates glutamatergic synaptogenesis, which might shorten the interval required for achieving functional significance in a behaving animal.

In our concept, “primed” GCs rapidly increase their excitatory drive while maintaining the enhanced excitability (high input resistance), resulting in immature neurons with expanded connectivity. This may favor their recruitment by incoming cortical signals and enhance their capacity for activity-dependent synaptic remodeling in response to novel challenges (4, 3537). Because priming is mediated by GABAergic interneurons and occurs in immature GCs lacking excitation, it may serve as a generalized mechanism that prepares new neurons to become recruited in a manner that is not restricted to the triggering experience, maximizing available resources. This cohort of highly excitable and highly connected GCs may be particularly prone to be recruited by incoming information, which would strengthen their connections and promote their permanence in the network to encode relevant features of the novel experience.

Supplementary Materials

www.sciencemag.org/content/354/6311/459/suppl/DC1

Materials and Methods

Figs. S1 to S7

References (3841)

References and Notes

Acknowledgments: We thank S. Arber and M. Soledad Espósito for providing AAV-hM4Di and AAV-PSAM-5HT3 particles and for invaluable advice on their use, M. C. Monzón Salinas for technical help, S. Sternson for PSEM308, B. Roth for the hM3Dq and hM4Di constructs, G. Davies Sala for preliminary experiments using retroviral expression of hM3Dq, J. Johnson for Ascl1CreERT2 mice, S. Arber for PVCre mice, members of the A.F.S. and G. Lanuza labs for insightful discussions, and V. Piatti for critical comments on the manuscript. D.G. and A.F.S. are investigators of the Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET). Supported CONICET fellowships (D.D.A., S.M.Y., M.F.T., and S.G.T.), Howard Hughes Medical Institute SIRS grant 55007652 (A.F.S.), Argentine Agency for the Promotion of Science and Technology grant PICT2013-1685, and NIH grant FIRCA R03TW008607-01. The data reported in this manuscript are tabulated in the main paper and the supplementary materials. Author contributions: D.D.A. and D.G. contributed to the concept, designed and performed the experiments, analyzed the data, and wrote the manuscript; S.M.Y. and S.G.T. performed electrophysiological recordings and analyzed the data; M.F.T. and K.A.B. contributed to experiments involving enriched environment and analyzed the data; N.B. prepared retroviruses; and A.F.S. contributed to the concept, designed the experiments, analyzed the data, wrote the manuscript, and provided financial support.
View Abstract

Navigate This Article