|Home | About | Journals | Submit | Contact Us | Français|
Learned associations between environmental stimuli and rewards play a critical role in addiction. Associative learning requires alterations in sparsely distributed populations of strongly activated neurons, or neuronal ensembles. Until recently, assessment of functional alterations underlying learned behavior was restricted to global neuroadaptations in a particular brain area or cell type, rendering it impossible to identify neuronal ensembles critically involved in learned behavior.
We used Fos-GFP transgenic mice that contained a transgene with a Fos promoter driving GFP expression to detect neurons that were strongly activated during associative learning, in this case context-independent and context-specific cocaine-induced locomotor sensitization. Whole cell electrophysiological recordings were used to assess synaptic alterations in specifically activated GFP-positive (GFP+) neurons compared to surrounding non-activated GFP-negative (GFP−) neurons 90 min after the sensitized locomotor response.
Following context-independent cocaine sensitization, cocaine-induced locomotion was equally sensitized by repeated cocaine injections in two different sensitization contexts. Correspondingly, silent synapses in these mice were induced in GFP+ neurons, but not GFP− neurons, following sensitization in both of these contexts. Following context-specific cocaine sensitization, cocaine-induced locomotion was sensitized exclusively in mice trained and tested in the same context (Paired), but not in mice that were trained in one context and then tested in a different context (Unpaired). Silent synapses increased in GFP+ neurons, but not in GFP− neurons from Paired mice but not from Unpaired mice.
Our results indicate that silent synapses are formed only in neuronal ensembles of the nucleus accumbens shell that are related to associative learning.
Learned associations between drugs of abuse and environment stimuli play a critical role in drug addiction. One of the key aspects of these drug-related memories is that one specific stimulus can induce a conditioned drug behavior such as relapse, while other unrelated stimuli do not. The neurobiological mechanism that underlies this type of learning must encode these drug-related memories with a degree of resolution sufficient to discriminate between memories activated by different sets of stimuli. Recent studies support the hypothesis that specific patterns of sparsely distributed Fos-expressing neurons act together as neuronal ensembles that mediate addiction-related learned behaviors, including context-specific expression of reinstatement of drug seeking (1, 2), incubation of drug craving (3) and context-specific locomotor sensitization (4); see (5, 6) for review. These Fos-expressing neuronal ensembles are formed by only a minority of neurons (~2–12%) that are selected by the relevant cues. The enormous number of possible permutations of these neuronal ensembles provides sufficiently high resolution to discriminate between different associative memories. Thus selective alterations within these neurons are likely to play a key role in learning and maintenance of these drug-related memories.
Synaptic plasticity is the key candidate neural mechanism for encoding learning and memory processes. Many years of research have shown that exposure to cocaine and other drugs of abuse can induce synaptic alterations in reward-related brain areas such as the nucleus accumbens and ventral tegmental area (7–11). These neuroadaptations include alterations in the balance of AMPAR-mediated and NMDAR-mediated transmission (7, 9), changes in AMPAR subunit composition (12, 13), alterations in synaptic plasticity (11, 14), and the development of silent synapses (15, 16). Synaptic alterations in addiction-relevant brain areas have also been shown to play an important role in the development and persistence of behavioral models in addiction research (17–19). However these synaptic alterations are observed in whole brain areas or in particular cell types regardless of their neural activity during the learned behaviors. Although these global alterations can play important general roles in learning and maintenance of memories, other mechanisms are required to encode the high-resolution information in cue-specific memories and discriminate them from other memories stored in the same brain area. Based on previous studies from our lab and others, we hypothesize that unique synaptic alterations induced within drug cue-activated Fos-expressing neuronal ensembles play a critical role in the expression of many learned associations in conditioned drug behaviors (5, 6).
We recently used Fos-GFP transgenic mice to examine synaptic alterations in Fos- and GFP-expressing neurons following context-specific sensitization of cocaine-induced locomotion sensitization (20). The small proportion of GFP-expressing neurons that was strongly activated during sensitized cocaine-induced locomotion exhibited large percentage increases of silent synapses and related synaptic alterations that were not present in the majority of surrounding GFP-negative neurons that were less activated during behavior. We hypothesized that the emergence of silent synapses might represent a cellular mechanism contributing to the development of learned associations, but were unable to determine this because mice were trained and tested in the same context. Under these circumstances, the development of silent synapses could result from a compensatory response of neurons that were particularly strongly activated by repeated exposure to cocaine with no role in associative learning. To address whether silent synapses are induced in a context-specific manner, we examined synaptic alterations in GFP+ and GFP− neuronal populations following two different forms of cocaine-induced locomotor sensitization: (1) context-independent form of sensitization in which animals displayed a sensitized locomotor response regardless of their previous training context; (2) context-specific form of sensitization in which mice were able to distinguish between two different contexts, and only displayed a sensitized locomotor response when tested in the context in which they were trained.
We used 163 male and female Fos-GFP transgenic mice in our experiments. These mice were initially obtained in 2008 from Alison Barth at Carnegie Mellon University. We bred male heterozygous Fos-GFP mice with female wild-type C57BI/6 mice (Charles Rivers Laboratories) for more than 15 generations at NIDA IRP facilities. An additional 24 male wild-type C57BI/6 mice were obtained from Charles Rivers Laboratories. Mice were maintained in a temperature and humidity controlled facility on a 12-h light-dark cycle. Mice were separated and individually housed with ad libitum food and water for 3–5 days prior to training. All experiments were conducted during the light phase. Animal protocols were approved by the Animal Care and Use Committee of the National Institute on Drug Abuse Intramural Research Program and were carried out according to US National Institutes of Health Guidelines.
Male and female Fos-GFP mice (n=47) were habituated 3X once daily for 60 min each in locomotor activity chambers (43 X 43 cm, Med Associates) and then divided randomly into two groups with approximately equal numbers of male and female mice. Mice in the Paired group were injected once daily for 5 days with cocaine (15 mg/kg i.p.) or saline (10 μl/g) in the locomotor activity chamber that we call Context A. Mice in the Unpaired group were injected once daily for 5 days with cocaine (15 mg/kg i.p.) or saline (10μl/g) in a different context that we call Context B-Independent (BInd), which was circular and contained a grid floor. On test day, following 6–11 days in their home cages, all mice were given a single injection of cocaine (20 mg/kg i.p.) in Context A and perfused 90 minutes later (see experimental timeline in Figure 1A).
Male and female Fos-GFP mice (n=116) were divided randomly into two groups with approximately equal numbers of male and female mice. Mice in the Paired group were injected once daily for 5 days with cocaine (15 mg/kg i.p.) or saline (10 μl/g) in the locomotor activity chamber that we call Context A. Mice in the Unpaired group were injected once daily for 5 days with cocaine (15 mg/kg i.p.) or saline (10 μl/g) in a different context that we call Context B-Specific (B-Sp), which was a round bowl with bedding and a toy, lights were dimmed and music (No Doubt, “Tragic Kingdom”) was played continuously. On test day, following 6–11 days in their home cages, all mice were given a single injection of cocaine (20 mg/kg i.p.) in Context A and perfused 90 minutes later (see experimental timeline in Figure 4A). In a separate experiment, 24 male C57/Bl6 mice received identical training, but on test day half the mice received cocaine injections (20mg/kg) and half received saline injections (10uL/g).
Fos-GFP mice were transcardially perfused 90 minutes following the final injection of cocaine on test day with 4% paraformaldehyde (PFA). Fos immunohistochemistry and Fos+NeuN double-labeling were performed as previously described (Koya 2009). For detailed information, see supplemental methods.
On test day, Fos-GFP mice were deeply anesthetized with isoflurane (60–90 seconds) 90 minutes after cocaine injections and transcardially perfused with ice-cold cutting solution. Coronal brain slices containing nucleus accumbens shell were prepared and whole cell voltage-clamp recordings were performed in GFP+ and GFP− medium spiny neurons of the nucleus accumbens shell as previously described (20). Spontaneous EPSC (sEPSC) data were collected using WinEDR software (J. Dempster, University of Strathclyde) and analyzed using MiniAnalysis (Synaptosoft). Minimal stimulation assays were performed as previously described (16, 20). For additional information, see supplemental methods.
Group data are presented as mean +/− SEM. For detailed information, see supplemental methods.
We used a sensitization procedure (Figure 1A) similar to that in our previous study (20) to assess cocaine-induced locomotor sensitization in Fos-GFP mice. Mice in the Paired group received repeated injections in the locomotor activity chamber (Context A) during training and then injected in the same Context A on test day. Mice in the Unpaired group received repeated injections in a different Context B-Ind during training and then injected in Context A on test day. Using two-way ANOVA, we found that repeated cocaine injections induced robust locomotor sensitization, regardless of training context (F1,43=30.69, p<0.0001 for Repeated drug; not significant for Training context or Interaction) (Figure 1B). When locomotor activity was assessed over time, cocaine-induced locomotion was significantly enhanced by previous repeated cocaine injections (F1,46=19.51, p<0.001 for Repeated drug) following test injections. Overall, sensitization was context-independent using these experimental conditions.
Ninety minutes following cocaine test injections, we obtained coronal slices for slice electrophysiology from the mice that were repeatedly injected with cocaine (n=8–10 cells from 6–9 mice/group [n=22 mice in total]) and compared synaptic alterations in GFP+ and GFP− neurons of nucleus accumbens shell (Figure 2A). We began by assessing the percentage of silent synapses using the minimal stimulation assay (for additional information see supplemental methods). Example experiments and traces are shown in Figure 2B. In mice that received repeated cocaine injections, two-way ANOVA indicated that the percentage of silent synapses was higher in GFP+ neurons than in GFP− neurons, regardless of training context (F1,31=70.43, p<0.0001 for GFP expression; not significant for Training context or Interaction) (Figure 2C). In mice that received repeated saline injections, a t-test indicated that the percentage of silent synapses was not different between GFP+ and GFP− neurons (t15=0.79, p=0.44, n=8–9 cells/group, GFP− = 6.52 +/− 3.06, GFP+ =9.83 +/− 2.87 with Paired/Unpaired groups collapsed together). Thus under these experimental conditions, Training context had no effect on development of silent synapses following cocaine-induced locomotor sensitization.
We also compared spontaneous EPSCs in GFP+ and GFP− neurons from Paired and Unpaired mice (example traces are shown in Figure 3A). Two-way ANOVA indicated a significantly lower frequency of spontaneous EPSCs in GFP+ than in GFP− neurons (F1,38=15.75, p < 0.001 for GFP expression; not significant for Training context or Interaction). The amplitude of spontaneous EPSCs were not different between GFP+ and GFP− neurons in Paired or Unpaired mice (Figure 3B). We also performed an alternative analysis of cumulative probability of spontaneous EPSC frequency and amplitude data (Figure 3C). Using the Kolmogorov-Smirnov test, we determined that there were significant differences in the cumulative probability of spontaneous EPSC frequencies between GFP+ and GFP− neurons from both Paired (p<0.0001) and Unpaired (p<0.05) groups, but no differences between groups for the cumulative probability of spontaneous EPSC amplitudes.
We used a different sensitization procedure to induce a context-specific form of sensitization in a separate group of Fos-GFP mice (Figure 4A). We removed the habituation period, and made the alternate context (Context B-Sp) more distinct from the locomotor activity chamber (Context A). Context B-Sp was a large bowl containing bedding and a toy; lights in the room were dimmed and music was played throughout the session. On test day, mice were given a saline injection followed 60 minutes later by a test injection of cocaine (20 mg/kg) in Context A. Locomotor activity data was divided into 15-min bins (Figure 4B) and analyzed using 4-factor ANOVA, with Training drug (cocaine, saline) and Training context (A, B-sp) as between-subjects factors and Test drug (cocaine, saline) and Time (−45 to 60 min) as within-subjects factors. There was a significant 4-way interaction (F3,138=3.99, p<0.01), as well as a significant three-way interaction between Training drug, Training context, and Time (F3,138=7.33, p<0.001). Post-hoc analyses revealed a conditioned locomotion effect in the Paired cocaine-treated group following the saline test injection (timepoint −45 minutes: F1,49=5.60, #p < 0.05). The Paired cocaine-treated group also displayed enhanced locomotion 15–30 min following the cocaine test injection (15 minutes: F1,49=7.31, 30 minutes: F1,49=4.19, #p < 0.05, *p < 0.01). We also analyzed locomotor data taken from the first 15 minutes following cocaine test injections (Figure 4C) using 2-way ANOVA and found that there was a significant effect of Training drug (F1,49= 15.17, p < 0.001) post-test revealed a as well as a significant interaction (F1,49= 5.24, p < 0.05). Bonferroni s significant difference between mice trained with saline versus cocaine in the Paired group only (*p < 0.0001). Thus mice expressed the learned behaviors context-specific sensitization and conditioned locomotion.
Ninety minutes following cocaine test injections, we obtained coronal slices for slice electrophysiology from mice following context-specific sensitization (n=7–8 cells from 5–7 mice/group, n= 17 mice in total) and compared synaptic alterations in GFP+ and GFP− neurons of nucleus accumbens shell of Paired and Unpaired mice following context-specific sensitization. Example experiments and traces are shown in Figure 5B. In mice that received repeated cocaine injections, two-way ANOVA indicated a significant main effects for GFP expression (F1,25=21.78, p<0.0001) and Training context (F1,25=16.94, p<0.0001), and a significant interaction (F1,25=24.15, p<0.0001). Bonferroni s post hoc tests confirmed that the percentage of silent synapses was greater in GFP+ neurons than in GFP− neurons from Paired mice but not from Unpaired mice (Figure 5C). In mice that received repeated saline injections, a t-test indicated that the percentage of silent synapses was not different between GFP+ and GFP− neurons (t8=0.41, p=0.69 GFP+: 9.21% +/− 4.10%, GFP−: 11.01% +/− 1.44% with Paired/Unpaired groups collapsed together).
We also examined spontaneous EPSCs in GFP+ and GFP− neurons from Paired and Unpaired mice (example traces are shown in Figure 6A). Two-way ANOVA of spontaneous EPSC frequency data using a 2-way ANOVA indicated a significant interaction between GFP expression and training context (F1,32=8.22, p<0.01). Post hoc tests confirmed that spontaneous EPSC frequency was lower in GFP+ neurons than in GFP− neurons from Paired mice but not in Unpaired mice (Figure 6B). The amplitude of spontaneous EPSCs were not different between GFP+ and GFP− neurons in Paired or Unpaired mice (Figure 6B). We also performed an alternative analysis of cumulative probability of spontaneous EPSC frequency and amplitude data (Figure 6C). Using the Kolmogorov-Smirnov test, we observed a rightward shift in cumulative probability of spontaneous EPSC frequency in the GFP+ neurons of the Paired mice (p < 0.0001). No differences were observed between groups for the cumulative probability of spontaneous EPSC amplitudes from Paired or Unpaired mice.
We examined Fos expression in the nucleus accumbens shell of (n=6/group, 24 total) that underwent context-specific sensitization. Two-way ANOVA revealed no significant differences for main effects or interaction. An acute control experiment confirmed that Fos expression was greater following acute cocaine injections than following saline injections on test day for both Paired and Unpaired group mice (F1,12=9.10, p=0.011 for Acute drug injection; not significant for Training context). For descriptive purposes only, we assessed the percentage of all neurons that expressed Fos by double-labeling for Fos and the neuronal marker NeuN (Figure 5A). We used an additional set of sections from mice that received cocaine test injections following context-specific sensitization. Approximately 5% of NeuN-labeled neurons co-expressed Fos (Paired: 5.07+/−0.44%, Unpaired: 4.50+/−0.39%). In the case of context-independent sensitization, 2–3% of NeuN-labeled neurons co-expressed Fos (4).
We examined whether silent synapses were induced in context-specific Fos-expressing neuronal ensembles in mouse nucleus accumbens shell 90 min after expression of the sensitized locomotor response. Mice that had undergone context-specific sensitization exhibited sensitized cocaine-induced locomotion when tested in the same context in which they were trained, but not when tested in a different context. Psychostimulant-induced locomotor sensitization has been reported for decades, and may arise due to a pharmacological effect of repeated drug exposure, or due to the formation of an association between environmental stimuli and the rewarding effects of drug (for reviews see (21–23). Context-specific sensitization in mice required very distinct contexts since less separable differences in context did not produce context-specific sensitization (24–26). For both forms of sensitization, silent synapses were increased exclusively in the small number of GFP+ neurons that were activated on test day, and not in the surrounding GFP− neurons. In mice that had undergone context-specific sensitization, silent synapses were induced in GFP+ neurons of mice in the paired group, but not in the unpaired group, which corresponds with context-specific sensitization of cocaine-induced locomotion.
Our context-independent sensitization procedure is similar to the protocol used in which we discovered more silent synapses in GFP+ neurons following repeated cocaine injections (20). In this study, all mice were trained and tested in the same context, rendering it impossible to determine whether silent synapses in GFP+ neurons arise as a result of pharmacological effects of repeated cocaine experience or whether they are induced selectively in neurons determined by the drug-associated context. In our current context-independent experiment, silent synapses increased in the small proportion of GFP+ neurons that were strongly activated on test day for both paired and unpaired groups. Based on this experiment alone, either explanation is possible. However, an alternative explanation is that the mice in our context-independent sensitization experiment failed to distinguish between the two contexts, likely activating a similar neuronal ensemble in the accumbens shell and expressing sensitized locomotion in both contexts. The context-specific sensitization procedure allowed us to test this hypothesis. If the specific set of neurons that is repeatedly activated during training is determined by the context, then changing the context on test day should activate a different set of neurons. Since silent synapses require repeated activation during sensitization (20), then we should observe silent synapses in the paired group, but not in the unpaired group. We confirmed this in our context-specific sensitization experiment where silent synapses were not induced in GFP+ neurons from unpaired group mice. Thus silent synapses are induced only in context-selected neurons that were repeatedly activated during sensitization. The population of neurons expressing Fos following cocaine sensitization is small, representing ~3–5% of nucleus accumbens shell neurons (4). In vivo recording studies have examined nucleus accumbens firing following cocaine self-administration and have found that ~25% of cells sampled show a phasic increase in firing following response-contingent cocaine administration (27, 28). Based on previous studies of Fos induction mechanisms, we hypothesize that Fos-expressing neurons in our study represent the most strongly activated subset of neurons undergoing increased phasic activity during repeated cocaine injections (5).
Context-specific selection of a small population of sparsely distributed neurons for silent synapse formation corresponds with the hypothesis that Fos-expressing neuronal ensembles encode memories to mediate learned behaviors (5, 6), which is based on Hebb s cell assembly hypothesis (29). A previous study from our lab had empirically shown that the same neuronal ensembles in nucleus accumbens are reactivated following re-exposure to the same context following context-specific sensitization (30). We used double-labeling with deltaFosB immunohistochemistry to label neurons that were activated during repeated cocaine injections in one context and c-fos mRNA in situ hybridization to label neurons that were activated later in the same or different context – 87% of the same neurons were reactivated following re-exposure to the same context. We also used the Daun02 inactivation procedure to inactivate Fos-expressing neuronal ensembles in the nucleus accumbens that were activated by cocaine injections in the drug-paired context and found reduced expression of sensitization when the rats were tested in the same drug-paired context three days later (4). This indicates that the same neuronal ensembles were reactivated by exposure to the same context. We have also shown that exposure to the same context reactivates the same neuronal ensembles in the ventromedial prefrontal cortex and accumbens to play causal roles in the expression of context-specific reinstatement of heroin and cocaine seeking respectively (1, 2), as well as neuronal ensembles in orbitofrontal cortex in incubation of cue-induced heroin seeking (3). Based on these studies, silent synapses appear to be induced in Fos-expressing ensembles that mediate expression of the learned association between the drug-paired context and cocaine. This is further supported by the finding that repeated cocaine injections, but not a single cocaine injection, in the same context are required to induce silent synapses in Fos expressing ensembles (20). Overall, our findings indicate that silent synapses are induced only in neurons that were repeatedly activated during training and then reactivated on test day in the same context. Global alterations of silent synapses and other synaptic alterations, such as AMPA/NMDA ratio (9), AMPA receptor long-term depression (31) and A-type potassium currents (32), have also been reported following cocaine sensitization. However, these global alterations do not have the required degree of resolution to encode the high-resolution information that defines the different contexts in context-specific sensitization. These global alterations may instead play important roles in preparing neurons for learning-specific alterations or generally increasing or decreasing the expression of the learned behaviors.
The detailed mechanism underlying the formation of silent synapses in these Fos and GFP-expressing neurons is not known. Previous studies have demonstrated increased silent synapse formation in the general population of nucleus accumbens neurons following cocaine locomotor sensitization (15, 16). In these studies, silent synapses were increased in randomly selected neurons of the nucleus accumbens shell for 3–5 days following cocaine sensitization due to the insertion of new NR2B-containing spines. In contrast, the appearance of new NR2B-containing spines does not underlie the formation of silent synapses that we observed in GFP+ neurons following 6–11 days of withdrawal from cocaine sensitization (20). A possible alternative explanation is that silent synapses in GFP+ neurons following context-specific sensitization may involve removal of AMPA receptors from dendritic spines (33–36). It has been previously shown that the ratio of surface to intracellular GluR1 is decreased 24 hours following cocaine challenge in sensitized animals, lending support to this idea (37).
The appearance of silent synapses exclusively in this activated neuronal population, only in mice that learned the context-reward association, suggests that silent synapses could represent a mechanism for encoding learned associations. However, it is possible that silent synapses may be induced in strongly activated neurons only as a compensatory mechanism. Strong activation of GFP+ neurons by convergent glutamatergic input as well as dopaminergic modulation (38, 39) could cause activated cells to tune down their response by internalizing AMPA receptors (40). In this case, the functional alterations that encode learned associations could precede the development of silent synapses. Indeed, one prior study determined that there was no relationship between AMPAR surface expression in nucleus accumbens and the expression of cocaine sensitization (41). Silent synapses could potentially represent a metaplastic mechanism (42, 43) through which synapses may be primed for subsequent experience-induced plasticity, most likely through AMPA receptor insertion. Other metaplastic mechanisms have been previously identified (44, 45) in several brain regions, which may be modulated by drug exposure (46, 47). There is evidence in support of this scenario from recent studies of the neural correlates of incubation of craving. Rats allowed to self-administer cocaine show an increase in silent synapses on withdrawal day 10, but by day 45 of withdrawal, when animals are showing increased cue-induced cocaine seeking, synapses are unsilenced by the insertion of CP-AMPARs (35, 36). Reversal of silent synapse maturation decreases cue-induced craving, however the question remains as to the specific contribution of silent synapses.
Until recently, most studies examined the role of randomly selected neurons in a particular brain area, or restricted their focus to a certain cell type, since it was not possible to identify the critical population of neurons that encode learned associations (6). Using Fos as an indicator of strong neuronal activation has proven reliable in identifying behaviorally activated ensembles that play a causal role in the expression of drug-related learned behavior (4). As mentioned previously, there is a high degree of overlap between the neuronal ensemble that is activated during training and during test in the animals that express sensitization. One limitation is the need for acute cocaine in the test context on test day to reactivate this population of neurons. A second limitation is the temporal separation between the induction of the Fos promoter and the expression of GFP, which is maximal at 90–120 minutes post-induction. Future studies will employ the Fos-Tet-Cre transgenic rat system to identify Fos-expressing neuronal ensembles without the need for acute cocaine to induce Fos, and will allow us to follow the time course of the development of silent synapses as well as determine if silent synapses play a causal role in context-specific sensitization (6).
We would like to acknowledge Yavin Shaham for helpful suggestions regarding behavioral procedures, statistics, and the manuscript, Jennifer Bossert for assistance with immunohistochemical procedures and statistics, Brandon Warren for assistance with animal perfusions, and Fabio Cruz, Rodrigo Leao, and Klil Babin for assistance in running behavioral experiments.
This research was supported by the National Institute on Drug Abuse Intramural Research Program, NIH. P.E.Carneiro de Oliveira was supported by a Capes fellowship from Brazil. All authors report no biomedical financial interests or potential conflicts of interest.
Author contributionsExperiments were planned by LRW, AB, BTH and carried out by LRW, KBM, PECO, RVF. LRW, AB, and BTH wrote the manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.