4. Results
4.1 Behavioral model of heroin self-administration
We trained mice to intravenously self-administer heroin under a fixed-ratio 1 schedule (Figure 1a, see Experimental Procedures) for 12 days (Figure 1b). The dose was decreased from 50 to 25
mg/kg/infusion during acquisition (Figure 1c). The animals quickly learned to discriminate between an active and an inactive lever (after 6 days of training, 144.9 ± 26 active lever presses versus 8.28 ± 2.5 inactive ones. After 12 days it was 283.4 ± 28 versus 20.9 ± 9.3. Figure 1d-f) and readily reached a robust number of heroin infusions (50.6 ± 6.9 infusions after 6 days of training during a 6h session, Figure 1g). After 6 days, we halved the dose but allowed 150 infusions. Mice then increased their number of infusions to finally take almost the maximum allowed (138.1 ± 5.1 infusions after 12 days of training) in about two to three hours at the end of the acquisition (Figure 1g). After 30d of withdrawal, mice were brought back into the apparatus in the absence of heroin injections and significantly differentiated between active and inactive lever during cue (Figure 1h, 1i). Taken together this experiment shows that heroin was highly reinforcing and induced seeking behavior, used as a model for relapse 107,123.
Figure 1. Behavioral model of heroin self-administration. a, Schematic of experiment for d-e. b, Schedule of experiment for d–e. c, Detail of the sequence of events following a press on the active lever. An active lever press triggers the illumination of a cue-light just above the lever and an infusion of heroin. The infusion is followed by a time-out period (10 or 40s depending on the session) where heroin is no longer available despite presses on the correct lever. d, Raster plot for infusions and
inactive lever presses as a function of time during acquisition of daily self-administration session of 6h for a mouse that self-administered saline (top) or heroin (bottom). e, Mean total lever presses (top) and infusions (bottom) during the acquisition phase of saline (n=10) or heroin (n=14) self-administration. Infusion rate was very robust in mice which self-administered heroin and animals quickly learned to discriminate between the active and inactive lever (number of heroin infusions versus number of saline infusions, number of active versus inactive lever presses for heroin treatment and number of active lever presses for heroin versus active lever presses for saline, two-way repeated measures ANOVA and multiple comparison post-hoc Bonferonni correction, *P<0.05,
**P<0.005, ****P<0.0001). f, Raster plot for active and inactive lever presses as a function of time during cue-associated seeking test at day 30 of withdrawal for a mouse that self-administered either saline (top) or heroin (bottom) during acquisition phase. g, Mean total lever presses at 30 days of withdrawal for mice trained for heroin (n=6) self-administration. After 30 days of forced withdrawal seeking was robust in mice which self-administered heroin (active versus inactive lever, two-way repeated measures ANOVA and multiple comparison post-hoc Bonferonni correction, *P<0.05). Error bars, s.e.m.
4.2 Heroin increases NAc DA levels and VTA DA activity in drug-naïve animals
To test whether heroin, when administered by drug-naïve mice, causes a DA increase in the NAc as observed with other drugs of abuse 102,220, we measured the fluorescence of a genetically encoded DA sensor (D-light) in freely moving mice with fiber photometry. D-light is a fusion protein of the human Dopamine D1 (DRD1) where the third intracellular loop has been replaced with the circularly permutated GFP module from GCaMP6 and then emits green fluorescence upon dopamine binding (Patriarchi et al. , personal communication, March 2018). DAT-Cre + animals were also injected with Cre-dependent Chrimson virus in the VTA (Figure 2a, 2b). Both optogenetic stimulation of VTA DA cell bodies, as well as consumption of a natural reward (lipofundin 4%) reliably increased Dlight
fluorescence (Figure 2c, 2d). Importantly, within less than a minute of the intraperitoneal heroin infusion, fluorescence increased and reached the maximum after three minutes (Figure 2e, 2f, dF/F saline: -0.008 +- 0.007, dF/F heroin: 0.133 +- 0.03, p = 0.0062, t(6) = 4.117, Paired Students T-Test ,n = 7). This experiment demonstrates that already a first injection of heroin leads to an increase of DA in the NAc within minutes.
We then expressed a Cre-dependant GCaMP6m in DAT-Cre + mice to monitor the activity of VTA DA neurons after an i.v heroin injection, again using fiber photometry in freely moving mice 219,222,223. (Figure 2g, 2h). A single dose of heroin increased the fluorescence within minutes (Figure 2i, 2j, dF/F saline compared to baseline: -0.0063 +- 0.016, dF/F heroin compared to baseline: 0.22 +- 0.085, n = 11, p = 0.001, Wilcoxon signed rank test). A kinetics similar to the DA transient suggests a tight correlation between neural activity and DA levels.
Figure 2. Heroin increases NAc DA levels and VTA DA activity in drug-naive animals. a, Schematic of the experiment for b-f; b, Top, medial NAc shell of DAT-Cre+ mice was bilaterally injected with the DRD1-based DA sensor (dLight) while the red light–drivable channelrhodopsin Chrimson was unilaterally injected in the VTA (right). c, Dlight-mediated fluorescence changes following optogenetic activation of VTA DA neurons by Chrimson (mean of n=3 ). d, D-light-mediated
fluorescence change following onset of consumption of a natural reward (5% lipofundin). Black line indicates group mean and grey lines indicate individual responses (n=4). Both optogenetic
stimulation of VTA DA cell bodies, as well as consumption of a natural reward reliably increased dLight fluorescence. e, Example trace from single animal, showing D-light-mediated fluorescence change in the NAc following intraperitoneal heroin (8mg/kg) or saline injections. Tick mark indicates injection. f, Average fluorescence after heroin or saline treatment compared to pre-infusion baseline (n=7). Intraperitoneal injection of heroin significantly increased fluorescence signals (as compared to control injections, p= 0.0062, t(6) = 4.117, Paired Students T-Test). g, Schematic of the experiment for h-j; h, Left, VTA of DAT-Cre+ mice was bilaterally injected with the floxed version of the calcium indicator GCAMP6m. Right, coronal confocal images of infected VTA. i, Average GCaMP6m fluorescence in VTA DA neurons following first intravenous infusion of heroin (100 μg/kg/inf) or saline. Black tick marks indicate infusion onset. j, Average fluorescence after heroin or saline treatment compared to pre-infusion baseline (n=11). Calcium transients significantly increased after heroin infusions ( dF/F saline compared to baseline: -0.0063 ± 0.016, dF/F heroin compared to baseline: 0.22 ± 0.085, n = 11, p = 0.001, Wilcoxon signed rank test, **P<0.005). Error bars, s.e.m.
4.3 Ventromedial VTA DA neurons projecting to the NAc drive heroin-induced increased level in DA
To map the neurons activated, we prepared brain slices staining for the immediate early gene cFos after the very first heroin self-administration session. cFos positive cells that were also
TH-immunoreactive were found most prominently in the ventromedial part of the VTA (Figure 3b-d, 21.5
± 3.43% TH+/cFos+ in heroin group, 1.4 ± 0.93% TH+/cFos+ in saline group). Taken together, in drug-naive animals, DA in the NAc increases after a first injection of heroin, which is most likely caused by an enhanced activity of a subpopulation of VTA DA neurons.
To identify the projections of VTA DA neurons activated after heroin self-administration we designed an experiment combining injection of CTB tracers in the lateral and medial NAc shell (Figure 3e, 3f), for retrograde labeling of neurons, with staining for the immediate early gene cFos after a single heroin self-administration session. As already described in a previous study 218 we first observed that VTA DA neurons are projecting to separate NAc shell subdivisions. Indeed we found that 52.6 ± 6.7%
of the cells of the VTA were red-CTB-expressing, i.e neurons projecting to the lateral NAc (NAcLat) shell, 21 ± 4.4% were grey-CTB-expressing, i.e projecting to the medial NAc (NAcMed) shell and only 2.83 ± 0.83% were projecting to both NAc subdivisions. The remaining 23.6 ± 4.1% were thus not NAc-projecting cells (Figure 3f, 3g). We then again prepared brain slices staining for the immediate
early gene cFos after the very first heroin self-administration session. cFos positive cells that were NAcLat seedings represent 45.5 ± 7.5 % of all the cFos positive cells while 22.3 ± 4.2% of the cFos positive cells were NAcMed seedings. A small fraction of cells (6.6 ± 2.73%) expressed the three colors and finally 25 ± 7.05% of the cells were only cFos positive. Taken together these data indicate that in drug naïve animals, the increased level of DA observed after the very first session of heroin self-administration is driven by DA neurons located in the ventromedial part of the VTA and projecting mostly to the lateral NAc shell.
Figure 3. Ventromedial VTA DA neurons projecting to the NAc drive heroin-induced increased level in DA. a, Schematic of experiment for b-d; b, TH (left, red),cFos (middle, green) staining of VTA DA neurons and co-localization of TH- and cFos- expressing neurons (right) after one day of either saline (top) or heroin (bottom) administration. Mice were perfused 60 min after the end of the self-administration session. Cell nuclei are stained with Hoechst (not shown). D, dorsal; L, lateral. c, Location within the VTA of histologically identified DA neurons expressing cFos after one day of heroin self-administration. Each color of the markers represents one animal. d, Quantification of the TH positive VTA DA neurons also expressing cFos after one day of saline or heroin self-administration (saline: 2102 cells from four mice, heroin: 1902 cells from four mice). First exposure to heroin
treatment significantly increased cFos expression in VTA DA neurons (cFos expression after saline and heroin in TH positives neurons, one-way ANOVA and multiple comparison post-hoc Bonferonni correction, **P<0.005). e, Schematic of experiment for f-h. The retrograde tracers CTB, conjugated to either the fluorescent dye AlexaFluor ® 555 (red) or AlexaFluor® 488 (green) were injected in the medial NAc shell or the lateral one, respectively. f, Coronal image of infected lateral and medial NAc shell. g, coronal confocal image of NAc seedings in the VTA. NAcMed seedings (left), NAc Lat seedings (middle) and both populations (right) in the VTA. h, Quantification of the of the NAc seedings
subpopulations in the VTA (3475 cells from 5 animals). NAcLat projecting neurons and NAcMed ones are well segregated in the VTA (52.6% of cells from the lateral shell, 21% from the medial one, 2.84%
from both NAc subnuclei and 23.7% of non-NAc-projecting neurons. One-way ANOVA and multiple comparison post-hoc Bonferonni correction, * p<0.5, **p<0.005, p<0.0005***, p<0.0001****). i, Schematic of experiment for j-l. The retrograde tracers CTB, conjugated to either the fluorescent dye AlexaFluor ® 555 (red) or AlexaFluor® 647 (violet) were injected in the medial NAc shell or the lateral one, respectively. In addition a catheter implantation was performed (see methods) in order to allow heroin self-administration. j, Coronal image of infected lateral and medial NAc shell. k, NAcMed seedings (left, red), NAcLat seedings (violet), cFos (green) and merge (right) in VTA neurons after one day of heroin self-administration. Mice were perfused 60mn after the end of the self-administration session and cell nuclei have also been stained with Hoechst (not shown). l, Quantification of the cFos positive VTA neurons also expression red or grey CTB (1200 cells from 5 animals). 45.5% of cFos positive cells are also CTB 555 positive, 22.3% are CTB 647 positive, 6.6% expressed the three colors and finally 25% are only cFos postitive (One way ANOVA and multiple comparison post-hoc
Bonferonni correction, **p<0.005). Error bars, s.e.m.
4.4 Chemogenetic inhibition of VTA DA neurons decreases heroin self-administration
To probe for causalities, we tested whether inhibiting VTA DA neurons during the initial sessions would impact on heroin SA. To this end, we injected DAT-Cre mice that expressed hSyn-DIO-hM4D(Gi) in VTA DA neurons (Figure 4a, 4b) with CNO 1 h prior to a heroin self-administration session. This regimen was efficient to hyperpolarize DA neurons in acute midbrain slices 224. Chemogenetically silencing VTA DA neurons in animals where self-administration was well
established significantly decreased the number of active lever presses and ensuing heroin infusions (Figure 4c-e, from 223 +- 60 LP to obtain 111 +- 25 infusions to 22.8 +- 9 LP to get 15.4 +- 6 infusions after 4 days of treatment). To establish the necessity of VTA DA signaling during the very early acquisition session, we silenced VTA DA neurons from the first session, which prevented the acquisition of the behavior (Figure 4f-h). After CNO was stopped, the mice quickly acquired the task and reached a number of lever presses and infusion close to the control animals (DAT-Cre + mice on session 9 pressed 304.5 +- 57.5 times to obtain 120.5 +- 13.9 infusions versus control mice on session 9 which pressed 178.8 +- 40.2 times to obtain 107 +- 19.5 heroin infusions). CNO had no effect on self-administration in DAT-Cre – mice. All together these results suggest that VTA DA activity is required for the initial reinforcing properties of opioids from the very early stage of drug exposure.
Figure 4. Chemogenetic inhibition of VTA DA neurons decreases heroin self-administration. a, Schematic of the experiment for b-h; b, Left, VTA of DTA-Cre+ mice was bilaterally injected with the inhibitory DREADD hM4D. Right, coronal confocal images of infected VTA. c, Raster plot for infusions and inactive lever presses during the daily acquisition sessions of heroin self-administration for a DAT-Cre+ mouse. Twenty minutes prior to the sessions highlighted in pink, CNO (2mg/kg) was
injected intraperitoneally. d, Mean total lever presses and e, infusions during the acquisition phase of heroin self-administration for DAT-Cre+ (n=5, closed circle ) and DAT-Cre- mice (n=6 ,open circle).
When the self-administration behavior was well established CNO was injected prior to the session and the activation of the inhibitory DREADD dramatically decreased the number of lever presses and infusions (session highlighted in grey and red, respectively) in the DAT-Cre+ animals (condition (DAT-Cre + versus DAT-(DAT-Cre –) x CNO (present, absent), two-way repeated measures ANOVA and multiple comparison post-hoc Sidack test, *P<0.05, **P<0.005, ****P<0.0001). f, Raster plot for infusions and inactive lever presses during the daily acquisition session of heroin self-administration for a DAT-Cre+ mouse. Twenty minutes prior to the sessions highlighted in pink, CNO (2mg/kg) was injected
intraperitoneally. g, Mean total lever presses and h, infusions during the acquisition phase of heroin self-administration for DAT-Cre+ (n=6, closed circle ) or DAT-Cre- mice (n=4 ,open circle). CNO
injection prevented the establishment of heroin self-administration behavior in the DAT-Cre+ animals (condition (DAT-Cre + versus DAT-Cre –) x CNO (present, absent) two-way repeated measures ANOVA and multiple comparison post-hoc Sidack test, *P<0.05, **P<0.005, ****P<0.0001). Error bars, s.e.m.
4.5 Heroin occludes optogenetic self-stimulation of VTA DA neurons
We next tested whether heroin would occlude optogenetic VTA DA neuron self-stimulation 87. We injected various doses of heroin in a random order intraperitoneally (i.p.) immediately prior to the self-stimulation sessions (Figure 5c-f). To further avoid the development of tolerance each heroin session was followed by two sessions with free access to laser stimulation (LS) (Figure 5e, 5f). At baseline, the mice pressed up to 291+- 39 times to obtain 134 +- 1.15 LS in 60 min under a fixed-ratio 1 (FR1, followed by 20s time out period) schedule. With heroin injection, the performance decreased significantly in a dose-dependent fashion (Figure 5i), suppressing lever pressing completely at the highest dose (Figure 5d, 5e, 5f, 5i). In order to rule out any sedative effects of heroin at these doses, a separate set of mice were tested over a 30-min free exploration period in an open field (see Experimental Procedures). We observed that heroin actually increases locomotor activity in open field, even at the highest dose tested (Figure 5j). This experiment indicates that reinforcement by optogenetic self-stimulation and reinforcement by heroin share underlying neural circuits.
Figure 5. Heroin occludes optogenetic self-stimulation of VTA DA neurons. a, Schematic of the experiment for b-f; b, Left, VTA of DAT-Cre+ mice was bilaterally injected with a floxed version of the excitatory opsin ChR2. Right, coronal confocal images of infected VTA. c, Schedule of the experiment.
d, Raster plot for laser stimulation during the daily acquisition session of 1h for a DAT-Cre+ mouse.
Right before the sessions highlighted in blue, heroin (mg/kg, dose administrated in a random order) was injected intraperitoneally. For a matter of clarity, only the three last baseline sessions are shown and the heroin sessions are arranged from the lowest dose to the highest. e, Active, inactive lever presses and f, laser stimulation during each session for an example DAT-Cre+ mouse. Heroin
dose-dependently reduced active lever pressing and the number of laser stimulations. g, Active, inactive lever presses and h, laser stimulation during the acquisition sessions of self-stimulations for either DAT-Cre + mice (n=11, closed circles) or DAT-Cre - mice (n=6, open circles). Establishment of self-stimulation behavior was present only in mice with expression of eYFP-ChR2 in VTA DA neurons (DAT-Cre + but not DAT-Cre- mice (two-way repeated measures ANOVA and multiple comparison post-hoc Bonferonni correction, ****p<0.0001), i, Dose-response and fitting curve for the effect of heroin i.p. injection on laser self-stimulation for DAT-Cre+ (n=11, closed circles) or DAT-Cre- (n=6, open circles) mice. The values of IC50 and Hill coefficient are 6.5 mg/kg and 3.9 respectively. j, Distance traveled an open field after daily injections of increasing doses of saline or heroin (n=6). At the highest doses used (16 and 32 mg/kg), heroin increased the distance traveled in an open field (saline versus heroin injection, two-way repeated measures ANOVA and multiple comparison post-hoc Bonferonni correction, *p<0.05). Error bars, s.e.m.
4.6 Heroin occludes reinforcing effects of inhibition of VTA GABA neurons
Finally to test for the involvement of VTA GABA neurons that may inhibit VTA DA neurons 225,226, we expressed the light-gated inhibitory proton pump eArch-3.0 in the VTA of GAD-Cre mice (Figure 6a, 6b, O’Connor et al. 2015) and gave the mice control over the laser switch. The idea was to test whether self-inhibition of VTA GABA neurons is reinforcing and whether heroin exposure would occlude this behavior. The mice quickly learned to press the lever (293.1 +- 40 times to obtain 118.6 +- 14.7 LS in 180 min under the FR1 schedule). Heroin, injected intraperitoneally just prior to the self-inhibition session significantly decreased the operant behavior in a dose-dependent fashion (Figure 2f) again abolishing the behavior at the highest dose (Figure 4c, 2d, 4e). In fact, the IC50 was very similar to the IC50 calculated based on the occlusion with DA neuron self-stimulation (6.9 vs 6.4 mg/kg). This experiment indicates that reinforcement by optogenetic self-stimulation and reinforcement by heroin share underlying neural circuits and is a confirmation of a disinhibitory mechanism where heroin targets GABA neurons leading to an increase of DA neurons activity.
Figure 6. Heroin occludes reinforcing effects of inhibition of VTA GABA neurons. a, Schematic of the experiment for b-f; b, Left, VTA of GAD-Cre+ mice was bilaterally injected with a floxed version of the inhibitory opsin Arch3.0. Right, coronal confocal images of infected VTA. c, Schedule of the
experiment. d, Raster plot for laser inhibition during the daily acquisition session of 3h for a GAD-Cre+ mouse. Right before the sessions highlighted in yellow, heroin (mg/kg) was injected intraperitoneally.
e, Mean total lever presses and f, infusions during the acquisition phase of laser self-inhibition for GAD-Cre+ mice (n=7). Heroin injection resulted in a dose-dependent decrease in laser self-inhibition.
g, Dose-response and fitting curve for the effect of heroin i.p injection on laser self-inhibition for GAD-Cre+ mice. The values of IC50 and Hill coefficient are 6.2 mg/kg and 2.8 respectively. h, Summary diagram. After self-administration heroin, metabolized in morphine, binds to the MORs located on GABA neurons and activates GIRKs channels. It results in the inhibition of these neurons and the disinhibition of the DA neurons located in the ventrolateral VTA. Disinhibition of these neurons leads to an increase in DA release in the NAc shell. Error bars, s.e.m.
5. Discussion
Our study revealed a crucial role of mesolimbic DA transmission in mediating the reinforcing properties of opioids. Specifically, we found that the administration of heroin to drug-naïve animals increased both the activity of dopaminergic neurons in the VTA, as well as DA levels in the NAc shell.
We postulate that the heroin-induced changes in mesolimbic DA transmission occurred through a disinhibitory mechanism, whereby heroin targeted GABAergic neurons in the VTA that then
increased the activity of DA neurons in the NAc shell. Finally, we demonstrated that this mechanism appeared engaged at the earliest stage of drug exposure (i.e., after the first self-administration session), before mice were dependent.
5.1 Function of subpopulation located in the medioventral VTA
Several studies have revealed a critical role for VTA afferents in mediating the effects of drugs 228–230; however, the afferents responsible for mediating these effects, as well as the specific neuronal population which drive addiction-like behaviors, remain unclear. The VTA is canonically thought of as a heterogeneous structure comprised of different neuronal populations, with dopaminergic neurons (60-65%) contributing the most to this structure, followed by GABAergic (30-35%), glutamatergic (around 2%), and mixed neuron groups 67,69,70,72,73. Since Schultz’s experiments 78,209,231, mesolimbic
Several studies have revealed a critical role for VTA afferents in mediating the effects of drugs 228–230; however, the afferents responsible for mediating these effects, as well as the specific neuronal population which drive addiction-like behaviors, remain unclear. The VTA is canonically thought of as a heterogeneous structure comprised of different neuronal populations, with dopaminergic neurons (60-65%) contributing the most to this structure, followed by GABAergic (30-35%), glutamatergic (around 2%), and mixed neuron groups 67,69,70,72,73. Since Schultz’s experiments 78,209,231, mesolimbic