|Home | About | Journals | Submit | Contact Us | Français|
Arc/Arg3.1 is an immediate-early gene whose expression levels are increased by strong synaptic activation, including synapse-strengthening activity patterns. Arc/Arg3.1 mRNA is transported to activated dendritic regions, conferring the distribution of Arc/Arg3.1 protein both temporal correlation with the inducing stimulus and spatial specificity. Here, we investigate, the effect of increased Arc/Arg3.1 levels on synaptic transmission. Surprisingly, Arc/Arg3.1 reduces the amplitude of synaptic currents mediated by AMPA-type glutamate receptors (AMPARs). This effect is prevented by RNAi knockdown of Arc/Arg3.1, by deleting a region of Arc/Arg3.1 known to interact with endophilin 3 or by blocking clathrin-coated endocytosis of AMPARs. In the hippocampal slice, Arc/Arg3.1 results in removal of AMPARs composed of GluR2 and GluR3 subunits (GluR2/3). Finally, Arc/Arg3.1 expression occludes NMDAR-dependent long-term depression. Our results demonstrate that Arc/Arg3.1 reduces the number of GluR2/3 receptors leading to a decrease in AMPAR-mediated synaptic currents, consistent with a role in the homeostatic regulation of synaptic strength.
Brain processes like learning and memory are thought to involve plastic changes in synaptic strength. It is therefore reasonable to postulate that cellular mechanisms supporting these plastic changes would show both spatial specificity and temporal correlation with the stimulus that triggered the plastic change. Despite the abundant evidence to connect synaptic plasticity with gene expression (Deisseroth et al., 2003), causal evidence to relate known gene products synthesized in the mammalian cell soma to changes in specific sets of synapses has been missing. The discovery of Arc/Arg3.1 in 1995 (Link et al., 1995; Lyford et al., 1995) revealed the only known activity-induced gene that correlates both temporally and spatially with the stimulus that induced its transcription.
Arc/Arg3.1 (activity-regulated cytoskeleton-associated) is an effector immediate-early gene rapidly induced by different forms of neuronal activity, e.g., LTP-inducing stimuli (Lyford et al., 1995; Steward et al., 1998), pharmacological stimulation (Fosnaugh et al., 1995), and behavioral stimulation (Guzowski et al., 1999; Montag-Sallaz et al., 1999). Increases in Arc/Arg3.1 mRNA can be observed 5 min after stimulation and may last over 8 hr (Cirelli and Tononi, 2000; Guzowski et al., 1999; Lyford et al., 1995). The most striking characteristic of this immediate-early gene is not only that its mRNA is transported to dendrites, but also that it accumulates in the region of the dendritic tree where the inducing stimulus arrived (Steward et al., 1998; Steward and Worley, 2001). Such a mechanism ensures the localization of Arc/Arg3.1 protein to the region of dendrite that had received strong synaptic activation. It should be noted, however, that there is no evidence that Arc/Arg3.1 is targeted specifically to activated synapses.
Similar to LTP induction, Arc/Arg3.1 transcription and mRNA localization depend on the activation of NMDA-type glutamate receptors (NMDARs). Indeed, NMDAR activation can also relocalize Arc/Arg3.1 mRNA which had been previously induced (Lyford et al., 1995; Steward et al., 1998; Steward and Worley, 2001). Following induction and mRNA localization, Arc/Arg3.1 protein is found in dendritic spines (Moga et al., 2004), where it can interact with components of the endocytic machinery involved in AMPA receptor trafficking (Chowdhury et al., 2006 [this issue of Neuron]).
Arc/Arg3.1 is expressed in glutamatergic neurons in the hippocampus and neocortex. Its mRNA can be first detected in rats at the eighth postnatal day (P8), and basal levels increase steadily until P21 (Lyford et al., 1995). This developmental profile correlates with a period of synaptogenesis and synaptic rearrangement. In spite of the interesting properties of Arc/Arg3.1 in terms of activity induction, as well as mRNA and protein localization, there has been no study to directly address the effect of increased levels of Arc/Arg3.1 on synaptic transmission.
Here, we increase Arc/Arg3.1 expression in hippocampal CA1 pyramidal neurons to assay its effects on synaptic properties. We find that Arc/Arg3.1 reduces the amplitude of currents mediated by AMPARs, and this is prevented by RNAi knockdown of Arc/Arg3.1. After Arc/Arg3.1 protein has been localized to the synapse, further synaptic activity is not required to mediate its effect. We also show that this effect of Arc/Arg3.1 occurs through endocytosis of AMPARs containing GluR2 but not GluR1 subunits in our system. The region of Arc/Arg3.1 that interacts with endophilin 3 (Chowdhury et al., 2006), a component of the endocytic machinery, is necessary for the effect. The Arc/Arg3.1-mediated decrease in AMPAR transmission is sensitive to phosphatase inhibitors, and Arc/Arg3.1 expression occludes long-term depression (LTD). Keeping in mind that Arc/Arg3.1 is induced by strong stimuli, including LTP-inducing stimuli, these results suggest that Arc/Arg3.1 expression may lead to a delayed decrease in AMPAR transmission following strong activity, suggesting a homeostatic function.
Strong synaptic activation increases levels of Arc/Arg3.1 mRNA and protein between 2- and over 10-fold over a period of 2 to more than 8 hr (Ons et al., 2004; Steward et al., 1998; Wallace et al., 1998). To mimic this increase, we expressed green fluorescent protein (GFP)-tagged recombinant Arc/Arg3.1 in CA1 pyramidal neurons in organotypic hippocampal slice cultures using a Sindbis virus expression system (Malinow et al., 1999). We analyzed the time course of recombinant Arc/Arg3.1 expression by GFP fluorescence and western blots (Figures 1A and 1C). We observed GFP fluorescence in pyramidal cell bodies starting 6 hr after virus injection and in dendritic spines after 7 hr (Figure 1B and see Figure S1 in the Supplemental Data available with this article online; see Experimental Procedures). The recombinant protein can be first detected by western blot 6 hr after virus injection. We made electrophysiological recordings 10 to 20 hr after virus injection to match the time window of increased Arc/Arg3.1 expression reported with activity-induced increases of endogenous Arc/Arg3.1 mRNA and protein (Montag-Sallaz et al., 1999; Wallace et al., 1998).
We also investigated the levels of recombinant Arc/Arg3.1 expression. Our expression system induced a median 8-fold increase in recombinant Arc/Arg3.1 expression compared to basal expression levels of endogenous Arc/Arg3.1 (Figures 1D and 1E; median, 7.89; range, 1.89 to 23.14). This increase in Arc/Arg3.1 expression is comparable to previously reported increases in mRNA and protein for endogenous Arc/Arg3.1, as well as other activity-induced genes (Castren et al., 1993; Guzowski et al., 2001; Matsuo et al., 2000; Ons et al., 2004; Steward et al., 1998; Ying et al., 2002).
In order to test whether increased Arc/Arg3.1 expression regulates synaptic transmission, we recorded evoked postsynaptic currents in CA1 pyramidal cells using whole-cell voltage clamp. We recorded simultaneously from two neighboring cells: a control uninfected cell and a cell expressing recombinant Arc/Arg3.1. This recording technique allows the paired comparison of the amplitudes of the responses of neighboring neurons in response to the same stimulus, and controls for differences between slices that arise from variations in stimulation electrode placement and stimulus intensity.
Arc/Arg3.1 induced a significant reduction (32%) in AMPAR-mediated synaptic currents (Figures 2A, 2B, and 2H; control median, −49.81 pA; IQR, −77.86 to −33.07; Arc/Arg3.1 median, −33.36 pA, IQR, −47.58 to −21.25). In contrast, expression of GFP alone had no impact on AMPA-evoked excitatory postsynaptic currents (EPSCs; Figures 2D and 2E; control median, −44.77 pA; IQR, −65.95 to −26.28; GFP median, −41.76 pA; IQR, −77.60 to −25.90). The effect of Arc/Arg3.1 on AMPA EPSCs was significantly different from the effect of GFP expression alone (Figure 2H). We did not observe changes in rise or decay times (Figure S2). Moreover, input resistance (Ri) at all recorded holding potentials, as well as resting potentials (Vrest) were unaffected in cells expressing recombinant Arc/Arg3.1 compared to control neighbors (Mann-Whitney test: Ri, −60 mV, p > 0.24; Ri, +40 mV, p > 0.79; Vrest, p > 0.23).
The amplitude of NMDAR-mediated currents was not affected by increased Arc/Arg3.1 expression (Figures 2A, 2C, and 2I; control median, 19.38 pA; IQR, 14.05 to 31.08; Arc/Arg3.1 median, 19.06 pA; IQR, 12.37 to 27.25). The effect of Arc/Arg3.1 on NMDAR-mediated currents was indistinguishable from the effect of expressing GFP alone. NMDAR-mediated currents in GFP-expressing cells were not significantly different from that of control neighboring cells (Figures 2D and 2F; control median, 16.75 pA; IQR, 10.82 to 23.03; GFP median, 14.08 pA; IQR, 9.85 to 24.47). In addition, the decay time constant of NMDAR-mediated currents in Arc/Arg3.1-expressing cells did not differ from that of control neighbors (Figure S2; control median, 78.9 ms; IQR, 68.71 to 93.32; Arc/Arg3.1 median, 74.11ms; IQR, 63.1 to 97.38; Mann-Whitney, p > 0.39). Finally, GABAR-mediated currents, rise and decay times were not significantly affected by Arc/Arg3.1 expression (Figures 2G and S1; control median, 157 pA; IQR, 94.79 to 286.21; Arc/Arg3.1 median, 153.7 pA; IQR, 103.54 to 217.78). These experiments indicate that Arc/Arg3.1 specifically reduces AMPAR-mediated synaptic currents.
We also recorded spontaneous AMPAR miniature EPSCs (mEPSCs). Increased Arc/Arg3.1 expression decreased the amplitude of mEPSCs by a comparable amount to that observed in the evoked EPSCs (Figures 2J and 2K; 32% averaged across mEPSC events; 30% averaged across cells). We found a nonsignificant decrease in mEPSC frequency (Figure 2L), likely due to the fact that the mEPSC amplitude distribution for the Arc/Arg3.1 group is closer to the event detection limit (Figure 2J).
Basal levels of Arc/Arg3.1 increase during the first three postnatal weeks in vivo (Lyford et al., 1995). During the equivalent period in culture, the organotypic slices also show a comparable developmental increase in Arc/Arg3.1 basal levels (Figure 3A). In addition, organotypic slice cultures parallel other aspects of in vivo development. For instance, both in vivo and in vitro, there is an increase in synapse number, a decrease in the rate at which synaptogenesis occurs, and changes in the ability to induce plasticity (Buchs et al., 1993; De Simoni et al., 2003; Hsia et al., 1998; Muller et al., 1993).
To test whether these developmental changes influence the effect of increased Arc/Arg3.1 expression upon AMPAR-mediated currents, we recorded from neurons expressing recombinant Arc/Arg3.1 in slices maintained for 3 weeks in culture (WIC). Similar to 1 WIC slices, increased Arc/Arg3.1 expression also reduced AMPAR-mediated currents in 3 WIC slices (Figures 3B–3F; 34%).
These previous experiments established that increased levels of recombinant Arc/Arg3.1 reduce the amplitude of AMPAR-mediated EPSCs. We tested this finding further using the opposite approach. Arc/Arg3.1 expression is normally low in unstimulated slices. Instead of increasing Arc/Arg3.1 levels by viral expression of recombinant protein, we increased endogenous Arc/Arg3.1 through synaptic activation of the slice with a brief application of picrotoxin (see Experimental Procedures). In addition, we blocked the increased Arc/Arg3.1 expression in single cells using biolistic transfer of a small interfering RNA (siRNA). Picrotoxin induces strong recurrent synaptic activity in the hippocampal slice and produces a 7- to 10-fold increase in Arc/Arg3.1 protein (data not shown). This protocol allows comparison of AMPAR-mediated transmission in cells expressing high levels of endogenous Arc/Arg3.1 with nearby cells in which Arc/Arg3.1 expression is blocked by siRNA.
Control untransfected cells showed reduced AMPAR-mediated EPSC amplitude compared to adjacent cells transfected with Arc/Arg3.1 siRNA (Figures 4D and 4E; control median, −33.3 pA; IQR, −43.96 to −19.57; Arc/Arg3.1 siRNA median, −61.55 pA; IQR, −83.92 to −36.75). This effect was significantly different from the effect of transfecting with control siRNA (Figure 4B), which did not affect AMPAR-mediated currents (Figures 4G and 4H; control median, −48.75 pA; IQR, −69.61 to −28.74; control siRNA median, −43.3 pA; IQR, −77.43 to −22.79). Neither Arc/Arg3.1 siRNA nor control siRNA affected NMDAR-mediated transmission (Figures 4C, 4D, 4F, 4G, and 4I).
Although it is known that Arc/Arg3.1 transcription and mRNA localization are regulated by activity, it is unclear whether the function of Arc/Arg3.1 protein is regulated by activity after it has reached dendritic spines. To address this question, we exposed slices to treatments that modify activity levels during the time of recombinant Arc/Arg3.1 expression. As expected, recombinant Arc/Arg3.1 expression is not susceptible to regulation by neuronal activity in our system. Only the open reading frame of Arc/Arg3.1 mRNA is expressed by the Sindbis virus, and the virally driven mRNA synthesis does not depend on cellular transcription factors. Incubation of slices with the NMDAR antagonist APV, which prevents endogenous Arc/Arg3.1 transcriptional induction and mRNA localization (Steward and Worley, 2001), did not prevent recombinant Arc/Arg3.1 localization to dendritic spines (Figure 5B). As shown in Figure 5, the reduction in AMPAR-mediated transmission persisted under treatments that blocked NMDARs (200 μM DL-APV), AMPARs (10 μM NBQX), generally reduced activity levels (12 mM Mg2+, 2 μM TTX), blocked L-type Ca2+ channels (1 μM nifedipine), or metabotropic glutamate receptors (mGluRs; 200 μM MCPG).
Considerable evidence suggests that synaptic depression results from the endocytosis of AMPARs (Kim et al., 2001; Lee et al., 2002; Man et al., 2000; Seidenman et al., 2003). Interaction of the cytoplasmic carboxy-terminal tail of the AMPAR subunit GluR2 (GluR2 c-tail) with the clathrin adaptor complex AP2 seems necessary for the expression of LTD (Lee et al., 2002). The effect of increased Arc/Arg3.1 expression differs from other types of depression in terms of induction mechanism and temporal window of expression. It is possible, however, that part of the molecular mechanisms involved in the expression of transcription-independent forms of depression could also participate in the effect induced by increased levels of Arc/Arg3.1.
To test whether GluR2-containing AMPAR endocytosis is required for the reduction in AMPA EPSCs induced by Arc/Arg3.1, we coexpressed Arc/Arg3.1 with pep- AP2 (previously termed pepΔA849–Q853; [Lee et al., 2002]), a peptide that prevents the interaction between GluR2 and the AP2 complex and abolishes clathrin-coated vesicle endocytosis of GluR2-containing AMPARs. As previously shown (Lee et al., 2002), pep-AP2 has no significant effect on basal synaptic transmission (Figures 6A and 6B). However, its coexpression with Arc/Arg3.1 abolished Arc/Arg3.1-induced reduction in AMPAR-mediated currents (Figures 6A and 6C). Synaptic transmission under this condition is indistinguishable from that recorded with pep-AP2 alone or GFP alone (Figure 6D), and it is significantly different from transmission recorded when Arc/Arg3.1 alone is expressed (Mann-Whitney, p < 0.05).
Amino acids 91 to 100 of Arc/Arg3.1 interact with the SH3 domain of endophilin 3 (Chowdhury et al., 2006), a postsynaptic component of the clathrin-coated vesicle endocytosis machinery. Chowdhury and coworkers also demonstrated that a deletion mutant of Arc/Arg3.1 lacking those amino acids does not interact with endophilin 3 and does not induce vesicle formation. To test if the physiological effect of Arc/Arg3.1 we have recorded relies on its molecular interaction with endophilin 3, we expressed Arc/Arg3.1(Δ91–100) (deletion mutant lacking amino acids 91 to 100). Arc/Arg3.1(Δ91–100) does not reduce AMPAR-mediated transmission (Figures 6E and 6F). Furthermore, its effect is indistinguishable from that of GFP expression alone (Figure 6G), and it is significantly different from transmission recorded when Arc/Arg3.1 alone is expressed (Mann-Whitney, p < 0.05). The lack of effect of Arc/Arg3.1(Δ91–100) is not due to exclusion of the recombinant protein from dendritic spines (Figures 6H and S3).
In the CA3-to-CA1 synapse, there are two major subpopulations of AMPARs: heteromers composed of GluR2 and GluR3 subunits (GluR2/3) and heteromers composed of GluR1 and GluR2 subunits (GluR1/2) (Wenthold et al., 1996). In the hippocampal slice, these receptor pools behave differently regarding trafficking and membrane life times. We reasoned that if Arc/Arg3.1 acts on GluR2/3 but not GluR1/2, then reducing the number of synaptic GluR2/3 would occlude the effect of Arc/Arg3.1 on AMPAR-mediated transmission. Expression of the GluR2 c-tail prevents delivery and synaptic stabilization of GluR2/3 AMPARs, thereby reducing their number and depressing basal transmission (Shi et al., 2001). The effect of Arc/Arg3.1 coexpressed with the GluR2 c-tail is indistinguishable from the effect of GluR2 c-tail expression alone or that of Arc/Arg3.1 alone (Figures 7A–7D). The occlusion and lack of summation between the effects of expressing Arc/Arg3.1 and the GluR2 c-tail demonstrate that the mechanism by which Arc/Arg3.1 reduces synaptic transmission affects AMPAR heteromers composed of GluR2 and GluR3 subunits.
To further test this selectivity for GluR2/3 AMPARs, we biotinylated surface proteins in slices expressing either recombinant Arc/Arg3.1 or GFP. The slices were homogenized, and 20% of the total homogenate was used to estimate total GluR2 or GluR1. The remaining 80% of the same homogenate was incubated with streptavidin beads to precipitate the biotinylated surface proteins. The samples were analyzed by western blot for GluR2 or GluR1 in the total and biotinylated fractions of the same homogenate. This protocol allows us to compare changes in surface expression of the protein induced by Arc/Arg3.1 expression, using GFP expression as a control. Slices with increased Arc/Arg3.1 expression have a smaller surface-to-total GluR2 ratio than GFP-expressing slices (Figures 7E and 7F). Arc/Arg3.1 expression did not affect the surface-to-total ratio of the GluR1 subunit (Figures 7E and 7G).
We did not detect any GluR2 or GluR1 in the precipitate when the biotinylation reagent was omitted, indicating that the precipitation of biotinylated GluR subunits was specific (Figure 7E, GFP no biot. lanes). Finally, probing for the intracellular proteins PICK1 and β-tubulin in each experiment showed that the biotinylation was specific for surface proteins (Figure 7E).
Our data indicate that Arc/Arg3.1 reduces AMPAR-mediated currents through a mechanism involving GluR2/3 AMPAR endocytosis. Some forms of LTD also result in AMPAR endocytosis (Man et al., 2000; Wang and Linden, 2000). If the Arc/Arg3.1-mediated decrease in AMPAR transmission shares the same mechanisms of expression as LTD, then Arc/Arg3.1 should occlude LTD. Indeed, cells with increased levels of Arc/Arg3.1 failed to express LTD (Figure 8A). Simultaneous recordings from recombinant Arc/Arg3.1-expressing cells and adjacent control cells demonstrate occlusion. While the amplitude of transmission in cells expressing Arc/Arg3.1 did not change after LTD induction, transmission onto control cells decreased to the same amplitude as recorded from the Arc/Arg3.1 cells (Figure 8B).
Although the distinct mechanisms for induction of Arc/Arg3.1 and LTD clearly differentiate the two processes, they appear to share some common mechanisms of expression. To test this possibility further, we incubated the slices in two compounds commonly used to block NMDAR-dependent LTD (Morishita et al., 2001; Mulkey et al., 1994). Okadaic acid (1 μM) and FK506 (50 μM), inhibitors of Protein Phosphatase 1 and Calcineurin, respectively, blocked the effect of Arc/Arg3.1 on AMPAR-mediated transmission but did not affect the localization of the recombinant protein to spines (Figures 8C–8J).
In this study, we showed that increased Arc/Arg3.1 expression, that mimics the levels and time window of activity- dependent endogenous Arc/Arg3.1 upregulation (Figure 1), reduces AMPAR-mediated synaptic currents without affecting NMDAR-mediated current amplitudes or decay time constants. Furthermore, synaptic currents mediated by the GABAR were unaffected by increased Arc/Arg3.1 expression (Figure 2). The observation that increased Arc/Arg3.1 levels had similar effects on evoked and miniature AMPAR-mediated currents, but no impact on evoked NMDAR-mediated currents, suggests that Arc/Arg3.1 reduces AMPAR-mediated transmission through a postsynaptic mechanism affecting AMPA receptor number and, together with input resistances and resting potentials identical to those of control cells, argues against nonspecific effects of the expression system. Reducing activity levels did not prevent the Arc/Arg3.1-induced reduction of AMPAR EPSCs (Figure 5), showing that activity does not regulate the effect of Arc/Arg3.1 protein after it has been localized, consistent with the fact that the localization of Arc/Arg3.1 is activity regulated at the transcriptional and mRNA localization level. Moreover, we observed the same magnitude of Arc/Arg3.1 effect in slices cultured for 1 or 3 weeks (Figure 3). This not only shows that the effect of increased Arc/Arg3.1 expression is constant throughout development, but it also argues against a negative impact of Arc/Arg3.1 on synaptogenesis that could have also resulted in reduction of the evoked currents. If that had been the case, we should have seen a bigger effect of Arc/Arg3.1 in the younger slices, when synaptogenesis is occurring at a more rapid rate (De Simoni et al., 2003).
We also performed the reverse experiment, in which we stimulated synaptic activity in slices to induce endogenous Arc/Arg3.1 upregulation in control cells. In those slices, cells transfected with an siRNA to reduce Arc/Arg3.1 expression showed higher AMPAR-mediated transmission levels than the control cells in which endogenous Arc/Arg3.1 expression was not blocked (Figure 4). In other words, preventing Arc/Arg3.1 up-regulation caused the opposite effect of increasing its expression. This demonstrates that cells with increased levels of Arc/Arg3.1, either recombinant or endogenous, exhibit reduced AMPAR EPSC amplitude compared to neurons with low Arc/Arg3.1 levels. It also shows that the increases in Arc/Arg3.1 levels induced by our recombinant expression system are functionally comparable to those occurring during activity-dependent Arc/Arg3.1 induction.
A previous study linked Arc/Arg3.1 expression to LTP maintenance showing reduced potentiation and learning deficits after injection of antisense oligos against Arc/Arg3.1 (Guzowski et al., 2000). Additionally, genetic removal of Arc/Arg3.1 shows similar effects on learning and potentiation (Plath et al., 2006 [this issue of Neuron]). In our hands, preventing Arc/Arg3.1 expression with RNAi leads to an absolute increase of AMPA transmission in RNAi-transfected cells compared to controls. Considering that increased basal AMPA transmission reduces the potentiation levels after LTP induction (Shi et al., 2001) and impairs performance in the Morris Water Maze (Moser et al., 1998), the result by Guzowski and colleagues can be explained by an increase in basal transmission levels rather than a deficit in LTP maintenance.
Previous work proposed a role for clathrin-coated vesicle endocytosis in synaptic depression (Carroll et al., 2001). The cytoplasmic tail of the GluR2 subunit interacts with the adaptor protein complex AP2, a component of the clathrin-coated vesicle endocytosis machinery that promotes the assembly of the clathrin-coated pit (Kirchhausen, 1999). Lee and coworkers (2002) characterized this interaction and generated a peptide that interferes with it (termed here pep-AP2). They showed that, despite its lack of effect in basal transmission, the peptide blocks NMDAR-dependent LTD. This peptide completely abolished the effect of Arc/Arg3.1 on AMPAR-mediated synaptic transmission (Figure 6), suggesting that clathrin-coated vesicle endocytosis of GluR2-containing AMPARs is necessary for the effect of Arc/Arg3.1 upon synaptic transmission. In addition, the region of Arc/Arg3.1 that mediates its interaction with endophilin 3 (Chowdhury et al., 2006), a postsynaptic component of the clathrin-coated vesicle endocytosis machinery, is necessary for the electrophysiological effect of Arc/Arg3.1 (Figure 6). This result further confirms the involvement of AMPAR endocytosis and indicates a molecular mechanism.
Regulated trafficking of AMPAR into and out of the synapse governs some forms of synaptic plasticity (Bredt and Nicoll, 2003; Malinow and Malenka, 2002). In the hippocampus, AMPARs are mainly divided in two subpopulations: GluR1/2 heteromers, and GluR2/3 heteromers (Wenthold et al., 1996). Several studies have attempted to distinguish the cell biological properties of GluR1/2 heteromers from those of GluR2/3 heteromers. Synaptic depression appears to rely on endocytosis of AMPARs containing the GluR2 subunit (Chung et al., 2003; Kim et al., 2001; Man et al., 2000; Seidenman et al., 2003), while exocytosis of GluR1/2 mediates synaptic potentiation (Shi et al., 1999, 2001).
GluR2/3 AMPARs have a shorter life time in the plasma membrane than GluR1/2 heteromers in primary cultures (Passafaro et al., 2001). In the hippocampal slice culture, recombinant GluR1/2 receptors do not replace a significant proportion of endogenous receptors in 24 hr, but that is not the case for recombinant GluR2/3. Furthermore, ectopic expression of the GluR1 c-tail has no effect on basal transmission but prevents the activity- dependent delivery of GluR1/2. Conversely, ectopic expression of the GluR2 c-tail depresses basal synaptic transmission, prevents GluR2/3 insertion, but does not affect the activity-dependent insertion of GluR1/2 (Shi et al., 2001). These results illustrate that GluR2/3 receptors cycle into and out of the synapse faster and are more labile than GluR1/2 in the hippocampal slice.
Decreasing synaptic GluR2/3 receptors by GluR2 c-tail expression occluded the effect of Arc/Arg3.1 on synaptic transmission (Figure 7), supporting our interpretation that Arc/Arg3.1 acts primarily on the labile AMPAR pool that is mainly composed of GluR2/3 heteromers. Finally, biotinylation of surface receptors showed that surface expression of GluR2 is reduced in Arc/Arg3.1-expressing slices without an effect on surface GluR1 (Figure 7). Wenthold and coworkers (1996) estimated that approximately 42% of the AMPA-binding receptors in the hippocampus are GluR2/3 heteromers, a slightly higher proportion are GluR1/2 receptors, and about 8% are homomeric GluR1. In the light of those results, reduced surface GluR2 in the absence of a reduction in surface GluR1 indicates that the effect of Arc/Arg3.1 is mainly upon GluR2/3 heteromers. This is consistent with our interpretation of the electrophysiological data and suggests that the magnitude of our effect (~32% reduction in AMPA response) results from removal of a significant proportion of the GluR2/3-containing receptors from the cell surface.
It is worth noting that the slow recycling of GluR1/2 compared to GluR2/3 in the hippocampal slice culture (Kolleker et al., 2003; Shi et al., 2001) probably caused the observed selectivity of Arc/Arg3.1 for GluR2/3. In primary cultured neurons, where GluR1/2 recycling is much faster and takes place in the minute time scale, Arc/Arg3.1 also induced a reduction in surface GluR1 (Chowdhury et al., 2006). The results in both experimental systems suggest that the selectivity of the effect of Arc/Arg3.1 for a particular receptor pool may not be governed by Arc/Arg3.1 itself, but by the rate at which the different receptor pools become available for endocytosis.
Our results indicate that Arc/Arg3.1 and LTD converge at the same end point, the removal of synaptic AMPAR. This also suggested that at least some of the mechanisms underlying this receptor removal could be shared. Consistent with this idea, we found that increased Arc/Arg3.1 expression occluded LTD (Figure 8). The effect of Arc/Arg3.1 was also blocked by phosphatase inhibitors that block NMDAR-dependent LTD. This indicates that the pathways leading to AMPAR removal used by Arc/Arg3.1 and NMDAR-dependent LTD share the same sensitivity to phosphatase inhibitors. The molecular intermediaries between the activity of these phosphatases and AMPAR removal remain unknown.
Although increased Arc/Arg3.1 expression and different forms of LTD result in endocytosis of GluR2-containing AMPARs (Kim et al., 2001; Man et al., 2000; Wang and Linden, 2000), several distinctions between Arc/Arg3.1 and LTD suggest that they have different functions with respect to cellular and circuit plasticity. Low-frequency stimulation leading to LTD does not induce Arc/Arg3.1 transcription, mRNA localization, or an increase in Arc/Arg3.1 protein levels. High-frequency stimulation, which results in LTP, induces Arc/Arg3.1 transcription and transport of the mRNA to activated dendritic regions (Steward et al., 1998; Steward and Worley, 2001) and increased protein levels, as detected by electron microscopy (Moga et al., 2004). The transport of Arc/Arg3.1 mRNA to dendrites takes 30 to 60 min, and it may be longer before Arc/Arg3.1 protein accumulates to functional levels in spines (Steward et al., 1998), while the induction of LTD takes 15 min or less (Dudek and Bear, 1992). Finally, NMDAR or mGluR antagonists do not block the effect of Arc/Arg3.1 (Figure 4), but they do block two different forms of LTD (Oliet et al., 1997). These differences in induction mechanism and time course suggest that the function of Arc/Arg3.1 expression following strong synaptic activation may serve a different cellular purpose.
Since its discovery in screens using massive electroconvulsive shock or drug-induced seizures to induce activity- regulated genes (Link et al., 1995; Lyford et al., 1995), it has been speculated that Arc/Arg3.1 could participate in the stabilization of long-lasting synaptic changes induced by activity, leading to learning and memory. Consistent with this idea, LTP-inducing, but not LTD-inducing, stimuli increase Arc/Arg3.1 expression (Steward et al., 1998). In addition, experience can upregulate Arc/Arg3.1 expression, as has been shown in CA1 pyramidal cells from animals exposed to environmental novelty (Guzowski et al., 1999). However, Arc/Arg3.1 expression is also increased in response to injuring stimuli like experimentally induced ischemia (Berger et al., 2003; Kunizuka et al., 1999), electroconvulsive shock (Lyford et al., 1995), and drug-induced seizures (Link et al., 1995), where the postulated role for Arc/Arg3.1 in stabilizing enhanced synaptic strength may not hold. Moreover, while the ability to induce LTP in the visual cortex decreases with age after the critical period (Kirkwood et al., 1995), robust Arc/Arg3.1 mRNA induction can be detected in response to visual activity in 13-week-old mice (Tagawa et al., 2005). The function of Arc/Arg3.1 in the refinement of orientation selectivity is also inconsistent with a role in synaptic strengthening (Wang et al., 2006). The direct relationship between Arc/Arg3.1 expression levels and learning has also been challenged (Kelly and Deadwyler, 2002; Kelly and Deadwyler, 2003). Taken together, the published literature on Arc/Arg3.1 indicates that Arc/Arg3.1 is induced in response to any strong enhancement of synaptic activity; however, until now, no study has tested the functional consequences of increased Arc/Arg3.1 expression with regard to synaptic transmission. An intriguing possibility, compatible with the literature on Arc/Arg3.1, is that Arc/Arg3.1 functions in the homeostatic regulation of synaptic strength.
In addition to mechanisms of synaptic plasticity, such as LTP and LTD, neuronal cells possess mechanisms to maintain synaptic strength within a range that allows further increases or decreases in synaptic transmission (Burrone et al., 2002; Desai et al., 2002; O’Brien et al., 1998; Rabinowitch and Segev, 2006). It is thought that without such homeostatic mechanisms, long-lasting potentiation and depression would drive synapses toward either saturation or silence (Miller, 1996; Miller and Mackay, 1994). Strengthening the idea that synaptic homeostasis is necessary for the proper coding of information, Moser and coworkers (1998) showed that saturation of synaptic strength is detrimental for learning. It is, therefore, reasonable to postulate that activity-induced genes, like Arc/Arg3.1, whose protein products are localized to activated dendritic regions, may play a role in local synaptic homeostasis after being upregulated by strong synaptic activation.
Indeed, homer 1a, an activity-dependent gene whose induced expression correlates with that of Arc/Arg3.1 in CA1 neurons (Vazdarjanova et al., 2002), has also been shown to reduce synaptic currents (Sala et al., 2003). Furthermore, the activity-dependent gene cpg-2 has been recently implicated in the regulation of glutamate receptor endocytosis (Cottrell et al., 2004). Another activity- inducible gene, the polo-like kinase SNK, contributes to the loss of dendritic spines after being up-regulated (Pak and Sheng, 2003). SNK also reduces the synaptic levels of PSD-95, a scaffolding molecule involved in the regulation of AMPAR-mediated currents (El-Husseini et al., 2000).
As in our system Arc/Arg3.1 upregulation leads to a reduction in AMPA transmission through GluR2/3 removal, recently potentiated synapses with its higher GluR1/2 content would be less affected. As the model in Figure 9 suggests, in such a scenario Arc/Arg3.1 would impose a threshold to potentiation, it would increase the potentiated-to-nonpotentiated synaptic-strength ratio, and at the same time it would homeostatically regulate total transmission levels. This model is consistent with data from Wang and colleagues (2006), who showed that, in orientation selective neurons in the visual system, Arc/Arg3.1 is involved in suppressing responses to nonpreferred orientations without affecting preferred orientation-driven responses. In that case, the synapses being strengthen by the preferred orientation responses could induce Arc/Arg3.1 transcription and localization with the consequent reduction in the AMPA content of less-potentiated synapses belonging to the nonpreferred orientations.
The reduction we observed in the amplitude of AMPAR-mediated synaptic current amplitudes upon increased Arc/Arg3.1 expression is consistent with a homeostatic role for activity-induced Arc/Arg3.1 upregulation, as also suggested by Shepherd et al. (2006; this issue of Neuron). Our results are supported by the biochemical analysis of Arc/Arg3.1 (Chowdhury et al., 2006) and by the role of Arc/Arg3.1 in the refinement of orientation selectivity (Wang et al., 2006). Our data, and data from other activity-induced genes, open the possibility for a concerted action of such genes to maintain synaptic strengths within a functional operating range.
The Arc/Arg3.1 ORF fused to EGFP was obtained from Dr. Chowdhury in pEGFP-C3 (Clontech). Arc/Arg3.1(Δ91–100) was obtained from Dr. Chowdhury in pBluescript and fused to GFP using pEGFPC3 (Clontech). Coexpression vectors were constructed in pCITE (Invitrogen). Pep-AP2 (pepΔA849–Q853) was generously provided by Dr. Morgan Sheng and has been previously described (Lee et al., 2002). The GluR2 c-tail has also been described (Shi et al., 2001). After the plasmids containing the genes of interest were prepared, the inserts were transferred into pSinRep5 (Invitrogen) to produce replication- deficient Sindbis virus as described (Malinow et al., 1999).
Organotypic hippocampal slices were prepared from postnatal 5- to 7-day-old rats as previously described (Stoppini et al., 1991) and cultured for 6 to 9 days, unless noted, in culture medium (Musleh et al., 1997). Recombinant protein expression in CA1 pyramidal neurons was achieved by pressure injection of the viral stock with 0.1% fast green using a sharp glass pipette and a Picospritzer (General Valve Co.). When drugs were used, they were added to the culture medium immediately after virus injection. With the exception of the experiment in DL-APV, the extracellular recording solution did not contain the drug, and the recording time was limited to less than 40 min. Recombinant protein distribution and localization to dendritic spines was unaffected by drug treatments.
We addressed the time course of Arc/Arg3.1 expression qualitatively by performing 160 identical virus injections in ten groups of four slices each. Consistently with the Sindbis literature, no fluorescent protein expression could be observed before 4 hr. At 4 hr, some fluorescent glial cells could be observed, but no pyramidal neurons. At 6 hr, all 16 injection sites examined showed a few fluorescent pyramids. At 7 hr, the first recognizable spines appeared, although not fully loaded. Distal dendrites were not visible or very faint. After 9 hr, the density of fluorescent pyramids per injection site was substantial, distal dendrites, and spines were visible (Figure S1).
Electrophysiological recordings were obtained 10 to 20 hr after virus injection, from CA1 pyramidal neurons located ~200 μm from the injection site to ensure an infection density low enough to identify non-infected cells with ease. Slices were perfused with artificial cerebrospinal fluid (pH 7.4) at 29°C containing 119 mM NaCl, 2.5 mM KCl, 4 mM CaCl2, 4 mM MgCl2, 26 mM NaHCO3, 1 mM NaH2PO4, 11 mM D-glucose, 0.1 mM picrotoxin, and 0.004 mM 2-chloroadenosine gassed with 5% CO2-95% O2. The only exception was the AMPA mEPSC recordings (1 mM MgCl2 was used instead of 4 mM, and 2 μM TTX was added) and the recordings of GABAR-mediated currents, in which the picrotoxin was omitted. Patch recording pipettes (3–6 MΩ) were filled with intracellular solution (pH 7.25), containing 115 mM cesium methanesulfonate, 20 mM ClCs, 10 mM HEPES, 2.5 mM MgCl2, 4 mM Na2ATP, 0.4 mM Na3GTP, 10 mM sodium phosphocreatine, and 0.6 mM EGTA, with the exception of the resting potential recordings, in which a potassium-based intracellular solution was used.
Whole-cell voltage-clamp recordings (or current clamp to measure the resting potentials) were made simultaneously with Axopatch-1D amplifiers (Axon Instruments) in infected and control uninfected cells separated by less than 5–10 μm, but in most cases adjacent. Synaptic responses were evoked with single voltage pulses (200 μs, 1–10V) using bipolar electrodes placed in stratum radiatum, 100 to 200 μm from the pyramidal layer. Responses were evoked every 3 s with a stimulus intensity set to evoke AMPAR-mediated responses of around −30 to −50 pA. This range allows us to reliably detect increases and decreases in amplitudes. AMPAR-mediated responses were recorded at −60 mV, NMDAR-mediated responses at +40 mV, and GABAR-mediated responses at 0 mV. The average of 50 to 100 responses at each holding potential was used to calculate the current amplitudes. AMPAR- and GABAR-mediated current amplitudes were calculated at the peak of the response; NMDAR-mediated current amplitudes were computed 50 ms after the stimulation as the average over 5 ms of recording. Cells were visualized using oblique illumination in a Zeiss Axioskop microscope with a 60× water-immersion objective. Infected cells were identified using epifluorescence illumination with filter settings suitable for GFP. LTD was induced with 1 Hz stimulation for 5 min at−10 mV of holding potential. Test responses were evoked every 3 s, and five responses were averaged to obtain every data point. The data were digitized with an ITC-18 board (Instrutech) and acquired with an IgorPro-based (Wavemetrics) custom software.
For all experiments, holding current, series, and input resistance were monitored, and no significant difference was observed between infected and uninfected or transfected cells at any holding potential.
Data analysis and statistics were performed in Matlab (Mathworks). All decay time constants were obtained by fitting the responses to a single exponential. The data sets were probed for normality using Shapiro-Wilk’s test. Because all data sets could not be assumed to be normally distributed, nonparametric statistics were used and medians with interquartile ranges (IQR) are shown instead of mean with standard error of the mean. The only exception is the LTD data, which satisfied normality and distributional symmetry criteria. Normalization of the data using Box-Cox transformation and testing using parametric methods did not change significance in any case. Wilcoxon paired test was used to compare infected against uninfected evoked EPSC amplitudes in each experiment. When comparisons were made across different experiments, the infected/uninfected amplitude ratio for every cell pair was used. Mann-Whitney test was used to compare two groups and Kruscal-Wallis to compare more than two groups. The significance threshold was set at p = 0.05 for all tests. For all the experiments in which the Wilcoxon test p value is less than 0.05, comparison against the effect of GFP expression alone using Mann-Whitney also shows significance and vice versa. The percent reduction in AMPAR-mediated currents reported is the median of the reduction for every pair in the data set referred to.
Fifty picomoles of predesigned Arc/Arg3.1 siRNAs (Ambion) or a control siRNA bearing no homology to known genes (Ambion cat#4611) were cotransfected with pCMV-Arc/Arg3.1 (a plasmid containing the whole coding sequence for Arc/Arg3.1 under the CMV promoter) into Cos-1 cells (60% confluent in 60 mm diameter plates) in a 100:1 molar ratio using Lipofectamine 2000 (Invitrogen) following manufacturer’s specifications. The following day, the cells were scraped in PBS with 1% Triton X-100, 0.2% SDS, and protease inhibitor cocktail (CompleteMini, Roche), homogeneized, and centrifuged at 4°C for 4 min at 1000 × g. The supernatant was used for western blot against Arc/Arg3.1 and reprobing for β-tubulin as explained below. One Arc/Arg3.1 siRNA (Ambion ID#199057; GCU GAUGGCUACGACUACA) produced more than 90% knockdown compared to sister plates transfected with the control siRNA. Gene Gun bullets were prepared using 12.5 mg of 1.6 μm diameter gold spheres and 3 μg of that siRNA + 10 μg pEGFP (Clontech) following the procedure detailed in the Helios Gene Gun (Bio-Rad) manual. Hippocampal cultured slices were transfected using the biolistics method. The following day the slices were exposed for 5 min to 100 μM picrotoxin, washed, and incubated with fresh culture medium for 3 to 10 hr before recording following the procedure outlined above. Eight slices exposed to the picrotoxin treatment and eight control slices were processed for Arc/Arg3.1 western blot, the protein content of the extracts was measured by the Bradford method, and equal amount of protein of picrotoxin-treated and control extracts were loaded into adjacent lanes of a 7% gel for PAGE-SDS. The western blot was performed as outlined below and the picrotoxin-treated-to-control Arc/Arg3.1 ratio ranged from 7 to 10.
Organotypic slices were homogenized in glass-teflon homogenizer (1200 rpm) on ice-cold Tris-HCl 25 mM(pH 7.4) with 320 mM sucrose and protease inhibitors (CompleteMini, Roche) and centrifuged at 13,000× g for 15 min at 4°C. The pellet was resuspended and loaded in a 7%gel for PAGE-SDS. The proteins were transferred to nitrocellulose membranes and incubated with anti-Arc/Arg3.1 rabbit polyclonal antibody at 1:5000. Horseradish peroxidase-conjugated secondary antibody (Bio-Rad 170-6515) at a 1:3000 dilution and ECL chemoluminescence (Amersham) were used for detection.
For the recombinant Arc/Arg3.1 to endogenous Arc/Arg3.1 ratio western blots, slices received multiple viral injections, spaced 100 to 150 μm, spanning all CA1 area. These multiple injections and spacing ensures a very high infection efficiency (see below). Fourteen hours later, only the area of CA1 showing the highest infection rate was dissected under a fluorescent magnifying scope. The dissected regions were processed as described above. For the time course western blot, every slice received an identical injection of viral stock with the same pipette, injection duration, and pressure. The slices were processed at the indicated time points and stored at −80°C until used.
Quantification of the western blots to calculate recombinant Arc/Arg3.1 to endogenous Arc/Arg3.1 ratios was done using Fluorchem scanner and software (Alpha Innotech). The exposure times and scanning parameters were optimized to minimize signal saturation. Rectangular sections of the same area were used to calculate integrated density values (IDV) over a band, and background above and below the band were averaged. IDV values minus background were used to calculate the ratios.
Slices received multiple viral injections, spaced 100 to 150 μm, spanning the entire CA1 area to ensure a high density of infection. The percent infected cells was calculated by counting the fraction of DAPI-stained nuclei showing green fluorescence in the same field of view (median infected cells, 85.3%; range, 65.7% to 97.1%; n = 19). Surface biotinylation in slices was performed as described (Rivera et al., 2004; Thomas-Crusells et al., 2003) with minor modifications. Slices were washed in ice-cold ACSF three times for 5 min and incubated on ice for 45 min in ACSF gassed with 5% CO2-95% O2 containing 100 μM of NHS-SS-biotin (Sigma B5161). To block the remaining reactive biotinylation reagent, the slices were washed twice for 5 min in 100 μM lysine. The CA1 region showing intense fluorescence was dissected and homogenized in teflon-glass homogenizer (1200 rpm) in 20 mM Tris-HCl (pH 7.5), 50 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS and protease inhibitors (Complete Mini EDTA-free, Roche). The homogenate was centrifuged 5 min at 4°C > 10000 × g. The supernatant volume was measured and 20% of it separated to load in the total protein lane. The remaining 80% of the homogenate was taken to 1 ml with ACSF and rotated overnight at 4°C with 40 μl of streptavidin beads (Ultralink, Pierce). The beads were precipitated at 3500 × g for 1 min, the supernatant discarded, and 1× Laemmli buffer added. The samples were boiled 5 min and electrophoresed as described for the Arc/Arg3.1 western blot. The antibodies used were as follows: GluR2 1:500 (Chemicon AB1768), GluR1 1:200 (Chemicon AB1504), PICK1 1:50 (Upstate 07-293), and β-tubulin 1:2000 (Santa Cruz Biotechnology sc-9104). Band intensities were quantified as described for the Arc/Arg3.1 western blots. The surface-to-total ratios, (surface density − background)/(total density − background), were normalized to the median value for the GFP control. To ensure linearity, several exposures were quantified for every experiment. The surface-to-total ratio was constant over the chosen exposures with a mean percent variation between the largest and smallest ratio of under 2.6% in all cases (three different exposures). The middle exposure was used for the quantifications and examples shown. We also normalized total GluR1 or GluR2 to β-tubulin in the same lane in order to measure total amounts of each subunit. We did not detect changes in total GluR1 or GluR2 in Arc/Arg3.1-expressing slices compared to GFP-expressing slices.
The images were obtained on a custom two-photon excitation laser scanning microscope, acquired, and the stacks flattened with Fluoview software (Olympus).
We thank A. Chowdhury (P. Worley Lab.) for pEGFPC3-Arc/Arg3.1, pBSK-Arc/Arg3.1, pBS-Arc/Arg3.1(Δ91–100), and Arc/Arg3.1 antibody, M. Sheng for pepΔA849-Q853, L. Manganas (G. Enikolopov Lab.) for β-tubulin antibody, and Fujisawa Healthcare for FK506. We also thank members of the Cline lab for their comments on the manuscript. This work was supported by Howard Hughes Medical Institute Predoctoral Fellowship, David and Fanny Luke Fellowship (E.M.R.V.), and National Institutes of Health (R.M. and H.T.C.).
The Supplemental Data for this article can be found online at http://www.neuron.org/cgi/content/full/52/3/461/DC1/.