|Home | About | Journals | Submit | Contact Us | Français|
Synaptic plasticity, the capacity of neurons to change the strength of their connections with experience, provides a mechanism for learning and memory in the brain. Long-term plasticity requires new transcription, indicating that synaptically generated signals must be transported to the nucleus. Previous studies have described a role for importin nuclear transport adaptors in mediating the retrograde transport of signals from synapse to nucleus during plasticity. Here, we investigated the possibility that stimulus-induced translocation of importins from synapse to nucleus involves activity-dependent anchoring of importins at the synapse. We show that importin a binds to a nuclear localization signal (NLS) present in the cytoplasmic tail of NR1-1a. This interaction is disrupted by activation of NMDA receptors in cultured neurons and by stimuli that trigger late-phase, but not early phase LTP, of CA3-CA1 synapses in acute hippocampal slices. In vitro PKC phosphorylation of GST-NR1-1a abolishes its ability to bind importin α in brain lysates, and the interaction of importin α and NR1 in neurons is modulated by PKC activity. Together, our results indicate that importin α is tethered at the PSD by binding to the NLS present in NR1-1a. This interaction is activity dependent, with importin α being released following NMDA receptor and phosphorylation rendering it available to bind soluble cargoes and transport them to the nucleus during transcription-dependent forms of neuronal plasticity.
We, and others, have previously shown that the classical nuclear import pathway provides one manner by which synaptically generated signals can reach the nucleus (Thompson et al., 2004; Otis et al., 2006; Lai et al., 2008; Dieterich et al., 2008; Jordan and Kreutz, 2009; Perry and Fainzilber, 2009). Proteins bearing a nuclear localization signal (NLS), are recognized by a nuclear transport adaptor protein, importin α, which forms a heterotrimeric complex with the nuclear transporter importin β1 (Goldfarb, 2004). This complex docks at the nuclear pore and undergoes facilitated transport into the nucleus. We found that importin α1 and 2 were present in post-synaptic density (PSD) fractions of mouse brain and that activation of NMDA receptors triggered translocation of importins α and β1 into the nucleus of cultured hippocampal neurons. In Aplysia sensory neurons, inhibition of importin-mediated transport blocked long-term facilitation without affecting basal synaptic transmission or short-term facilitation (Thompson et al., 2004). Collectively, these data indicate that importin-mediated signaling is required for long-term synaptic plasticity.
How do importins localize to the synapse to be available to transport stimulus-activated cargoes? Here, we explored the possibility that synaptic localization is mediated by importin binding to a resident PSD protein. Towards this end, we compared a proteomic PSD database (Husi H, 2001) with a database of putative NLSs (http://cubic.bioc.columbia.edu/predictNLS/ (Cokol et al., 2000) and identified 28 NLS-containing PSD proteins (Table S1). Of these, NR1-1a, a splice variant of the NMDA receptor NR1 subunit, was particularly interesting given our earlier finding that importin translocation to the nucleus was triggered by NMDA receptor activation (Thompson et al., 2004). Of further interest, the NR1-1a splice variant of NR1 has been shown to be specifically required for efficient transcriptional responses to NMDA receptor activation (Bradley, 2006). Finally, the NLS in NR1-1a is known to be functional, as it can direct the nuclear import of a normally cytosolically-restricted protein (Holmes, 2000). Together, these findings suggested that importin α binding to the NLS in NR1-1a might serve to localize importins to the PSD.
The NLS in NR1-1a is flanked by three protein kinase C (PKC) and one cAMP activated protein kinase (PKA) phosphorylation sites (Tingley et al., 1997). Phosphorylation of residues flanking NLSs can change the affinity of importin for its cargo proteins (Poon and Jans, 2005). Phosphorylation could thus modulate the binding of importin α to NR1, thereby regulating the anchoring of importins at synapses.
In the present study we show that importin α binds specifically to a NLS present in the cytoplasmic tail of the NR1-1a subunit of the NMDA receptor. This interaction is regulated by activity; binding is significantly reduced by stimuli known to produce long-lasting synaptic plasticity. Phosphorylation of residues within and flanking the NLS in NR1 interferes with the binding of importin α to NR1. Together, our findings indicate that importin α is anchored at synapses by binding to the NLS in NR1-1a, and that this binding regulated in an activity- and phosphorylation-dependent manner during transcription-dependent plasticity.
Antibodies used include: rabbit anti-importin α1 and importin α2, gifts from Marian Waterman (UC Irvine, CA); rabbit anti-Rch1, Bethyl Labs (Montgomery, TX); custom made rabbit polyclonal anti-isoform-specific importin α (described in SI); rabbit anti-synaptophysin and anti-MAP2, Chemicon (Temecula, CA); mouse anti-MAP2, Sigma (St. Louis, MO, clone HM-2); mouse anti-NR1 c-terminus, Millipore (Billerica, MA), rabbit anti-NMDAR1 C1 cassette, AbCam (Cambridge, MA); rabbit anti-NR1 pSer896 (Calbiochem San Diego, CA); rabbit anti-NR1 pSer890, pSer896, pSer897, Cell Signaling (Danvers, MA); mouse anti-FLAG, Sigma (St. Louis, MO); mouse anti-calmodulin, Upstate (Billerica, MA); mouse anti-PSD-93 (Chapsyn-110), NeuroMab (Davis, CA); and mouse anti-GAPDH, AbCam (Cambridge, MA).
Primary neuronal cultures from P0 Sprague Dawley rats grown in defined media were prepared as previously described (Lai, 2008); detailed protocol is available at http://www.biolchem.ucla.edu/labs/martinlab/. For glutamate stimulation, neurons (21DIV) were silenced for 6h with 1μM TTX then stimulated by replacing the media with media containing 40μM glutamate + 1μM TTX for 5 min. For PKC manipulation, neurons (21DIV) were silenced for 6h with 1μM TTX then incubated with 20nM PMA for activation or preincubated for 30 min with 5μM chelerythrine for inhibition or vehicle in 37°C incubator for 5min before fixation. For PKA manipulations after silencing as above were activated with 25μM forskolin or inhibited with 60μM KT5720 for 30 min before glutamate. For biochemistry, after 6h silencing with 1 μM TTX, PKA or PKC were activated with 25μM forskolin or 20nM PMA for 10 min then lysed in 1% Triton X HNTG.
Synaptosomes and PSD fractions were prepared from P21-28 Sprague Dawley rats as described at http://www.biolchem.ucla.edu/labs/martinlab/.
CA1 “mini-slices” from 8–10-week-old male C57BL/6 mice were used as previously described (Ho et al.). For experiments involving APV, 100 μM D-APV was bath applied for 30 min prior to stimulation. High-frequency stimulation (HFS) consisted of either one or two trains of 100 Hz stimulation (1 sec. in duration) with an inter-train interval of 10 sec. A single pulse of presynaptic fiber stimulation was delivered 5 seconds post-HFS to ensure that HFS induced a 2-fold or greater increase in population spike amplitude.
Cultures were processed as previously described (Thompson, 2004) except that 0.5% Triton X-100 was used.
Proteins were resolved by 10% SDS-PAGE, transferred to a polyvinylidene fluoride membrane, blocked in TBS-T 5% milk or 5% BSA with phosphatase inhibitors and probed with primary antibodies, HRP-conjugated secondary antibodies and detected by chemiphospholuminescence.
Images were obtained using a Zeiss Pascal scanning laser confocal microscope with 40x or 63x oil immersion objectives. All images within a single experiment were acquired using equivalent settings by an individual who was blind to treatment conditions. Images were analyzed using LSM Pascal or Universal Imaging Metamorph software. Neurons were selected at random from each quadrant of the coverslip by MAP2 staining. Nuclei were manually outlined based on absence of MAP2 signal. Nuclear to cytoplasmic ratio was determined using Slidebook (Intelligent Imaging Innovations, Inc, Santa Monica, CA) software by subtracting nuclear signal from total somatic signal. Colocalization of NR1 and importin α was determined using Slidebook software. Colocalization in NR1 containing puncta was determined using Slidebook by thresholding all NR1 signal to include true synaptic puncta in size and brightness. Importin a signal was then collected from these puncta after thresholding above background. Signal intensity of IBs were measured with ImageQuant software (General Electric) and normalized to control bands.
All statistical analysis was done using SPSS software (Chicago, IL).
GST was fused to the N terminus of NR1-1a and NR1-2a cytoplasmic tails in pGEX vector (Amersham, Piscataway, NJ). FLAG-importin α1 were previously described (Lai, 2008).
Rat forebrains were homogenized in HNTG lysis buffer (50mM HEPES pH 7.3, 150mM NaCl, 1.5mM MgCl2, 10% Glycerol, 1mM EDTA, 1% Triton-X, 100 pH 7.5) containing protease inhibitor cocktail (Roche, Nutley, NJ), NaF (10mM) and NaOrthovanadate (1mM), spun at low speed (1000 × g) for 10 min and high speed (10,000 × g) for 10 min. Lysates were precleared with protein A (for rabbit antibodies) or protein G (Amersham, for mouse antibodies) for 2h at 4°C and incubated with 10μg/mL concentration of antibodies and 25μL of a 50% slurry of either protein-A or G beads at 4°C overnight. IPs were washed with lysis buffer and eluted by boiling in Laemlli buffer. Protein concentration determined by Bradford Assay (BioRad, Hercules, CA).
GST pulldowns in transfected HEK cells and brain lysates were performed as previously described (Lai, 2008) except for the use of modified RIPA as wash buffer (50 mM Tris pH 7.9, 100 mM KCl, 0.25 % Na deoxycholate, 0.6 mM EDTA, 1% NP-40). Competition with NLS peptide was carried out as previously described (Lai, 2008). For in vitro phosphorylation, GST fusion proteins were incubated in kinase buffer (250mM MES, 1mM EGTA, 10mM MgCl2, pH 6.0), 20mM ATP, with or without 1μL (7.5 units) recombinant active PKCM (Calbiochem), in an agitated 30°C water bath for 30 min.
To determine whether importin α and NR1 colocalize at synapses, and whether colocalization is activity-dependent, we performed double label immunocytochemistry (ICC) of cultured rat hippocampal neurons (21 DIV) with anti-importin α1 and anti-NR1 antibodies. As shown in Fig 1a, importin α1 was present in 55% of NR1-immunoreactive dendritic spines in cultured neurons in which activity had been silenced with tetrodotoxin (TTX). A brief stimulation with glutamate (40 μM, 5 min or 15 min) significantly decreased both the percent of NR1-immunoreactive spines containing importin α1, and the concentration of importin α1 in NR1-immunoreactive spines (Fig 1b). As we have previously reported (Thompson et al., 2004), and as shown in supplemental figure 3, 25 min after glutamate stimulation (40 μM, 5 min), importin α1 accumulates in the nucleus, and is largely absent from synapses.
In complementary biochemical assays, both importin α and NR1 were present in synaptosome and PSD fractions prepared from adult rat forebrain (Fig 1c). We investigated interactions between the two proteins at synapses by co-immunoprecipitation (co-IP) from synaptosomes. As shown in Fig 1d, importin α was present in NR1 immunoprecipitates (IPs), and NR1 was present in importin α IPs. Neither importin α nor NR1 were present in parallel control IPs with mouse anti-GFP antibodies and rabbit pre-immune serum, respectively.
The NR1 subunit is alternatively spliced from a single gene into eight splice variants that differ in the composition of their cytoplasmic tails (Hollmann et al., 1993). The NR1-1a splice variant is characterized by inclusion of exon 21, encoding the C1 cassette, which contains the NLS. To determine whether importin α specifically binds the NLS in NR1-1a, we generated GST fusion proteins of the cytoplasmic tail of NR1-1a and NR1-2a, which are identical except that NR1-1a contains the NLS-containing C1 cassette and NR1-2a does not (Fig 2a). GST-pulldown experiments in rat forebrain lysates revealed that GST-NR1-1a, but neither NR1-2a nor GST alone, bound importin α (Fig 2c). In control experiments, both GST-NR1 fusion constructs bound calmodulin (Fig S1a), consistent with published results (Ehlers MD, 1996). In a complementary set of experiments, we found that GST-NR1-1a pulled down FLAG-tagged importin α heterologously expressed in HEK 293-T cells (Fig S1b).
To confirm that importin binds the NLS in NR1-1a, we conducted competition experiments with peptides containing the NLS of NFκB (VQRKRQKLM), which has been shown to bind importin α and block nuclear translocation of importin α cargoes (Torgerson et al., 1998; Lai, 2008). A mutant peptide (mNLS) with two amino acid changes in the NLS (VQRNGQKLMP) does not bind importin α and served as a negative control. Preincubation of lysates with the NLS peptide effectively interfered with importin binding to NR1 whereas mNLS peptides did not (Fig 2d), providing evidence that importin α binds NR1-1a through its NLS binding sites.
To determine whether the interaction between importin α and NR1 is regulated by activity, we monitored importin α-NR1 interactions in primary cultured cortical neurons following stimuli known to trigger nuclear translocation of importins to the nucleus (Thompson, 2004). Cortical cultures (21 DIV) were silenced with TTX (1μM, 6 hrs), incubated with 40uM glutamate (or vehicle) for 5 min, lysed and processed for co-IP with NR1 or importin α antibodies. In TTX-silenced neurons, importin α was present in NR1 IPs (Fig 3a) and NR1 was present in importin α IPs (Fig S5). This interaction was significantly decreased following glutamatergic stimulation. Glutamate did not alter the total concentration of either importin α or NR1 (input lanes, Fig 3a).
We next asked whether stimuli that elicit long-lasting, transcription-dependent forms of hippocampal plasticity altered binding of importin α to NR1. Experiments were performed in CA1 minislices (Nayak et al., 1996) using a large bipolar stimulating electrode that spanned the full extend of stratum radiatum to activate Schaeffer collateral fiber inputs onto CA1 pyramidal cells. Presynaptic fibers were stimulated with a single train of high-frequency stimulation (HFS, 100 Hz for 1 s) to induce early LTP or with two trains of HFS (inter-train interval=10 s), a protocol that induces protein synthesis-dependent late-phase LTP (Osten et al., 1996) as well as transcription-dependent potentiation (R. Blitzer, personal communication). These lysates were used for co-IP experiments. As shown in Fig 3b, NR1 antibodies co-immunoprecipitated importin α in unstimulated slices and in slices receiving a single train of HFS, but not in slices receiving two trains of HFS. As a control, we also examined the concentration of PSD93, since PSD93 has been shown to be present in NR1 IPs (Al-Hallaq et al., 2007) and since its concentration in PSDs is stable with activity (Ehlers, 2003). Equal amounts of PSD93 were present in NR1 IPs from stimulated and unstimulated slices (Fig 3b). Tetanus-induced disruption of importin α-NR1 binding required NMDA receptor activation, since it was blocked by preincubation with APV (Fig 3c).
Taken together, these experiments indicate that the binding of importin α to NR1 is regulated by activity. Binding occurs in unstimulated synapses, is maintained following stimuli that produce short-term neuronal plasticity, but is lost following stimuli that produce long-term, transcription-dependent neuronal plasticity. Further, the loss of binding is specifically dependent on NMDA receptor activation.
We next asked whether importin α binding was regulated by phosphorylation of NR1. Phosphorylation of residues flanking the NLS has been shown to either inhibit or promote binding of the NLS-bearing cargo to importin α (Poon and Jans, 2005). The NLS in NR1-1a is flanked by three experimentally demonstrated PKC phosphorylation sites (ser 889, ser 890 and ser 896) and one experimentally demonstrated PKA phosphorylation site (ser 897) (Tingley et al., 1997). Using phospho-specific antibodies, we found that phosphorylation at ser 890 was increased following 2×100 Hz stimulation of acute hippocampal slices (Fig S7). This finding, together with a report that PKC-mediated phosphorylation of ser 896 is significantly increased following stimuli that elicit LTP of CA3-CA1 synapses (Moriguchi et al., 2006), further suggested that PKC and/or PKA phosphorylation of the cytoplasmic tail of NR1-1a could regulate its interaction with importin α.
We initially focused on PKC phosphorylation by in-vitro phosphorylating GST-NR1-1a with recombinant PKC and using these constructs for GST pulldown assays in brain lysates. As shown in Fig 4a, significantly less importin α bound PKC-phosphorylated GST-NR1-1a than non-phosphorylated GST-NR1-1a. We then used recombinant PKA to in-vitro phosphorylate GST-NR1 constructs and conducted GST-pulldown assays in brain lysates. As shown in figure 4b, phosphorylation of NR1 by PKA did not interfere with the binding between importin α and NR1. Immunoblotting with phospho-specific revealed that serines 890, 896 and 897 were phosphorylated following in vitro incubation with PKC, and that only serine 897 was phosphorylated following incubation with PKA (Fig S6a). Together, these results indicate that PKC phosphorylation of serines 890 and/or 896 (and possibly serine 889, for which phospho-specific antibodies are not available) disrupts the association of importin α with NR1-1a.
To determine whether and how endogenous PKC activation affects the interaction between importin α and NR1 at synapses, we blocked PKC during glutamatergic stimulation of rat hippocampal neurons (21 DIV) and monitored synaptic localization of importin α by double label ICC with anti-importin α1 and anti-NR1 antibodies. As shown in Fig 5b, importin α1 was present in 44% of NR1-immunoreactive puncta in cultured neurons in which activity had been silenced with tetrodotoxin (TTX). Consistent with data shown in figure 1, a brief stimulation with glutamate (40 μM, 8 min) significantly decreased both the percent of NR1-immunoreactive spines containing importin α1 (Fig 5b), and the concentration of importin α1 in NR1-immunoreactive spines (Fig S8b). Incubation of cultures with the specific PKC inhibitor chelerythrine (5 μM) for 30 minutes prior to and during stimulation with glutamate blocked the decrease in importin α immunoreactivity in NR1-containing spines, indicating that glutamate-induced loss of importin α from synapses requires PKC activity (Fig 5b).
In a complementary set of experiments, we incubated dissociated rat cortical cultures (21 DIV) with TTX for 6 hrs, and then applied phorbol esters (20nM, 5 min) to activate PKC. Cultures were lysed and used for co-IP experiments. As shown in Fig 5c, incubation with phorbol esters decreased the amount of importin α present in NR1 IPs. Of note, no difference in binding between importin α and NR1 was seen following activation of PKA with forskolin (25μM, 5 min) (Fig S6b). Together with the in vitro phosphorylation data (fig 4a), these data support the idea that activity-dependent PKC phosphorylation of serines flanking the NLS disrupts the interaction between importin α and NR1.
Long-lasting activity-dependent changes in synaptic efficacy require new transcription and thus involve the transport of signals from synapse to nucleus. While electrochemical signaling allows neurons to rapidly communicate between compartments, soluble signaling molecules have also been shown to translocate from synapse to nucleus to trigger changes in gene expression (Otis et al., 2006; Cohen and Greenberg, 2008). We, and others, have described a role for importin nuclear transporters in relaying signals from distal processes to the nucleus (Hanz, 2003; Thompson, 2004; Dieterich, 2008; Lai, 2008; Jordan and Kreutz, 2009). In this study, we present data showing that importin α binds to a bipartite NLS in the NR1-1a subunit of the NMDA receptor. Our findings are consistent with this interaction serving to tether importin α at the PSD in an activity-dependent manner. Specifically, our results suggest that NMDA receptor stimulation activates PKC which phosphorylate residues surrounding the NR1 NLS, decreasing the affinity of the importin-NR1 interaction and thereby releasing importin α so that it is available to transport soluble cargo proteins from the synapse to the nucleus. Of note, the NLS and flanking serines in the NR1 subunit are highly conserved across species (Fig S4).
A precedent for stimulus-dependent anchoring of importins at the membrane through interactions with membrane proteins exists in hepatocytes, where importin α2 binds the Glut2 transporter in a manner that is modulated by glucose concentrations (Guillemain et al., 2002). We also considered the possibility that the NR1-1a cytoplasmic tail was proteolytically cleaved with subsequent importin-mediated nuclear translocation of the cytoplasmic domain. In fact, another transmembrane protein present in both the PSD proteome and the putative NLS database, the L-type voltage gated calcium channel Ca(V)1.2, has been reported to be cleaved by calpain (De Jongh et al., 1994) and to translocate to the nucleus (Gomez-Ospina et al., 2006). Our finding that activity decreases the interaction between NR1 and importin α (Fig 3) argues against cleavage of NR1 and subsequent importin-mediated nuclear transport. Two additional experiments indicate that NR1 does not undergo activity-dependent cleavage. First, we immunoblotted lysates of hippocampal slices with NR1 antibodies and did not detect smaller products or decreases in full length NR1 following 2×100Hz stimulation (Fig S2a). Second, ICC of cultured hippocampal neurons with antibodies directed against the C1 or C0 cassette of NR1 did not reveal any nuclear immunoreactivity in the presence or absence of glutamate stimulation (Fig S2b). This is consistent with an earlier report that full-length NR1 with a c-terminal GFP tag expressed in neurons localized to the cytosol and plasma membrane, with no nuclear GFP signal (Marsh et al., 2001).
The NMDA receptor plays a critical role in learning-related synaptic plasticity (Malenka and Bear, 2004), and as such is an ideal candidate anchor for importin α. The NR1 subunit undergoes alternative splicing to generate receptors with distinct cytoplasmic tails, which endow the receptor with distinct properties (Durand et al., 1992, 1993; Sugihara et al., 1992) for review see (Zukin and Bennett, 1995). NR1 splicing is regulated developmentally (Monyer et al., 1994; Sheng et al., 1994) and by activity (Mu et al., 2003). Alternative NR1 splicing has been shown to regulate membrane expression of the receptor (Perez-Otano and Ehlers, 2005). Thus, the C0 cassette has been reported to mediate endocytosis and degradation (Scott et al., 2004) and sequences in the C1 cassette, contained within the NLS, have been reported to function as ER retention signals (Scott et al., 2001; Xia et al., 2001). NR1 ER retention is regulated by phosphorylation of residues flanking the ER retention signal and by sequences within the C2 cassette, leading to changes in the way that the NMDA receptor is trafficked and targeted (Standley et al., 2000; Scott et al., 2001; Carroll and Zukin, 2002; Scott et al., 2003). Further, distinct NR1 splice variants have been reported to interact with distinct proteins: the C0 cassette has been shown to bind calmodulin (Wyszynski et al., 1997; Zhang et al., 1998), α-actinin (Wyszynski et al., 1997) and CamKIIα (Bayer and Schulman, 2001; Leonard et al., 2002); and the C1 cassette has been reported to bind neurofilament L (Ehlers et al., 1998) and yotiao (Lin et al., 1998). We do not know whether importin α binding to NR1 modulates membrane trafficking of the receptor, and/or whether it competes with or regulates the binding of other NR1-interacting proteins. It is probable, however, that the interaction of NR1 with its binding partners is highly regulated. The stimulus-induced release of importin α from NR1 likely allows other NR1-interacting proteins to bind. In this regard, it is appealing to speculate that the regulated importin α–NR1 interaction provides a mechanism for tagging synapses during transcription-dependent plasticity (Frey, 1997), with stimulated, “tagged” synapses being containing NR1 subunits that no longer bind importin α.
Activity-dependent tethering of importin α at the synapse by regulated binding to NR1 provides an elegant mechanism for integrating synaptic events with nuclear signaling (for our model see Fig 5). We expect, however, that it represents only one of multiple synapse-to-nucleus signaling pathways. Thus, importin α isoforms may interact with additional receptors and thereby be responsive to a range of stimuli that elicit transcription, including neuromodulators such as dopamine and inhibitory inputs. Future studies aimed at detailed molecular dissection of the importin α–NR1 interaction, at identification of the synaptically localized cargoes that importins carry to the nucleus, and at characterization of additional synapse-to-nucleus transport pathways promise to elucidate the cell biological mechanisms whereby synaptic stimulation leads to transcription-dependent, persistent changes in neuronal function.
We thank K. Olofsdotter Otis for initial comparison of PSD and NLS databases, J. Boulter for full length GFP-NR1-1a and 2a cDNAs, M. Honsberger and M. Mason for advice on statistical analyses, D. Black, F. Schweizer, L. Zipursky and members of the Martin lab for helpful discussions. This work was supported by NIH grant R01MH077022 (to KCM), R01MH609197 (to TJO), NIH training grant T32GM007185 (to RAJ) and a University of California Dissertation Year Award (to RAJ).