eIF2α-mediated translational control regulates the persistence of cocaine-induced LTP in midbrain dopamine neurons

Recreational drug use leads to compulsive substance abuse in some individuals. Studies on animal models of drug addiction indicate that persistent long-term potentiation (LTP) of excitatory synaptic transmission onto ventral tegmental area (VTA) dopamine (DA) neurons is a critical component of sustained drug seeking. However, little is known about the mechanism regulating such long-lasting changes in synaptic strength. Previously, we identified that translational control by eIF2α phosphorylation (p-eIF2α) regulates cocaine-induced LTP in the VTA (Huang et al., 2016). Here we report that in mice with reduced p-eIF2α-mediated translation, cocaine induces persistent LTP in VTA DA neurons. Moreover, selectively inhibiting eIF2α-mediated translational control with a small molecule ISRIB, or knocking down oligophrenin-1—an mRNA whose translation is controlled by p-eIF2α—in the VTA also prolongs cocaine-induced LTP. This persistent LTP is mediated by the insertion of GluR2-lacking AMPARs. Collectively, our findings suggest that eIF2α-mediated translational control regulates the progression from transient to persistent cocaine-induced LTP. DOI: http://dx.doi.org/10.7554/eLife.17517.001


Introduction
Drug addiction is a complex behavioral disorder that starts with recreational use and, in some people, progresses to compulsive drug-seeking (Hyman, 2005). The precise molecular and cellular mechanism underlying this transition remains unclear. In addicts, repeated drug use leads to longlasting changes in neuronal structure and function in key reward areas (Koob and Volkow, 2010), which have emerged as cellular correlates of drug addiction (Chen et al., 2008;Lüscher and Malenka, 2011). Of particular interest are excitatory synaptic afferents to dopaminergic neurons in the ventral tegmental area (VTA). These activity-dependent changes in synaptic strength in the VTA constitute the initial synaptic adaptations observed after drug exposure (Lüscher and Malenka, 2011;Schilstrom et al., 2006). For instance, a single injection of cocaine (or other drugs of abuse) given 24 hr before recording induces an LTP that is manifested as an increase in the ratio of the amplitude of a-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR)-to N-methyl D-aspartate receptor (NMDAR)-mediated excitatory postsynaptic currents (EPSCs) onto DA neurons in the VTA (Ungless et al., 2001). These early changes in synaptic strength are believed to facilitate persistent alterations in response to repeated exposure to drugs of abuse.
LTP induced by an acute injection of cocaine typically lasts up to five days, but it returns to baseline after 10 days (Borgland et al., 2004). This short-lasting LTP is associated with the insertion of AMPARs lacking the GluR2 subunit, as demonstrated by an increased inward rectification of postsynaptic AMPAR currents (Bellone and Lüscher, 2006;Engblom et al., 2008). Interestingly, in contrast to passive cocaine administration, LTP persists for several weeks following self-administration (Chen et al., 2008). Thus, these drug-induced persistent changes in synaptic strength in the VTA may represent the cellular process driving the progression from recreational to compulsive drug use. Furthermore, pharmacological activation of metabotropic glutamate receptors (mGluRs) with dihydroxyphenylglycine (DHPG), a group I mGluR agonist, inhibits cocaine-induced LTP in VTA DA neurons (Bellone and Lüscher, 2006). This inhibition is attributed to the opposing effect of mGluRmediated long-term depression (mGluR-LTD), which reverses drug-evoked LTP and inward rectification in VTA DA neurons by internalization of AMPARs (Bellone and Lüscher, 2006;Lüscher and Huber, 2010;Lüscher and Malenka, 2011).
Our recent work provided a unifying model that explains how translational control by phosphorylation of the eukaryotic translation initiation factor eIF2 alpha subunit (eIF2a) regulates these two opposing forms of plasticity (cocaine-induced LTP and mGluR-induced LTD) in the VTA . First, cocaine induces LTP and reduces p-eIF2a levels in the VTA , whereas DHPG elicits mGluR1/5-induced LTD and increases p-eIF2a (Di Prisco et al., 2014;Trinh et al., 2014). Second, both genetic and pharmacological inhibition of p-eIF2amediated translational control facilitates the induction of LTP by blocking mGluR-LTD in VTA DA neurons . Finally, pharmacologically increasing p-eIF2a levels induces mGluR-LTD and prevents cocaine-induced LTP in VTA DA neurons .
Given that mGluR-LTD blocks the persistence of cocaine-induced LTP (Mameli et al., 2009), and our previous finding that p-eIF2a-mediated translational control regulates both forms of long-lasting plasticity , we examined whether the same translational control program regulates the shift from a relatively transient cocaine-induced LTP to a more persistent one in VTA DA neurons.

Results
Single or multiple injections of cocaine reduce eIF2a phosphorylation in the VTA Single or repeated exposure to cocaine in rodents induces LTP that lasts for five days, but decays to baseline after 10 days (Borgland et al., 2004;Kauer and Malenka, 2007;Mameli et al., 2009;Ungless et al., 2001). To examine the effects of multiple doses of cocaine on p-eIF2a in the VTA, we injected mice with either saline or cocaine (10 mg/kg) once a day for five days and measured the phosphorylation levels of eIF2a in the VTA (Figure 1a). Interestingly, both acute and repeated exposure to cocaine reduced p-eIF2a levels in the VTA (Figure 1b and c).

Cocaine elicits persistent LTP in mice with reduced eIF2a phosphorylation
We previously showed that the translational program controlled by p-eIF2a regulates the induction of LTP by a single injection of cocaine . To test whether the same process is involved in the persistence of LTP, we first employed Eif2s1 S/A heterozygous mutant mice, in which a single phosphorylation site in eIF2a (encoded by the Eif2s1 gene) at Serine-51 is replaced by alanine. These mice have reduced p-eIF2a levels in the VTA .
We injected control WT (Eif2s1 S/S ) and Eif2s1 S/A mice with saline or cocaine (10 mg/kg) once a day for three days and recorded glutamate-mediated excitatory postsynaptic currents (EPSCs) from VTA DA neurons in midbrain slices. We measured AMPAR/NMDAR ratios (recorded at +40 mV as we previously described Placzek et al., 2016) 5 and 14 days after cocaine withdrawal. As previously reported (Borgland et al., 2004), in control slices, cocaine (10 mg/kg) elicited an LTP that lasted five days (5d) but not 14d after withdrawal (Figure 2a and c). By contrast, in VTA slices from Eif2s1 S/A mice, LTP was greater at five days compared to WT controls and was still present and not significantly changed 14 days after the last cocaine injection (Figure 2b and c). Indeed, cocaine-induced LTP in the VTA persisted for at least 40 days in Eif2s1 S/A mice, which was as long as we continued the withdrawal period (Figure 2-figure supplement 1). Thus, a decrease in eIF2a phosphorylation facilitates the progression from transient to persistent cocaine-induced LTP.
Conditional reduction in eIF2a phosphorylation specifically in the VTA leads to persistent cocaine-induced LTP To assess the regional and temporal specificity of our findings, we used a new eIF2a transgenic mouse line (Eif2s1 A/A ;ftg) (Di Prisco et al., 2014) characterized by conditional expression of homozygous Eif2s1 (Ser51Ala) mutants upon cell-type specific recombination mediated by Cre recombinase (Back et al., 2009). Cre-mediated deletion of the exogenous Eif2s1 coding region coordinately induces the expression of green fluorescent protein (GFP) in a small population of VTA neurons, enabling us to manipulate p-eIF2a levels in single cells and record the consequences at the single-cell level ( Figure 2-figure supplement 2). In midbrain slices from these mice, we recorded from GFP-positive (GFP+) neurons (in which eIF2a cannot be phosphorylated) and GFP-negative (GFP-) control neurons. As expected, in GFP-(control) VTA DA neurons, cocaine-evoked LTP lasted five days, but returned to baseline by 14 days after withdrawal (Figure 2d  DA neurons in the VTA, we expect the effect of inputs from Cre-GFP expressing non-DA neurons to be minimal towards the persistence of LTP in GFP+ VTA DA neurons.
(d-f) Similarly, cocaine (10 mg/kg, i.p.) increased AMPAR/NMDAR ratio in GFP-positive cells (in which eIF2a cannot be phosphorylated), both at five and at 14 days post-injection (5d, p<0.001; 14d, p<0.001 , n = 5/6/6 saline/5d cocaine/14d cocaine, F 2,14 = 8.30), compared to control GFP-negative cells (5d, p<0.05, n = 5/8/5 saline/5d cocaine/14d cocaine, F 2,15 = 4.78) from Eif2s1 A/A ;ftg mice. DOI: 10.7554/eLife.17517.003 The following figure supplements are available for figure 2: receptors that lack GluR2 subunits and are consequently inwardly rectifying (manifested as smaller AMPAR EPSCs measured at positive holding potentials versus those measured at corresponding negative potentials). In addition, GluR2-lacking AMPARs are sensitive to blockade by intracellular polyamines (Iino et al., 1996). We therefore sought to determine whether the persistent cocaineinduced LTP that we observed in Eif2s1 S/A mice was accompanied by a similarly long-lasting shift in AMPAR subunit composition. By recording EPSCs in VTA DA neurons at À70, 0, and +40 mV to calculate the rectification index, we found that cocaine increased inward rectification in control VTA DA neurons (Eif2s1 S/S ; Figures 3a,c and e) and to a greater degree in slices from Eif2s1 S/A mice (Figures 3b,d and e) at five days post-injection. However, 14 days after cocaine treatment, EPSCs were non-rectifying in WT controls (Figures 3a,c and e), whereas Eif2s1 S/A mice retained their inwardly-rectifying properties (Figures 3b,d and e). Moreover, potentiated EPSCs in cocaine-treated Eif2s1 S/A mice measured at 14 days were sensitive to two blockers of GluR2-lacking AMPARs, Joro spider toxin (JST (Iino et al., 1996), 500 nM; Figure 3-figure supplement 1a) and 1-Naphthyl acetyl spermine (NASPM (Tsubokawa et al., 1995), 50 mM; Figure 3-figure supplement 1b). We conclude that postsynaptic insertion of GluR2-lacking AMPARs underlies the persistent LTP in VTA DA neurons with reduced p-eIF2a levels. Selective pharmacological reduction of p-eIF2a-mediated translational control facilitates persistent cocaine-induced LTP and inward rectification While phosphorylation of eIF2a inhibits general protein synthesis by inhibiting the guanine nucleotide exchange factor (GEF) eIF2B, it also increases the translation of specific mRNAs Ron and Harding, 2007). In our previous study, we used the small molecule inhibitor ISRIB, which promotes eIF2B activity resulting in selective reversal of p-eIF2a-mediated translational events (Sekine et al., 2015;Sidrauski et al., 2013Sidrauski et al., , 2015 and, consequently, makes adult mice more susceptible to the synaptic and behavioral effects of a sub-threshold dose of cocaine . To determine whether blocking p-eIF2a-mediated translation facilitates the switch from transient to persistent LTP, we administered WT mice with ISRIB (i.p. 2.5 mg/ kg) followed by cocaine (10 mg/kg) once a day for three days. Consistent with our findings with Eif2s1 S/A mice, cocaine induced both persistent LTP (Figure 4a-c) and inward rectification (Figure 4d-h) only in VTA DA neurons from ISRIB-injected mice. Given that p-eIF2a-mediated synthesis of oligophrenin-1 (OPHN1) blocks cocaine-induced LTP , we next asked whether a reduction in OPHN1 levels in the VTA would also prolong cocaine-induced LTP. To answer this question, we knocked-down OPHN1 selectively in the VTA with a specific shRNA (Ophn1-shRNA) (Nadif Kasri et al., 2011). As expected, in OPHN1-deficient, but not in control VTA DA neurons, cocaine-induced LTP persisted for at least 14 days ( Figure 4i). These data support the notion that cocaine-evoked LTP becomes persistent when p-eIF2a-mediated translation of OPHN1 is repressed.

Discussion
Drug addiction is a disorder involving maladaptive plasticity in key reward areas of the brain (Bowers et al., 2010;Hyman et al., 2006;Kauer and Malenka, 2007;Lüscher and Malenka, 2011). In the VTA, the origin of the mesolimbic dopamine system, single or multiple injections of cocaine evoke synaptic potentiation that persists for about five days. This LTP is generated by the insertion of GluR2-lacking AMPARs into the postsynaptic membrane. Two studies revealed that mGluR-LTD in VTA DA neurons blocks cocaine-evoked LTP in these same cells by replacing GluR2lacking calcium permeable AMPARs with GluR2-containing receptors (Bellone and Lüscher, 2006;Mameli et al., 2007). Unlike NMDA-induced LTD, mGluR-LTD depends on new protein synthesis (Lüscher and Huber, 2010). We have previously shown that eIF2a-mediated translational control is necessary and sufficient for mGluR-LTD in both the hippocampus and the VTA (Di Prisco et al., 2014;Huang et al., 2016). Translational control by the mammalian target of rapamycin complex 1 (i.e., mTORC1) Costa-Mattioli and Monteggia, 2013) has also been implicated in mGluR-LTD. Treatment with the mTORC1 inhibitor rapamycin blocks mGluR-LTD in the VTA (Mameli et al., 2007). While both translational control mechanisms (mTORC1 and p-eIF2a) could be required for mGluR-LTD, it is possible that rapamycin treatment could affect eIF2a-mediated translational control. During mGluR-LTD, activation of mGluRs by DHPG triggers the synthesis of OPHN1 (Nadif Kasri et al., 2011) in a p-eIF2a-dependent manner (Di Prisco et al., 2014). If rapamycin exerted its effects on mGluR-LTD by blocking OPHN1 translation, which is required to induce mGluR-LTD (Di Prisco et al., 2014;Nadif Kasri et al., 2011), these seemingly contradictory findings would be easily reconciled.
Inhibition of mGluR-LTD by genetic disruption of Homer1C-mGluR1 interaction or by treatment with an antagonist of group I mGluRs [1-aminoindan-1,5-dicarboxylic acid (AIDA)] has been shown to prolong cocaine-induced LTP from five days to seven days (Mameli et al., 2009). By suppressing p-eIF2a-mediated mGluR-LTD, we found that cocaine induces an LTP that lasted up to 40 days and was mediated by the insertion of GluR2-lacking receptors. Thus, like self-administration of cocaine (Chen et al., 2008), passive injections of cocaine in mice with reduced p-eIF2a-mediated translational control elicit a synaptic potentiation that lasts for weeks.
Recreational drug use is clinically separable from escalated drug use and compulsive drug-seeking behavior that characterize addiction. Our findings indicate that p-eIF2a-mediated translation prevents the progression from transient to persistent cocaine-induced LTP in VTA DA neurons, thus regulating synaptic potentiation that is believed to underlie compulsive drug seeking (Chen et al., 2008;Lüscher and Malenka, 2011;Mameli et al., 2009). In future studies, it will be interesting to examine whether this translational control mechanism regulates drug-seeking behaviors. In addition, it will important to assess whether cocaine-induced LTP could prime mice for enhanced behavioral output when challenged with a low dose of cocaine (or other drug of abuse) within the persistence window. Finally, it will be intriguing to study whether blocking p-eIF2a-mediated translation with ISRIB during extinction of cocaine memory is sufficient to prevent reinstatement.

Mice
All experiments were conducted using C57Bl/6J male and female mice. Eif2s1 S/A and Eif2s1 A/A ;ftg mice were previously described (Di Prisco et al., 2014). Mice were kept on a 12 hr/12 hr light/dark cycle (lights on at 7:00 am) and had access to food and water ad libitum. Animal care and experimental procedures were approved by the institutional animal care and use committee (IACUC) at Baylor College of Medicine, according to NIH Guidelines.
No statistical methods were used to predetermine sample sizes. All sample sizes meet the criteria for corresponding statistical tests-our sample sizes are similar to those reported in previous publications (Argilli et al., 2008;Bellone and Lüscher, 2006;Koo et al., 2012;Saal et al., 2003;Ungless et al., 2001).
Data were obtained with a MultiClamp 700B amplifier, digitized at 20 kHz with a Digidata 1440A, recorded by Clampex 10 and analyzed with Clampfit 10 software (Molecular Devices), and filtered online at 4 kHz with a low-pass Bessel filter. A 2 mV hyperpolarizing pulse was applied before each EPSC to evaluate the input (R i )and access resistances (R a ). Data were not included if R a was either unstable or greater than 25 MW, holding current was >200 pA, R i dropped >20% during the recording, or EPSC baselines changed by >10%. The representative traces illustrated in Figures are averages of 10-15 consecutive recorded sweeps.
After establishing a gigaohm seal (>2 GW) and recording stable spontaneous firing in cellattached, voltage clamp mode (À70 mV holding potential), cell phenotype was determined by measuring the width (>1.0 ms) of the cell-attached action potential (Chieng et al., 2011;Ford et al., 2006). AMPAR/NMDAR ratios were calculated as previously described (Ungless et al., 2001). Briefly, neurons were voltage-clamped at +40 mV until the holding current stabilized (at <200 pA), which usually occurred after a period of 5 to 15 min. A bipolar stimulating electrode placed 50-150 mm rostral to the lateral VTA was used to evoke stable monosynaptic EPSCs at 0.05 Hz. Picrotoxin (100 mM) was added to the recording ACSF to block GABA A R-mediated IPSCs. After recording the dual-component EPSC, bath-application of DL-AP5 (100 mM for 10 min) was used to remove the NMDAR component, which was then obtained by offline subtraction of the remaining AMPAR component from the original dual-component EPSC. The peak amplitudes of the isolated currents were used to calculate the AMPAR/NMDAR ratios. Rectification indices were calculated as the ratio of the chord conductance of evoked EPSCs at a negative holding potential (À70 mV) to the chord conductance obtained from recordings made at a positive holding potential (+40 mV) in the presence of 100 mM DL-AP5, as previously described (Bellone and Lüscher, 2006). Joro spider toxin (JST) was obtained from Sigma-Aldrich (St. Louis, MO), Picrotoxin and DL-AP5 were purchased from Tocris Bioscience, and all other reagents and experimental compounds were obtained from Sigma-Aldrich.

Virus injection
AAV5-Cre (Titer: 1.0e13GC/ml) was purchased from Vector Biolabs (Cat#7012, Philadelphia, PA); Lentiviral constructs expressing Ophn1 shRNA and scrambled shRNA were generously provided by Dr. Linda van Aelst (Nadif Kasri et al., 2011) and viruses were produced by Gene Vector Core Laboratory (Baylor College of Medicine). Viral injections were performed as previously described (Di Prisco et al., 2014). Briefly, mice were anaesthetized with isoflurane (2-3%) and viruses (1-2 ml/ site) were injected bilaterally at the rate of 0.1 ml/min, and an additional 10 min to allow for diffusion of viral particles. Injection coordinates targeting the VTA were as follows (with reference to bregma): À2.50 AP, ± 0.45 ML, À4.50 DV. The incision was sutured after injection and mice were returned to home cages. Mouse body weight and signs of illness were monitored until full recovery from surgery (~1 week). Drug treatment and experiments were all performed at least three weeks after viral injection.

Statistical analyses
All data are presented as mean ± s.e.m. Statistical analyses were performed using SigmaPlot (Systat Software). Data distribution normality and homogeneity of variance were assessed using the Shapiro-Wilk and Levene tests, respectively. The statistics were based on the two-sided Student's t test, or one-or two-way ANOVA with Tukey's HSD (or HSD for unequal sample sizes where appropriate) to correct for multiple post hoc comparisons. Within-groups variation is indicated using standard errors of the mean of each distribution, which are depicted in the graphs as error bars. p<0.05 was considered significant (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001).