|Home | About | Journals | Submit | Contact Us | Français|
Notch signaling in the nervous system has been most studied in the context of cell fate specification. However, numerous studies have suggested that Notch also regulates neuronal morphology, synaptic plasticity, learning and memory. Here we show that Notch1 and its ligand Jagged1 are present at the synapse, and that Notch signaling in neurons occurs in response to synaptic activity. In addition, neuronal Notch signaling is positively regulated by Arc/Arg3.1, an activity-induced gene required for synaptic plasticity. In Arc/Arg3.1 mutant neurons, the proteolytic activation of Notch1 is disrupted both in vivo and in vitro. Conditional deletion of Notch1 in the postnatal hippocampus disrupted both long-term potentiation (LTP) and long-term depression (LTD), and led to deficits in learning and short-term memory. This work shows that Notch signaling is dynamically regulated in response to neuronal activity, that Arc/Arg3.1 is a novel context-dependent Notch regulator, and that Notch1 is required for the synaptic plasticity that contributes to memory formation.
Notch receptors and ligands are highly conserved transmembrane proteins, which are expressed in the developing mammalian nervous system and in the adult brain (Givogri et al., 2006; Stump et al., 2002). The function of Notch signaling in the nervous system has been most studied in the context of neural stem/progenitor cell regulation, and neuronal/glial cell fate specification (Louvi and Artavanis-Tsakonas, 2006). However, numerous reports have suggested that Notch also plays a role in neuronal differentiation (Breunig et al., 2007; Eiraku et al., 2005; Redmond et al., 2000; Sestan et al., 1999), neuronal survival (Lutolf et al., 2002; Saura et al., 2004), and in neuronal plasticity (Costa et al., 2003; de Bivort et al., 2009; Ge et al., 2004; Matsuno et al., 2009; Presente et al., 2004; Saura et al., 2004; Wang et al., 2004).
While studies in both vertebrates and invertebrates suggest that Notch signaling regulates neuronal plasticity, learning, and memory, it remains unclear where and how Notch is activated in mature neurons, how it affects synaptic plasticity, and whether it interacts with known plasticity genes. Here we provide evidence that Notch signaling is induced in neurons by increased activity, and that this signaling is heavily dependent upon the activity-regulated plasticity gene Arc/Arg3.1 (Arc hereafter)(Chowdhury et al., 2006; Link et al., 1995; Lyford et al., 1995; Shepherd et al., 2006). Furthermore, disruption of Notch1 in CA1 of the postnatal hippocampus reveals that Notch signaling is required to maintain spine density and morphology, as well as to regulate synaptic plasticity and memory formation.
Using an antibody that recognizes the active form of Notch1 (NICD1, S3 fragment), we found it present in the cell soma and dendrites of neurons in many regions of the brain, including the cerebral cortex and hippocampus (Figure 1A, and not shown). We also found that NICD1 and the activity-induced Arc were present in many of the same cells, suggesting that Notch1 signaling occurs in active neurons. Indeed, most EGFP+ neurons in a transgenic Notch reporter (TNR) mouse line (Mizutani et al., 2007) expressed Arc (e.g, 73% of EGFP+ cells in the cortex, see Figure 3A). In cultured neurons, Notch1 was enriched in the dendrites and cell soma, while the ligand Jagged1 (Jag1) was enriched in axons (Figures 1B, S1A). Notch1, NICD1 and Jag1 co-localized with synaptic proteins (Figure 1C,D,E Figure S1), and NICD1 was enriched in synaptosomal fractions derived from cortical extracts (Figure 1F).
Increased neuronal activity after treatment with NMDA (Figure 2A,B) or bicuculline (Figure 2C) led to higher NICD1 levels, while treatment with the NMDA-receptor blocker AP5 led to reduced NICD1 levels (Figure 2B). Neuronal activity also increased Notch1 protein levels (Figure 2C-E, see also Figure 3E), including the pre-processed form of the receptor (Figure 2D,E), and Jag1 expression (Figure 2F). Activity-induced Notch1 expression occurred in the presence of the transcriptional inhibitor actinomycin-D, suggesting that pre-existing Notch1 transcript is translated in response to synaptic activity (Figure 2D,E).
To test the effect of synaptic activity on Notch expression and processing further, we used acute hippocampal slices. The Schaffer collateral pathway was activated to induce LTP (Figure 2I), and increased Arc expression was observed in both CA3 and CA1 neurons (Figure 2G). In addition, somal Notch1 expression was increased in CA3 and CA1 (6.1-fold by pixel count, n=6, p<0.02)(Figure 2G), as was NICD1 staining (Figure 2H, and not shown). The increase in Notch expression in CA1 could be reduced by AP5 (Figure S2).
We next evaluated Notch expression and signaling in response to neuronal network activity in vivo after exploration of a novel environment. Spatial exploration activates specific ensembles of hippocampal pyramidal neurons that can be identified by expression of Arc (Guzowski et al., 1999). TNR mice were allowed to explore a novel environment for 5 min, and were sacrificed 1.5 or 8 hours later. Consistent with prior work (Ramirez-Amaya et al., 2005), the number of Arc+ hippocampal CA1 neurons was increased ~3-fold at 1.5 hours, and ~2-fold at 8 hours (Figure S3A). In addition, the number of Notch1+ CA1 neurons was elevated at both time points (~3-fold, Figure S3A,C), as was EGFP expression (indicative of Notch activity) at 8 hours (Figure S3B,C). Notably, nearly all (94–97%) of the Arc+ neurons also had Notch1 signal in the nucleus (e.g. see Fig, 3C), indicating that Arc induction and Notch signaling occur in the same neuronal networks in response to exploration. Some Notch1+ neurons did not express Arc (18% in controls and at 1.5 hours, and 29% at 8 hours). Thus, the temporal dynamics of Notch1 and Arc may be different, with Notch1 persisting longer than Arc, or not all neuronal Notch signaling occurs in Arc+ networks.
Both Notch signaling (Fortini and Bilder, 2009; Vaccari et al., 2008) and Arc function (Chowdhury et al., 2006; Shepherd et al., 2006) engage Dynamin-mediated endocytosis, raising the possibility that they might interact. Thus, we examined Notch activity in the adult brain of Arc mutants using the TNR mouse line. Of fifteen Arc mutants, twelve (80%) had reduced EGFP expression (Notch activity) throughout the cerebral cortex as compared to twenty-two non-mutants (Figure 3A,B). Arc mutants also had reduced NICD1 levels, consistent with less Notch signaling in the absence of Arc (Figure 3B).
To test if Arc is required for Notch pathway recruitment in response to network activity in vivo, we compared Notch1 expression in the hippocampus of wild type and Arc mutants after exploration of a novel environment. In controls, we observed elevated expression of both Arc and Notch1, the latter of which was localized both to the cell soma and nucleus, in CA1 (not shown) and CA3 (Figure 3C). In contrast, no change in Notch1 expression or subcellular localization was observed in Arc mutants (Figure 3D).
We next examined the status of Notch1 processing in Arc mutant neuronal cultures. In the absence of Arc there was a reduction in the S3 cleaved form of Notch1 (NICD1)(Figure 3E), indicating that Arc positively regulates the γ-secretase-mediated cleavage of Notch1 in neurons. Treatment with bicuculline led to elevated Notch1 and NICD1 levels in control neurons but not in Arc mutant neurons (Figure 3E), indicating that Arc is required for the activity-mediated recruitment of neuronal Notch signaling. No change in Jag1 expression was observed in Arc mutant cultures (Figure S4) in line with the idea that receptor processing, and not ligand availability, is defective in mutant cells.
In an effort to rescue Notch1 processing in Arc mutant cells, we used Sindbis virus to introduce functional or non-functional Arc into mutant neurons in vitro. Restoration of Arc expression rescued Notch1 processing (2.9-fold increase, n=3, p<0.001)(Figure 3F), suggesting that the Notch1 cleavage defect in Arc mutant neurons is not caused by aberrant neuronal differentiation. A form of Arc lacking the ability to bind Endophilin and participate in endocytic trafficking (Δ91–100)(Chowdhury et al., 2006) was unable to restore Notch1 processing in Arc mutants neurons (Figure 3F).
Next, we found that Arc and Dynamin co-immunoprecipitated (IP) with Notch1 in protein preparations from adult cortical extracts (Figure 3G). In addition, Notch1 co-IP’d with Arc in protein extracts from wild type, but not Arc mutant, cortical tissue (Figure 3H). Thus, Arc-mediated Dynamin-driven endocytosis of Notch1 may be important for activity-dependent Notch signaling in neurons. Interestingly, Arc is not required for Notch activation in embryonic forebrain progenitors (Figure S5), indicating that Arc regulates Notch in a context-dependent manner.
Having shown that Arc-dependent Notch signaling is activated in neuronal ensembles after spatial exploration, we next tested the function of Notch in such ensembles. To conditionally knock out (cKO) Notch1 in the postnatal hippocampus, we crossed Notch1flox/flox (Radtke et al., 1999) mice with the CamKII-cre (T29-1) driver line (Tsien et al., 1996), and Notch1 deletion was confirmed at both the mRNA and protein levels (Figure S6 and Figure 4A, respectively; n=6 each). Golgi-Cox staining of CA1 pyramidal neurons revealed that loss of Notch1 postnatally did not affect dendritic length (Figure 4B,C). However, spine density on secondary and tertiary dendrites was reduced (Figure 4D,F), and spine morphologies were altered (Figure 4E,F).
To test the role of Notch in synaptic plasticity, the electrophysiological properties of Notch1 cKO animals were tested using hippocampal slices and field recordings. Basal transmission was the same for mutants and controls (10–11 slices)(Figure 4G), and paired-pulse facilitation (PPF), revealed that Notch1 cKO slices had presynaptic strength comparable to controls (Figure 4H). However, when we induced LTP in the Schaffer collateral pathway, the magnitude of LTP in the CA1 region was uniformly higher in controls (188.5 ± 23.1, n=6) than in Notch1 cKO slices (140.9 ± 20.6, n=5, p<0.05)(Figure 4I). Similarly, after low frequency stimulation, LTD in CA1 was uniformly reduced in Notch1 cKO mice (83.4 ± 11.2, n=6 slices) compared to controls (70.0 ± 11.5, n=5, p<0.05)(Figure 4J). Thus, Notch1 influences the magnitude of both the potentiation and depression of synaptic efficacy.
Next we performed behavioral tests to evaluate the cognitive abilities of Notch1 cKO mice. During novel object recognition testing, mutants initially had a lower novel object preference than controls, and the next day, in contrast to controls, mutants had no preference (Figure 5A). Similarly, in a social interaction test unlike controls Notch1 cKO mice did not interact more with a new subject (Figure 5B), although like controls mutants preferred a subject to an object (not shown). In Y-maze testing, Notch1 cKO mice chose alternating arms at the same frequency as controls (not shown), but showed no preference for a previously hidden arm (Figure 5C,D).
Next, spatial reference memory was investigated using the Morris water maze. Performance improved over 5 days of learning in both Notch1 cKO and control mice (p<0.0001), although latency was greater in the mutants (Figure 5E, p<0.01), despite the average swim speed being comparable (p=0.4). A learning deficit was also seen in the Notch1 cKO mice when subjected to reversal learning (Figure 5F). In both cases, 24 hours after the last learning session, mutant and control mice spent more time in the target quadrant (Figure S7A,B). Thus, Notch1 cKO mice can learn using spatial cues, although they do so more slowly than wild types.
In line with the previous report on the Notch1+/− mice (Costa et al., 2003), we could not detect any difference between Notch1 cKO and controls in contextual fear-conditioning 24 hours after a shock was delivered (Figure S7C). In addition, Notch1 cKO mutant mice displayed normal motor coordination (rotarod test), motor activity (open field test), and anxiety levels (elevated plus maze)(Figure S7C).
We have shown that Notch1 co-localizes with PSD95 in cultured neurons, and that the transcriptionally active form of the receptor, NICD1, is present at the synapse. In addition, we have shown that Jag1 is present in axons, localizes to synapses, and is upregulated in response to neuronal activity. Stimulation of neurons in culture, in hippocampal slices, or in vivo after exposure to a novel environment all lead to increased Notch1 expression and signaling. The notion that activity-dependent γ-secretase-mediated Notch receptor activation can occur at the synapse is consistent with recent work showing that synaptic γ-secretase activity cleaves EphA4 in response to neuronal activity (Inoue et al., 2009).
The activity-regulated neuronal Notch signaling we have identified both in vitro and in vivo is heavily dependent upon Arc. In Arc mutant neurons we observe a drastic reduction in the S3 cleaved form of Notch1, indicating that the γ-secretase-mediated processing is disrupted in the absence of Arc function. Furthermore, our rescue and co-IP experiments indicate that the role of Arc in mediating Notch1 activation requires association with Endophilin, and that Arc exists in a protein complex with Notch1 and Dynamin. Thus, in addition to its role in AMPA receptor trafficking (Chowdhury et al., 2006; Shepherd et al., 2006), Arc appears to regulate synaptic plasticity through interactions with the Notch pathway.
We next probed the potential function of activity-induced Notch signaling by conditionally deleting Notch1 in CA1 of the adult hippocampus. This model is an improvement over the Notch1+/−, CBF1+/− (Costa et al., 2003) and Notch1 antisense mice (Wang et al., 2004), because deletion occurs after development is complete. Ablation of Notch1 in pyramidal CA1 neurons affects both spine density and morphology, and the electrophysiological properties of mutants are altered, with both synaptic potentiation and depression reduced. Our LTP result is consistent with reduced potentiation resulting from decreased Notch1 expression (Wang et al., 2004), or conditional γ-secretase disruption (via ablation of Presenilin 1/2)(Saura et al., 2004). However, our LTD result differs from those previous studies, the former of which found enhanced LTD, and the latter of which found no change in LTD. However, the prior studies were confounded by possible developmental defects (Wang et al., 2004), and by lack of specificity with respect to Notch signaling (Saura et al., 2004).
Finally, to assess the effect of Notch disruption on learning and memory processes in hippocampal networks, we tested the Notch1 cKO mice using numerous behavioral paradigms. In the absence of Notch1, learning and rapid memory retrieval of newly presented cues are affected, whereas memory after repetitive learning is not. A function for Notch in rapid processing is consistent with the increase in Notch activation in hippocampal networks that occurs shortly after sensory input.
In summary, we have shown that Notch signaling is highly dynamic in mature neurons, and that it is induced in response to neuronal activity both in vitro and in vivo. In addition, we have identified the activity-regulated gene Arc as a novel context-dependent regulator of Notch signaling, and have shown that Arc is required for the γ-secretase-mediated activation of Notch1 in response to neuronal activity. Finally, using conditional disruption we have shown that Notch1 is required for normal spine morphology, synaptic plasticity, and memory processing.
All mice were maintained in accordance with the Institutional Animal Care and Use Committee (IACUC) at Johns Hopkins University School of Medicine. Generation of Arc mutant mice has been previously described (Plath et al., 2006). Notch1cKO and wild-type littermate control (Notch1flox/+, Notch1flox/flox and CamKII-Cre), mice were obtained by crossing Notch1flox/flox mice on CD1 background to the CamKII-Cre (T29-1) mouse line on C57BL6/129 background (Tsien et al., 1996).
Novel spatial exploration: Cage control mice (t=0 hr) were killed directly from their home cages, whereas the experimental mice performed a five-minute exploration session, and were returned to their home cage prior to analysis at the given time point. Novel object recognition was done accordingly to a published protocol (Bevins and Besheer, 2006). In the Y-maze mice were videotaped and scored for time spent in each arm and number of entries in each arm using the stopwatch+ software. The Social Interaction testing was carried out in three sessions using a three-chambered box with openings between the chambers. The Morris Water Maze test was done accordingly to a published protocol (Vorhees and Williams, 2006). Details for all behavioral tests are provided in the SI.
Neuronal cultures were prepared from the hippocampus of E17.5 embryos and plated on poly-L-lysine coated 60 mm dishes or 18 mm glass cover slips. Neurons were exposed to pharmacological manipulations after 14 days in vitro (DIV). For Sindbis virus infection, the pSinRep5 vector (Invitrogen) was used to generate viruses expressing either full-length Arc, or a non-functional form with residues 91–100 deleted (Chowdhury et al., 2006).
Synaptosomal fractions were prepared as previously described (Blackstone et al., 1992). Standard Western blot protocols were used. Details regarding fractionation, IP and WB protocols is provided in the SI. Quantitation of individual protein bands was done using ImageJ software. Values were averaged between experiments, and Student’s T-test was used to compare samples.
A complete list of the antibodies used can be found in the SI. Brain tissue and neuronal cultures were fixed in 4% PFA, and post-fixed in ice-cold acetone-methanol (1:1) at −20°C for 10 minutes. The immunostainings with rabbit anti-Arc and anti-Notch1 antibodies were performed using the TSA fluorescence amplification kit (Perkin Elmer). ImageJ software (NIH) was used to quantify fluorescence intensity of immunostainings with NICD1 (Figure 2A), EGFP (Figure S3B) and Notch1 (see Figure 3C,D legend). Student’s T-test was used to determine p values.
Golgi-cox staining (FD NeuroTechnologies) was performed according to the manufacturer instructions. Dendrite and spine lengths/widths were measured using Reconstruct software by the Neural Systems Laboratory (http://www.bu.edu/neural/Reconstruct.html). Spine length and width data was analyzed using the Kolmogorov-Smirnov statistical test.
Transverse hippocampal slices (350 μm) were prepared from Notch1 cKO and control mice, and maintained in artificial cerebrospinal fluid at room temperature. Data were collected using an Axopatch 1D amplifier (Molecular Device); signals were filtered at 2 kHz, digitized at 10 kHz and analyzed using pCLAMP 8 software (Molecular Device).
The authors thank Jason Shepherd, Richard Flannery, Marlin Dehoff, Vera Goh, and Keejung Yoon for technical and intellectual input during the course of this project. We also thank Ted Dawson and Jay Baraban for critically reading the manuscript. Funding for this work came from the Institute for Cell Engineering at Johns Hopkins University (N.G), a NARSAD Young Investigator Award (N.G), the James S. McDonnell Foundation (N.G.) and from the National Institute of Mental Health (P.F.W.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.