Electrical activity regulates the manner in which neurons mature and form connections to each other. However, it remains unclear whether increased single-cell activity is sufficient to alter the development of synaptic connectivity of that neuron or whether a global increase in circuit activity is necessary. To address this question, we genetically increased neuronal excitability of in vivo individual adult-born neurons in the mouse dentate gyrus via expression of a voltage-gated bacterial sodium channel. We observed that increasing the excitability of new neurons in an otherwise unperturbed circuit leads to changes in both their input and axonal synapses. Furthermore, the activity-dependent transcription factor Npas4 is necessary for the changes in the input synapses of these neurons, but it is not involved in changes to their axonal synapses. Our results reveal that an increase in cell-intrinsic activity during maturation is sufficient to alter the synaptic connectivity of a neuron with the hippocampal circuit and that Npas4 is required for activity-dependent changes in input synapses.
The addition of new granule cell (GC) neurons to the adult dentate gyrus (DG) of the hippocampus may serve as a substrate for memory throughout life (Imayoshi et al., 2008). These excitatory neurons display enhanced synaptic plasticity (Schmidt-Hieber et al., 2004) and integrate into existing circuitry (Jessberger and Kempermann, 2003). The production, survival, and wiring of GCs into the DG circuit are affected by activity (Kempermann et al., 1997; van Praag et al., 1999; Kee et al., 2007). Importantly, seizures alter the maturation and connectivity of adult-born DG GCs (Parent et al., 1997; Jessberger et al., 2007).
Understanding the factors underlying activity-dependent maturation and connectivity of adult-born neurons is important to understand the physiological basis of learning and the pathological basis of epilepsy. In a seizure or in behavioral paradigms used to stimulate DG activity, general levels of activity in the brain are increased, so it is unclear whether the observed changes in maturation and connectivity result directly from the increased activity of an individual new neuron, indirectly via elevated activity of other neurons in the circuit in which the new neurons are embedded, or a combination of both.
To investigate how the level of neuronal activity of a single developing neuron affects its maturation and integration into an unperturbed circuit, we have developed a system to genetically increase excitability in individual neurons by introducing a voltage-gated sodium channel (Kelsch et al., 2008; Lin et al., 2010). We recently used this system to investigate the effects of genetically increased excitability on the maturation and integration of the GCs of the olfactory bulb, a type of inhibitory neuron produced during adulthood. We demonstrated that genetically increased intrinsic excitability was sufficient to enhance the survival of the new granule neurons of the bulb but, surprisingly, did not affect their synaptic organization (Kelsch et al., 2009; Lin et al., 2010). The plasticity of excitatory and inhibitory neurons differs in many respects (Bi and Poo, 1998), and it is plausible that electrical hyperexcitability affects the morphology of excitatory but not inhibitory neurons.
To test this hypothesis, we genetically raised the intrinsic excitability of individual new GCs in the DG, a type of excitatory neuron also produced during adulthood. We observed that elevating neuronal excitability of individual new neurons during their maturation is sufficient to induce changes in synaptic connectivity such as aberrant localization of synapses and enlarged spines. Cell-autonomous hyperexcitability leads to both input and output connectivity alterations that increase inhibition on the hyperexcitable neuron and dampen its excitatory influence on its downstream targets. We then examined the genetic basis for these alterations by deleting the transcription factor Npas4 in individual new neurons in conditional NPas4 knock-out mice. We observed that the transcription factor Npas4 is required for the activity-induced changes in synaptic inputs to these neurons but not for changes to output synapses in their axons. These observations indicate that cell-autonomous increases in excitability during neuronal maturation can effect profound changes in neuronal connectivity and that separate genetic programs regulate activity-dependent changes in input and output synapses.
Materials and Methods
Cloning of the different constructs was performed using standard molecular techniques. The cDNA for NaChBac was obtained from David Clapham (Howard Hughes Medical Institute, Children's Hospital, Harvard Medical School, Boston, MA). NaChBac E191K was generated by PCR based on previously published sequences (Ren et al., 2001; Yue et al., 2002). Retroviral vectors were derived from a Moloney leukemia virus with an internal promoter derived from the Rous sarcoma virus (Molar) (Kelsch et al., 2007). Retroviral particles were produced and stored as described previously (Lois et al., 2002). The viral titers were ∼107 infectious units/μl. Viral constructs were generated as follows. For NaChBac–EGFP, the stop codon of NaChBac was eliminated by PCR and fused in-frame to the cDNA of EGFP. For NaChBac–Cre, the stop codon of the NaChBac–EGFP fusion was eliminated by PCR and linked by a foot-and-mouth disease (FMDV) virus 2A sequence to the cDNA of Cre recombinase. For PalmEGFP–NaChBac, the palmitoylation sequence from the GAP43 gene was first added to the N terminus of EGFP. The stop codon of the palmitoylated version of EGFP was eliminated by PCR and linked by an FMDV 2A picornavirus sequence to the cDNA of NaChBac. For Synaptophysin–GFP, the stop codon of Synaptophysin was eliminated by PCR and fused in-frame to the cDNA of EGFP. For Synaptophysin–GFP–NaChBac, the stop codon of the Synaptophysin–GFP fusion was removed by PCR and linked by a 2A sequence to the cDNA of NaChBac. For PSD-95–GFP, the stop codon of PSD-95 was eliminated by PCR and fused in-frame to the cDNA of EGFP. For PSD-95–GFP–NaChBac, the stop codon of the PSD-95–GFP fusion was removed by PCR and linked by a 2A sequence to the cDNA of NaChBac. For invertible palmitoylated mCherry (PalmMCherry), the double-floxed inverse open reading frame vector was obtained from Karl Deisseroth (Stanford University, Palo Alto, CA). The ChR2–EYFP cDNA was excised from the vector and replaced with the cDNA of PalmMCherry, which was made by adding the palmitoylation sequence from the GAP43 gene to the N terminus of the mCherry cDNA.
Retroviral labeling in vivo.
Nine- to 12-week-old female BL6 mice (Charles River), floxed NMDA receptor subunit 1 mice (Tsien et al., 1996), floxed Npas4 mice (Lin et al., 2008), and their respective wild-type littermates were stereotaxically injected at two sites per DG with 0.5 μl/site retroviral vectors, after anesthesia with avertin solution. The stereotaxic coordinates were 2.0 mm posterior from bregma, 1.5 mm lateral from the midline, and 1.95 mm ventral from the brain surface, and 2.7 mm posterior from bregma, 1.9 mm lateral from the midline, and 2.05 mm ventral from the brain surface.
Mice were administered an overdose of avertin before they were perfused intracardially, first with PBS and then with 3% paraformaldehyde (PFA). The brains were incubated with 3% PFA overnight and cut with a Leica vibratome into 40 μm frontal sections. For immunocytochemistry, the sections were first blocked with blocking solution containing bovine serum albumin (3 mg/ml PBS) and 0.25% Triton X-100 in PBS and incubated overnight in the relevant primary antibody diluted in blocking solution: polyclonal rabbit anti-GFP antibody (AB3080; 1:2000; Millipore Bioscience Research Reagents), rabbit anti-RFP antibody (LS-C60076; 1:200; Lifespan), mouse anti-vesicular GABA transporter (VGAT) antibody (catalog #131 002; 1:500; Synaptic Systems), mouse anti-parvalbumin (P3088; 1:500; Sigma), and rabbit anti-Npas4 (1:10,000; gift from Y.L.). Sections were washed four times in PBS, for 10 min each time, before a 2 h incubation at room temperature with Alexa Fluor 488 or 555 goat anti-rabbit or anti-mouse secondary antibody (Invitrogen) diluted 1:700 in blocking solution. The sections were washed four times in PBS, for 10 min each time, before being mounted on slides with mounting medium (Vectashield; Vector Laboratories).
Survival ratio analyses.
Two viruses were mixed at an approximate 1:1 ratio for survival analysis. One of the viruses carried the construct encoding mCherry, whereas the other carried either NaChBac or NaChBac E191K fused to EGFP (NaChBac–EGFP or NaChBacE191K–EGFP). Fluorescently labeled cells were quantified with the aid of the Neurolucida software (MicroBrightField). The survival ratio is defined as the total number of EGFP-positive cells (including double-labeled cells) divided by the number of single-labeled mCherry+ cells. The ratio of EGFP+ to mCherry+ neurons at 7 d post-infection (dpi) was used to normalize all data at subsequent time points for comparison; hence, ratios at all subsequent time points were relative to the 7 dpi ratio. Ten to 20 entire sections per DG were analyzed to collect at least 100 counted cells in each DG. The mean survival ratio from each DG was treated as a single sample.
For electrophysiology, viruses in which NaChBac is directly fused to GFP were used because they produce strong fluorescent signals in the soma, which is useful for targeting neurons for fluorescence-guided whole-cell recordings. Animals were given an overdose of ketamine/xylazine and then perfused intracardially with ice-cold slicing solution containing the following (in mm): 212 sucrose, 3 KCl, 1.25 NaH2PO4, 26 NaHCO3, 7 MgCl2, and 10 glucose, pH 7.3 (308 mOsm). Brain slices were incubated in ice-cold cutting solution and cut into 350 μm frontal slices with a Leica microtome at a speed of 0.08 mm/s. Slices were incubated for 30 min at 35°C, for recovery, in carbogenated recording solution containing the following (in mm): 125 NaCl, 2.5 KCl, 1.25 NaH2PO4, 26 NaHCO3, 1 MgCl2, 2 CaCl2, and 20 glucose, pH 7.3 (312 mOsm). Fluorescent-guided whole-cell patch-clamp recordings were performed with a MultiClamp 700B amplifier (Molecular Devices). The pipette solution contained the following (in mm): 2 NaCl, 4 KCl, 130 K-gluconate, 10 HEPES, 0.2 EGTA, 4 Mg-ATP, 0.3 Tris-GTP, and 14 Tris-phosphocreatine, pH 7.3. Successful patching onto the target cell was confirmed by identifying a fragment of fluorescent membrane trapped inside the pipette tip during or after the recording. Pipette resistance ranged from 5 to 8 MΩ, and the pipette access resistance was always <16 MΩ after series resistance compensation. The junction potential was not corrected throughout the study. For spontaneous EPSC (sEPSC) recording, the neurons were held at −77 mV, and synaptic events were collected at 25°C. sEPSCs contributed to the majority of spontaneous events because ∼98% of events could be blocked by 100 μm d,l-AP-5 and 20 μm NBQX (Sigma) at the end of the recording. Furthermore, we observed that the spontaneous synaptic events recorded with bicuculline and TTX in both control and NaChBac+ neurons have a reversal potential at approximately −4 mV. These values are consistent with those of sEPSCs mediated by the opening of glutamate receptors (Cull-Candy and Usowicz, 1989), as reported previously for sEPSPs in GCs of the rat DG (−5.5 ± 1.1 mV; Crunelli et al., 1984).
Inhibitory blockers, such as bicuculline, were not included during sEPSC recording because they triggered frequent EPSC bursting input in granule neurons, which precluded additional analysis. To record sIPSCs, intracellular 130 K-gluconate was replaced with 130 CsCl and included 20 μm NBQX and 50 μm AP-5 in the recording bath to increase the driving force for chloride efflux, enabling us to record spontaneous GABAergic input at −77 mV.
Analysis of electrophysiological data.
Data were acquired and analyzed with pClamp9 software (Molecular Devices), and 2 min traces of sIPSCs and sEPSCs were analyzed with Mini Analysis Program (Synaptosoft). Overall current was calculated by multiplying the average charge area per spike of each individual neuron by frequency of spikes of the same neuron.
For confocal microscopy, confocal image stacks of 40-μm-thick DG sections were acquired by using an Olympus Fluoview confocal microscope (60× oil-immersion lens; numerical aperture 1.4; pixel size, 0.23 × 0.23 μm) and with z-step 0.25 μm. Ten to 0–30 neurons were analyzed in each DG for dendritic length, density, spine size, and perisomatic inhibitory analysis, and data from four to seven DGs were collected for each experimental condition. A typical image stack consisted of ∼80–150 image planes each of 1024 × 1024 pixels.
For image processing and quantification, after acquisition, maximal intensity projections were prepared for each image stack by using the MetaMorph analysis software (Universal Imaging).
Measurements of PSD-95:GFP clusters were performed as described in our previous publication (Kelsch et al., 2007). To attribute the GFP puncta to a particular neuron, we took advantage of the presence of low levels of diffuse PSD-95:GFP protein in the cytoplasm not detectable by its endogenous fluorescence. This diffuse PSD-95:GFP protein could be visualized by amplifying its signal with antibodies raised against GFP (coupled to a red fluorophore to distinguish it from the intrinsic green fluorescence of PSD+ puncta) and allowed us to attribute PSD+ puncta to the neurites belonging to a particular neuron. For the projection images, the threshold was set so that any possible diffuse GFP fluorescence at the dendritic shaft was below this threshold. The number of PSD-95:GFP+ clusters in a region of interest was counted by using the integrated morphometry analysis function of the MetaMorph software. The length of the respective segment of the dendritic arbor was then measured, and the density of PSD-95:GFP+ clusters was determined. All datasets were manually supervised to prevent the inclusion of nonspecific green specks.
Spine size in palmitoylated EGFP (PalmG) neurons were measured as follow. For each stack, laser intensity and detector sensitivity were set so that the fluorescence signal from the spines occupied the full dynamic range of the detector. This meant that some pixels in the dendritic branch were saturated, but no pixels were saturated within the spines. Maximum density projections of the confocal stacks were prepared. Only the areas of the spine heads flanking the dendrites were measured. Spines above or below the respective dendrite were not included in the analysis. The cross-sectional area of spine heads in a region of interest was obtained from the integrated morphometry analysis function in MetaMorph and expressed in square micrometers. To calculate spine density, we divided the number of spines by the length of the respective segment of the dendritic arbor.
Presynaptic sites in PalmG neurons were measured as follows. First, we traced (with Neurolucida) the axons present in each section and marked their presynaptic sites. On each axon, we identified these presynaptic sites as boutons or large mossy fiber terminals (LMTs) based on their characteristic morphology. Second, we divided the number of boutons or LMTs by the length of the axons traced and expressed their density as “number of presynaptic sites per micrometer” or “number of LMTs per micrometer.”
The Mann–Whitney test from OriginPro 8 (OriginLab) was used for comparing the frequency of spontaneous firing in NaChBac+ and control neurons at resting membrane potential to determine statistical significance. All other data were analyzed with the two-sample two-tailed Student's t test in Prism 5 (GraphPad Software). Data were reported as mean ± SEM.
Expression of NaChBac in adult-born DG GCs elevates neuronal excitability
Changes in brain activity, such as those triggered by increased behavioral demands or seizures, affect the maturation and connectivity of new neurons born in the adult DG (Overstreet-Wadiche et al., 2006; Kron et al., 2010). These changes in connectivity could be attributable to the increased activity of the new neurons, of the circuit in which the neurons are embedded, or a combination of both. To isolate the contribution of elevated activity in new neurons, we increased the activity of individual new neurons cell autonomously with the ion channel NaChBac. NaChBac is a bacterial voltage-gated sodium channel that has both a more negative activation threshold than native sodium channels in GCs (∼15 mV more negative) and a longer inactivation time (hundreds of milliseconds compared with <1 ms in mammalian sodium channels) (Ren et al., 2001; Bean, 2007). Because of its unique electrical properties, NaChBac was used previously to induce hyperexcitability in Drosophila pacemaker neurons (Nitabach et al., 2006). More recently, we took advantage of NaChBac-induced depolarization to investigate the maturation of the GCs of the olfactory bulb, a type of inhibitory neuron generated throughout life. Genetically increasing intrinsic excitability by NaChBac expression is sufficient to enhance the survival of the new granule neurons of the bulb but, surprisingly, does not affect their synaptic organization (Kelsch et al., 2009; Lin et al., 2010). The plasticity responses of excitatory and inhibitory neurons differ in many respects (Bi and Poo, 1998), and it is plausible that electrical hyperexcitability by NaChBac affects the morphology of excitatory but not inhibitory neurons.
To investigate whether cell-autonomous increases in excitability are sufficient to alter neuronal connectivity of excitatory neurons, we used oncoretroviruses to introduce NaChBac into individual adult-born DG GCs, a type of excitatory neuron. Because this class of retroviruses cannot transport their genetic material across the intact nuclear envelopes of nondividing cells (Lewis and Emerman, 1994), they can selectively infect dividing cells in the hilus region of the DG, labeling and effectively birthdating new GCs. We used a titer of oncoretrovirus that sparsely labeled GCs [∼100 hyperexcitable cells among ∼400,00 wild-type cells of the DG in the adult B6/57mouse (Abusaad et al., 1999)] in the DG, thus keeping the vast majority of the circuit unaltered. We performed patch-clamp electrophysiological recordings of labeled GCs and found that control, wild-type cells (mCherry+) did not exhibit any type of spontaneous activity at any times tested, between 9 and 36 d post-injection (dpi). In contrast, NaChBac+ cells exhibited spontaneous, long-lasting depolarizations starting as early as 7 dpi (Fig. 1A,B). NaChBac expression did not affect the resting membrane potential or input resistance of the neurons at any times, but it increased their capacitance (Fig. 1B).
Current injection failed to elicit action potentials in control cells at 9 dpi but triggered trains of action potentials between 24 and 36 dpi (Fig. 1C,D). In contrast, current injection triggered long-lasting depolarizations in NaChBac+ cells at all times examined (9, 28, and 36 dpi) (Fig. 1C,D). Finally, whereas the threshold to fire action potentials in wild-type, mCherry+ cells between 26 and 36 dpi is 40 pA, 20 pA is sufficient to trigger a full depolarization in NaChBac+ neurons (Fig. 1D).
To investigate whether, as in the olfactory bulb, NaChBac expression enhances the survival of new neurons, we measured the ratio of mCherry+ (wild-type) to NaChBac+ (hyperexcitable) cells at different time points after viral injection. However, we did not find any deviations from the 1:1 ratio at any time point (7–36 dpi; data not shown). Similarly, we did not find any deviations in the ratio of mCherry+ to E191K+ [a nonconducting variant of NaChBac (Yue et al., 2002)] cells at any time points after injection. These observations suggest that NaChBac expression does not significantly affect the proliferation of neuronal progenitors or the survival advantage of young neurons in the DG, in contrast with our observations in the olfactory bulb (Lin et al., 2010).
We did not observe any seizures or any other gross behavioral changes in the animals injected with NaChBac viruses, which is not surprising given the very low percentage of cells expressing NaChBac (approximately one NaChBac+ cell for every 4000 wild-type cells in the DG). These results confirm that expression of NaChBac in small numbers of new DG GCs increases their excitability in a cell-autonomous manner, without grossly perturbing the function of the DG.
NaChBac-induced excitability results in additional GABAergic input to the cell body
In many circuits, surrounding neurons react to individual neuron activity to keep circuit activity within a range that prevents disruption of function (Turrigiano and Nelson, 2004). DG GCs only start receiving glutamatergic input late in development, ∼21 d after birth, but receive GABAergic input much earlier. These cells first receive extrasynaptic input by ambient (extrasynaptic) GABA starting 3 d after they are generated, followed by GABA-mediated synaptic inputs as early as 7 d after their birth (Ge et al., 2006). For this reason, we hypothesized that, when hyperexcitable adult-born GCs are introduced into the DG circuit, one of the earliest responses of the surrounding circuit would be to alter the GABAergic input targeted to NaChBac+ neurons. To test this hypothesis, we performed immunostaining against the VGAT, which is present in the vast majority of the presynaptic terminals of inhibitory interneurons (Chaudhry et al., 1998). We quantified VGAT+ puncta on cell bodies, because this measurement was more reliable than counting the number of contacts on dendrites.
At 9 dpi, there was no significant difference in the density of VGAT+ contacts with the soma of either control or NaChBac+ neurons (Fig. 2A). However, by 13 dpi, there were significantly more VGAT+ contacts on the soma of NaChBac+ neurons compared with neurons expressing NaChBac E191K, a nonconducting variant of NaChBac (Yue et al., 2002), and this effect persisted until at least 28 dpi (Fig. 2A). To further investigate this increased innervation, we then used antibodies against parvalbumin and GAD65. Parvalbumin+ cells are a subset of inhibitory interneurons that preferentially synapse onto the cell bodies of DG granule neurons (Freund and Buzsáki, 1996), whereas GAD65 is an isoform of glutamic acid decarboxylase (GAD), an enzyme present in a large proportion of inhibitory interneurons (Erlander and Tobin, 1991). We confirmed the trend of increased GABAergic contact in NaChBac+ neurons using parvalbumin (Fig. 2B,C) and GAD65 immunolabeling (Fig. 2C). To verify whether our observations regarding perisomatic GABAergic contact corresponded to a functional increase in inhibitory input, we performed electrophysiological recordings to measure sIPSCs of individual neurons. We coinjected a mixture of retroviruses, one carrying the construct for NaChBac fused to GFP and the other carrying the construct for mCherry, into the DG and recorded from control neurons (mCherry-only) and NaChBac+ neurons in the same DG at 17 dpi. mCherry was used to label control neurons because these cells would appear red and could be easily distinguished from the GFP-expressing NaChBac+ neurons. Indeed, there was an increase in both the frequency and amplitude of sIPSCs received by NaChBac+ GCs relative to control GCs (Fig. 3A). These results indicate that individual adult-born DG GCs with elevated neuronal excitability receive more GABAergic inputs than age-matched wild-type GCs.
GABAergic innervation to adult-born GCs is initially depolarizing as a result of high levels of expression of the NA+/K+/Cl− cotransporter NKCC1 relative to the K+/Cl− cotransporter KCC2 (Plotkin et al., 1997; Clayton et al., 1998). The subsequent upregulation of KCC2 as cells mature lowers the intracellular concentration of Cl− and eventually makes the GABA reversal potential more negative than the resting membrane potential, rendering GABAergic innervation hyperpolarizing (Rivera et al., 1999; Wang et al., 2002). This switch from depolarizing to hyperpolarizing GABAergic inputs occurs after 14 dpi in adult-born GCs (Ge et al., 2006). As mentioned above, we observed an increase in inhibitory input to NaChBac+ GCs that occurs by 13 dpi (Fig. 2A). At 13 dpi, we also observed an increase in the number of KCC2-positive NaChBac+ GCs compared with controls (Fig. 3B). The increase in the percentage of KCC2+ cells induced by NaChBac expression suggests a premature reduction in Cl− concentration, which would in turn result in an earlier switch to inhibition by GABA. This accelerated maturation could enable the increase in GABAergic input to prematurely become inhibitory and thus dampen the hyperexcitable neurons earlier in development. In agreement with the notion of hyperexcitability accelerating the maturation of new GCs, we observed that polysialylated neural cell adhesion molecule (PSA-NCAM), a marker for immature neurons (Seki and Arai, 1993), is also downregulated earlier in NaChBac+ GCs compared with control neurons (Fig. 3C).
Our results indicate that NaChBac activity impacts the maturation and early synapse formation of DG GCs. From an early developmental stage, hyperexcitable GCs start receiving more GABAergic input from surrounding interneurons. In addition, cell-autonomous hyperexcitability speeds up development of newly born GCs.
Increased excitability leads to changes in excitatory glutamatergic input
Having discovered that NaChBac-induced hyperexcitability induces marked changes in neuronal maturation and an increase in inhibitory inputs early on, we proceeded to study how NaChBac affects the next phase of development of these GCs when they start to receive excitatory input synapses. DG GCs normally begin receiving excitatory inputs from 21 d after birth via spines along their apical dendrites that acquire mature morphology by ∼28 d. To examine the changes in excitatory input received by a neuron rendered hyperexcitable by NaChBac, we infected neural progenitors in the DG with a bicistronic retroviral vector that expresses both PalmG and NaChBac. PalmG is localized into the membranes of infected neurons, allowing us to visualize the detailed morphology of the neurons, including dendritic spines.
Using the bicistronic PalmG:NaChBac construct, we observed that NaChBac+ GCs exhibited some connectivity changes that were similar to those observed in immature DG GCs after seizures. DG GCs migrate a small distance, ∼5–10 μm, from the hilar border in the DG, in which neural progenitors reside, to the granule layer of the GC, in which they settle and integrate into the DG circuit. Seizures induce the ectopic migration of GCs to either the outer third of the GC layer or the hilar region (Parent et al., 1997; Jessberger et al., 2009). The majority of NaChBac+ neurons had cell bodies correctly localized within the granule layer of the DG. We observed no NaChBac+ neurons in the outer third of the GC layer but found occasional neurons in the hilar region (Fig. 4A), whereas no wild-type neurons were ever found there. The morphology of these ectopic neurons was similar to their counterparts in the GC layer. They were polarized and had dendrites extending in the opposite direction of their axon. The neurons were entirely in the hilus, and their dendrites were also located solely in the hilus instead of the molecular layer. Because of the location of their dendrites, the connectivity of these neurons is likely to be perturbed. As mentioned above, NaChBac activity induces premature downregulation of PSA-NCAM (Fig. 3C), which has been implicated in neuronal migration either through its role in decreasing cell–cell adhesion (Johnson et al., 2005) or in sensing growth factor gradients (Muller et al., 2000). Thus, it is possible that the downregulation of PSA-NCAM may be responsible for the ectopic location of some of these NaChBac+ neurons. In addition to cells found in the hilus, we also observed that cell bodies of NaChBac+ GCs reside closer to the hilar border than wild-type controls at 28 dpi (Fig. 4B). These observations suggest that hyperexcitability in young neurons affects the migration and settling of the GCs within the GC layer.
Mature GCs in the DG display apical dendrites that branch into the molecular layer of the DG. During their development, newly generated GCs in the adult DG transiently display basal dendrites extending into the hilus, which disappear by 4 or 5 d after the birth of the neuron (Shapiro and Ribak, 2006). Approximately 20% of NaChBac+ neurons displayed these aberrant basal dendrites, whereas control neurons expressing the E191K nonconducting channel never did (Fig. 4C). Interestingly, the presence of spiny basal dendrites emanating from the cell bodies of fully mature GCs and extending into the hilus is one of the hallmarks of seizure-related changes in the DG (Shapiro and Ribak, 2006; Jessberger et al., 2007). To investigate whether the basal dendrites of NaChBac+ neurons contained postsynaptic sites, we infected new DG GCs with viral vectors encoding both NaChBac and a fusion between GFP and PSD-95, a scaffolding protein selectively localized to the postsynaptic density of glutamatergic input synapses (Kelsch et al., 2008). We performed immunocytochemistry against the diffuse, unclustered GFP that filled the cytoplasm with a red secondary antibody to visualize the dendritic morphology, whereas PSD-95+ clusters were identified by the direct green fluorescence from GFP (Kelsch et al., 2008). We observed that the basal dendrites of NaChBac+ cells had PSD-95:GFP+ clusters (Fig. 4C). The persistence of basal dendrites suggests that hyperexcitable neurons receive additional synaptic inputs to their cell bodies, and this input is likely to be excitatory (Ribak et al., 2000; Thind et al., 2008).
Increase in neuronal activity during seizures also affects the formation of apical dendrites and their synapses. When we examined the morphology of NaChBac+ GCs in the granule layers, we observed that they had shorter apical dendrites on average (Fig. 5A). In addition, the density of protrusions on apical dendrites of NaChBac+ neurons was half of the spine density of control neurons expressing the pore-dead NaChBac E191K channel at 28 dpi (Fig. 5B). There was an increase in spine density from 28 to 42 dpi for control neurons (Fig. 5B, far right panel) but no additional change for NaChBac+ neurons. The average spine size at 28 dpi of NaChBac+ neurons was twice of that of NaChBac E191+ neurons (Fig. 5C, far right panel). Interestingly, the increase in spine size of NaChBac+ neurons resembles the increased proportion of mushroom spines observed in GCs after seizure (Jessberger et al., 2007).
To investigate whether the large protrusions on NaChBac+ neuron apical dendrites were indeed synaptic spines, we infected new DG GCs with GFP constructs fused to PSD-95 (Kelsch et al., 2008). Control neurons were infected with a virus expressing only the PSD-95–GFP fusion, whereas hyperexcitable neurons were infected with PSD-95–GFP:NaChBac, a bicistronic construct encoding both GFP-tagged PSD-95 and NaChBac. All protrusions present on the dendrites of labeled neurons expressed GFP:PSD-95 (Fig. 5C, left panels), confirming that the larger protrusions in NaChBac+ cells are postsynaptic sites. Furthermore, larger protrusions exhibited larger GFP:PSD-95+ clusters, suggesting that any observed change in spine size could possibly indicate larger postsynaptic densities and, in effect, larger synapses.
The morphological alterations we report here suggest that NaChBac+ neurons experience an overall decrease in the number of excitatory inputs. However, although fewer in number, each individual spine in NaChBac+ neurons was larger on average than those of control neurons. To examine how these morphological changes translated into functional differences, we measured the sEPSCs of individual NaChBac+ neurons by electrophysiological recording and found that overall frequency of sEPSCs is significantly reduced in NaChBac+ neurons (Fig. 6A, left graph), whereas the average amplitude of sEPSCs was increased (Fig. 6A, middle graph). These results are consistent with NaChBac+ neurons having fewer but larger synapses. We observed that these spontaneous synaptic events, in both control and NaChBac+ neurons, could be blocked by glutamate blockers such as d,l-AP-5 and NBQX and had a reversal potential at approximately −4 mV (Fig. 6B). These values are consistent with those of sEPSCs mediated by the opening of glutamate receptors (Cull-Candy and Usowicz, 1989), as reported previously for sEPSPs in GCs of the rat DG (−5.5 ± 1.1 mV; (Crunelli et al., 1984).
Because the frequency and amplitude of sEPSCs in NaChBac+ neurons changed in opposing directions, to find out what the resultant current was, we calculated the overall excitatory current that the neurons received by multiplying the average charge area per spike of each individual neuron by the frequency of spikes of the same neuron. The overall excitatory current received by NaChBac+ neurons was not significantly different from that received by controls (Fig. 6A, right graph), hinting at the existence of a mechanism that maintains a set level of excitatory drive into these neurons.
Elevated excitability leads to changes in excitatory outputs at CA3
To quantify the changes in outputs of NaChBac-expressing DG GCs, we examined the morphology of presynaptic terminals on their axons in the CA3 region, in which their main output is.
The axons of DG GCs synapse on multiple targets on CA3, on both excitatory pyramidal cells and inhibitory interneurons. The axon collaterals of dentate GCs form specialized presynaptic sites called LMTs. LMTs measure between 3 and 8 μm in their greatest dimension and form complex interdigitating connections with CA3 pyramidal cells. DG axons also have two other types of smaller output synapses that contact inhibitory neurons at CA3 called en passant boutons and filopodial terminals (Acsády et al., 1998). En passant boutons are varicosities 0.5–2 μm in diameter distributed along the axons of GCs, and filopodial terminals are thin protrusions emanating from the LMT. We focused on the effects of hyperexcitability on LMTs because, as a result of their characteristic morphology, these presynaptic sites can be unambiguously identified by membrane-bound GFP labeling. Expression of NaChBac decreased the overall density of presynaptic terminals present on the axons of adult-born dentate GCs at CA3 (Fig. 7A,B), which suggests that the hyperexcitable neurons downregulated their overall output to CA3. We observed that the overall density of LMTs in the axons of NaChBac+ neurons was significantly decreased compared with control neurons (Fig. 7A,C). In addition, NaChBac expression also appears to reduce the size of the LMTs (Fig. 7A). Finally, he proportion of total presynaptic sites that are LMTs is also significantly reduced in NaChBac+ neurons (Fig. 7D). These observations indicate that the output from hyperexcitable NaChBac+ DG GCs onto CA3 is significantly decreased.
To confirm that we were quantifying actual presynaptic sites in our measurements, we injected adult mice with retroviral vectors expressing Synaptophysin–GFP, a protein selectively localized to presynaptic neurotransmitter vesicles (Wiedenmann and Franke, 1985). Neurons were infected with a bicistronic virus encoding both Synaptophysin–GFP and NaChBac to visualize the presynaptic sites on axons of hyperexcitable cells. Cells in a separate DG infected with a virus encoding Synaptophysin–GFP were used as controls. Using a red fluorescent secondary antibody against the diffuse, unclustered GFP that filled the cytoplasm, we were able to visualize the full morphology of the axons at CA3, as well as determine the location of Synaptophysin-positive presynaptic sites, which showed up as green GFP-positive clusters. All structures resembling presynaptic terminals as labeled by PalmG were positive for GFP:synaptophysin in both wild-type neurons and those expressing NaChBac (Fig. 7E). This confirms that the structures we quantified corresponded to presynaptic terminals on the axons of these adult-born GCs.
Our observations of output connectivity at CA3 suggest that an increase in intrinsic excitability in adult-born GCs leads to a decrease in excitatory output at CA3. However, the output does not depend exclusively on the number of synaptic contacts but also on the properties of each of these contacts. For instance, the properties of voltage-gated potassium channels in the terminal are critical in determining the amount of release (Geiger and Jonas, 2000). Additional experiments involving electrophysiological recordings from the postsynaptic neurons in CA3 may allow to investigate whether the output is indeed reduced.
Activity-induced changes in input connectivity are dependent on cell-autonomous Npas4 signaling
Two of our observations in NaChBac+ neurons led us to hypothesize that the early increase of GABAergic synapses triggered by hyperexcitability could be related to the later changes in synaptic connectivity observed in dendrites and axons. First, one of the earliest changes observed in the development of NaChBac+ neurons was the increase in perisomatic GABAergic inputs at ∼13 dpi (Fig. 2). At 17 dpi, the overall current of sIPSCs received by NaChBac+ neurons was 10 times that of controls (Fig. 3A, bottom panel, right). An alteration of this magnitude early in neuronal development could have a significant impact on subsequent integration. Second, the premature upregulation of KCC2 (Fig. 3B) suggests that the action of GABA could be hyperpolarizing earlier in the maturation of NaChBac+ neurons. Rendering GABA hyperpolarizing on immature neurons by altering chloride concentration is known to affect the dendritic development of adult-born GCs in the DG (Ge et al., 2006).
The transcription factor Npas4 was a likely candidate underlying the increase of inhibition in NaChBac+ neurons, because it is regulated by activity and is involved in the activity-dependent regulation of inhibitory synapses in hippocampal neurons (Lin et al., 2008). We hypothesized that the increased inhibition observed in hyperexcitable new DG GCs could be attributable to the expression of Npas4.
To study the effects of expressing NaChBac in the absence of Npas4, we expressed Cre recombinase, NaChBac, and a fluorescent protein in individual neurons in the DG of Npas4 conditional knock-out mice. The expression level of a tricistronic vector containing the three abovementioned genes was too low for visualization of the labeled neurons. To achieve stronger expression of fluorescent proteins, we injected a mixture of viruses into the DG of Npas4 conditional knock-out mice. The first virus carried a bicistronic construct expressing GFP and Cre recombinase, and the second virus carried an invertible cassette with a bicistronic construct encoding both PalmMCherry and NaChBac. The invertible cassette is in the reverse 3′ to 5′ orientation with respect to the retroviral promoter except in the presence of Cre recombinase when it flips to the correct 5′ to 3′ orientation and expresses both mCherry and NaChBac (Fig. 8). In this manner, the presence of Cre leads to the expression of mCherry and NaChBac and, simultaneously, to the deletion of the Npas4 locus in individual GCs in the Npas4 conditional knock-out mice (Fig. 8). The PalmMCherry protein localizes to the membranes of such neurons, enabling the identification of fine structural features, such as synaptic spines. In this experiment, we used the same dual virus strategy in both wild-type mice and conditional knock-out mice and compared results between them.
As described above, expression of NaChBac in wild-type adult mice results in an increase in VGAT+ perisomatic inhibitory terminals on new DG GCs (Fig. 2A). In contrast, the deletion of Npas4 in individual NaChBac+ new GCs blocked the increase in VGAT terminals triggered by NaChBac at both 17 and 28 dpi (Fig. 9A). Knocking out Npas4 alone in DG GCs, using a virus carrying only a GFP–Cre recombinase construct, has no effect on the number of VGAT+ puncta at 17 dpi and leads to a very small increase at 28 dpi (Fig. 9A). Furthermore, Npas4 signaling within individual adult-born neurons in the DG is necessary to trigger the changes in dendritic morphology induced by NaChBac+ neurons (Fig. 9B,D,E). Deletion of Npas4 in new GCs blocked the changes caused by NaChBac expression in dendritic length, spine density, and size. Dendritic length and spine density in NaChBac+Npas4− neurons are indistinguishable from those of control neurons expressing the pore-dead version of NaChBac (Fig. 9B,D).
Interestingly, Npas4 was not required for activity-dependent changes in the output synapses, because NaChBac+Npas4− neurons still exhibited a dramatic reduction in density of LMTs at CA3 (Fig. 9C). Knocking out Npas4 alone has no effect on either input or output synapses (Fig. 9B–E). These results demonstrate that Npas4 signaling specifically regulates activity-dependent formation of synaptic inputs to DG GCs.
Increase in electrical activity of a single new neuron in the DG is sufficient to induce changes in maturation and connectivity
Global manipulations of brain activity via behavioral paradigms or seizures have demonstrated the influence of neuronal activity on the maturation, integration, and connectivity of adult-born neurons in the DG (Kee et al., 2007; Kron et al., 2010). However, it is unclear whether these changes resulted directly from cell-autonomous increased firing of new neurons, indirectly through the elevated activity in the surrounding circuit, or from a combination of both. Here, we genetically modulate the electrical activity in individual adult-born DG GCs and show that an increase in cell-intrinsic activity of new neurons is sufficient to cause dramatic changes in their maturation and connectivity. NaChBac activity-induced connectivity changes in the adult DG appear to be homeostatic, because NaChBac induces an increase in inhibitory inputs and decrease in excitatory outputs. Furthermore, at 13 dpi, when NaChBac+ neurons display a significantly higher number of perisomatic VGAT+ puncta, more NaChBac+ neurons express KCC2 than control neurons (Fig. 3B). This observation is consistent with previous reports indicating that the timing of KCC2 expression is activity dependent (Ganguly et al., 2001). Because the upregulation of KCC2 relative to NKCC1 is correlated with the switch between GABAergic inputs being depolarizing to hyperpolarizing (Rivera et al., 1999; Wang et al., 2002), this suggests that the GABAergic input to NaChBac+ neurons becomes inhibitory earlier in their development than in control cells, which could serve to dampen the heightened excitability of these neurons.
The results of a previous study in vitro seemed to suggest that global activity alterations are necessary to affect changes in GABAergic terminals because suppression of single-cell activity in dissociated hippocampal cultures using the potassium channel Kir2.1 did not alter GABAergic inputs (Hartman et al., 2006). Our results show that, in new DG neurons in vivo, elevating single-cell activity is sufficient to induce changes in GABAergic input (Fig. 2). It is not possible to directly compare the results from these two experiments because they were produced in different conditions and systems, namely adult-generated DG neurons rendered hyperexcitable in vivo versus embryonic hippocampal neurons silenced in vitro. However, in agreement with this previous study, when we silenced adult-generated DG neurons in vivo via retroviral expression of the Kir2.1 channel, we did not observe any changes in either dendritic or axonal morphology (data not shown). These observations suggest that the regulation of synaptic input to a single adult-born DG GC may be modulated by increases, but not decreases, in intrinsic activity.
In our study, we observed that sEPSCs received by NaChBac+ neurons were of a lower frequency but of increased amplitude than those received by control neurons, and this was consistent with the decreased spine density but increased spine size on their dendrites (Fig. 5B,C). Interestingly, the overall excitatory current received was not significantly different compared with controls (Fig. 6A, right graph). In contrast, the overall GABAergic current received by NaChBac+ neurons was ∼10 times that of control neurons (Fig. 3A, right graph). These findings suggest that there may exist mechanisms to ensure that new DG neurons receive a set value of excitatory input and that modulating inhibition may be the primary method by which the activity of an adult-born neuron in the DG is regulated.
An activity-dependent genetic program involving immediate early gene Npas4 governs neuronal connectivity of adult-born neurons to the mature DG circuit
Our results reveal an intermediate step between neuronal activity and changes in synaptic connectivity, which involves a transcription factor whose expression is activity dependent. The connectivity changes triggered by an increase in cell-intrinsic excitability in neurons are dependent on the immediate early gene Npas4. The role of Npas4 in these connectivity changes is activity dependent; deletion of Npas4 does not affect the formation of synapses in control cells that have baseline excitability, but it blocks synaptic alterations triggered by NaChBac-induced hyperexcitability. Although Npas4 has been shown to regulate the formation of inhibitory inputs to CA1 pyramidal cells in vitro (Lin et al., 2008), our results in DG GCs suggests that knocking out Npas4 alone in individual neurons does not decrease the number of inhibitory VGAT+ contacts on its soma (Fig. 9A). Npas4 is expressed at extremely low levels at baseline in DG GCs, but it is upregulated by strong stimuli, such as during kainic acid-induced seizures (Ramamoorthi et al., 2011). These observations suggest that, in young DG GCs, the role of Npas4 in inducing an increase in inhibitory contacts may require exceeding a threshold of activity that is only achieved by hyperexcitable neurons.
Finally, Npas4 is involved in activity-dependent changes to input connectivity to adult-born DG GCs but not to their output connectivity in CA3 (Fig. 9C). This observation suggests that there are independent programs governing input and output synapses, and this finding could have important implications for the structural alterations triggered by epilepsy.
This work was supported by a NIDCD RO1 Grant to C.L. and funded in part with a grant from the MIT Simons Initiative on Autism and the Brain (Y.L. and C.L.). We thank Alberto Stolfi and Drew Friedman for help with the engineering of the viral constructs, David Clapham (Harvard Medical School, Boston, MA) for providing us with the NaChBac cDNA, and Karl Deisseroth (Stanford University, Palo Alto, CA) for providing the double-floxed inverse ORF vector.
The authors declare no competing financial interests.
- Correspondence should be addressed to Carlos Lois, Neurobiology Department, University of Massachusetts Medical School, 364 Plantation Street, Worcester, MA 01605.