BDNF Is a Negative Modulator of Morphine Action

See allHide authors and affiliations

Science  05 Oct 2012:
Vol. 338, Issue 6103, pp. 124-128
DOI: 10.1126/science.1222265

Regulating Opioid Responses

Different drugs of abuse are thought to highjack similar reward systems in the brain using common mechanisms. However, Koo et al. (p. 124) now observe that some of the neural mechanisms that regulate opiate reward can be both different and even opposite to those that regulate reward by stimulant drugs. While knockdown of brain-derived neurotrophic factor (BDNF) in the ventral tegmental area in mice antagonized the response to cocaine, the same manipulation strengthened the potential of opiates to increase dopamine neuron excitability. Optogenetic stimulation of dopaminergic terminals in the nucleus accumbens could counteract the effects of BDNF on morphine reward blockade.


Brain-derived neurotrophic factor (BDNF) is a key positive regulator of neural plasticity, promoting, for example, the actions of stimulant drugs of abuse such as cocaine. We discovered a surprising opposite role for BDNF in countering responses to chronic morphine exposure. The suppression of BDNF in the ventral tegmental area (VTA) enhanced the ability of morphine to increase dopamine (DA) neuron excitability and promote reward. In contrast, optical stimulation of VTA DA terminals in nucleus accumbens (NAc) completely reversed the suppressive effect of BDNF on morphine reward. Furthermore, we identified numerous genes in the NAc, a major target region of VTA DA neurons, whose regulation by BDNF in the context of chronic morphine exposure mediated this counteractive function. These findings provide insight into the molecular basis of morphine-induced neuroadaptations in the brain’s reward circuitry.

Brain-derived neurotrophic factor (BDNF) is a positive modulator of many forms of neural plasticity throughout the adult nervous system (1, 2). In the context of drug addiction, BDNF is best characterized for its role in promoting the neural and behavioral plasticity induced by cocaine or other stimulants via actions on the mesolimbic dopamine (DA) system, a key reward circuit in the brain, where the BDNF pathway is engaged in a feed-forward loop that promotes further actions of stimulant drugs (38).

Opiates also act on the ventral tegmental area (VTA) and nucleus accumbens (NAc) to produce reward acutely and addiction chronically. However, there are differences in the cellular actions of opiates versus stimulants on this reward circuit. Stimulants promote DA signaling in the NAc primarily by acting on DA terminals in this region to increase extracellular DA levels. In contrast, opiates promote DA signaling to the NAc by inhibiting local γ-aminobutyric acid (GABA) interneurons in the VTA, which then disinhibit (activate) VTA DA neurons (9); opiates also act via DA-independent mechanisms (10). Chronic opiates induce some unique biochemical and morphological alterations in the VTA. Although the effect of opiates on BDNF expression in the VTA is inconsistent (1012), opiates down-regulate intracellular BDNF signaling cascades and reduce the soma size of VTA DA neurons (1316), effects not seen with stimulants. Some of these biochemical and morphological changes in the VTA are reversed by direct administration of BDNF into this brain region (13, 16). This suggests that, opposite to the situation for cocaine and other stimulants, opiate and BDNF actions might converge by producing counteractive effects on VTA DA neurons. This led us to hypothesize an antagonistic role for endogenous BDNF-TrkB signaling in modulating adaptive responses of the VTA-NAc pathway to chronic opiate exposure.

We first demonstrated that chronic morphine, whether given to mice by subcutaneous pellets or intermittent intraperitoneal injections, decreased BDNF expression in the VTA (fig. S1). Next, we examined the role of BDNF-TrkB signaling in the VTA-NAc in morphine action by performing morphine-conditioned place preference (CPP), which provides an indirect measure of drug reward. Morphine doses were selected on the basis of an initial dose-response analysis (fig. S2). Infusion of an adeno-associated virus (AAV) vector encoding Cre recombinase fused to green fluorescent protein (GFP) (AAV-CreGFP) into the VTA of mice homozygous for bdnf or trkB genes flanked by loxP sites [floxed BDNF (flBDNF) or floxed TrkB (flTrkB) mice] produced highly localized CreGFP expression in this brain region (fig. S3A). This induced a 40 to 60% reduction of BDNF or TrkB mRNA levels in the VTA, as compared with control animals injected with AAV-GFP (fig. S3, B and C). Localized knockdown of BDNF enhanced the rewarding effect of morphine at both subthreshold doses [5 mg per kilogram of body weight (mg/kg); fig. S3D] and higher doses (15 mg/kg; Fig. 1A) as compared with AAV-GFP controls. Equivalent effects were seen for local VTA knockdown of TrkB (Fig. 1B, 15 mg/kg). There were no differences in baseline preference scores before conditioning among the flBDNF and flTrkB groups.

Fig. 1

Effects of BDNF-TrkB signaling within the VTA-NAc on morphine reward. Localized knockout (KO) of BDNF (A) or TrkB (B) from VTA neurons enhances morphine CPP [15 mg/kg, subcutaneous (sc)]. Student’s t test, *P < 0.05, n = 8 to 12 mice. (C) DAT-Cre/flTrkB (TrkBlx/lx;DATcre/wt) mice also displayed enhanced morphine CPP [10 mg/kg, intraperitoneal (ip)]. One-way analysis of variance (ANOVA), Fisher's protected least significant difference (PLSD) post-hoc test, *P < 0.05 compared to controls; #P < 0.05 compared with TrkBlx/lx;DATcre/wt mice, n = 6 to 11 mice. (D) A single infusion of BDNF into the VTA (0.25 μg per side) suppressed morphine CPP (15 mg/kg, sc). PBS, phosphate-buffered saline. t test, *P < 0.05, n = 7 or 8 mice. (E) Localized TrkB KO in the NAc and (F) intra-NAc BDNF infusion (1.0 μg per side) had no effect on morphine CPP (15 mg/kg, sc), n = 8 or 9 mice.

Because the viral-mediated knockdown procedure affects both DA and non-DA VTA neurons, we used a complementary approach to knock out TrkB selectively from DA neurons by crossing flTrkB mice with DA transporter (DAT)–Cre mice (TrkBlx/lx;DATcre/wt) (17). Selective ablation of TrkB from DA neurons in TrkBlx/lx;DATcre/wt mice increased morphine reward (Fig. 1C). There were no differences in baseline or morphine CPP among several control groups examined, which included wild-type mice (TrkBw/w;DATwt/wt), floxed control mice (TrkBlx/lx;DATwt/wt), and Cre control mice (TrkBw/w;DATcre/wt), indicating that the increase in morphine reward seen upon selective TrkB ablation in DA neurons does not result from allele-specific effects. In contrast, a single intra-VTA infusion of BDNF (0.25 μg per side) decreased morphine CPP as compared with vehicle-infused control animals (Fig. 1D).

BDNF synthesized in VTA DA neurons can undergo anterograde transport and release in the NAc to activate TrkB receptors on NAc neurons (18). We therefore tested the effect on morphine reward of localized deletion of TrkB receptors in the NAc of flTrkB mice and of intra-NAc infusions of BDNF (1.0 μg per side). In contrast to the VTA, TrkB knockdown (Fig. 1E) and BDNF infusion (Fig. 1F) in the NAc had no effect on morphine CPP. It is thus local BDNF signaling in VTA DA neurons that is responsible for the regulation of morphine reward.

We have recently shown that chronic morphine decreases the expression of certain K+ channels in the VTA, such as kcnab2, kcnj2, and girk3, and that such adaptations are associated with increased excitability of DA neurons (14). Viral-mediated BDNF knockdown in the VTA similarly suppressed mRNA levels of these and some additional K+ channels (fig. S4A). These findings raise the possibility that morphine increases VTA DA neuron excitability via down-regulation of BDNF and the subsequent reduction in K+ channel expression. To test this hypothesis, we obtained extracellular single-unit recordings from VTA DA neurons in three groups of anesthetized mice: controls, chronic morphine, and chronic morphine+intra-VTA BDNF-infused (Fig. 2A). Consistent with previous ex vivo findings from brain slices (14), morphine increased the in vivo spontaneous firing rates of VTA DA neurons by 44%. Intra-VTA infusion of BDNF normalized this morphine-induced firing rate increase (Fig. 2B). Analysis of burst phasic firing, which substantially increases DA release as compared to single-spike tonic firings (19, 20), revealed that overall bursting events were increased by chronic morphine and restored by intra-VTA BDNF (Fig. 2, C and D; fig. S4, B and C). Conversely, localized VTA BDNF knockdown alone (in morphine-naïve mice) increased the spontaneous burst firing of DA neurons (Fig. 2, G to I).

Fig. 2

Regulation of VTA DA neuron excitability by morphine and BDNF. (A) Sample traces of in vivo recordings from VTA DA neurons from control (top), morphine-treated (middle), and BDNF+morphine–treated mice (bottom). (B to E) Morphine (25-mg pellet, sc; animals were analyzed 48 hours later) increases (B) basal firing rate, (C) burst firing rate, and (D) burst duration in VTA DA neurons, which were normalized by intra-VTA infusion of BDNF (0.25 μg per side). One-way ANOVA, Fisher's PLSD test, *P < 0.05, ***P < 0.001 compared with control; ##P < 0.01, ###P < 0.001 compared with the morphine group. n = 4 to 6 mice. (E) Sample traces of K+ conductance recorded from VTA DA neurons in brain slices from control, morphine-treated, and BDNF+morphine–treated mice. (F) Morphine treatment as in (A) significantly decreased both peak and sustained phases of K+ currents in VTA DA neurons, an effect that was reversed by BDNF. Two-way ANOVA, Fisher's PLSD test, *P < 0.05, **P < 0.01, ***P < 0.001 compared with control; #P < 0.05, ##P < 0.01, ###P < 0.001 compared with the morphine group. n = 5 to 9 mice. Localized BDNF KO from VTA increases (G) basal firing rate, (H) burst firing rate, and (I) burst duration in VTA DA neurons. t test, *P < 0.05, **P < 0.01 compared with AAV-GFP controls, n = 7 mice.

Next, we studied possible ionic mechanisms underlying these changes, using standard whole-cell voltage-clamp recordings. Different components (peak and sustained) of K+ currents in VTA DA neurons were reduced by morphine treatment, which were blocked by intra-VTA BDNF (Fig. 2, E and F). These data demonstrate that morphine and BDNF differentially regulate K+ currents in these neurons, which is consistent with our molecular findings above. The ability of chronic morphine to excite VTA DA neurons may be mediated via decreased AKT signaling (14), which would be an expected downstream consequence of the withdrawal of BDNF support. Prior work suggested as well that down-regulation of BDNF signaling in the VTA would excite VTA DA neurons further by reducing GABAA receptor responses, also downstream of reduced AKT signaling (21). Our findings demonstrate that BDNF additionally controls the intrinsic excitability of VTA DA neurons via altered K+ channel expression, and thereby opposes morphine-induced DA neuron excitability through a homeostatic scaling mechanism (22, 23).

Given these direct links between VTA BDNF expression and VTA DA neuron excitability in morphine action, we next determined whether BDNF-regulated activity of VTA DA neurons is important for BDNF’s influence on behavioral responses to morphine. We stereotaxically delivered AAV-ChR2 (channel rhodopsin)–EYFP (enhanced yellow fluoresceent protein) or AAV-EYFP into the mouse VTA as described (24). Two to 3 weeks later, when AAV expression was maximal, we infused BDNF into the VTA and implanted cannulae in the NAc for optical fiber placement (Fig. 3A and fig. S5A). Animals were studied 1 week later. 86% of ChR2-EYFP–positive cells in the VTA colocalized with TH, a marker of DA neurons (fig. S5B). ChR2-EYFP immunoreactivity in the NAc colocalized with DAT, a marker of DA nerve terminals (Fig. 3, B to D), but not with GAD67 (Fig. 3, E to G) or VGLUT2 (Fig. 3, H to J), markers of GABA or glutamate terminals, respectively, showing selective expression of ChR2-EYFP in DA nerve terminals in the NAc. Mice that expressed AAV-ChR2-EYFP or AAV-EYFP in the VTA were studied in the morphine CPP model by conditioning them with subthreshold morphine doses (10 mg/kg) plus 20-Hz phasic pulses delivered to the NAc in one chamber, and with saline and no light in the opposite chamber. In non–BDNF-infused animals, this optogenetic protocol induced significant morphine CPP in mice expressing ChR2-EYFP but not EYFP alone (Fig. 3K). Such stimulation also completely prevented the inhibitory effect of intra-VTA BDNF infusion on morphine CPP (Fig. 3K, compare to Fig. 1D and fig. S6A). In this model, light stimulation of DA terminals in NAc, at 20 Hz or several other frequencies, in the absence of morphine did not induce a CPP regardless of intra-VTA BDNF infusion (fig. S6, C to E). The effect of light stimulation was mediated by D1 DA receptors in the NAc, because intra-NAc injection of a D1 receptor antagonist (SCH 23390, 1 μg), at a dose known to be behaviorally active (25, 26), completely blocked the ability of light stimulation to enhance morphine CPP regardless of intra-VTA BDNF infusion (Fig. 3L). In contrast, intra-NAc injection of behaviorally active doses (2729) of a D2 (eticlopride, 1 or 4 μg) or glutamate [6,7-nitroquinoxaline-2,3-dione (DNQX), 1 or 4 μg; or 6-nitro-7-sulfamoylbenzo[f]quinoxaline-2,3-dione (NBQX), 400 ng] receptor antagonist had no effect (fig. S7). These data demonstrate that BDNF impairment of morphine reward can be rescued by phasic stimulation of the VTA-NAc DA pathway, specifically through D1 receptors in the NAc. This is consistent with evidence showing that DA released by burst-like phasic firing selectively binds to D1 receptors (19, 30). In contrast to the NAc, optical stimulation of DA terminals in the medial prefrontal cortex had no effect on morphine CPP (fig. S8).

Fig. 3

Modulation of morphine reward by optogenetic activation of DA terminals in the NAc. (A) Experimental model for optogenetic stimulation during morphine CPP. Mice were conditioned to a morphine/light chamber and a saline/no-light chamber for 30 min. 470-nm phasic light pulses (20 Hz, five pulses, 40 ms duration) were delivered during the 30-min conditioning session. (B to J) Immunostaining for ChR2-EYFP [(B), (E), and (H), green], DAT [(C), red], GAD67 [(F), red], and VGLUT2 [(I), red] in the NAc. (D), (G), and (J) Confocal microscopy shows that ChR2-EYFP puncta in NAc co-label for DAT, but not GAD67 or VGLUT2. Scale bar, 10 μm. (K) In vivo optogenetic stimulation of VTA DA nerve terminals in the NAc enhances morphine reward (10 mg/kg, ip) and prevents VTA BDNF-induced impairment of morphine reward, (L) whereas D1 receptor antagonism (SCH 23390, 1 μg) blocks light potentiation, and VTA BDNF-induced impairment, of morphine reward. t test, **P < 0.01, **P < 0.001, n = 7 to 11 mice.

We investigated the downstream consequences of VTA BDNF and chronic morphine on gene expression in the NAc. We performed microarray analysis on the NAc from mice in which VTA BDNF was virally knocked down, with half of the animals then treated chronically with morphine. We identified clusters of NAc genes that are regulated by morphine or by knockdown of VTA BDNF and analyzed interactive effects between the two to investigate the molecular mechanisms underlying BDNF regulation of morphine responses. Three main categories of genes were identified (table S1): category A, genes regulated by morphine, in which that regulation is lost upon knockdown of VTA BDNF (Fig. 4A); category B, genes whose regulation by morphine was uncovered upon knockdown of VTA BDNF (Fig. 4A); and category C, genes regulated by morphine regardless of VTA BDNF knockdown (Fig. 4B) (see table S2 for complete gene lists). Among the genes significantly regulated in the NAc under these conditions are several that have previously been implicated in morphine action, such as zfp40, xdh, nt5e, sult1a1, gadd45g, rbm3, and zbtb16 (3134). We also found a significantly greater transcriptional effect of chronic morphine in the NAc of mice lacking BDNF in the VTA: About three times more genes (151 versus 46) were regulated by morphine in VTA BDNF knockdown mice than in control mice (fig. S9, A to C). Similarly, about five times more genes (185 versus 39) in the NAc were regulated upon knockdown of BDNF in the VTA in morphine-treated mice than in sham mice (fig. S9, D to F) (see table S3 for complete gene lists). These findings extend our behavioral and electrophysiological evidence that BDNF in the VTA antagonizes chronic morphine actions on the VTA-NAc circuit.

Fig. 4

Morphine-regulated NAc gene expression after VTA BDNF deletion: Identification of NAc mediators of BDNF-morphine interactions. Microarray analysis was performed on the NAc of sham- and morphine-pelleted mice under control or VTA BDNF knockdown conditions. (A) Heat map of up-regulated (red) or down-regulated (green) NAc genes upon knockdown of VTA BDNF. (B) Heat map of up- or down-regulated NAc genes by morphine regardless of knockdown of VTA BDNF. (C) Venn diagrams of genes that were regulated by morphine (red) or by knockdown of VTA BDNF (blue), and of genes that were regulated by morphine and knockdown of VTA BDNF in an interactive manner (green). (D to G) Alterations of sox11 [(D) and (E)] and gadd45g [(F) and (G)] expression in the NAc from a heat map of microarray analysis [(D) and (F)] and RT-PCR (RT-PCR) validation [(E) and (G)]. One-way ANOVA for RT-PCR validation, Fisher's PLSD test, τP < 0.1, *P < 0.05, ***P < 0.001 compared to sham+AAV-GFP controls; $P < 0.1, #P < 0.05, ###P < 0.001 compared to sham+AAV-CreGFP mice, n = 9 to 12 mice. (H) Reduction of sox11 expression using AAV-shRNA-Sox11 increases morphine CPP (10 mg/kg, sc). Fisher's PLSD test, *P < 0.05 compared with AAV-GFP controls; #P < 0.05 compared with AAV-scrambled controls, n = 11 or 12 mice. (I) Sox11 overexpression in NAc using HSV-Sox11 decreases morphine CPP (15 mg/kg, sc). t test, *P < 0.05, n = 8 mice. (J) Enhancement of morphine reward (15 mg/kg, sc) induced by knockdown of VTA TrkB is counteracted by sox11 overexpression in the NAc. Fisher's PLSD test, *P < 0.05 compared with HSV-tomato (TMT) (NAc)+AAV-GFP (VTA) controls; #P < 0.05 compared with HSV-TMT (NAc)+AAV-CreGFP (VTA) mice. (K) Enhancement of morphine reward (12.5 mg/kg, sc) induced by knockdown of VTA TrkB is further enhanced by gadd45g overexpression in the NAc. Fisher's PLSD test, *P < 0.05 compared to HSV-GFP+AAV-GFP controls; #P < 0.05, ###P < 0.001 compared to HSV-Gadd45g+AAV-CreGFP.

We selected two genes, sox11 and gadd45g, from categories A and B, respectively, for further study and validated their expression patterns using reverse transcription polymerase chain reaction (RT-PCR) analysis (Fig. 4, D to G). SOX11 is a transcription factor involved in embryonic neurogenesis and tissue remodeling (35). The function of the gene in the adult nervous system remains unknown, although its regulation was observed in a previous microarray study on the NAc (36). We found that sox11 gene expression levels in the NAc were induced by chronic morphine and that this increase was prevented by the deletion of BDNF in the VTA (Fig. 4, D and E).

To directly test whether alterations in sox11 expression in the NAc influence morphine reward, we generated an AAV vector that expresses a short-hairpin RNA (shRNA) against sox11 to knock it down in the NAc selectively and a herpes simplex virus (HSV) vector to overexpress sox11 in this region (fig. S10A). After first validating the AAV-shRNA-Sox11 vector in cultured Neuro2A cells (fig. S10, B and C), we demonstrated that intra-NAc infusion of this vector reduced sox11 expression in the NAc (fig. S10D). In contrast, intra-NAc infusion of HSV-Sox11 robustly augmented sox11 mRNA levels in this region (fig. S10E). We observed that AAV-shRNA-Sox11 in the NAc increased morphine reward: A subthreshold dose of morphine (10 mg/kg) induced significant CPP, an effect not seen in animals treated with control AAV-GFP or AAV-scrambled shRNA vectors (Fig. 4H). Conversely, HSV-Sox11 in the NAc decreased CPP to a higher dose of morphine (15 mg/kg) as compared to HSV-tomato–infused control animals (Fig. 4I). Furthermore, the ability of locally knocking down BDNF-TrkB signaling in the VTA to enhance morphine’s rewarding effects was completely normalized upon HSV-mediated sox11 overexpression in the NAc (Fig. 4J). No changes were observed in baseline levels of place preference in these experiments.

Another gene implicated in these interactions by our microarray data is gadd45g, a stress-responsive immediate early gene, which is involved in DNA repair, cell growth arrest, and apoptosis (37). Here, gadd45g gene expression levels were more robustly and consistently suppressed in the NAc by chronic morphine in mice with deletion of VTA BDNF (Fig. 4, F and G). We then tested whether such regulation of gadd45g expression in the NAc influences the rewarding effects of morphine using HSV-Gadd45g that we developed and validated for gadd45g overexpression (fig. S11). HSV-Gadd45g infusion into the NAc of normal mice did not affect the rewarding actions of a moderate dose of morphine (12.5 mg/kg). However, gadd45g overexpression in the NAc of mice with local VTA knockdown of TrkB enhanced morphine CPP as compared to HSV-GFP treated mice (Fig. 4K). No changes were observed in baseline levels of place preference.

We observed the opposite effect of BDNF-TrkB signaling in the VTA-NAc on morphine reward to that seen in earlier observations with cocaine and other stimulants. Prior work has shown that knockdown of BDNF from the VTA, or knockdown of BDNF or TrkB from the NAc, antagonizes behavioral responses to cocaine, whereas knockdown of TrkB in the VTA has no effect (3). These findings identified the NAc as the primary site of action of BDNF-TrkB signaling in regulating cocaine reward. In contrast, we show here that the primary site of action of BDNF-TrkB signaling in regulating morphine reward is the VTA, because knockdown of either BDNF or TrkB in the VTA promotes behavioral responses to morphine, whereas knockdown of TrkB in the NAc and BDNF administration into the NAc were without effect. On the other hand, our findings that optical stimulation of VTA DA nerve terminals in the NAc completely reverses the ability of BDNF, acting in the VTA, to impair morphine reward demonstrate that BDNF’s influence on VTA DA neuron excitability is responsible for its behavioral effects reported here (fig. S12). These optogenetic experiments also identify the NAc as the key neural site where VTA BDNF’s influence on morphine reward is ultimately mediated, and we showed that this occurs via DA activation of D1 receptors on NAc neurons. For this reason, we analyzed morphine-induced changes in gene expression in the NAc as a function of VTA BDNF and demonstrated the importance of two genes, sox11 and gadd45g, among numerous other putative genes regulated in a similar fashion, in behavioral responses to morphine (fig. S12). The results of the present study differ from those in an earlier report (10), which found that exogenous BDNF increases opiate reward in rats by altering GABAA receptor function in VTA GABAergic neurons (fig. S12). The different findings could be due to the different species, drug treatment regimens, or behavioral models used.

Consistent with the differences in BDNF’s influence on cocaine versus morphine action are the very different ways in which the drugs initially affect the VTA-NAc pathway. In addition, although opiate and stimulant drugs of abuse induce many common molecular and cellular adaptations in both the VTA and NAc (38), some notable differences are seen as well with respect to synaptic plasticity (39, 40) and to VTA BDNF expression (fig. S1B). Further studies are now needed to directly investigate whether the opposite interactions between BDNF and stimulants versus opiates are related to these adaptations. Given the substance-specific features of drug addiction syndromes, it will be important to further explore the downstream functional consequences of the BDNF-stimulant feed-forward loop versus the BDNF-opiate negative feedback loop in addiction and to study how they influence polydrug use.

Supplementary Materials

Materials and Methods

Figs. S1 to S12

Tables S1 to S3

References (4153)

References and Notes

  1. Acknowledgments: We thank R. E. Burger-Caplan and N. Mensah for excellent technical assistance. This work was supported by grants from the National Institute on Drug Abuse (E.J.N. and M.S.M.-R.) and a Rubicon Grant from the Dutch Scientific Organization (C.S.L.).

Stay Connected to Science

Navigate This Article