|Home | About | Journals | Submit | Contact Us | Français|
Homeostatic synaptic scaling alters the strength of synapses to compensate for prolonged changes in network activity, and involves both excitatory and inhibitory neurons. The immediate-early gene termed Narp (Neuronal activity-regulated pentraxin) encodes a secreted synaptic protein that can bind and cluster AMPA receptors (AMPARs). Here, we report that Narp prominently accumulates at excitatory synapses on Parvalbumin-expressing interneurons (PV-INs). Increasing network activity results in a homeostatic increase of excitatory synaptic strength onto PV-INs that increases inhibitory drive, and this response is absent in neurons cultured from Narp knock-out (Narp−/−) mice. Activity-dependent changes in the strength of excitatory inputs on PV-INs in acute hippocampal slices are also dependent on Narp, and Narp−/− mice display increased sensitivity to kindling-induced seizures. We propose that Narp recruits AMPARs at excitatory synapses onto PV-INs to rebalance network excitation/inhibition dynamics following episodes of increased circuit activity.
Long-lasting changes in synaptic strength underlie information storage within the central nervous system. Within the hippocampus, Hebbian long-term potentiation (LTP) and long-term depression (LTD) provide neurons with an effective use-dependent means for modification of individual synapses. However, the positive feedback nature of these processes makes them inherently unstable1. Additionally, for LTP or LTD to occur, basal synaptic strength must be maintained within an optimal range to prevent occlusion of further increases or decreases in activity2, 3. Therefore, bidirectional homeostatic feedback mechanisms are critical to provide long-term stability of networks and to ensure their potential for plasticity.
Immediate-early genes (IEGs) are dynamically regulated by forms of synaptic activity that underlie information processing and storage, making them excellent candidates to contribute to both Hebbian and homeostatic plasticity. For example, Activity-regulated cytoskeleton-associated protein (Arc, also known as Arg3.1) is a cytosolic protein that associates with Endophilin and Dynamin and increases the rate of endocytosis of AMPA receptors (AMPARs) at synapses on pyramidal neurons4. Steady state levels of Arc increase or decrease in parallel with changes in neuronal activity and contributes to bidirectional control of homeostatic scaling of AMPAR on pyramidal neurons5. Arc also contributes to synapse-specific mGluR-LTD in a process that involves the rapid de novo translation of Arc mRNA6.
Neuronal activity-regulated pentraxin (Narp, also known as Neuronal pentraxin 2) is another IEG that can alter synaptic function. Narp is a member of the neuronal pentraxin (NP) family of calcium-dependent lectins that includes Neuronal pentraxin 1 (NP1) and Neuronal pentraxin receptor (NPR)7. Of these, only Narp is regulated as an IEG8. Narp and NP1 are secreted proteins, while NPR possesses an N-terminal transmembrane domain9. On the extracellular surface, these NPs form large, organized heteromeric complexes, stabilized via disulfide bond linkages8. NPs localize to excitatory synapses where their conserved, C-terminal pentraxin domains can interact with the N-terminal extracellular domain of AMPARs10. These features underlie the contribution of NPs in various forms of synaptic plasticity. For example, axonally derived NP1 and NPR are critical for the recruitment of AMPARs to both artificial and native synapses10. Additionally, NPR plays an essential role in mGluR-LTD in a process that involves activation of the extracellular metalloprotease TACE (TNF-α converting enzyme), cleavage of NPR near the transmembrane domain, and rapid endocytosis of NPR and AMPAR11. At the systems level, NPs are important for the activity-dependent segregation and refinement of eye-specific retinal ganglion cell projections to the dorsal lateral geniculate nucleus12.
Here, we found that Narp was highly enriched at excitatory synapses present specifically on Parvalbumin-expressing interneurons (PV-INs) and its expression was dynamically regulated by network activity. Accumulation of Narp at these synapses resulted from its secretion from presynaptic excitatory neurons and required the presence of perineuronal nets surrounding PV-INs. Narp increased synaptic strength at PV-IN excitatory synapses, both in culture an in the acute hippocampal slice, by regulating levels of GluR4-containing AMPARs in an activity-dependent manner. Mice lacking Narp displayed a marked increase in sensitivity to kindling-induced seizure. Together, these results demonstrate that Narp contributes to homeostatic plasticity of interneurons and suggests a key role in the activity-dependent recruitment of PV-IN-mediated inhibition.
We examined Narp protein expression by surface labeling primary hippocampal cultures prepared from embryonic day 18 (E18) mice after 14–17 days in vitro (DIV). Narp immunocytochemical (ICC) staining was markedly enriched on a small subpopulation of large neurons with complex dendritic branches (Fig. 1a). Lower levels of Narp were distributed broadly on the majority of neurons. Based on its expression pattern, we asked if Narp preferentially accumulated onto interneurons. Interneurons represented ~10% of neurons within our hippocampal culture preparations and included distinct subtypes (unpublished observation). We performed ICC with antibodies against the calcium-binding proteins Parvalbumin (PV), Calretinin, and CAMKIIα, which represent non-overlapping neuronal subpopulations13. Pyramidal neurons expressing CAMKIIα, as well as Calretinin-expressing interneurons, displayed similar, low levels of Narp on the surface of their dendrites, while dendrites of PV-expressing interneurons (PV-INs) exhibited 10-fold higher levels of surface Narp (Fig. 1b,c). A similar enrichment of Narp was seen in PV-INs within the hippocampus in vivo. (Fig. 1d)
Narp is present at excitatory synapses on cultured spinal cord and hippocampal interneurons14, 15. To assess if Narp is at excitatory synapses on PV-INs, we labeled cultures with antibodies against Narp, Parvalbumin, and the excitatory synaptic scaffolding protein PSD95. Narp co-localized with PSD95, indicating that Narp localized to excitatory synapses on PV-INs (Fig. 1e). Narp did not co-localize with the inhibitory synaptic marker GAD65 (Supplementary Fig. 1).
While Narp expression is activity-regulated in vivo7, we sought to confirm its activity-dependent expression in hippocampal cultures. ICC revealed that bicuculline treatment increased Narp protein levels ~6-fold on PV-INs within 24 hours and ~25-fold by 48 hours (Fig. 2a,b). Reciprocally, treatment of cultures with tetrodotoxin (TTX) led to an ~50% reduction of Narp on PV-INs by 4 hours and an ~80% decrease by 12 hours (Fig. 2a,b). Analysis by Western blot detected a prominent decrease in surface Narp in response to TTX, relative to control or bicuculline treatment (Fig. 2c,d). Bicuculline did not substantially increase surface Narp levels, consistent with the fact that PV-INs represent a small percentage of the total number of neurons, and that basal activity is high in the high density cultures used for biochemical analysis (Unpublished observation. Compare ~1×105 cells/cm2 for biochemistry vs ~1.4×104 cells/cm2 for immunocytochemistry. See methods). Consistent with previous results7, 16, Narp mRNA levels also underwent activity-dependent changes (Fig. 2e).
To examine how Narp becomes enriched at excitatory synapses on PV-INs, we asked whether synaptic Narp is preferentially derived from pre- or postsynaptic neurons. Using a co-culture approach, we labelled a small number of either WT or Narp−/− neurons with CM-DiI and plated them amongst a much larger number of unlabeled neurons of the other genotype. CM-DiI is a water-soluble and fixable derivative of DiI, a non-toxic, lipophilic, fluorescent dye that internalizes over time and persists in cultured neurons for up to 3.5 weeks17. When axons of presumed WT neurons contacted CM-DiI labeled Narp−/− PV-INs (“WT pre/−/− post”), Narp levels on Narp−/− interneurons were comparable to Narp levels on WT PV-INs within the same culture (“WT pre/WT post”) (Fig. 3a,c). This suggests that a presynaptic source of Narp is sufficient for Narp accumulation at PV-IN excitatory synapses. Moreover, when presumed Narp−/− axons contacted CM-DiI labeled WT PV-INs (“−/− pre/WT post”), there was a significant reduction of Narp on PV-INs (Fig. 3b,c). The residual Narp present on these WT PV-Ins was not due to non-specific background since it was absent in Narp−/− cultures (data not shown), and instead could be attributed to postsynaptic secretion or to a small number of WT axons in the culture innervating these interneurons. The latter possibility is supported by the observation that Narp−/− PV-INs within the same culture (“−/− pre/−/− post”) exhibited comparable levels of Narp on their dendrites. Taken together, these data indicate that Narp originates primarily from presynaptic neurons, consistent with previously published results14, 15, 18.
PV-INs comprise ~2.5% of the total neuronal population in our cultures (unpublished observation) and in the intact hippocampus19. Accordingly, excitatory synapses on PV-INs represent a small fraction of the total synapses in the culture, raising the question of how Narp could selectively accumulate at these synapses. One distinguishing trait of PV-INs is the presence of dense perineuronal nets, a specialized extracellular matrix consisting of chondroitin sulfate proteoglycans, glycoproteins, and glycosaminoglycans, which ensheath the soma, dendrites, and axon initial segment of PV-INs20. The polysaccharide side chains present on these molecules make them substrates for numerous plant-derived lectins including Vicia villosa agglutinin (VVA), Wisteria floribunda agglutinin, and soybean agglutinin21, 22.
Because Narp is a calcium-dependent lectin7, we speculated that these nets could provide a postsynaptic target to enhance accumulation of Narp. To test whether perineuronal nets play a role in Narp accumulation on PV-INs, we treated neuronal cultures with Chondroitinase ABC (ChABC), which degrades the glycosaminoglycan side chains of chondroitin sulfate proteoglycans and is used to disrupt perineuronal nets23 (Supplementary Fig. 3). Treatment with ChABC for 48 hours led to a ~50% reduction in surface Narp levels on PV-INs (Fig. 4a,b). Even with addition of bicuculline, we found a significant reduction of surface Narp in cultures that were simultaneously treated with ChABC (Fig. 4b). ChABC did not reduce surface Narp detected by western blot (Fig. 4c,d), consistent with the notion that Narp at PV-INs represent a small fraction of the total secreted Narp that is also present at lower levels on other neurons (Fig. 1b,c). Taken together, this suggests that surface Narp expression still occurs in the presence of ChABC. However, Narp is unable to accumulate onto PV-INs in the absence of perineuronal nets.
Since Narp binds and clusters AMPARs10, 15 and was highly enriched on PV-INs, we asked whether Narp could contribute to the regulation of either the GluR1 or GluR4 AMPAR subunits which PV-INs predominantly express24. A model in which Narp, an activity-regulated gene, regulates AMPAR levels on PV-INs, predicts that AMPAR levels should scale in direct relation with network activity. To test this hypothesis, we treated neuronal cultures with either TTX or bicuculline, prior to staining for AMPARs. Bicuculline treatment resulted in a two-fold increase in PV-IN GluR4 levels by 48 hours, while TTX resulted in a 65% decrease (Fig. 5a,b). The kinetics for these activity-dependent changes in GluR4 followed that of Narp, where increases in GluR4 levels progressed much slower than reductions (Fig. 5d). GluR4 levels from untreated Narp−/− PV-INs were ~50% lower than WT and failed to change in response to perturbations in activity (Fig. 5b,c). In contrast, GluR1, GluR2, and the NMDA Receptor subunit NR1, were not significantly different between WT and Narp−/− PV-INs (Supplementary Fig. 4).
Based on our finding that Narp accumulates on PV-IN synapses from presynaptic elements, we hypothesized that Narp from WT neurons might be sufficient to rescue GluR4 levels when deposited onto PV-INs lacking Narp. We tested this hypothesis by co-culturing WT and Narp−/− neurons and staining for GluR4. When we plated CM-DiI labeled Narp−/− PV-INs amidst surrounding WT neurons (“WT pre/−/− post”), we were able to partially rescue the reduced GluR4 levels on Narp−/− PV-INs (Fig. 5e,f). Importantly, when we reversed the genotypes and plated labeled WT PV-INs on surrounding unlabeled Narp−/− neurons (“−/− pre/WT post”), GluR4 levels were indistinguishable from Narp−/− PV-INs (Fig. 5f). In sum, these observations indicate that activity-dependent changes in Narp mediate parallel changes in GluR4 levels on PV-INs and implicate Narp in homeostatic scaling of these synapses.
To evaluate how changes in total AMPAR levels functionally relate to synaptic strength, we performed patch clamp analysis of spontaneous miniature excitatory postsynaptic currents (mEPSCs). We initially identified PV-INs via live application of fluorescein-conjugated VVA and then confirmed their identity through their characteristic non-accommodating, fast-spiking response to current injection25 (Supplementary Fig. 5a–c). Consistent with our ICC data, mEPSC amplitudes recorded from WT PV-INs changed in direct correlation with activity, and these changes were largely absent in Narp−/− PV-INs (Fig. 6a,b). In contrast, mEPSC frequency was not affected by changes in activity in either WT or Narp−/− PV-INs (Fig. 6c). Additionally, we found no changes in excitatory synapse number between WT and Narp−/− PV-INs as measured by double ICC labeling (Supplementary Fig. 6), indicating that PV-IN synaptogenesis proceeds independent of Narp. Together, these observations indicate that activity-dependent changes in Narp mediate parallel changes in synaptic strength of excitatory synapses on PV-INs and implicate Narp in homeostatic scaling of these synapses.
A model in which Narp regulates homeostatic adaptations of network activity via modulation of excitatory synapses on PV-INs predicts that the frequency of network-driven firing of PV-INs will be dependent on Narp. Accordingly, we compared spontaneous action potential frequency between WT and Narp−/− PV-INs. Consistent with our model, WT PV-INs spontaneously fired at more than double the rate of Narp−/− PV-INs (Fig. 6d,e). This difference was not attributable to changes in any membrane or action potential property we measured (Supplementary Fig. 7). Importantly, when we blocked synaptic transmission with gabazine, NBQX, and D-APV, we observed a complete elimination of action potentials (Fig. 6d), implying that all the spontaneous firing events were synaptically driven. These results suggest that the recruitment of GluR4 to excitatory synapses by Narp translates into increased activity of PV-INs.
To assess if the synaptic scaling deficit seen in Narp−/− cultures occurs in intact circuits in vivo, we examined excitatory input to PV-INs in acute hippocampal slices from WT and Narp−/− mice. We monitored spontaneous excitatory postsynaptic currents (sEPSCs) during whole-cell voltage clamp recordings from PV-INs situated in the dentate gyrus26. We initially selected presumptive PV-INs based on morphology and position at the dentate hilar border as well as fast-spiking properties. Then, we anatomically recovered and probed each recorded cell for PV expression, using biocytin processing combined with immunohistochemistry (Supplementary Fig. 5d,e). In contrast to cultured PV-INs, basal sEPSC amplitude and frequency were not significantly different between Narp−/− and WT PV-INs (Fig. 7). We hypothesized that this was due to low basal Narp levels in WT mice. Thus, in order to increase network activity, we administered maximal electroconvulsive seizure (MECS) to the mice ~12–16 hours prior to obtaining slices for recording. Consistent with prior observations15, 27, MECS significantly increased Narp levels in WT mice (Supplementary Fig. 8a). Moreover, in support of our working hypothesis, MECS administration to WT mice significantly increased the amplitude of sEPSCs recorded from PV-INs when compared to unstimulated controls (Fig. 7). This MECS-induced change was selective for PV-INs (Supplementary Fig. 8b–f) and was not associated with any other effects on sEPSC or firing properties examined (Fig 7b). We also probed hippocampal PV-IN GluR4 levels in vivo by immunohistochemistry and found a significant increase in GluR4 levels after MECS (Supplementary Fig. 9). Although the group average sEPSC kinetics did not change following MECS, we did observe the emergence of a negative correlation between sEPSC amplitude and decay time constant following MECS, consistent with recruitment of a fast GluR4-containing receptor population (Supplementary Fig. 10)28. Most importantly, MECS-induced plasticity of sEPSCs and GluR4 accumulation was absent in PV-INs from Narp−/− mice (Fig. 7 and Supplementary Figs. 9 and 10), confirming a role for Narp in activity-induced upregulation of excitatory input selectively onto PV-INs.
The importance of inhibitory networks in the suppression of seizures is well documented29. Thus, we hypothesized that the ability of Narp to regulate synaptic GluR4 levels on PV-INs, in response to activity, could be a compensatory mechanism for the suppression of epileptogenesis. To test this hypothesis, we stimulated awake, behaving 3–4 month old WT or Narp−/− mice twice daily in the basolateral amygdala and recorded the evoked afterdischarge (AD) from the same stimulating electrode. This stimulation protocol was sufficient to drive a significant increase of Narp protein expression in both the hippocampus and cortex of WT mice (Supplementary Fig. 11). With each evoked AD, we observed a gradual increase in seizure severity. We reached the first class III/IV seizures, in both WT and Narp−/− mice, after identical numbers of ADs. However, we began evoking more severe class V seizures after an average of 10 ADs in WT mice, while an average of only 6 ADs were required in Narp−/− mice (Fig. 8a). This difference was also evident in the number of ADs required to fully kindle the mice (i.e. evoke three class V seizures). Additionally, the minimum current required to evoke an AD decreased more rapidly in Narp−/− mice, relative to WT mice, over the course of the experiment (Fig. 8b). These data suggest that Narp functions to inhibit kindling-evoked progression of circuits involved in the development of chronic and long-term seizure plasticity.
Excitatory synapses onto PV-INs are capable of undergoing plasticity30-32, and this study provides the first molecular mechanism for a postsynaptic form of plasticity at these synapses. Activity-dependent expression of Narp by presynaptic excitatory neurons regulates homeostatic adaptations of circuit activity by enhancing the strength of excitatory synapses on PV-INs, concomitantly increasing their network-driven firing rate. PV-INs are the most abundant subtype of interneuron within the hippocampus19 and are implicated in processes such as gamma oscillations33, visual cortical plasticity34, and fear memory resilience35. Additionally, the dysfunction and/or loss of PV-INs may underlie several neurological disorders such as temporal lobe epilepsy36 and schizophrenia37. Therefore, uncovering the molecular mechanisms of how these neurons regulate the strength of their synapses has implications for understanding plasticity and cognitive disorders.
The present model of Narp-dependent synaptic plasticity is consistent with its regulation as an IEG, and its ability to bind AMPARs and sugars7, 15. Increases in activity increase Narp expression in excitatory neurons and Narp is subsequently secreted, and preferentially accumulates, at excitatory synapses on PV-INs. Narp is required for activity-regulated changes in GluR4-mediated synaptic strength at these synapses. By this process, an IEG can evoke transsynaptic effects to modulate circuit activity. It is interesting to contrast the respective roles of Narp and Arc in homeostatic scaling. With sustained increases in activity, Arc is rapidly expressed and functions to promote the endocytosis and downregulation of GluR1 in pyramidal neurons4, 5. At the same time, Narp targets to PV-INs where it functions to cluster GluR4 and potentiate excitatory synapses on these neurons. Using independent mechanisms, Narp and Arc thereby function in a complementary manner, at two separate populations of glutamatergic synapses, to reset pyramidal neuron activity back to baseline levels.
Despite these molecular insights, several important questions arise. First, how is Narp expression regulated by activity? Recent evidence demonstrates that the activity-dependent transcription of Narp mRNA requires Ca2+ influx through L-type voltage-gated calcium channels (VGCCs) and subsequent activation of Ca2+/calmodulin (CaM), CaM-dependent kinases, and Extracellular signal-regulated kinase (ERK) 1/238. Additionally, Narp is misregulated by knockdown of the neuronal IEG transcription factor Npas4 via RNAi expression39. It is interesting to note that Npas4, and Narp share many similar features: both are activity-regulated genes, require Ca2+ influx through L-type VGCCs for their induction, are expressed primarily in excitatory neurons, and yet are both key regulators of the inhibitory network39. Whether Npas4 is required for Narp expression remains to be studied.
Second, how does Narp selectively accumulate on PV-INs? Glycoproteins appear to be important for the targeting of Narp to excitatory synapses on PV-INs, consistent with the general property of pentraxins to bind sugars40. It is notable, however, that perineuronal nets are not localized precisely at synapses41. One model consistent with current data envisions that secreted Narp accumulates locally within the glycoprotein network, aided by lectin-based interactions. Narp might then diffuse along pre or postsynaptic membranes and localize to the synapse by interactions with other sugars or proteins, or form disulfide linked complexes with NPR11. The lectin properties of pentraxins are linked to the Ca2+ binding pocket, which in turn, is important for proper folding of the pentraxin domain40. Accordingly, it is difficult to selectively disrupt lectin properties by mutagenesis. Simple addition of recombinant Narp to cultures does not result in selective binding to PV-INs (unpublished observation), suggesting that targeting may involve processes beyond simple lectin-dependent binding.
Important questions also arise regarding how Narp can evoke an increase of GluR4, but not other AMPARs, on PV-INs. Binding of Narp to AMPARs does not require sugar adducts to the receptor, but rather, appears to be dependent on protein sequences within the N-terminal X-domain10. Previous reports demonstrate the capacity for NPs to bind and cluster AMPARs at sites of cell-cell contact8, 9, 14. It is possible that Narp-containing pentraxin complexes preferentially retain GluR4 on the cell surface, and this conjecture is consistent with the observation that NPs cluster homomeric AMPARs consisting of GluR4 subunits better than any other AMPAR subunit9. However, the avidity of Narp binding for GluR1 and GluR4 is not dramatically different in binding assays (unpublished observation) suggesting that the difference in activity-dependent accumulation may be due to factors in addition to their association with Narp, such as the level of GluR expression in PV-INs or other selective protein interactions. Currently, assays of native GluR4 trafficking are technically limited due to lack of appropriate antibodies, and our attempts at expressing an N-terminal-tagged GluR4 transgene resulted in similar elevated levels in both WT and Narp−/− PV-INs which were unresponsive to activity.
The observation that Narp−/− mice showed accelerated kindling to class V seizures provides insight into conditions in which Narp may contribute to the suppression of network excitability. It is notable that the baseline properties of excitatory synapses on hippocampal PV-INs in the dentate gyrus were not different in Narp−/− mice; only after MECS did we observe a change in excitatory input strength. The activity level in vivo was much less than in our primary cultures and this difference was consistent with a minimal role of Narp under basal conditions in the hippocampus in vivo. Narp−/− mice showed identical initial responses to kindling stimuli but clearly diverged in their responses as the kindling process evolved. Together, this suggests that Narp becomes part of the physiological adaptation only under conditions of intense activity, and is consistent with culture models that suggest the inability to recruit GluR4 levels to excitatory synapses on PV-INs in response to activity may underlie enhanced kindling. These studies do not exclude a role for Narp in scaling of PV-INs in physiological plasticity, since recordings of sEPSCs may not detect the small subset of Narp-associated excitatory synapses relevant for proper network function under basal conditions. Moreover, studies of Narp mRNA expression indicate that Narp is induced by non-epileptiform activity in models of cocaine administration42, monocular deprivation7, and in vivo LTP7. Because of the striking complexity of interneuron populations in vivo and their importance to integrated neural function it will be compelling to assess the contribution of Narp-dependent homeostatic plasticity in broader studies of physiological plasticity and models of disease.
All wild-type and Narp−/− mice were of the genetic background C57/Bl6. The use of vertebrate animals was regulated and approved by the Johns Hopkins University Animal Care and Use Committee.
We used the following primary antibodies in this study: Narp rabbit polyclonal15 (1:1,000), Parvalbumin mouse monoclonal (1:2,000, Sigma, St. Louis, MO) or goat polyclonal (1:2,000, Swant, Bellinzona, Switzerland), Calretinin mouse monoclonal (1:1,000 BD Biosciences, San Jose, CA), αCAMKII mouse monoclonal (1:1,000 Boehringer Mannheim/Roche,Indianapolis, IN), GluR2 mouse monoclonal (1:200, Chemicon/Millipore, Billerica, MA), GluR4 rabbit polyclonal (1:100, Chemicon/Millipore, Billerica, MA), NR1 mouse monoclonal (BD Pharmingen, 1:500, San Jose, CA), GAD65 mouse monoclonal (1:2,000, Chemicon/Millipore, Billerica, MA), Transferrin receptor mouse monoclonal (1:1,000, Zymed/Invitrogen, Carlsbad, CA), and PSD95 mouse monoclonal (1:500, Affinity Bioreagents/Thermo Scientific, Rockford, IL).
For immunocytochemical studies, we used fluorescent secondary antibodies (Molecular Probes/Invitrogen, Carlsbad, CA) at 1:400. We used peroxidase-conjugated secondaries (Pierce/Thermo Scientific, Rockford, IL) for western blot detection at 1:10,000
We plated hippocampal preparations of E17.5 WT or Narp−/− mice on 25 mm coverslips coated with poly-D-lysine (1 mg/mL in 0.1M Trizma buffer pH 8.5) at a density of 4×105 cells per 60 mm dish for immunocytochemistry, 1×106 cells per well (6-well plate) for biochemistry, and 1.5×106 per 60 mm dish for electrophysiology. We initially cultured neurons in Neurobasal (Invitrogen, Carlsbad, CA) media containing 5% horse serum (Hyclone/Thermo Scientific, Rockford, IL), 1% penicillin/streptomycin (Invitrogen, Carlsbad, CA), 2% Glutamax-I (Invitrogen, Carlsbad, CA), and 2% B-27 supplement (Invitrogen, Carlsbad, CA). We added Cytosine arabinoside (AraC) on DIV 4 to inhibit glial proliferation. After 7 days, we fed neurons by replacing half the media with Neurobasal media containing 1% fetal bovine serum, 1% penicillin/streptomycin, 2% Glutamax-I, and 2% B-27 supplement after which we fed neurons twice weekly.
For wild-type/Narp−/− co-culture, we dissociated and labeled either Narp−/− or wild-type hippocampal neurons with 2 μM Vybrant® CM-DiI (Invitrogen, Carlsbad, CA) for 20 minutes at 37°C. We then pelleted cells and washed them 3 times with pre-warmed growth medium prior to plating at 1.5×104 cells per 60 mm dish alongside unlabeled neurons of the opposing genotype, plated at 4×105 cells per 60 mm dish.
For GluR2, GluR4, and NR1 staining, we fixed DIV 14–17 cells in ice-cold methanol for 20 minutes at −20°C prior to staining. Otherwise, we fixed DIV 14–17 cells with 4% paraformaldehyde/4% sucrose in PBS for 20 minutes at 25°C followed by permeablization with 0.2% Triton-X 100 for 10 minutes at 25°C. We blocked nonspecific binding of antibodies with 10% goat serum (Colorado Serum Company, Denver, CO) in PBS. Following application of primary and secondary antibodies, we mounted coverslips with ProLong® Gold Antifade Reagent (Invitrogen, Carlsbad, CA).
For surface labeling of Narp, we incubated neurons with Narp antibody live for 20 minutes at 10°C in serum-free medium prior to fixation.
To label perineuronal nets, we added fluorescein-conjugated Vicia villosa agglutinin (2 μg/mL, Vector Laboratories, Burlingame, CA) to neurons for 10 minutes prior to fixation. For chondroitinase ABC treatments, we incubated cells in 0.2 U chondroitinase ABC (Sigma, St. Louis, MO) for 48 hours prior to fixation.
We anesthetized mice with pentobarbitol prior to perfusion fixation with 4% paraformaldehyde. We made 40 μm free-floating sections using a VT-1000S vibratome (Leica Microsystems, Bannockburn, IL,). For Narp staining, we incubated sections in 10 mM sodium citrate buffer (pH 6) for 30 minutes at 80°C, prior to blocking in 10% goat serum, 1% BSA, and 0.3% Triton-X 100 in PBS for 1 hour and followed by overnight incubation with primary antibody at 4°C. For GluR4 staining, we incubated sections in PBS for 30 minutes at 37°C, followed by incubation with pepsin (1mg/mL, DAKO, Glostrup, Denmark) in 0.2 M HCl for 5 minutes at 37°C, prior to blocking and staining43. Following application of secondary antibodies, we mounted slices with ProLong® Gold Antifade Reagent (Invitrogen, Carlsbad, CA)
We added Sulfo-NHS-SS-Biotin (1 mg/mL, Thermo Scientific, Rockford, IL) to high-density hippocampal neuronal cell cultures for 30 minutes at 4°C prior to quenching with 100mM glycine and solubilization in 1% Triton, 0.5% DOC, and 0.1% SDS. We sonicated lysates and mixed NeutrAvidin Agarose resin (Thermo Scientific, Rockford, IL) with total lysates overnight. We ran immunoprecipitated protein on a NuPAGE 4–12% Bis-Tris gel (Invitrogen, Carlsbad, CA), transferred onto PVDF membrane (Thermo Scientific, Rockford, IL), and visualized with chemiluminescence (Thermo Scientific, Rockford, IL).
We isolated total RNA from treated or untreated high-density hippocampal neuronal cell cultures using TRIzol® Reagent (Invitrogen, Carlsbad, CA). We performed reverse-transcription of the RNA using a Thermoscript™ RT-PCR system (Invitrogen, Carlsbad, CA). The forward and reverse primer sequences used for amplification were 5′-CTCCGCACAAATGTGTCTAAC-3′ and 5′-CTTCACAGGTCTCCACAGGC-3′, respectively, which yielded a 844 bp product. We set up reactions using an RT2 SYBR® Green qPCR Master Mix (SABiosciences, Frederick, MD) and carried out amplification and analysis using an iCycler iQ Real-Time PCR System (Bio-Rad Laboratories, Hercules, CA).
We captured Z-stacks of each neuron with a Zeiss LSM 510 confocal microscope at 0.5 μm intervals, 1024×1024 pixels, 4-frame averaging, and 8-bit color without binning. Prior to quantification, we made maximum intensity projections of each neuron. We selected puncta, with an intensity above a set threshold, and residing on primary dendrites within 100 μm from the soma, for quantification. We measured puncta using ImageJ (NIH) imaging software and performed all statistics using Prism® software (GraphPad, La Jolla, CA). We expressed all values as mean ± s.e.m. We used Student's t-test with Welch's correction for all statistical comparisons between any two groups. Otherwise, for comparisons between multiple groups, we used nonparametric one-way ANOVA tests with Bonferroni analysis. We regarded P < 0.05 as statistically significant.
We performed whole-cell patch-clamp recordings from high density (1.5 × 106 cells/ 60 mm culture dish) hippocampal cultures at DIV 14–17. To isolate AMPAR-mediated mEPSCs, we continuously perfused neurons with 30–32°C artificial cerebral-spinal fluid (aCSF) at a flow rate of 2 ml/min. The composition of aCSF was as follows (in mM): 124 NaCl, 2.5 KCl, 2.5 CaCl2, 1.3 MgCl2, 1 NaH2PO4, 26.2 NaHCO3, and 10 glucose equilibrated with 95% O2 and 5% CO2. We adjusted the osmolarity of aCSF to 300 ± 5 mOsm and pH to 7.4. The pipette solution consisted of the following (in mM): 100 K-gluconate, 0.6 EGTA, 5 MgCl2, 8 NaCl, 2 Na-ATP, 0.3 Na-GTP, and 40 HEPES. We adjusted this solution to 290 ± 5 mOsm and pH to 7.2. We pulled patch pipettes from borosilicate glass (4–5 MΩ) using a horizontal puller (Sutter Instruments, Novato, CA). We visually identified PV-INs via binding of fluorescein-tagged Vicia villosa agglutinin, then confirmed their fast-spiking, non-accommodating action potentials following somatic current injection25. We monitored passive properties of voltage clamped neurons throughout the experiments. Uncompensated series resistance (Rs) was ~10–13 MΩ and in the event of a change in either Rs or input resistance (Ri) >15% during the course of a recording, we excluded the data from the set. We acquired mEPSCs through a MultiClamp 700B amplifier (Molecular Devices, Sunnyvale, CA), filtered at 2 kHz, and digitized at 10 kHz using a Digidata 1332A. We recorded data continuously only after allowing the cell to stabilize for 10 minutes. We acquired sweeps of 15 s with zero latency until we recorded a sufficient number of events (a minimum of 3 and no longer than 10 min). For mEPSC recording, bath solution contained both 1 μM TTX and 10 μM GABAzine to block action potential-dependent EPSCs and GABAA receptors, respectively. We manually detected mEPSCs with MiniAnalysis software (Synaptosoft Inc., Decatur, GA) by setting the amplitude threshold to RMS × 3 (usually 8 pA). Once we collected a minimum of 200 events from a neuron, we measured the amplitude, frequency, rise time (time to peak), decay time (10%–90%), and passive properties. We then averaged data from each group, and performed statistical comparison by the independent t-test, ANOVA. We purchased all drugs from Tocris (Ellisville, MO) except for TTX (Ascent Scientific LLC, Princeton, NJ).
We prepared hippocampal slices (300–350 μm thick) from 3–4 week old WT and Narp−/− mice as described previously44. We dissected control and MECS-administered mouse littermates and interleaved recordings from slices obtained from each mouse. We anesthetized mice with isoflurane, and dissected the brain in ice-cold saline solution (in mM): 130 NaCl, 24 NaHCO3, 3.5 KCl, 1.25 NaH2PO4, 1 CaCl2, 5.0 MgCl2, and 10 glucose, saturated with 95% O2 and 5% CO2, pH 7.4. We cut transverse slices using a VT-1000S vibratome (Leica Microsystems, Bannockburn, IL) and incubated them in the above solution at 35°C for 30 minutes following which we kept them at room temperature until use. We transferred slices to a recording chamber and perfused them (3–5 ml/min, 32–35°C) with extracellular solution (in mM): 130 NaCl, 24 NaHCO3, 3.5 KCl, 1.25 NaH2PO4, 2.5 CaCl2, 1.5 MgCl2, 10 glucose, 0.05 ± dl-AP5, and 0.005 bicuculline methobromide saturated with 95%O2/5%CO2, pH 7.4. We targeted putative PV-INs in the dentate gyrus visually identified within the first 100 μm of slices using a 40× objective and IR-DIC video microscopy (Zeiss Axioskop) for whole-cell recording using a multiclamp 700A amplifier (Axon Instruments, Foster City, CA). We filled recording electrodes (3–5 MΩ) pulled from borosilicate glass (WPI, Sarosota, FL) with (in mM): 130 K-gluconate, 5 KCl, 0.6 EGTA, 2 MgCl2, 2 Na2ATP, 0.3 GTPNa, and 10 HEPES, pH 7.2–7.3, 290mOsm. We routinely added biocytin (0.2%) to the recording electrode solution to allow post hoc morphological processing of recorded cells for confirmation of basket cell anatomy using fluorescently conjugated avidin to visualize the biocytin-filled cell (Supplementary Fig. 6d,e). In some cases, we also processed slices for PV immunoreactivity. We rigorously monitored uncompensated series resistance, 5–15 MΩ, and discarded recordings if changes of >10% occurred. We performed data acquisition and analysis with PC computer equipped pClamp 9.2 (Axon Instruments, Foster City, CA). We monitored firing properties of recorded cells in current clamp mode by delivery of electrotonic current injections through the recording pipette with the resting membrane potential biased to approximately −60 mV, and monitored sEPSC in voltage clamp mode at a holding potential of −60 mV. All cells included for analysis exhibited fast spiking behavior, and basket cell anatomy (Supplementary Fig 6d,e). We detected sEPSCs and analyzed them offline in Clampfit using a template event detection strategy for 30 s of gap-free recording for each cell. For each recorded cell, we averaged all events collected and used this average sEPSC to determine the amplitude, 10–90% rise time, and decay time constant (monoexponential fit). We used Student's t-test or Kolmogorov-Smirnov (K-S) tests to evaluate significance as indicated.
We administered MECS to 3–4 week old WT or Narp−/− mice, 12–16 hours prior to obtaining slices for electrophysiology, using a 7801 ECT unit (Ugo Basile, Comerio, Italy) (100 pulses/second, 0.4 ms pulse width, 1 second shock duration). We initially gave mice 6 mA current with successive 2 mA current increases until we observed tonic hindlimb extension45.
We placed adult mice (3–4 months old) in a stereotaxic apparatus and anesthetized them with 2% isoflurane. We lowered a bipolar, insulated stainless-steel stimulating-recording electrode (PlasticsOne, Roanoke, VA) into the right basolateral nucleus of the amygdala (1.2 mm posterior, 2.9 mm lateral, 4.6 mm ventral) and placed a ground electrode (i.e., a jeweler's screw) in the skull over left frontal cortex. We used dental acrylic to secure the electrodes according to standard chronic methods. After a week-long recovery following surgery, we focally stimulated awake mice in the amygdala twice daily (5 days per week) with a one second train of 60 Hz biphasic constant current 1 ms square wave pulses46. We determined the AD threshold by a standard protocol designed to identify the minimum current necessary to evoke ADs (100–1100 μA). We scored induced seizures by standard behavioral classes as follows: (1) behavioral arrest, eye closure, vibrissae twitching, sniffing; (2) facial clonus and head bobbing; (3) forelimb clonus; (4) rearing with continued forelimb clonus; and (5) rearing with a loss of motor control and falling47. We evoked repeated ADs until each mouse experienced three Class V seizures and were considered fully kindled.
We thank M.S. Perin for the Narp−/− mouse, R.L. Huganir (Johns Hopkins University) for antibodies to GluR1, M. Dehoff and P. Chuang for animal care, M. Papapavlou for administrative assistance, and X. Yuan, and B. Jeffries for expert technical assistance. Work in the lab of C.J.M is supported by an NIH intramural award. Work in the lab of P.F.W is supported by NIH grants PAR-02-059 and NS 39156.
Author Contributions: M.C.C. designed the experiments, carried out the in vitro immunocytochemical and biochemical studies and analysis and wrote the bulk of the manuscript. J.M.P. performed the in vitro electrophysiological experiments and analysis and contributed to the manuscript. K.A.P performed the slice electrophysiological experiments and analysis and contributed to the manuscript. H.L.G. performed the kindling experiments and analysis and contributed to the manuscript. D.X. provided reagents and helped with the discussion. P.F.W., D.J.L, T.P.S., and C.J.M directed the project.