|Home | About | Journals | Submit | Contact Us | Français|
Chromatin remodeling by histone deacetylases (HDACs), is a key mechanism regulating behavioral adaptations to cocaine use. We report here that cocaine and cyclic-adenosine monophosphate (cAMP) signaling induce the transient nuclear accumulation of HDAC5 in rodent striatum. We show that cAMP-stimulated nuclear import of HDAC5 requires a novel signaling mechanism that involves transient, protein phosphatase 2A (PP2A)-dependent dephosphorylation of a novel Cdk5 site (S279) found within the HDAC5 nuclear localization sequence. Dephosphorylation of HDAC5 increases its nuclear accumulation, by accelerating its nuclear import rate and reducing its nuclear export rate. Importantly, we show that dephosphorylation of HDAC5 S279 in the nucleus accumbens suppresses the development, but not expression, of cocaine reward behavior in vivo. Together, our findings reveal a new molecular mechanism by which cocaine regulates HDAC5 function to antagonize the rewarding impact of cocaine, likely by putting a brake on drug-stimulated gene expression that supports drug-induced behavioral changes.
Histone deacetylation by chromatin modifying enzymes plays a critical role in shaping transcriptional responses to experience. Drug addiction is thought to represent a long-lasting, maladaptive change in the function of the brain reward circuitry, and drug-induced transcriptional responses contribute to behavioral adaptations relevant to addiction (Hyman et al., 2006; Kalivas, 2004; McClung and Nestler, 2003). Several recent studies have reported an important role for HDAC activity in the regulation of cocaine-induced behaviors in rodent models of addiction (Hui et al., 2010; Kumar et al., 2005; Renthal et al., 2009; Renthal et al., 2007; Sanchis-Segura et al., 2009; Wang et al., 2010). However, how cocaine regulates HDAC function in brain reward circuitry, and whether regulation is important for its ability to modulate addiction-related behavioral responses, is poorly understood.
The class IIa HDACs emerged recently as important modulators of cocaine-induced behavioral responses in vivo (Kumar et al., 2005; Renthal et al., 2007; Wang et al., 2010). The class IIa HDACs (HDAC4, 5, 7, and 9) are unique among the HDAC family proteins in that they shuttle between the nucleus and the cytoplasm in cells (Belfield et al., 2006; Bertos et al., 2001; McKinsey et al., 2000a; McKinsey et al., 2001). Nucleocytoplasmic shuttling is governed by a basic residue-rich nuclear localization sequence (NLS) located within the N-terminal half of the proteins and a nuclear export sequence (NES) located within the C-terminal region (McKinsey et al., 2000a; McKinsey et al., 2001). Numerous studies have reported that CaMK superfamily proteins, in response to an intracellular calcium rise, increase phosphorylation at two conserved sites, S259 and S498, which serve to: (1) increase binding of HDAC5 to the 14-3-3 cytoplasmic anchoring proteins, (2) disrupt binding between HDAC5 and myocyte enhancer factor 2 (MEF2) transcription factors in the nucleus, and (3) promote cytoplasmic localization of HDAC5 (Chawla et al., 2003; McKinsey et al., 2000a; McKinsey et al., 2000b, 2001; Sucharov et al., 2006; Vega et al., 2004).
HDAC5 in the nucleus accumbens (NAc) was shown recently to reduce the rewarding impact of cocaine and inhibit cocaine experience-dependent reward sensitivity (Renthal et al., 2007), suggesting that it plays an active role in the nucleus to repress genes expression that promotes cocaine reward behavior. One of the only known HDAC5 interacting proteins in the nucleus is the MEF2 family of transcription factors, and HDAC5 is known to antagonize MEF2-dependent transcription (Lu et al., 2000). Consistently, expression of active MEF2 in the NAc enhances cocaine reward behavior (Pulipparacharuvil et al., 2008), which is opposite to the effects of HDAC5 expression in the NAc (Renthal et al., 2007). Activation of D1-class dopamine receptors (D1-DARs), or elevation of cAMP levels, reduces basal and calcium-stimulated MEF2 activity in striatal or hippocampal neurons (Belfield et al., 2006; Pulipparacharuvil et al., 2008), which motivated us to explore the possibility that cocaine and cAMP signaling might regulate HDAC5’s nuclear localization and/or function in the striatum in vivo.
In the present study, we uncover a novel signaling mechanism by which cocaine and cAMP signaling promote transient nuclear accumulation of HDAC5 through dephosphorylation-dependent regulation of NLS function in striatal neurons in vitro and in vivo, and demonstrate that this regulatory process is essential for the ability of HDAC5 to limit cocaine reward in the NAc in vivo. Taken together with previous work, our findings reveal that transient and dynamic regulation of this epigenetic factor plays an important role in limiting the rewarding impact of cocaine after repeated drug exposure.
To test whether cAMP signaling regulates striatal HDAC5, we transiently transfected a plasmid expressing HDAC5-EGFP fusion protein into cultured primary striatal neurons, and then analyzed the basal and cAMP-stimulated steady-state subcellular distribution. Under basal culture conditions we observed that a majority of HDAC5 is localized in the cytoplasm or is evenly distributed between the nucleus and cytoplasm (Figure 1A–B). However, elevation of cAMP levels with the adenylyl cyclase activator, forskolin (10 μM), induced the rapid nuclear import of HDAC5 (Figure 1A–B) where it accumulated in a predominantly punctate pattern (Figure 1A). The cAMP-induced steady-state nuclear accumulation of HDAC5 occurred over a time course of ~1–2 hrs in striatal neurons (Figure 1B, bottom). Interestingly, the nuclear HDAC5 puncta co-localized with endogenous MEF2 proteins (Figure S1A), suggesting that the nuclear HDAC5 is associated with transcriptional complexes on genomic DNA and that previously noted cAMP-dependent suppression of MEF2 activity is likely mediated by HDAC5 (Belfield et al., 2006; Pulipparacharuvil et al., 2008).
We speculated that cAMP signaling might regulate nuclear accumulation by regulating HDAC5 phosphorylation. By in silico analysis of the HDAC5 primary amino acid sequence, we identified a highly conserved serine (S279) that was a candidate substrate for Protein Kinase A (PKA) or Cyclin-dependent kinase 5 (Cdk5), both of which are implicated in drug addiction-related behavioral adaptations (Benavides et al., 2007; Bibb et al., 2001; Pulipparacharuvil et al., 2008; Self et al., 1998). Since S279 resides within the HDAC5 nuclear localization sequence (NLS), which is characterized by a high density of basic residues (Figure 1C, noted by asterisks), we speculated that phosphorylation at this site may modulate nucleocytoplasmic localization of HDAC5. The HDAC5 S279 site (and surrounding residues) was highly conserved from fish to humans (Figure 1C) and in both HDAC4 and HDAC9. Tandem mass spectrometry analysis of flag-epitope tagged HDAC5 in cultured cells revealed a singly-phosphorylated peptide (SSPLLR: 278–283 a.a.) (Figure S1B). Therefore, we generated a phosphorylation site-specific antibody against HDAC5 S279 to study its regulation by cAMP signaling. The P-S279 peptide antibody recognizes wild type (WT) HDAC5, but not a mutant form that cannot be phosphorylated at this site (HDAC5 S279A) (Figure 1D). It also recognizes endogenous P-HDAC5 after immunoprecipitation (IP) of total HDAC5 from cultured striatal neurons or adult striatal tissues, but not from anti-HDAC5 IPs using HDAC5 knockout mouse lysates (Figure S1C) (Chang et al., 2004), indicating that endogenous HDAC5 is basally phosphorylated at S279 in striatum in vitro and in vivo.
To determine whether Cdk5 or PKA can phosphorylate HDAC5 S279, we incubated full-length, dephosphorylated HDAC5 with recombinant Cdk5/p25 or PKA in vitro and found that either kinase can phosphorylate S279 in vitro (Figure S2A–B). However, when we incubated striatal neurons with specific kinase inhibitors for either Cdk5, PKA or p38 MapK (all potential kinases predicted for S279), we observed dramatically reduced P-S279 levels in the presence of Cdk5 inhibitors (Figure 2A and S2C), but observed no change in P-S279 in the presence of PKA or p38 MapK inhibitors (Figure S2C). Together, these findings indicate that while PKA is able to phosphorylate HDAC5 in vitro, it is not required for endogenous HDAC5 P-S279 in striatal neurons.
To test whether cAMP signaling regulates HDAC5 P-S279 levels, we cultured primary striatal neurons and elevated cAMP levels with forskolin (10 μM) or a nonselective phosphodiesterase inhibitor, 3-isobutyl-1-methylxanthine (IBMX, 200 μM), for various periods of time. We observed a rapid and robust dephosphorylation of HDAC5 S279 within 20 minutes of forskolin or IBMX treatment (Figure 2B and S2D), an effect that was stable for at least 3 hrs. In addition, forskolin induced robust dephosphorylation of endogenous HDAC5 S279 in cultured primary cortical neurons, COS7 cells (Figure S2E and S2F), as well as with over-expressed HDAC5-EGFP in HEK-293T cells (data not shown). These findings suggest that cAMP-stimulated dephosphorylation of HDAC5 S279 is a conserved mechanism across multiple cell types, including non-neuronal cells.
We next sought to identify the molecular mechanisms by which cAMP signaling stimulates HDAC5 dephosphorylation of P-S279. Elevation of cAMP levels increases the activity of the protein phosphatase, PP2A, in striatal neurons (Ahn et al., 2007; Ceglia et al., 2010). Consistent with this pathway, we found that okadaic acid, a potent inhibitor for PP2A and partial inhibitor of PP1, blocked cAMP-induced dephosphorylation of P-S279 in striatal neurons (Figure 3A), whereas the PP1-specific inhibitor, tautomycetin, had no effect (Figure S3). In addition, we observed that purified PP2A was sufficient to dephosphorylate endogenous HDAC5 P-S279 in vitro (Figure 3B). Together, these data reveal that PP2A activity is necessary and sufficient for cAMP-stimulated dephosphorylation of HDAC5 S279 in striatal neurons.
To test the role of PP2A activity on nucleocytoplasmic localization of HDAC5, striatal neurons were treated with okadaic acid or tautomycetin in the presence or absence of forskolin treatment. Okadaic acid treatment increased basal HDAC5 localization in the cytoplasm, and it blocked the cAMP-induced nuclear import of WT HDAC5-EGFP (Figure 4A). In contrast, tautomycetin altered neither basal nor cAMP-induced localization of WT HDAC5-EGFP (Figure S4A), indicating that PP2A activity is required for cAMP-induced nuclear accumulation.
To test whether the PP2A-dependent dephosphorylation of HDAC5 S279, specifically, was required for the cAMP-induced nuclear import of HDAC5, we generated a phospho-mimetic mutant at this site by changing S279 to a negatively-charged residue, glutamic acid (E), and then analyzed the subcellular localization pattern of the HDAC5 before and after elevation of cAMP in cultured striatal neurons. Compared to WT HDAC5, we observed that most of the HDAC5 S279E protein localized in the cytoplasm under unstimulated conditions (Figure 4B). However, unlike WT HDAC5, the HDAC5 S279E mutant failed to relocalize to the nucleus by 3 hrs after forskolin treatment (Figure 4B, middle). The S279E mutant did not simply disrupt the NLS function, since treatment with the Crm1-mediated nuclear export inhibitor, leptomycin B (LMB) (Harrison et al., 2004; Vega et al., 2004), led to near complete nuclear accumulation of both WT and S279E HDAC5 at similar rates (Figure S5 and 5A). These findings indicate that dephosphorylation of HDAC5 S279 is necessary for cAMP-induced nuclear accumulation.
To test whether dephosphorylation of S279 is sufficient to promote nuclear localization, we expressed in striatal neurons the non-phosphorylatable HDAC5 S279A mutant. Under basal conditions, localization of the HDAC5 S279A mutant was similar to WT HDAC5 (Figure 4B, right), indicating that dephosphorylation of S279 alone is not sufficient to confer nuclear localization of HDAC5. Similar to WT HDAC5, forskolin stimulated nuclear accumulation of HDAC5 S279A, which indicates that dephosphorylation of S279 is necessary, but not sufficient, for cAMP-induced nuclear accumulation of HDAC5. Similar basal subcellular distribution and responses to cAMP were observed with HDAC5 proteins lacking EGFP fusion protein (Figure S4B).
CaMK or PKD-dependent phosphorylation of HDAC5 P-S259 and P-S498 confers cytoplasmic localization of HDAC5 in non-neuronal cells (McKinsey et al., 2000a), mediates binding to 14-3-3 cytoplasmic anchoring proteins, and disrupts association with MEF2 transcription factors (Harrison et al., 2004; McKinsey et al., 2000b; Vega et al., 2004). Interestingly, forskolin treatment stimulated dephosphorylation of both S259 and S498 to a similar extent as S279 (Figure 4C), indicating that all three sites are negatively-regulated by cAMP signaling. Consistent with previous studies (McKinsey et al., 2000a; Vega et al., 2004), we found that HDAC5 S259A or S259A/S498A mutants were distributed evenly between the cytoplasm and nucleus or were concentrated in the nucleus (Figure 4D, left, and S4C), confirming a critical role for these phosphorylation sites in striatal neurons. However, we found that the HDAC5 S259A and S259A/S498A mutants had significantly reduced (~60%) P-S279 levels (Figure S4D), confounding a straightforward interpretation of the S259A and the S259A/S498A effects on nuclear/cytoplasmic localization and suggesting that P-S279 is sensitive to the phosphorylation status of S259. Interestingly, forskolin treatment of striatal neurons stimulated strong nuclear accumulation of HDAC5 S259A or S259A/S498A (Figure 4D and S4C), indicating that dephosphorylation of S259 and S498 alone cannot account for cAMP-induced nuclear import. To test the specific importance of P-S279 in this context, we generated compound HDAC5 mutants, S259A/S279E and S259A/S498A/S279E, and observed that the S279E mutation shifted the basal subcellular localization away from the nucleus in a pattern similar to WT HDAC5 (Figure 4D and S4C). Consistent with the single mutant (S279E, Figure 4B), forskolin-induced nuclear accumulation of HDAC5 was defective in either of the compound mutants, confirming an essential and independent function for dephosphorylation of HDAC5 S279 in cAMP-induced nuclear import.
We next sought to understand how dephosphorylation of HDAC5 S279 promotes nuclear accumulation in striatal neurons. As mentioned earlier, blocking Crm1-dependent nuclear export with LMB results in strong nuclear accumulation of both WT and S279E HDAC5 proteins (Figure S5), indicating that both of these proteins shuttle between the nucleus and cytoplasm under basal conditions. The steady-state, nucleocytoplasmic distribution of HDAC5 is determined by the balance of nuclear import and nuclear export kinetics. Therefore, the cAMP-induced accumulation of HDAC5 in the nucleus likely represents a change in the nuclear import rate, the nuclear export rate or both. To evaluate these parameters, we used conditions where HDAC5 nuclear export was blocked (LMB) with or without simultaneous elevation of cAMP. Compared to the LMB-only condition, we observed a dramatic increase in the nuclear import rate of WT HDAC5 after forskolin treatment, resulting in near complete disappearance from the cytoplasm by 20 minutes (Figure 5A); this condition showed similar kinetics to forskolin-induced dephosphorylation of S279 (Figure 2B). In contrast, the import rate of HDAC5 S279E after forskolin + LMB treatment is nearly indistinguishable from the rate of nuclear import of WT HDAC5 treated with vehicle + LMB (Figure 5A), which indicates that dephosphorylation of S279 accelerates the nuclear import rate. We next tested potential effects of P-S279 on HDAC5 nuclear export by first incubating striatal neurons with LMB to force accumulation of WT or S279E HDAC5 into the nucleus (Figure 5B). Following washout of LMB, we monitored the initial rate of nuclear export and observed that the HDAC5 S279E mutant disappeared from the nucleus more rapidly than WT HDAC5 (Figure 5B). Therefore, our findings suggest that cAMP increases the HDAC5 nuclear import rate and decreases the nuclear export rate by stimulating dephosphorylation of HDAC5 S279. In addition, we observed that the HDAC5 S279E mutant co-precipitates with a cytoplasmic chaperone protein, 14-3-3, to a significantly greater extent than WT HDAC5 in cultured cells (Figure 5C), suggesting that P-S279 enhances the affinity of 14-3-3 and HDAC5, potentially enhancing cytoplasmic retention and nuclear export of HDAC5. However, the HDAC5 S259A/S498A/S279E mutant, despite its enhanced cytoplasmic localization, fails to co-IP with 14-3-3 (data not shown), indicating that the primary cytoplasmic localizing function of P-S279 is not likely due to its enhancement of 14-3-3 binding.
Cocaine and dopamine signaling regulate cAMP levels in striatum. To test whether dopamine signaling regulates HDAC5 phosphorylation in striatum in vivo, we injected adult mice with a dopamine D1-class receptor agonist, SKF81297 (5 mg/kg), or a dopamine D2-class receptor agonist, quinpirole (5 mg/kg), and analyzed striatal HDAC5 P-S279 levels in vivo. We found that SKF81297 administration stimulates significant dephosphorylation of HDAC5 S279 at 30 minutes and 3 hrs after injections, whereas exposure to quinpirole stimulated a trend toward increased HDAC5 P-S279 levels (Figure S6A-B). We next injected adult C57BL/6 mice with cocaine (20 mg/kg) and analyzed both HDAC5 P-S279 levels and nuclear/cytoplasmic localization of endogenous HDAC5 in the striatum. We compared mice injected 7 days with saline (vehicle control), 7 days with cocaine (cocaine-experienced), or 6 days with saline and one cocaine injection on the 7th day (cocaine-naive), and analyzed HDAC5 P-S279 levels at 1, 4 and 24 hrs after the last injection (Figure 6A). By first immunoprecipitating total HDAC5, we were able to measure HDAC5-specific P-S279 levels as confirmed in HDAC5 KO mice (Figure S1C). We observed a significant dephosphorylation of HDAC5 at 1 and 4 hrs following the last injection in both the cocaine-naïve and cocaine-experienced mice, but phosphorylation at S279 had returned to baseline levels by 24 hrs after the last cocaine injection. We next analyzed the levels of P-S259 and P-S498 HDAC5 after cocaine, and similar to P-S279 regulation, we observed a significant reduction of all three sites (Figure 6B). Taken together, these findings reveal that cocaine stimulates the coordinated dephosphorylation of P-S259, P-S279, and P-S498 on HDAC5.
We next analyzed the effects of cocaine on nuclear/cytoplasmic distribution of endogenous striatal HDAC5 using a biochemical fractionation approach. Similar to the subcellular distribution of HDAC5 in primary striatal neurons in culture (e.g. Figure 1B), a majority of the striatal HDAC5 co-fractionated with cytoplasmic proteins (Figure 6C). Following the same dosing paradigm detailed above, administration of cocaine to naïve or cocaine-experienced mice resulted in a significant accumulation of HDAC5 in the nucleus at 4 hrs after the last injection, and like the regulation of P-S279, nuclear accumulation was transient and returned to saline control levels by 24 hrs after the last injection (Figure 6D). Taken together, these results reveal that cocaine administration stimulates the rapid and transient dephosphorylation of HDAC5 and subsequent nuclear accumulation of endogenous HDAC5 in vivo.
To test the importance of cocaine-induced dephosphorylation of HDAC5 S279 for the development of cocaine reward behavior, we utilized viral-mediated gene transfer to express full-length, HDAC5 WT or mutants (S279A or S279E) bilaterally in the NAc of WT, adult male mice (Figure 7A) prior to a cocaine conditioned place preference (CPP) assay. This assay involved pairing one of two distinct chambers with either cocaine or saline injections for two consecutive days. Subsequently, the mice were given equal access to both chambers and time spent in either the cocaine-paired or saline-paired chamber was measured. As expected, the control virus (GFP)-injected mice spent significantly more time in the cocaine-paired chamber (Figure 7C), indicating a clear positive preference for the context in which cocaine was experienced. Similar to a previous report (Renthal et al., 2007), the overexpression of WT HDAC5 reduced cocaine-associated place preference, but it did not reach significance. Mice expressing the HDAC5 S279E mutant protein had a cocaine place preference similar to the GFP-only control virus-injected mice, whereas mice expressing HDAC5 S279A dephosphorylation mutant showed significantly reduced cocaine place preference (Figure 7C; S279A, 81 sec vs. S279E, 246 sec). We observed similar expression levels of the HDAC5 WT, S279A and S279E mutants in striatal neurons (Figure 7B), indicating that the results are not likely due to differences in HDAC5 protein expression levels. As expected, we observed that mice injected with the lower dose of cocaine used in the CPP assay (5 mg/kg) showed a significant transient reduction of HDAC5 P-S279 levels (Figure 7D), although the magnitude and duration were somewhat attenuated when compared to the higher doses of cocaine (Figure 7D and data not shown). The absence of an effect by the HDAC5 S279E mutant is consistent with its localization in the cytoplasm in striatal neurons. These findings indicate that dephosphorylation of HDAC5 S279 in the NAc is required for HDAC5 to limit the rewarding impact of cocaine in vivo.
Since HDAC5 dephosphorylation was required for its ability to reduced cocaine reward behavior, we next asked whether HDAC5 dephosphorylation suppresses the dpevelopment of cocaine CPP, which is the period where regulation of P-HDAC5 is observed (Figure 6 and and7D),7D), or whether HDAC5 might be influencing the expression of CPP behavior during the test. To test this idea, we first performed cocaine vs. saline context pairing prior to bilateral expression of HDAC5 S279A or GFP-only vector in the NAc and then tested for the expression of cocaine CPP. Unlike expression of HDAC5 S279A during the cocaine/context pairings (development of CPP), we observed no significant differences between vector and HDAC5 S279A treatments during the expression of cocaine CPP behavior on the test day (Figure 7E), indicating that dephosphorylation of HDAC5 S279 resists the development of cocaine reward behavior, but does not reduce its expression.
Since HDAC5 dephosphorylation limits the development of cocaine reward, we next asked whether this mechanism might also regulate natural reward behavior, or whether the effect of HDAC5 is more specific for cocaine reward. To this end, we performed bilateral NAc injections of GFP-only control virus or the HDAC5 S279A virus and then measured a natural reward behavior, sucrose preference. When sucrose preference was measured daily for four consecutive days, we observed no differences in 1% sucrose preference between mice expressing HDAC5 S279A expression or GFP-only vector control expression (Figure 7F and S7), suggesting that HDAC5 does not regulate natural reward behavior and may have a more specific role for substance abuse.
Taken together, our findings reveal a novel molecular mechanism by which cocaine and cAMP signaling regulate HDAC5 nuclear accumulation to limit adaptations that increase the rewarding impact of cocaine (Figure 8). Our findings support the noted role for HDAC5 in limiting cocaine reward behavior (Renthal et al., 2007); however, our observations that cocaine induces transient, delayed dephosphorylation and nuclear import of HDAC5 to suppress cocaine reward is a significant departure from previous ideas of how cocaine regulates HDAC5 function in vivo (Renthal et al., 2007). We observed a significant regulation of HDAC5 phosphorylation and nuclear levels that strongly suggest that dynamic regulation of this epigenetic factor plays a crucial role in limiting the impact of cocaine reward in vivo.
Several studies have reported that cocaine exposure increases P-S259 HDAC5 levels by Western blotting or immunohistochemistry (Dietrich et al., 2011; Host et al., 2011; Renthal et al., 2007), but with the near perfect conservation of amino acids spanning the P-S259 site in HDAC4, HDAC5, HDAC7 and HDAC9, it is important to note that the P-S259 antibody recognizes multiple class IIa HDAC proteins, not only HDAC5. In contrast to these reports, our study revealed a robust decrease in P-S259 and P-S498 levels on HDAC5 (Figure 6B). Our analysis of the P-S279 HDAC5 site, which is also highly conserved in HDAC4 and HDAC9, revealed that total P-S279 immunoreactivity was not specific to HDAC5 (i.e. HDAC5 knockout mouse tissues had significant residual P-S279 immunoreactivity). To achieve HDAC5-specific analysis of these conserved sites, we had to first IP total HDAC5 protein prior to Western blotting with the phosphorylation site-specific antibodies (e.g. Figure S1C). In the future, it will be important to determine whether the reported increases in P-S259 signal after cocaine exposure reflect specific regulation of HDAC5 or might instead represent regulation of other class IIa HDAC(s).
The binding of HDAC5 to 14-3-3 proteins is mediated by phosphorylation of S259 and S498 sites, and this association is thought to be important for HDAC5 cytoplasmic localization (Chawla et al., 2003; McKinsey et al., 2000a; McKinsey et al., 2000b, 2001; Sucharov et al., 2006; Vega et al., 2004). Similar to previous work, we observe that the HDAC5 S259A/S498A mutant protein is largely localized within the nucleus or evenly distributed between nucleus and cytoplasm. However, this mutant has significantly reduced P-S279 levels (Figure S4D), which suggests that the increase in nuclear localization of this mutant may be due, at least in part, to reduced P-S279 levels. This conclusion is strengthened by the observation that combining the S279E phosphomimetic mutation with the S259A/S498A mutations results in increased cytoplasmic distribution of HDAC5 and resistance to cAMP-induced nuclear import. The S259A/S498A/S279E HDAC5 mutant does not bind to 14-3-3 (data not shown), which strongly suggests that P-S279 exerts its effect on HDAC5 nuclear import through a 14-3-3-independent mechanism. Several studies have reported that phosphorylation close to, or within, an NLS can mask a protein’s interaction with other proteins or inactivate its NLS function (Jans et al., 1991; Moll et al., 1991). Due to the strong concentration of positively charged residues within the HDAC5 NLS, we speculate that the introduction of three negative charges by organic phosphate at S279 might neutralize the NLS charge or induce a conformational change that reduces association with nuclear import proteins.
During review of our manuscript, a study reported regulation of P-S279 HDAC5 by PKA in COS7 cells (Ha et al., 2010), and provided evidence that P-S279 promoted nuclear retention in these cells. Similar to this study, we had also found that purified PKA phosphorylates HDAC5 S279 in vitro (Figure S2A); however, we found that basal phosphorylation at this site, at least in striatal neurons, did not require PKA activity (Figure S2C). In addition, our direct measurements of endogenous HDAC5 P-S279 levels revealed that forskolin treatment of COS7 cells, striatal neurons, cortical neurons, or acute, adult striatal slices actually decreased P-S279 HDAC5 levels (Figure 2B and S2, and data not shown), which seems incompatible with the proposed role for P-S279 in the COS7 cells. We speculate that the expression of constitutively-active PKA in COS7 cells may regulate additional HDAC5 sites that influence nuclear localization and require P-S279 or that over-expressed HDAC5-EGFP is regulated differently than endogenous HDAC5 in COS7 cells. Additional experiments will be required to help resolve the different conclusions drawn by these two studies, but in striatal neurons, it seems clear that HDAC5 P-S279 does not promote nuclear accumulation, but quite the opposite.
Our observations about the role and regulation of HDAC5 P-S279 in cocaine-induced behavioral plasticity raise a number of interesting questions for future study. For example, what is the nuclear function of HDAC5 that limits cocaine reward? Nestler and colleagues (Renthal et al., 2007) reported that the enzymatic HDAC domain of HDAC5 is required for reducing cocaine reward, suggesting that the ultimate substrate is histone deacetylation and indirect suppression of HDAC5 target genes. Indeed, many hundreds of genes were aberrantly increased or decreased by cocaine in the HDAC5 KO mice at 24 hrs after repeated cocaine injections. Since these were total HDAC5 KO mice, lacking HDAC5 expression throughout the lifetime of the animal, it is difficult to know whether these are direct effects of HDAC5 on the identified genes. Moreover, the time point analyzed (i.e. 24 hrs) is during a phase when HDAC5 phosphorylation and nucleocytoplasmic localization are similar to saline control conditions. In the future, it will be interesting to determine the target genes that are bound and regulated by HDAC5 after cocaine, particularly at those time points when enhanced HDAC5 nuclear function is observed following cocaine exposure. It is possible, and perhaps likely, that regulation of multiple HDAC5 gene targets contributes to the reduction of cocaine reward behavior, and dissecting out the relative contributions of each gene target will represent a major challenge going forward.
It is interesting to note that the HDAC5 S279A mutant suppressed cocaine reward to a greater extent than WT HDAC5 (Figure 7C). There are several possible explanations for this difference, including: (1) The HDAC5 S279A mutant in vivo resides constitutively in the nucleus, whereas the WT HDAC5 is only transiently localized in nucleus upon cocaine exposure. In this case, the levels of the P-S259/P-S498 would presumably be low such that P-S279 plays the dominant major role in subcellular localization (unlike the striatal cultures). (2) The HDAC5 S279A mutant has reduced nuclear export kinetics compared to WT HDAC5, and as a result, resides in the nucleus for a longer time after cocaine exposure. We found that P-S279 HDAC5 increases nuclear export kinetics (Figure 5B). If HDAC5 S279A cannot be rephosphorylated following nuclear import, then HDAC5 S279A may remain in the nucleus and exert longer-lasting effects following cocaine exposure. (3) HDAC5 P-S279 may regulate not only its nuclear/cytoplasmic localization, as documented in our study, but might also regulate its function as a transcriptional co-repressor in the nucleus. As such, the HDAC5 S279A mutant may be a more effective co-repressor via unknown mechanisms. Due to technical limitations, we were unable to visualize the subcellular distribution of the HDAC5 mutants in vivo. Nevertheless, our findings in this study reveal an important role for dephosphorylation of P-S279 HDAC5 in the regulation of cocaine reward behavior.
Our findings in striatal cultured neurons revealed a high degree of co-localization of HDAC5-EGFP with endogenous MEF2A and MEF2D, two of the well-studied transcription factor proteins that interact with HDAC5, suggesting MEF2 as a possible mediator of HDAC5 function in reducing cocaine reward sensitivity after repeated cocaine experience. Consistent with this idea, we reported recently that expression of constitutively-active MEF2 in the NAc enhances cocaine reward behavior (Pulipparacharuvil et al., 2008), which is opposite of the effect of HDAC5 expression in this region. In the future, it will be important to determine whether HDAC5 exerts its effects on cocaine reward through binding to MEF2 proteins, or whether the critical nuclear target of HDAC5 in the mediation of cocaine reward may be one or more previously undescribed transcription factors. The identification of HDAC5 target genes after cocaine exposure may help determine whether MEF2 and HDAC5 bi-directionally regulate cocaine reward through a common pathway or whether these proteins regulate cocaine behavior through distinct transcriptional mechanisms in vivo.
Similar to our observed regulation of HDAC5 P-S279, previous studies in striatal neurons have reported that cAMP signaling increases PP2A activity (Ahn et al., 2007), which then dephosphorylates the Cdk5 substrates, Wave1 (Ceglia et al., 2010) and dopamine- and cAMP-regulated neuronal phosphoprotein (DARPP-32) (Bibb et al., 1999; Nishi et al., 2000). Acute cocaine does not alter the levels or activity of Cdk5 or levels of p35 in striatum (Kim et al., 2006; Takahashi et al., 2005), suggesting that the decrease in P-S279 is due to increased phosphatase activity rather than decreased Cdk5 activity. Interestingly, cocaine and cAMP signaling has been shown to induce transient DARPP-32 nuclear accumulation via dephosphorylation in striatal neurons (Stipanovich et al., 2008). Similar to our findings with HDAC5, nuclear accumulation of DARPP-32 attenuates cocaine reward behavior, which is proposed to involve epigenetic gene regulation (Stipanovich et al., 2008). Together, these findings indicate that while cocaine induces rewarding effects, the striatum stimulates negative feedback processes, such as enhanced HDAC5 and DARPP-32 nuclear levels, to attenuate the reward impact of future cocaine exposures. As such, these proteins may represent critical intrinsic mechanisms for counteracting the maladaptive changes in reward circuit function, and understanding these negative feedback processes may reveal new avenues for the treatment of drug addiction.
Taken together, our findings reveal that cocaine regulates the transient nuclear accumulation of HDAC5, and this likely occurs through a newly discovered molecular mechanism involving PP2A phosphatase-dependent dephosphorylation of HDAC5 at three critical phosphoserines, S259, S279, and S498. The removal of phosphate from these sites likely increases the NLS function and decreases binding to 14-3-3 proteins, and promotes the repression of HDAC5 target genes in the nucleus. Importantly, our findings reveal that dephosphorylation of S279 HDAC5 is critical for its ability to limit the development of cocaine reward-related behavioral adaptations, but not natural reward behavior. Since cocaine-experienced HDAC5 KO mice have enhanced place preference to cocaine, and this effect is rescued by NAc expression of WT HDAC5 (Renthal et al., 2007), our combined findings suggest that HDAC5 provides a delayed braking mechanism on gene expression programs that support the development, but not expression, of cocaine reward behaviors. As such, deficits in this process may contribute to the development of maladaptive behaviors associated with addiction following repeated drug use in humans.
Herpes simplex virus (HSV)-flag-human HDAC5 plasmid was provided by Dr. Eric Nestler, and using PCR, we generated a non-tagged version of HDAC5 for subcloning. The PCR fragment was subcloned into Bluescript vector, and the insert region was confirmed by DNA sequencing. The non-tagged HDAC5 was then digested with XbaI and subcloned into HSV vector. Serine to alanine (S259A, S279A, S498A) and serine to glutamate (S279E) mutants of HDAC5 were generated by quick-change site-directed mutagenesis.
All C57BL/6 mice (Charles River) used in this study were adult males tested between 10-12 weeks old. They were housed on a 12 hr light-dark cycle with access to food and water ad libitum. All procedures were in accordance with the Institutional Animal Care and Use (IACUC) guidelines.
Rabbits were injected with a synthesized P-HDAC5 encompassing HDAC5 amino acids 274–285, where position S279 was phosphorylated (Covance). The injected peptide was conjugated to Keyhole Limpet Hemocyanin via an N-terminal cysteine residue. Anti-P-S279 antibodies were affinity-purified with peptide-conjugated Sepharose beads (SulfoLink, Pierce).
Embryonic striatal neurons (E18/19) were cultured from Long-Evans rats (Charles River) as described previously (Cowan et al., 2005; Pulipparacharuvil et al., 2008). Details can be found in the Supplemental Online Materials section.
Striatal neurons (E18 rat) were plated at 100,000 cells/well (24 well plate; Corning), grown on PDL/Laminin-coated glass coverslips for 8 days, and then transfected using calcium phosphate method (Pulipparacharuvil et al., 2008). Two days later, cells were stimulated with indicated agents. Neurons were fixed in 4% paraformaldehyde/2% sucrose in 1X PBS for 20 minutes at RT, permeabilized and stained with indicated primary and secondary antibodies (see Supplemental Online Materials). The localization of HDAC5 was categorized as cytoplasmic, nuclear, or both (evenly distributed across nucleus and cytoplasm) for each neuron under experimenter-blind conditions.
C57BL/6 mice (Charles River) were injected 1x/day (IP) with saline or cocaine (5 or 20 mg/kg) before rapid isolation of brain tissues at indicated times after injection. HDAC5 was IPed from diluted total striatal lysates and analyzed by standard Western blot analysis with indicated antibodies (see Supplemental Online Materials for dilutions and sources). Cytosolic and nuclear extracts were prepared with NE-PER nuclear and cytoplasmic extraction kit (Pierce Biotechnology) according to the manufacturer’s instructions.
HEK293T cells were cultured in Dulbecco’s modified Eagle medium containing 10% (v/v) FBS, penicillin-streptomycin (1X; Sigma) and L-glutamine (4 mM; Sigma). HEK293T cells were transfected with HSV-flag-hHDAC5 using calcium phosphate and harvested 2 days after transfection. Flag-HDAC5 was prepared from HEK293T cell extracts in RIPA buffer (50 mM Tris pH 7.4, 1 mM EDTA, 150 mM NaCl, 1% NP40, 0.1% SDS, 0.5% sodium deoxycholates, 10 mM NaF, 10 nM okadaic acid and complete protease inhibitor cocktail tablet (1X; Roche)) by IP with anti-flag antibody (M2) conjugated beads. The protein was separated by SDS-PAGE and stained with Coomassie Brilliant Blue. The HDAC5 band was excised from the gel, washed and then digested with trypsin. The tryptic digests were analyzed an EC-MS/MS system.
Flag-HDAC5 was prepared from transfected HEK293T cell extracts by IP with anti-flag antibody (M2) conjugated beads in RIPA buffer. For the PKA phosphorylation, IPed beads were washed and suspended in PKA phosphorylation buffer (50 mM PIPES, pH 7.3, 10 mM MgCl2, 1 mM DTT, 0.1 mg/ml BSA and protease inhibitor) and incubated with or without recombinant PKA catalytic subunit (Sigma) or alkaline phosphatase (Roche) in the presence of 1 mM ATP at 30°C. For the Cdk5 phosphorylation assay, the IPed flag-HDAC5 on the beads was washed and resuspended in alkaline phosphatase buffer (Roche) and incubated with alkaline phosphatase at 37°C for 2 hrs. Dephosphorylated beads were washed with RIPA buffer three times and Cdk5 kinase assay buffer (10 mM MOPS, pH 7.2, 10 mM MgCl2, 1 mM EDTA) three times, and the IPed HDAC5 was incubated with or without Cdk5-p25 (Sigma) in the presence of 1 mM ATP at 30°C. After boiling SDS sample buffer to elute from the beads, the incubated IP samples were subjected to Western blotting analysis.
IPed HDAC5 from striatal neurons in RIPA buffer was washed with dephosphorylation buffer (50 mM Tris-HCl, pH8.5, 20 mM MgCl2, 1 mM DTT, protease inhibitor cocktail (1X; Roche)) five times and incubated with or without 2.5 U of purified PP2A (Promega) at 30°C for 60 min. Proteins were subjected to western blotting analysis.
Expression plasmids for HDAC5 WT, S279A and S279E mutants in HSV vector were packaged into high-titer viral particles as described previously (Barrot et al., 2002). Stereotactic surgery was performed on mice under general anesthesia with a ketamine/xylazine cocktail (10 mg/kg: 1mg/kg). Coordinates to target the nucleus accumbens (shell and core) were +1.6 mm anterior, +1.5 mm lateral, and −4.4 mm ventral from Bregma (relative to dura) at a 10° angle. Virus was delivered bilaterally using Hamilton syringes at a rate of 0.1 μl/min for a total of 0.5 μl. Viral placements were confirmed by GFP signal, which was co-expressed in each virus.
Mice were conditioned to cocaine using an unbiased accelerated paradigm to accommodate the timing of transient HSV expression (Barrot et al., 2002; Renthal et al., 2007). Additional details can be found in the supplemental online materials.
Singly-housed mice were provided tap water in two identical double-ball bearing sipper-style bottles for two days followed by two days of 1% (w/v) sucrose solution to allow for acclimation and to avoid undesired effects of neophobia (Green et al., 2006). The next day mice underwent stereotactic injections of control or HDAC5 virus into the NAc (bilaterally, as described above for CPP assays). Forty-eight hrs after viral injection, mice were again given two bottles: one containing water and the other containing 1% sucrose solution. The consumption of water versus sucrose was measured after 24, 48, 72 and 96 hrs of access to the bottles to determine preference for sucrose (Renthal et al., 2007). Bottle positions of water and sucrose were swapped each day of testing to avoid potential drinking side bias.
One-, Two-way or repeated measures ANOVAs with Tukey’s multiple comparison post-hoc tests were used to analyze the following: Western blotting, phosphorylation level of S279, nuclear/cytoplasmic ratio of HDAC5 with cocaine exposure, conditioned place preference, sucrose preference and rates of nuclear export and import of HDAC5. Student’s t-tests were used to analyze HDAC5-EGFP localization, western blotting for phosphorylation level of S279 for samples treated with roscovitine, forskolin, okadaic acid, tautomycetin, and for the in vitro dephosphorylation assay with PP2A, cocaine-treated samples compared to saline controls and averaged sucrose preference data.
The authors would like to thank Darya Fakhretdinova, Katie Schaukowitch, Lindsey Williams and Marissa L. Baumgardner for the technical assistance, Dr. Li Yan in Protein Chemistry Technology Center lab for mass spectrometry analysis, Dr. Shari Birnbaum in UTSW Dept. of Psychiatry Behavioral Phenotyping Core for assistance with CPP data analysis, Dr. Timothy A. McKinsey for generously sharing the P-S259 class IIa HDAC antibody, and Dr. James Bibb (UTSW) for advice on in vitro Cdk5 assays. We also acknowledge Dr. Eric Olson (UTSW) for sharing HDAC5 KO mice and Yong-Chao Ma (CMRC/Northwestern) for critical reading of the manuscript. L.N.S. was supported by fellowships from NIDA (T32 DA07290 and F32 DA027265) and the FRAXA Foundation. C.W.C. acknowledges the generous support of the Whitehall Foundation, the Simons Foundation (SFARI), NIDA (DA008277 and DA027664) and NEI (EY018207 and a research supplement for underrepresented minorities to M.B.C.).
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.