Bidirectional synaptic structural plasticity after chronic cocaine administration occurs through Rap1 small GTPase signaling

Summary

Dendritic spines are the sites of most excitatory synapses in the CNS, and opposing alterations in the synaptic structure of medium spiny neurons (MSNs) of the nucleus accumbens, a primary brain reward region, are seen at early vs. late time points after cocaine administration. Here we investigate the time-dependent molecular and biochemical processes that regulate this bidirectional synaptic structural plasticity of NAc MSNs and associated changes in cocaine reward in response to chronic cocaine exposure. Our findings reveal key roles for the bidirectional synaptic expression of the Rap1b small GTPase and an associated local-synaptic protein translation network in this process. The transcriptional mechanisms and pathway-specific inputs to NAc that regulate Rap1b expression are also characterized. Collectively, these findings provide a precise mechanism by which nuclear to synaptic interactions induce “metaplasticity” in NAc MSNs, and we reveal the specific effects of this plasticity on reward behavior in a brain circuit-specific manner.

Introduction

Dendritic spines are the sites of most excitatory synapses on neurons and are classified as immature or mature based largely on spine head size. Spine structure and function are highly associated as spines with small heads typically have fewer functional AMPA receptors than spines with large heads (Tada and Sheng, 2006). Drugs of abuse induce alterations in spine density and morphology in medium spiny neurons (MSNs) of the nucleus accumbens (NAc), a key reward region in ventral striatum (Dietz et al., 2012; Shen et al., 2009). Dendritic spines of MSNs, the main output cell of NAc, undergo remodeling after repeated cocaine administration. At early times after cocaine exposure (e.g., up to 24 hr), NAc MSNs exhibit an increased density of immature spines with small heads, which is paralleled by a reduction in synaptic strength. In contrast, after longer times post-cocaine (e.g., several weeks to months), NAc MSNs bear an increased density of mature spines with large heads and concomitant synaptic strengthening (Dong and Nestler, 2014; Russo et al., 2010; Shen et al., 2009). Increasing evidence supports altered synaptic strength/connectivity of NAc MSNs in contributing to cocaine-mediated behavioral responses at different times after cocaine exposure (Dietz et al., 2012; Ma et al., 2014; Pascoli et al., 2012) and during cocaine relapse (Gipson et al., 2013).

Cocaine administration is associated with alterations in gene transcription mediated by several transcription factors and accompanying epigenetic mechanisms (Robison and Nestler, 2011). Although studies have found broad associations between transcription and dendritic spine alterations after cocaine exposure, in most instances the synaptic molecules regulated at the transcriptional level by cocaine remain to be identified. Furthermore, it is not clear if the divergent forms of plasticity occurring at short vs. longer times after cocaine administration are mediated by distinct signaling pathways or by the bidirectional regulation of a single pathway.

Here we identify a mechanism by which nuclear–dendritic spine interactions in NAc MSNs regulate specific synaptic and behavioral phenotypes after cocaine administration. The results of this study implicate the actions of a signaling hub that regulates local synaptic protein synthesis in triggering biphasic dendritic spine structural plasticity in NAc at short vs. longer times after cocaine exposure. The behavioral consequences of the altered activity of this signaling hub in NAc have important implications for understanding the underpinnings of reward regulation.

Results

Identification of PDZ-RhoGEF in NAc as a target for cocaine

Chronic cocaine exposure was previously shown to increase levels of the tandem microRNAs, miR132 and miR212, in dorsal striatum thereby limiting the reward induced by excessive cocaine consumption (Hollander et al., 2010). While the effect of cocaine on miR132/212 levels in NAc remains unknown, the identification of protein products whose expression is controlled by miR132/212 can be used to identify novel candidate cocaine-regulatory molecules. First, we determined if the loss of miR132/212 in NAc increases cocaine reward as previously shown for dorsal striatum. We injected adult homozygous loxP-flanked (floxed) miR132/212 mice (Magill et al., 2010) intra-NAc with an adeno-associated virus (AAV) expressing Cre and green fluorescent protein (AAV-Cre-GFP). Consistent with effects in dorsal striatum, knockdown of miR132/212 in NAc increases the rewarding effects of cocaine as measured by conditioned place preference (CPP) (Figures S1A and S1B).

The Arhgef11 gene, which encodes PDZ-RhoGEF, is a top predicted target of both miR132 and miR212 (MIT TargetScan, Weizmann Institute PITA). PDZ-RhoGEF is a guanine nucleotide exchange factor that selectively activates the RhoA small GTPase (Oleksy et al., 2006). AAV-Cre-mediated knockdown of miR132/212 in NAc induced PDZ-RhoGEF levels >3-fold (Figures S1C–S1E). These findings implicated PDZ-RhoGEF acting in NAc in regulating cocaine reward.

To determine if cocaine affects PDZ-RhoGEF levels, we injected mice intraperitoneally (IP) with saline or cocaine for 7 days; 24 hr after the last injection, NAc tissue was subjected to subcellular fractionation which yields P1 (crude nuclear), S2 (cytoplasm), and P2 (crude synaptoneurosomal) fractions. Actin was used as a loading control for all fractions as its levels are highly consistent between treatment conditions (Figures S2A–S2C). 24 hr after the last cocaine dose (“24-hr post-cocaine”), PDZ-RhoGEF levels were elevated in the P1 fraction (Figures 1A and 1D), and this effect was paralleled by a loss in the S2 fraction (Figures 1B and 1E). No change in the P2 fraction was detected (Figures 1C and 1F), and cocaine effects on PDZ-RhoGEF levels were not detected in whole-cell lysates (Figures S2D and S2E). Increased P1 levels of PDZ-RhoGEF were further validated in individual NAc neuronal nuclei by immunohistochemistry (Figures S2F and S2G). As PDZ-RhoGEF contains a predicted nuclear import sequence, we further examined the relationship between P1 and S2 PDZ-RhoGEF levels in individual mice. Linear-regression showed that in cocaine-treated mice, but not in saline-treated mice, P1 PDZ-RhoGEF levels are inversely related to S2 levels (Figures 1G and 1H).

Figure 1. Cocaine Alters RhoA Signaling in NAc.

(A–C) Western blots showing PDZ-RhoGEF levels in NAc P1 (crude nuclear) (A), S2 (cytoplasm) (B), and P2 (crude synaptoneurosomal) (C) fractions from saline- or cocaine- (20mg/kg IP × 7 days) treated mice studied 24 hr after last injection.

(D–F) PDZ-RhoGEF NAc protein levels were increased in P1 (D) [t₂₉=2.861, **p<0.01, n=15(s), 16(c)] and decreased in S2 (E) [t₂₉=2.558, *p<0.01, n=15(s), 16(c)] by cocaine, with no effect in P2 (F) [t₂₉=0.3074, *p<0.01, n=15(s), 16(c).

(G and H) Linear regression of NAc PDZ-RhoGEF protein levels in P1 vs. S2 for individual saline (G) and cocaine-treated (H) mice. Statistical correlation values are shown above each graph, n=15(s), 16(c).

(I–L) Blot shows increased RhoA NAc levels in P1 by cocaine (I,K) [t₅₂=2.268, *p<0.05, n=27(s), 27(c)], with no difference in S2 (J, L) [t₃₀=0.305, p>0.05), n=16(s), 16(c)].

(M and N) Blots show increased active RhoA (RhoA-GTP) NAc levels in P1 by cocaine [t₂₀=2.197, *p<0.05, n=11(s), 11(c)].

(O and P) Blots show levels of MAL in G-actin and F-actin subfractions of P1 in NAc of saline- and cocaine-treated mice, with elevated MAL F-actin to MAL G-actin ratios after cocaine [t₁₉=2.342, *p<0.05, n=10(s), 11(c)].

(Q and R) MAL was immunoprecipitated (IP) from NAc P1 fractions of saline- and cocaine-treated mice. Blots show increased SRF protein in MAL NAc P1 immunoprecipitates after cocaine; SRF was normalized to the amount of MAL immunoprecipitated [t₄=2.896, *p<0.05, n=3(s), 3(c)].

All summary data are the mean +/− SEM.

PDZ-RhoGEF activates RhoA, which is inactive when bound to GDP and active when bound to GTP; the binding of PDZ-RhoGEF to RhoA favors GDP to GTP exchange on RhoA leading to increased activity (Oleksy et al., 2006). At 24-hr post-cocaine, we found increased total RhoA levels in NAc P1, but not S2, fractions (Figures 1I–1L). To determine if these increased P1 RhoA levels are associated with altered RhoA activity, we immunoprecipitated RhoA-GTP and observed increased active RhoA levels in NAc P1 fractions 24-hr post-cocaine (Figures 1M and 1N).

Regulation of nuclear actin in NAc by cocaine

Studies in non-neuronal cells have uncovered a role for RhoA in regulating actions of the serum response factor (SRF) transcription factor. RhoA increases the transcription of SRF-target genes (Gineitis and Treisman, 2001), an action mediated by RhoA controlling the spatial localization of the SRF co-activator MAL (also known as MRTF-A or MKL1) (Cen et al., 2003). Actin exists within cells in two primary states: as a monomer (G-actin) or in a complex assembly of monomers that result in higher order filaments (F-actin) (Penzes and Cahill, 2012). MAL binds directly to G-actin, which blocks its ability to interact with SRF. RhoA polymerizes G-actin into F-actin and, by depleting the G-actin pool, RhoA relieves MAL of its inhibitory G-actin interaction thereby fostering MAL-SRF binding (Miralles et al., 2003). MAL is also capable of binding to F-actin (Posern et al., 2004), and actin mutants that favor polymerization increase SRF activity (Posern et al., 2002).

To determine if MAL interacts with actin in NAc P1 fractions and, if so, if cocaine affects this interaction, we isolated NAc P1 fractions from mice 24-hr post-cocaine and immunoprecipitated MAL. Cocaine increased the amount of actin recovered in MAL immunoprecipitates (Figures S2H and S2I), but as immunoprecipitates are subjected to denaturing/reducing conditions that depolymerize actin filaments prior to resolution via SDS-PAGE, the individual contributions of G-actin and F-actin cannot be determined in these experiments. To this end, we subfractionated NAc P1 fractions to separate F-actin and G-actin. Cocaine increased the ratio of MAL associated with F-actin vs. G-actin in P1 fractions (Figures 1O and 1P), with no effect on total MAL levels (Figures S2J and S2K). We also found increased SRF recovered in MAL immunoprecipitates from NAc P1 fractions 24-hr post-cocaine (Figures 1Q and 1R), consistent with a shift from G- to F-actin states. Actin polymerization drives gene transcription, possibly via interactions with RNA polymerase (Ye et al., 2008). As MAL facilitates RNA polymerase recruitment (Esnault et al., 2014), the increased association of MAL with F-actin 24-hr post-cocaine could enable this process.

Identification of Rap1b in NAc as a target of SRF and cocaine

A sequencing study of cultured cells recently identified gene promotors that bind MAL and SRF in an inducible manner and whose activity is bidirectionally controlled by fostering vs. inhibiting G-actin:MAL interactions (Esnault et al., 2014). Cytoskeleton regulatory genes were the most highly represented cluster that inducibly bind MAL/SRF in a manner controlled by MAL’s interaction with G-actin. We therefore determined how cocaine affects the expression levels of these cytoskeletal genes. To this end, we used RNA sequencing (RNA-seq) of NAc from saline- or cocaine-treated mice 24-hr post-cocaine (see Table S1 for full gene list; GEO: GSE76868), and overlaid the data with the list of previously identified MAL/SRF cytoskeletal target genes. This revealed 14 SRF/MAL target genes that are also induced in NAc 24-hr post-cocaine (Table S2). 3 of the identified genes, Mapk1, Grb2, and Rap1b, are key components of cytoskeletal regulatory pathways, of which Rap1b is the most robustly regulated by cocaine.

Rap1b encodes a small GTPase, which is highly enriched in dendritic spines (Xie et al., 2005). Rap1b belongs to the Rap subfamily of the Ras small GTPase superfamily and shares 95% sequence homology with Rap1a (Penzes and Cahill, 2012). Quantitative PCR (qPCR) in a separate cohort of mice validated increased Rap1b mRNA expression in NAc 24-hr post-cocaine, with no effect on other small GTPases studied, including Rap1a (Figure 2A). That Rap1b is induced in NAc by cocaine is supported by a ~2-fold increase in H3K4me3 levels, an epigenetic mark of transcriptionally active genes, at the Rap1b promoter under these conditions (Feng et al., 2014) (Figures S3A and S3B).

Figure 2. Cocaine-Mediated Rap1b Expression Controls Cocaine Reward.

(A) qPCR in mice treated with saline or cocaine (20mg/kg IP × 7 days) revealed that Rap1b transcript levels (t₁₅=3.724, **p<0.01) but not those of other small GTPases were elevated 24-hr post-cocaine (p>0.05). n=9(s), 8(c).

(B–D) Blots show Rap1b levels in NAc P2 (B), P1 (C), or S2 (D) fractions in cocaine-treated mice.

(E) Cocaine treatment increases Rap1b protein levels in NAc P2 fractions [[t₄₂=2.074, *p<0.01, n=21(s), 23(c)].

(F) Cocaine does not alter Rap1b protein levels in NAc P1 fractions [[t₃₀=0.1231, *p<0.01, n=16(s), 16(c)].

(G) No alterations in Rap1b protein levels occurred in response to cocaine in NAc S2 fractions [[t₃₀=0.4571, *p<0.01, n=16(s), 16(c)].

(H) HSV-GFP or HSV-Cre-GFP was infused into NAc of floxed SRF mice and mice were given 5 cocaine injections over a 3-day period. One NAc hemisphere was infused with HSV-GFP and the other with HSV-Cre-GFP.

(I and J) Blots show the effects of SRF knockdown in NAc on P2 Rap1b levels in cocaine-treated mice. SRF levels in the P1 fraction show SRF knockdown (See figure S3E for P1 quantification). HSV-Cre-GFP decreased Rap1b levels in NAc P2 fractions (paired t-test, t₇=2.696, *p<0.05, n=8 pairs).

(K) GGTI-298 or vehicle was infused into NAc. 24 hr post-infusion, saline/cocaine pairing began (7.5mg/kg cocaine); cocaine CPP behavior was assessed 3 days post-infusion.

(L) Cocaine CPP shows that GGTI-298 NAc infusion reduced cocaine preference [t₃₁=2.157, *p<0.05, n=17(v), 16(g)].

(M) GGTI-298 or vehicle was infused into the NAc. 24 hr post-infusion mice were tested for 8 consecutive days in a cocaine (7.5mg/kg) locomotor sensitization task.

(N) Reduced locomotor responses to cocaine occurred in GGTI-298 infused mice. There was a significant main effect for drug treatment across days 1–8 [two-way repeated measures ANOVA, F_1,14=5.800, *p<0.05, n=8(v), 8(g)].

(O) Images show the targeting of HSV-GFP and HSV-Rap1b-GFP to NAc. The anterior commissure landmark is labeled (a.c.). Scale bar=400µm

(P) HSV-GFP or HSV-Rap1b-GFP was infused into NAc. 3-days post-infusion saline/cocaine pairing began (5mg/kg cocaine).

(Q) Rap1b overexpression in NAc increases cocaine preference [t₂₁=2.142, *p<0.05, n=12(g), 11(r)].

(R) Rap1b or vehicle was infused into NAc. 3-days post-infusion mice were tested over 7 days for cocaine (5mg/kg)-induced locomotion.

(S) Increased locomotor responses to cocaine were seen in mice overexpressing Rap1b in NAc across all testing days [two-way repeated measures ANOVA, main effect of Rap1b expression, F_1,16=6.199, *p<0.05, n=9(g), 9(r)].

All summary data are the mean +/− SEM.

Rap1b is enriched in dendritic spines (Xie et al., 2005). To determine if induction of Rap1b mRNA is associated with corresponding induction of protein, we fractionated NAc derived from individual saline-or cocaine-treated mice and assessed Rap1b levels in multiple fractions. 24-hr post-cocaine, Rap1b levels were selectively increased in P2 fractions only (Figures 2B–2G), an effect lost 1–2 weeks after the last cocaine dose (Figures S3C and S3D).

To determine if cocaine regulates Rap1b P2 levels in an SRF-dependent manner, we infused HSV-GFP or HSV-Cre-GFP into NAc of floxed SRF mice (Ramanan et al., 2005) treated repeatedly with saline or cocaine (Figure 2H). To avoid between-mouse variability, we used a within-subjects design in which HSV-GFP and HSV-Cre-GFP were infused into opposite NAc hemispheres of individual mice. SRF knockdown in NAc decreased P2 levels of Rap1b in NAc of cocaine-treated animals (Figures 2I and 2J and Figure S3E). This effect did not occur in saline-treated mice (Figures S3F and S3G), consistent with cocaine’s ability to induce SRF activity in NAc (Vialou et al., 2012).

To determine directly if RhoA signaling controls Rap1b P2 levels as we would predict, we generated an HSV vector that expresses the tandem DH/PH GEF domains of PDZ-RhoGEF (HSV-PRG GEF-GFP). This construct has been shown to augment RhoA activity in cultured cells (Medina et al., 2013). We likewise found that PRG GEF overexpression in NAc (Figures S3H and S3I) increased levels of active RhoA in P1 fractions (Figures S3J and S3K). Overexpression of PRG GEF also increased Rap1b levels in P2 fractions of this brain region (Figures S3L and S3M).

Regulation of behavioral responses to cocaine by Rap1b in NAc

We used CPP to assess the effects of Rap1b signaling in NAc on cocaine-elicited behavioral reward. GGTI-298 is a GGTase I inhibitor that blocks the normal targeting of Rap1 to membrane regions (Le Borgne et al., 2014). We found that when GGTI-298 was infused into NAc, mice displayed reduced CPP for cocaine (Figures 2K and 2L). In separate mice we found that GGTI-298 NAc infusion also reduced cocaine-induced locomotor activity (Figures 2M and N).

A caveat of these findings is that GGTI-298 can affect other small GTPases (Tafazoli et al., 2003). Therefore, to more specifically study the effects of altered Rap1b on cocaine-mediated behavior, we generated an HSV that overexpresses Rap1b and GFP (HSV-Rap1b-GFP). HSV-Rap1b-GFP was infused into NAc and cocaine CPP assessed using a threshold dose of cocaine (5 mg/kg), which does not produce a significant preference in control mice. Rap1b overexpression in NAc increased cocaine CPP under these conditions compared to HSV-GFP control mice (Figures 2O–2Q). Rap1b overexpression also augmented cocaine-driven locomotor responses (Figures 2R and 2S). To determine if the loss of Rap1b produces the opposite effect, we infused HSV-Cre-GFP or HSV-GFP into NAc of floxed Rap1 mice (Pan et al., 2008) and assessed CPP behavior. Knockdown of Rap1 from this region rendered mice less responsive to the rewarding effects of cocaine (Figures S3N–P).

Akt-mTOR are downstream targets of Rap1b in NAc

A recent study found that Rap1 binds directly to, and activates, phosphoinositide-3-kinase (PI3K) in non-neuronal cells (Kortholt et al., 2010), and we found that endogenous Rap1b co-immunoprecipitates with PI3K in NAc P2 fractions (Figure S4A). PI3K converts PIP2 to PIP3, and PIP3 recruits PDK1 and the protein kinase Akt to the plasma membrane where PDK1 phosphorylates Akt at T308, thereby activating Akt. Active Akt, in turn, leads to the activation of mammalian target of rapamycin (mTOR) (Dibble and Cantley, 2015). mTOR exists in two independent protein complexes, mTORC1 and mTORC2, which are distinguished by mTOR’s interaction with Raptor or Rictor, respectively. Unlike mTORC1, mTORC2 phosphorylates Akt at S473. Phosphorylation of Akt at T308, but not S473, is essential for Akt activation (Dibble and Cantley, 2015). mTORC1 has an established role in facilitating the translation of mRNAs present in dendritic spines via its actions on cap-dependent translation regulators (Liu-Yesucevitz et al., 2011).

To determine if Rap1b activates the Akt-mTOR pathway, we overexpressed Rap1b in NAc and isolated P2 fractions. Rap1b overexpression increased levels of p-Akt T308 with no change detected at S473. We also examined levels of mTOR phosphorylated at S2448 as phosphorylation at this site correlates with mTORC1 activity (Rosner et al., 2010). Rap1b overexpression increased p-mTOR S2448 levels and levels of total mTOR (Figures 3A–3C). Collectively these findings suggest that Rap1b preferentially activates the mTORC1 complex. By contrast, we found that knockdown of Rap1b in NAc reduced both Akt T308 and mTOR S2448 phosphorylation (Figures S4B–S4D). We confirmed the stability of p-Akt T308 and p-mTOR S2448 levels in P2 fractions during the fractionation procedure (Figures S4E and S4F).

Figure 3. Cocaine-Mediated Regulation of Akt/mTOR Activity Controls Cocaine Reward.

(A) HSV-GFP or HSV-Rap1b-GFP was infused into NAc of mice. For each mouse one NAc hemisphere was infused with HSV-GFP and the other with HSV-Rap1b-GFP. 5-days post infusion, NAc was dissected and P2 fractions isolated.

(B and C) Blot shows the effects of Rap1b overexpression on NAc P2 protein levels. Pairwise analysis indicates that Rap1b NAc overexpression increased levels of p-Akt T308 (t₆=2.755, *p<0.05), p-mTOR S2448 (t₆=2.744, *p<0.05), and total mTOR (t₆=3.645, *p<0.05); the virus elevated Rap1b levels as well (t₆=2.698, *p<0.05). Levels of p-Akt S473 were not altered (t₆=0.5416, p>0.05). n=7 pairs.

(D) HSV-GFP or HSV-Cre-GFP was infused into NAc of floxed SRF mice. 2-days post-surgery, mice were given 5 cocaine injections over a 3-day period. NAc was dissected and P2 fractions isolated. For each mouse one NAc hemisphere was infused with HSV-GFP and the other with HSV-Cre-GFP.

(E and F) Blots show the effects of SRF knockdown on NAc P2 protein expression profiles in cocaine-treated mice. Data are derived from the same experiment in Figures 2I and 2J. Pairwise analysis indicates that SRF knockdown reduced levels of p-Akt T308 (t₇=2.657, *p<0.05), p-mTOR (t₇=2.008, *p<0.05), and total mTOR (t₇=2.410, *p<0.05), with no effect on p-Akt S473 (t₇=0.4725, p>0.05). n=8 pairs.

(G–I) Blots show Akt-mTOR P2 protein levels from mice injected with saline or cocaine (IP, 20mg/kg cocaine) for 7 days analyzed 24 hr (G), 1 week (H), or 2 weeks (I) after the last injection.

(J) At 24 hr, cocaine increased P2 levels of p-Akt T308 [t₂₈=2.063, *p<0.05, n=15(s), 15(c)] and p-mTOR S2448 [t₂₉=2.149, *p<0.05, n=16(s), 15(c)] with no effects on total Akt [t₂₉=0.4155, p>0.05, n=16(s), 15(c)] or total mTOR [t₂₉=0.9300, p>0.05, n=16(s), 15(c)].

(K and L) Cocaine did not alter P2 levels of p-Akt T308, p-mTOR S2448, total Akt, or total mTOR at 1 (K) or 2 (L) weeks [p>0.05; 1-week n=12(s), 13(c); 2-week n=10(s), 9(c)].

(M) Images show the targeting of the indicated virus to NAc, the anterior commissure (a.c.) landmark is labeled. Scale bar=400um

(N) HSVs were infused into NAc and 3 days post-infusion saline/cocaine pairing began (5mg/kg or 7.5mg/kg cocaine).

(O) Akt, but not DN-Akt, increased cocaine preference to a low dose of cocaine (5mg/kg) [One-way ANOVA, F_2,35=3.656, p<0.05; post hoc GFP vs. Akt, *p<0.05; post hoc GFP vs. DN-Akt, p>0.05, n=15(g), 13(akt), 10(dn-akt)].

(P) DN-Akt reduced cocaine preference to a higher dose of cocaine (7.5mg/kg); Akt overexpression had no effect [One-way ANOVA, F_2,64=4.020, p<0.05, post hoc GFP vs. Akt, p>0.05; post hoc GFP vs. DN-Akt, *p<0.05, n=26(g), 23(akt), 18(dn-akt)].

(Q) Graphs depict locomotor responses during the second day of saline/cocaine pairing with a 5mg/kg cocaine dose. No differences between groups were detected [One-way ANOVA, F_2,36=1.893, p>0.05, post hoc GFP vs. Akt and DN-Akt, p>0.05). One sample actual vs. theoretical mean analyses revealed while both GFP (t=2.198, *p<0.05) and Akt (t=2.206,* p<0.05) overexpression increased locomotor activity in response to cocaine as compared to saline (above a chance level of 0), DN-Akt overexpression locomotor activity did not differ from chance levels (t=0.3832, p>0.05). n=15(g), 14(akt), 10(dn-akt).

(R) Locomotor responses during the second day of saline/cocaine pairing with a 7.5mg/kg cocaine dose. Akt, but not DN-Akt, increased cocaine activity relative to control mice. [One-way ANOVA, F_2,63=3.275, p<0.05; post hoc GFP vs. Akt, *p<0.05; post hoc GFP vs. DN-Akt, p>0.05, n=25(g), 23 (akt), 18(dn-akt)].

All summary data are the mean +/− SEM.

As the loss of SRF reduced P2 Rap1b levels in NAc from mice shortly after cocaine treatment as presented earlier, we determined if this is associated with concomitant alterations in Akt-mTOR signaling. Using a within-subjects design, we found that SRF knockdown from NAc reduced levels of p-Akt T308 and p-mTOR S2448 in NAc P2 fractions of cocaine-treated mice, without affecting levels of p-Akt S473. Local SRF knockdown also reduced total mTOR expression in this NAc fraction (Figures 3D–3F), suggesting that SRF activity has multiple downstream consequences on mTOR signaling. SRF knockdown did not affect levels of p-Akt T308 or p-mTOR S2448 in P2 fractions derived from chronic saline-treated mice (Figures S4G and S4H), again consistent with cocaine’s induction of SRF activity in NAc (Vialou et al., 2012).

We next assessed if chronic cocaine alters phosphorylation levels of Akt and mTOR. We found augmented p-Akt T308 and p-mTOR S2448 levels in NAc of cocaine- vs. saline-treated animals 24-hr post-cocaine, an effect not apparent 1 or 2 weeks later (Figures 3G–L). In contrast, levels of p-Akt S473 were not affected at the 24 hr time point. Interestingly, p-Akt S473 was increased at 1-week post-cocaine, with a return to saline levels by 2 weeks (Figures S4I and S4J).

Regulation of behavioral responses to cocaine by Akt-mTOR in NAc

To determine the effects of Akt activity in NAc on behavioral responses to cocaine, we injected HSVs into NAc that encode dominant-negative Akt (DN-Akt-GFP) or wildtype Akt (WT-Akt-GFP), and assessed cocaine CPP (Figures 3M and 3N). Using a low dose of cocaine (5 mg/kg), WT-Akt, but not DN-Akt, overexpression in NAc increased CPP relative to control mice (Figure 3O). In separate mice tested with a higher cocaine dose (7.5 mg/kg), overexpression of DN-Akt in NAc reduced cocaine CPP scores; WT-Akt had no effect at this higher cocaine dose likely due to a ceiling effect (Figure 3P). We also analyzed locomotor responses during the saline/cocaine pairings on day 2 of CPP training and found that, at the 5 mg/kg dose, DN-Akt mice failed to display a cocaine-induced increase in locomotor activity over saline levels (Figure 3Q). At the 7.5mg/kg dose, overexpression of WT-Akt, but not DN-Akt, increased cocaine-induced locomotor activity (Figure 3R).

We next complemented our viral approach with a pharmacological one. We infused the mTORC1 inhibitor rapamycin into NAc and assessed single-session locomotor responses to 3 different doses of cocaine (naïve mice for each dose). Intra-NAc rapamycin reduced locomotor activity across cocaine doses, an effect most prominent at the lowest dose tested (Figures S5A and S5B). PTEN is a negative regulator of Akt activity and Rock1 is a protein kinase that activates PTEN. Rock1 inhibition increases Akt activity (Vemula et al., 2010), and we found that NAc infusion of Y27632, a Rock1 inhibitor, increased levels of p-Akt T308 in P2 fractions with no effect on p-Akt S473 (Figures S5C–S5E). Consistent with the effects of heightened Akt activity on cocaine-mediated behavior, Y27632 increased cocaine CPP when infused into NAc just prior to cocaine administration (Figures S5F and S5G), with increased locomotor activity observed as well (Figures S5H and S5I). Together, these pharmacological studies indicate that acutely manipulating NAc Akt-mTOR activity just prior to cocaine administration is sufficient to control cocaine-mediated behavioral responses.

Cocaine regulation of NAc dendritic spine morphology: Role of Rap1b and Akt

Alterations in NAc MSN dendritic spine morphology are believed to influence behavioral responses to cocaine (Introduction). While Rap1 alters the morphology of dendritic spines in cultured pyramidal neurons (Xie et al., 2005), its effect on dendritic spines in MSNs remains unknown. Increased density of immature, thin spines in NAc MSNs is seen 24-hr post-cocaine (Introduction). As we detected increased Rap1b levels in NAc P2 fractions at this time point, we determined if Rap1b overexpression in NAc MSNs in vivo phenocopies the spine effects of cocaine and, if so, determined if Akt activity is important for this effect. Three distinct spine subtypes were quantified: thin spines (have a neck and small spine head), stubby spines (no discernable neck), and mushroom spines (discernable neck and large spine head) (Figure 4A). Thin spines are typically considered immature, while stubby and mushroom spines are typically considered mature (Nimchinsky et al., 2002). We overexpessed HSV-Rap1b-GFP or HSV-DN-Akt-GFP into NAc either alone, or together, for 3-days (Figures 4B and 4C). Rap1b overexpression increased the total density of dendritic spines in NAc MSNs, an effect blocked by concurrent overexpression of DN-Akt; 3-day DN-Akt overexpression by itself did not affect total spine density (Figure 4D). Of the spine subytpes, Rap1b induced a selective increase in the density of thin spines in NAc, an effect blocked by concomitant overexpression of DN-Akt (Figures 4E–G).

Figure 4. Rap1b Regulates NAc MSN Dendritic Spine Morphogenesis Through Akt.

(A) Low magnification image of a NAc MSN with an adjacent high magnification image of a dendritic segment illustrating different spine subtypes. This neuron overexpresses HSV-GFP shown in black and white. Scale bar for left image=25um; for right image=5um.

(B) Experimental design. HSV-GFP, HSV-Rap1b-GFP, HSV-DN-Akt-GFP, or HSV Rap1b+HSV-DN-Akt-GFP were infused into NAc. 3-days post-infusion, brains were perfused.

(C) High magnification 3D reconstruction images of NAc MSN dendrites bearing dendritic spines for all viral conditions. Differences in brightness across a dendrite reflect different Z-plane depths. Images are cropped from larger dendrite segments. Scale bar=5um.

(D) Only neurons overexpressing Rap1b show altered total spine density on NAc MSNs relative to those overexpressing GFP only (One-way ANOVA, F_3,75=3.693, p<0.05; post hoc GFP vs. Rap1b, *p<0.05; post hoc GFP vs. DN-Akt and DN-Akt+Rap1b, p>0.05). n=21 cells from 4 brains (GFP), 22 cells from 3 brains (Rap1b), 12 cells from 3 brains (DN-Akt), and 24 cells from 3 brains (DN-Akt+Rap1b).

(E) Only Rap1b overexpression increased the density of thin spines on NAc MSN dendrites (One-way ANOVA, F_3,75=5.920, p<0.01; post hoc GFP vs. Rap1b, **p<0.01; post hoc GFP vs. DN-Akt and DN-Akt+Rap1b, p>0.05).

(F and G) No conditions altered stubby (F) or mushroom (G) spine density in NAc MSNs (Stubby one-way ANOVA, F_3,75=0.465, p>0.05; post hoc GFP vs. all conditions, p>0.05); (Mushroom one-way ANOVA, F_3,75=1.350, p>0.05; post hoc GFP vs. all conditions, p>0.05).

(H) Across all spine types only Rap1b overexpression alters mean spine head diameter relative to cells overexpressing GFP (One-way ANOVA, F_3,3316=10.92, p<0.0001; post hoc GFP vs. Rap1b, ***p<0.001; post hoc GFP vs. DN-Akt and DN-Akt+Rap1b, p>0.05). n=746 spines (GFP), 1211 spines (Rap1b), 408 spines (DN-Akt), 955 spines (Rap1b+DN-Akt).

(I) Cumulative plot of thin spine head diameter reveals no significant differences between Rap1b overexpression and control neurons (spine log-rank test, x²=1.596, p>0.05).

(J and K) Rap1b overexpression in NAc MSNs results in a leftward shift of the cumulative stubby spine (J) and mushroom spine (K) head diameter curve relative to GFP overexpression (stubby log-rank test, x²=7.164, **p<0.01; mushroom log-rank test, x²=4.362, *p<0.05).

All summary data are the mean +/− SEM.

To further assess spine morphology, we analyzed the head diameter of individual spines. First, we found that Rap1b overexpression alone, but not in combination with DN-Akt, reduced mean spine head diameter across all spine classes (Figure 4H). To determine if this reflects changes within particular spine types, we analyzed cumulative plots of spine head diameter separately for thin, stubby, and mushroom spines, and found that Rap1b overexpression produced a leftward shift in the cumulative curve of both mushroom and stubby spines, indicative of a reduction in head diameter for these spine types. No significant change in the cumulative curve of thin spines was detected (Figures 4I–4K). Collectively, these spine data indicate that the reduction in mean spine head diameter upon Rap1b overexpression in NAc MSNs stems from both an increase in thin spine density and a reduction in the head diameter of more mature spines.

Next, we examined the consequences of Rap1b knockdown on cocaine-mediated spine morphogenesis. Starting two days post-viral infusion, floxed Rap1b mice expressing either HSV-GFP or HSV-Cre-GFP in NAc were administered five doses of cocaine over a 3-day period to allow for multiple cocaine doses within the transient lifespan of HSVs, and spine morphology was assessed 4 hr later (Figures 5A and 5B). In GFP-expressing MSNs, cocaine increased total spine density (Figure 5C) and thin spine density (Figure 5D), with no effect on other spine subtypes (Figures 5E and 5F). Cocaine also decreased the mean head diameter across all spine types in GFP-expressing MSNs (Figure 5G), and caused a leftward shift in the thin spine head diameter curve (Figure 5H). Cre-expressing MSNs were unresponsive to the induction of total (Figure 5C) or thin (Figure 5D) spines following cocaine treatment. Notably, the loss of Rap1b increased mushroom spine density in both saline- and cocaine-treated mice as compared to GFP-expressing MSNs (Figure 5F). While Cre-expressing MSNs had a larger total spine head diameter across all spine types than those expressing GFP, similar to GFP-expressing MSNs, cocaine-treatment reduced mean spine head diameter across all spine types in Cre-expressing MSNs (Figure 5G). In Cre-saline mice, there was a rightward shift in the mushroom spine head diameter curve, with no effect on other spine subtypes (Figures 5H–J).

Figure 5. Rap1b and Akt are Necessary for Cocaine-Induced Spine Remodeling in NAc MSNs.

(A) HSV-GFP or HSV-Cre-GFP was infused into NAc of floxed Rap1 mice. 2-days post-surgery, mice were administered 5 saline or cocaine injections (IP, 20mg/kg cocaine) over a 3 day period and brains perfused 4 hr after the last injection.

(B) High magnification 3D reconstruction images of NAc MSN dendrites bearing dendritic spines for all four experimental conditions. Differences in brightness across a dendrite reflect different Z-plane depths. Images are cropped from larger dendrite segments. Scale bar=5um.

(C) Cocaine increased total spine density in GFP-expressing, but not Cre-expressing neurons (2×2 ANOVA main effect of drug treatment F_1,62=7.064, p=0.01; post hoc GFP saline vs. GFP cocaine, *p<0.05; Cre saline vs. Cre cocaine, p>0.05). n=19 cells from 3 brains (GFP saline), 16 cells from 3 brains (GFP cocaine), 12 cells from 3 brains (Cre saline), and 19 cells from 3 brains (Cre cocaine).

(D) Cocaine increased thin spine density in GFP-expressing, but not Cre-expressing, NAc MSNs (2×2 ANOVA main effect of drug treatment, F_1,62=11.30, p=0.01; post hoc GFP saline vs. GFP cocaine, ***p<0.001; Cre saline vs. Cre cocaine, p>0.05).

(E and F) Cocaine did not alter the density of stubby (E) or mushroom (F) spines in either GFP- or Cre-expressing NAc MSNs (p>0.05). There was a main effect of Rap1 level on mushroom spine density (2×2 ANOVA, F_1,62=17.56, ***p<0.0001).

(G) Analysis of total spine head diameter across all spine types revealed main effects of drug treatment (2×2 ANOVA, F_1,2569=37.75, ***p<0.0001) and Rap1b expression (2×2 ANOVA, F_1,2569=82.54, ***p<0.0001). In both GFP- and Cre-expressing NAc MSNs, cocaine reduced mean spine head diameters (post hoc, ***p<0.001 for both GFP and Cre saline vs. cocaine). n=670 spines (GFP saline), 642 spines (GFP cocaine), 490 spines (Cre saline), and 771 spines (Cre cocaine).

(H) Cumulative plot of thin spine head diameter indicates that only the GFP cocaine group differed from the GFP saline group, with a significant leftward shift in the curve (spine log-rank test, x²=8.305, *p<0.05).

(I) No difference in the cumulative frequency plot of stubby spine head diameter was detected between any groups and the GFP saline condition (p>0.05).

(J) Cumulative plot of mushroom spine head diameter indicates a significant rightward shift in the curve in the Cre saline group compared to the GFP saline group (spine log-rank test, x²=20.91, ***p<0.0001). No other groups differed from GFP saline (p>0.05).

(K) GFP or DN-Akt were infused into the NAc. 2 days post-surgery, mice were administered 5 saline or cocaine injections (IP, 20mg/kg cocaine) over a 3 day period and brains perfused 4 hr after the last injection.

(L) High magnification 3D reconstruction images of NAc MSN dendrites bearing dendritic spines for all experimental conditions. Differences in brightness across a dendrite reflect different Z-plane depths. Images are cropped from larger dendrite segments. Scale bar=5um.

(M) Cocaine increased total spine density in GFP-expressing, but not DN-Akt-expressing, NAc MSNs (2×2 ANOVA interaction effect, F_1,88=5.264, p<0.05; post hoc GFP saline vs. GFP cocaine, *p<0.05; DN-Akt saline vs. DN-Akt cocaine, p>0.05).

(N) Cocaine increased thin spine density in GFP-expressing, but not DN-Akt-expressing, NAc MSNs (2×2 ANOVA interaction effect, F_1,88=10.28, p<0.01; post hoc GFP saline vs. GFP cocaine, ***p<0.001; DN-Akt saline vs. DN-Akt cocaine, p>0.05). n=19 cells from 3 brains (GFP saline), 22 cells from 5 brains (GFP cocaine), 25 cells from 4 brains (DN-Akt saline), and 26 cells from 4 brains (DN-Akt cocaine).

(O and P) Cocaine did not alter stubby (O) or mushroom (P) spine density in GFP- or DN-Akt-expressing NAc MSNs (p>0.05). There was a main effect of DN-Akt expression on mushroom spine density (2×2 ANOVA, F_1,88=12.83, ***p<0.001).

(Q) Analysis of spine head diameter across all spine types revealed main effects of drug treatment (2×2 ANOVA, F_1,3148=9.630, **p<0.01) and DN-Akt expression (2×2 ANOVA, F_1,3148=68.15, ***p<0.0001). In GFP-, but not DN-Akt-expressing, NAc MSNs, cocaine reduced mean spine head diameters (post hoc, **p<0.01 for GFP saline vs GFP cocaine; p>0.05 for DN-Akt saline vs. DN-Akt cocaine). n=579 spines (GFP saline), 801 spines (GFP cocaine), 919 spines (DN-Akt saline), and 853 spines (DN-Akt cocaine).

(R) Cumulative frequency plot analysis of thin spine head diameters revealed no significant differences between groups (p>0.05).

(S) Cumulative plot analysis of stubby spine head diameters revealed a rightward shift in the curve in the DN-Akt saline group as compared to the GFP saline group (spine log-rank test, x²=9.978, **p<0.01); no other groups differed from GFP saline.

(T) Cumulative frequency plot of mushroom spine head diameter revealed a rightward shift in the curve in both the DN-Akt saline and DN-Akt cocaine groups compared to the GFP saline group (DN-Akt saline spine log-rank test, x²=16.47, ***p<0.0001; DN-Akt cocaine spine log-rank test, x²=10.96, **p<0.01 ).

All summary data are the mean +/− SEM.

Next, we determined if interfering with Akt signaling disrupts thin spine formation by cocaine in a manner similar to the loss of Rap1b by examining spine morphology in saline- and cocaine-treated mice expressing HSV-GFP or HSV-DN-Akt-GFP in NAc (Figures 5K and 5L). As before, cocaine increased both total and thin spine numbers in GFP-expressing NAc MSNs (Figures 5M and 5N), with no effect on other spine types (Figures 5O and 5P). Different from our results in floxed Rap1b mice, although cocaine reduced total spine head diameter in GFP-expressing MSNs (Figure 5Q), there was no alteration in the thin spine head diameter curve in the GFP-cocaine condition (Figure 5R); this difference is likely attributable to the different genetic backgrounds of the mice tested. DN-Akt increased both mushroom spine density (Figure 5P) and total spine head diameter (Figure 5Q) as compared to GFP-expressing NAc MSNs, regardless of drug treatment. While the increased density of mushroom spines in both the DN-Akt saline and DN-Akt cocaine groups likely contributes to the increase in mean spine head diameter across all spine types in these conditions, the rightward shift in the mushroom spine head diameter curve in both groups and the additional rightward shift in the stubby spine head diameter curve specifically in the DN-Akt saline group (Figures 5S and 5T), also likely contributes to an increase in mean spine head diameter across spine subtypes in these conditions.

Overall, the effects of DN-Akt on NAc MSN spines closely parallel those caused by the loss of Rap1, with both conditions inhibiting cocaine-mediated thin spine induction. Within saline-treated mice, both DN-Akt overexpression and the loss of Rap1b increased the density and head diameter of mushroom spines on NAc MSNs as compared to the GFP-saline condition. Further, in neurons overexpressing DN-Akt and in neurons with reduced Rap1, cocaine decreased mature spine head diameter as compared to their saline-treated counterparts; however, the mature spine subclass in which this effect occurred differed (mushroom spines for floxed Rap1b and stubby spines for DN-Akt).

Effects of longer times after cocaine exposure on Rap1b-Akt-mTOR in NAc

That loss of Rap1b and disruption of Akt activity facilitate spine maturation of NAc MSNs is intriguing given that increased spine maturation and synaptic strengthening are seen at longer time points after the last cocaine dose (“3–4-week post-cocaine”) (Dong and Nestler, 2014; Russo et al., 2010; Shen et al., 2009). Data presented above showed increased Rap1b-p-Akt T308-p-mTOR S2448 signaling in P2 fractions 24-hr post-cocaine with no differences observed at 1 or 2 weeks. To determine if this signaling pathway is altered at still longer times, we studied mice 3 weeks after the last cocaine dose (Figures S6A and S6B). Relative to saline-treated mice, decreased levels of Rap1b and of p-Akt T308, but not S473, were observed in NAc P2 fractions 3-week post-cocaine. Levels of p-mTOR S2448 showed a nonsignificant trend toward a reduction (Figures S6A and S6B). When the 24-hr and 1–3 week data are analyzed in concert, there is a significant interaction between drug treatment and time post-cocaine for Rap1b and p-mTOR S2448, and a strong trend for an interaction for p-Akt T308 (Figures S6C–E), supportive of a downregulation of this pathway at protracted times after the last cocaine dose.

Regulation of PDZ-RhoGEF, Rap1b, and Akt-mTOR signaling in NAc by cocaine self-administration

Drug self-administration paradigms, in which animals volitionally take drug infusions, model aspects of human addiction (Deroche-Gamonet et al., 2004). We thus determined if the biochemical changes identified in mice that received IP cocaine injections are also seen after drug self-administration. Rats self-administered saline or cocaine 10 days and NAc was dissected 24 hr or 4 weeks after the last self-administration session (Figures 6A and 6H). 24-hr post-cocaine PDZ-RhoGEF levels were decreased in the S2 fraction (Figures 6B and 6C), and increased in the P1 fraction (Figures 6D and 6E). Additionally, levels of Rap1b, p-Akt T308, and p-mTOR S2448 in P2 fractions of NAc were elevated in cocaine self-administering rats. Levels of total mTOR were also elevated, similar to what was seen upon Rap1b overexpression in this region (Figures 6F and 6G).

Figure 6. Self-Administered Cocaine Bidirectionally Regulates P1 and P2 Protein Expression Profiles as a Function of Time After Last Cocaine Exposure.

(A) Mice self-administered saline or cocaine during 6-hr sessions for 10 consecutive days. 24 hr after the last session, NAc tissue was dissected and cell fractionated. Graphs show that mice volitionally self-administered cocaine at high rates (one-way ANOVA cocaine vs. saline, ***p<0.0001).

(B and C) Decreased S2 levels of PDZ-RhoGEF 24-hr post-cocaine [(t₁₇=2.686, *p<0.05, n=8(s), 11(c)].

(D and E) Increased P1 PDZ-RhoGEF levels 24-hr post-cocaine [(t₁₇=2.461, *p<0.05, n=8(s), 11(c)].

(F and G) P2 levels of Rap1b (t₁₆=3.359, **p<0.01), p-Akt T308 (t₁₆=4.848, ***p<0.001), p-mTOR (t₁₆=3.807, **p<0.01), and total mTOR (t₁₆=4.135, ***p<0.001) were increased 24-hr post-cocaine. Levels of total Akt were not affected (t₁₆=0.5910, p>0.05). n=7(s), 11(c)

(H) Mice self-administered saline or cocaine as above, but were analyzed 4 weeks after the last session. Graphs show that mice volitionally self-administered cocaine at high rates (one-way ANOVA cocaine vs. saline, ***p<0.0001)

(I and J) Cocaine increased S2 PDZ-RhoGEF levels 4-weeks post-cocaine [(t₁₆=3.400, **p<0.01, n=8(s), 10(c)].

(K and L) Decreased P1 PDZ-RhoGEF levels 4-week post-cocaine (t₂₈=2.068, *p<0.05).

(M and N) P2 levels of Rap1b (t₂₈=2.141, *p<0.05), p-Akt T308 (t₂₈=2.578, *p<0.05), p-mTOR S2448 (t₂₈=2.765, **p=0.01), and total mTOR (t₂₈=3.053, **p<0.01) were decreased 4-week post-cocaine. Levels of total Akt were not affected (t₂₈=1.171, p>0.05). n=14(s), 16(c)

All summary data are the mean +/− SEM.

In contrast, opposite biochemical changes were observed 4-week post-cocaine. Levels of PDZ-RhoGEF were increased in the S2 fraction (Figure 6I and 6J), and decreased in the P1 fraction (Figures 6K and 6L). Levels of Rap1b, p-Akt T308, p-mTOR S2448, and total mTOR in P2 fractions were reduced at this time point (Figures 6M and 6N). Collectively, these self-administration data complement our non-contingent cocaine data by further indicating that the signaling pathways described show biphasic alterations at early vs. longer times after the last cocaine dose.

It is possible that the loss of p-mTOR S2448 in P2 fractions at 3–4-week post-cocaine is a compensatory response to the increase in Rap1b-mTOR signaling at the 24 hr time point. To test this, we overexpressed HSV-Rap1b-GFP in NAc and isolated P2 fractions 4 weeks later (HSV expression dissipates 1-week post infection) (Figure S6F). p-mTOR S2448 levels were not altered by the prior and transient overexpression of Rap1b in NAc (Figures S6G and S6H), indicating that an initial rise in NAc Rap1b expression and hence p-mTOR S2448 does not per se beget a subsequent decline in p-mTOR.

Regulation of local protein formation in P2 fractions of NAc

In forebrain pyramidal neurons, mTOR, when functioning in the mTORC1 complex, stimulates local synaptic formation of proteins that are components of spine postsynaptic densities such as PSD-95 and actin regulatory molecules such as Arc (Dong et al., 2003; Lee et al., 2005; Li et al., 2010). However, it is not clear how mTOR affects local synaptic protein formation in MSNs. To test this, isolated NAc P2 fractions were treated with vehicle or rapamycin using established methods (Lee et al., 2005). Protein levels of PSD-95 and Arc were assessed both just prior to, and 1 hr following, vehicle or rapamycin treatment. We found that rapamycin decreased Arc, but not PSD-95, levels in NAc P2 fractions relative to vehicle (Figures S7A and S7B). Levels of the spine regulatory protein Src were also assessed as an additional control and, as expected, rapamycin did not affect Src levels (Figures S7A and S7B). The efficacy of rapamycin in reducing p-mTOR S2448 levels in NAc MSN P2 fractions was confirmed in this in vitro system (Figures S7C and S7D).

As p-mTOR S2448 levels are elevated 24-hr post-cocaine, we reasoned that Arc expression in P2 fractions should parallel alterations in p-mTOR. As expected, in self-administration animals, Arc NAc P2 levels were increased at this time point (Figures S7E and S7F), with no effect on PSD-95 (Figures S7E and S7F). Moreover, levels of NAc Arc in P2 fractions, but not those of PSD-95, correlated with levels of p-mTOR S2448 in cocaine-treated, but not saline-treated, animals (Figures S7G and S7H).

Neural circuit regulation of PDZ-RhoGEF, Rap1b, and mTOR signaling in NAc

The NAc receives glutamatergic inputs from prefrontal cortex (PFC), ventral hippocampus (vHIPP), and basolateral amygdala (BLA), among other regions (Phillipson and Griffiths, 1985). To determine if the activation of glutamatergic input to NAc produces biochemical changes reminiscent of 24-hr post-cocaine conditions, we infused non-retrograding AAV5-CaMKIIa–ChR2–EYFP into PFC (targeting infralimbic PFC), vHIPP, or BLA of mice. NAc shell targeting optical fibers were implanted and ChR2-expressing axon terminals in NAc shell were burst activated using stimulation parameters previously shown to generate behavioral reward (Britt et al., 2012). Mice underwent a single 20-min stimulation session each day for 7 days; mock-stimulation was used as a control. 24 hr after the final stimulation session, NAc was dissected and fractionated (Figure 7A). Stimulation of infralimbic PFC terminals in NAc increased PDZ-RhoGEF levels in P1 fractions (Figures 7B–D) and levels of Rap1b and p-mTOR S2448 in P2 fractions (Figures 7E and 7F). Stimulation of vHIPP or BLA terminals in NAc did not affect these endpoints (Figures 7G–P).

Figure 7. Pathway-specific Inputs to the NAc Differentially Regulate Nuclear and Synaptic Signaling.

(A) AAV5-CaMKIIa–ChR2–EYFP was infused into ventral hippocampus (vHIPP), basolateral amygdala (BLA), or ventromedial prefrontal cortex (vmPFC; infralimbic targeting). NAc shell targeting optical fibers were implanted and mice underwent 7 days of 20-min burst stimulation sessions in which AAV5-expressing axon terminals in NAc shell were stimulated; mock-stimulation was used as a control. 24 hr after the final stimulation day, NAc shell tissue was dissected and fractionated.

(B) Image shows the placement of AAV5 to the vmPFC (infralimbic targeting).

(C and D) Increased P1 PDZ-RhoGEF levels after vmPFC terminal stimulation [(t₁₇=2.246, *p<0.05, n=9(mock), 10(stim)].

(E and F) Increased P2 levels of Rap1b (t₃₇=2.048, *p<0.05) and p-mTOR S2448 (t₃₇=2.298, *p<0.05) after vmPFC terminal stimulation. n=19(mock), 20(stim).

(G) Image shows the placement of AAV5 to the vHIPP.

(H–K) No alterations in P1 PDZ-RhoGEF levels (H and I) (t₁₁=0.050, p>0.05), or in P2 Rap1b or p-mTOR S448 levels (J and K) (Rap1b, t₁₁=0.300 p>0.05; p-mTOR, t₁₁=1.301, p>0.05), after vHIPP terminal stimulation. n=8(mock), 5(stim).

(L) Image shows the placement of AAV5 to the BLA.

(M–P) No alterations in P1 PDZ-RhoGEF levels (M and N) (t₁₁=0.368, p>0.05), or in P2 Rap1b or p-mTOR S2448 levels (O and P) (Rap1b, t₁₁=0.728, p>0.05; p-mTOR, t₁₁=0.0455, p>0.05), after BLA terminal stimulation. n=7(mock), 6(stim).

(Q) AAV5-CaMKIIa–ChR2–EYFP was infused into vmPFC of floxed Rap1 mice. After >6-weeks post-AAV5 infusion, HSV-GFP or HSV-Cre-GFP (Rap1 knockdown) was infused into NAc and NAc shell targeting optical fibers implanted. Mice were trained in a CPP paradigm in which one end-chamber was paired with vmPFC terminal NAc stimulation and the other with mock-stimulation, or in some mice both chambers were paired with mock stimulation.

(R) Stimulation increases CPP scores in mice expressing HSV-GFP, but not those expressing HSV-Cre-GFP in NAc. [2×2 ANOVA stimulation effect, F_1,33=6.55; post hoc GFP mock vs. GFP stim, *p<0.05; post hoc Cre mock vs. Cre stim, p>0.05. n=12(gfp mock), 10(gfp stim), 9(cre mock), 6(cre stim).

All summary data are the mean +/− SEM.

Next, we tested how stimulation of infralimbic PFC to NAc projections influences behavioral reward, and assessed the impact of altered Rap1 NAc levels in these effects. We infused AAV5-CaMKIIa–ChR2–EYFP into infralimbic PFC and implanted optical fibers targeting NAc shell in floxed Rap1 mice in which either HSV-GFP or HSV-Cre-GFP was infused into NAc. Mice were then trained in a CPP task where one chamber was paired with optical stimulation and the other with mock-stimulation. To assess baseline behavior, in some mice both chambers were paired with mock-stimulation. 24 hr after the last training session, the amount of time mice spent in the pre-assigned stimulated chamber vs. mock-stimulated chamber was determined (Figure 7Q). Mice infused with HSV-GFP or HSV-Cre-GFP in NAc that received mock-stimulation in both chambers exhibited no preference. Mice infused with HSV-GFP that received stimulation in one chamber showed a significant preference for the stimulated chamber, and this effect was blocked by Rap1 knockdown from NAc (Figure 7R).

Discussion

While several studies have shown that the actions of specific nuclear transcription factors, or the actions of specific regulators of spine plasticity, in NAc control behavioral responses to cocaine, in most instances nuclear and synaptic signal transduction pathways have been studied as discrete events. Here we investigate a potential means through which nuclear to synaptic influences are coordinated to regulate synaptic structural plasticity and associated behavioral adaptations in response to chronic cocaine administration. The nuclear and synaptoneurosomal signal transduction pathway identified in this study is biphasically regulated in a temporal manner after the last cocaine dose, and our findings indicate that this bidirectional regulation contributes to the dynamic spine structural changes that occur at early vs. longer times post-cocaine (Figure 8).

Figure 8. Summary Schematic.

Levels of PDZ-RhoGEF, an activator of the RhoA small GTPase, decrease in the cytoplasm and increase in the nucleus of NAc MSNs 24 hr after the last chronic cocaine dose. This is associated with increased levels of active RhoA in the nucleus, which promotes the formation of F-actin. These actions, in turn, activate the serum response factor (SRF) transcription factor by controlling the localization of the SRF co-activator MAL. MAL is unable to interact with SRF when bound to G-actin, but 24-hr post-cocaine MAL shifts away from G-actin and toward F-actin due to higher RhoA activity, with a resulting increase in MAL-SRF-dependent transcription. This PDZ-RhoGEF-RhoA-actin-MAL-SRF signaling cascade drives Rap1b expression in synaptoneurosomes, where it activates the PI3K-Akt-mTOR pathway. Activation of this pathway drives the formation of thin, immature spines 24-hr post-cocaine. Conversely, the activity of this signaling pathway is reversed 3–4 weeks after cocaine administration and results in the formation of mature, mushroom spines.

Opposing alterations in synaptic structure and function of NAc MSNs are seen at early vs. longer times after the last cocaine dose, such that synaptic weakening as evidenced by de novo thin spine formation is seen at early time points, while increased formation of mushroom spines and associated synaptic strengthening are seen at later time points (Dong and Nestler, 2014; Russo et al., 2010; Shen et al., 2009). Despite the seemingly disparate forms of plasticity occurring early vs. late, the synaptic reconfiguration during this process is a potential example of “metaplasticity” in which the initial formation of new thin spines provides a locus at which subsequent spine maturation can occur (Lee and Dong, 2011). Our findings suggest that Rap1b contributes to the time-dependent bidirectional synaptic structural plasticity during this process, as Rap1b levels are elevated 24-hr post-cocaine and induces de novo thin spine formation, while Rap1b is downregulated at 3–4-week post-cocaine causing spine maturation. While we view the 24 hr vs. 3–4 week differences as reflecting time-dependent changes that occur during cocaine withdrawal, future studies are needed to examine the alternative possibility that these differences reflect delayed effects of prior cocaine administration.

We found that Rap1b overexpression increases active forms of Akt and mTOR in NAc P2 fractions, while the loss of Rap1b in NAc has the opposite effects. Similar to Rap1b, active forms of NAc Akt and mTOR are also biphasically regulated as a function of time after the last cocaine dose, such that their levels are elevated 24-hr post-cocaine and reduced 3–4-week post-cocaine. In addition to alterations in mTOR activity, Rap1b overexpression increases total mTOR levels, and increases in total mTOR are evident 24-hr post-cocaine in self-administering animals, with a reduction seen at the 4 week time point. While Rap1b binds to and activates the upstream Akt-mTOR activator PI3K, the observed alterations in total mTOR suggest that the effects of Rap1b on mTOR signaling are more complex than a simple linear pathway. One complicating factor is that, while total Rap1b levels in the P2 fraction are increased 24-hr post-cocaine and decreased at 3–4-week post-cocaine, it is not clear how strongly the activity of the protein is altered in these instances. Insight into this question will require techniques to reliably measure Rap1b activity in dendritic spines.

While we found that manipulating Rap1b levels affects levels of p-mTOR S2448, the effects of Rap1b on mTOR could occur through multiple pathways, and not solely through PI3K-PDK1-Akt-mTOR signaling. Rap1 activates the MEK and ERK activating protein B-Raf (York et al., 1998), and like Akt, ERK can also increase mTOR activity (Huang and Fingar, 2014). A better understanding of the specific pathways that lead to altered mTOR signaling after cocaine exposure requires further investigation, including a more detailed time course of protein level changes that occur during the withdrawal process.

Studies in cortical and hippocampal neurons have shown that PI3K-Akt-mTOR activity increases spine density downstream of multiple signaling cascades; however, some studies have reported that activation of this pathway increases mature spine formation, while other studies point to stimulation of de novo thin spine formation (Enriquez-Barreto et al., 2014; Lee et al., 2011; Li et al., 2010). As the effects of PI3K-Akt-mTOR signaling on spine morphogenesis likely occur in large part through the regulation of local protein synthesis in spines, the varying effects of manipulating this pathway on spine morphogenesis could depend on which locally synthesized proteins are sensitive to mTOR in the brain region, cell type, and age at which experiments are conducted. PSD-95 and Arc are two of the most commonly studied locally synthesized targets of mTOR in spines. In some studies, upregulation of PSD-95 in response to mTOR activation is associated with spine maturation (Li et al., 2010), while other studies have shown that upregulation of Arc, an established facilitator of thin spine formation (Peebles et al., 2010), underlies the ability of PI3K-Akt signaling to induce de novo thin spines in some neuronal populations (Enriquez-Barreto et al., 2014). We found that short-term disruption of Akt activity inhibits Rap1b’s ability to induce de novo thin spine formation in NAc MSNs, indicating that Akt is a key mediator of Rap1b’s effects on synaptic structural plasticity in NAc. Moreover, longer-term disruption of Akt activity in NAc increased spine maturation that closely parallels the effects of prolonged Rap1 loss. Also, we found that rapamycin treatment of isolated NAc MSN synaptoneurosomes reduced protein levels of Arc with no effect on PSD-95 levels. Together, these data are consistent with studies indicating that the PI3K-Akt pathway is important for thin spine formation (Enriquez-Barreto et al., 2014) and that inhibition of PI3K-Akt signaling can lead to increased spine maturation (Kumar et al., 2005).

Recent evidence indicates that increased glutamatergic activity in infralimbic PFC to NAc synapses occurs 24 hr after the last cocaine dose with non-contingent or contingent drug administration (Suska et al., 2013). We found that stimulation of infralimbic PFC terminals in NAc, but not those of other input regions, increased Rap1b and active mTOR levels in NAc P2 synaptoneurosomes 24 hr post-stimulation. Moreover, infralimbic PFC terminal stimulation in NAc increased reward behavior, and this effect was blocked by the loss of Rap1 in NAc. That increased Rap1b and active mTOR levels in NAc occurred specifically in response to infralimbic PFC terminal stimulation is intriguing given evidence indicating that the basal function of this circuit promotes drug seeking and that this circuit eventually gains a new anti-relapse function in concert with increased synaptic maturation during protracted cocaine withdrawal (Ma et al., 2014). As we found decreased Rap1b-mTOR signaling in NAc 3–4-week post-cocaine, and that the loss of Rap1b or Akt activity increases spine maturation and decreases cocaine reward, it is possible that the loss of Rap1-mTOR signaling within MSN synapses of the infralimbic PFC projection contributes to the acquired anti-relapse properties of this circuit. Support for this comes from our findings indicating that Rap1 downregulation in NAc occludes the rewarding properties of infralimbic PFC to NAc stimulation. Whether the effects of short-term and long-term cocaine withdrawal on the nuclear and synaptoneurosomal protein expression profiles identified in this study occur via cocaine’s action of infralimbic PFC to NAc circuits remains a worthy endeavor for further investigation.

Experimental Procedures

For biochemistry experiments, mice were intraperitoneally (IP) injected with saline or cocaine (in 0.9% sterile saline) at 20mg/kg body weight daily for 7-consecutive days and bilateral NAc punches were taken from each animal 24-hr or 1-, 2-, or 3-week post-cocaine. For cocaine experiments done in mice infused with HSVs, 5 intraperitoneal cocaine injections (20mg/kg) were administered during a 3-day period to allow for sufficient cocaine exposure during the transient HSV lifespan. Specifically, 2-days after HSV NAc infusion, mice were given a single saline or cocaine injection during the early evening. The next day, mice were given 3 temporally spaced saline or cocaine injections. The following day (5-days post viral-infusion), mice were injected with a single dose of saline or cocaine, and NAc punches taken 4-hours following the last injection. For self-administration experiments, rats were implanted with intravenous catheters and allowed to self-administer saline or cocaine during daily 6-hr sessions for 10 consecutive days. Standard procedures were used for subcellular fractionations and behavioral assays (see Supplemental Experimental Procedures). Table S3 contains five-number summary descriptive statistics for all biochemical data.