Restoration of a paraventricular thalamo-accumbal behavioral suppression circuit prevents reinstatement of heroin seeking

SUMMARY

Lack of behavioral suppression typifies substance use disorders, yet the neural circuit underpinnings of drug-induced behavioral disinhibition remain unclear. Here, we employ deep brain two-photon calcium imaging in heroin self-administering mice, longitudinally tracking adaptations within a paraventricular thalamus to nucleus accumbens behavioral inhibition circuit from the onset of heroin use to reinstatement. We find select thalamo-accumbal neuronal ensembles become profoundly hypoactive across the development of heroin seeking and use. Electrophysiological experiments further reveal persistent adaptations at thalamo-accumbal parvalbumin interneuron synapses, whereas functional rescue of these synapses prevents multiple triggers from initiating reinstatement of heroin seeking. Finally, we find an enrichment of μ-opioid receptors in output- and cell type-specific paraventricular thalamic neurons, which provide a mechanism for heroin-induced synaptic plasticity and behavioral disinhibition. These findings reveal key circuit adaptations that underlie behavioral disinhibition in opioid dependence, and further suggest that recovery of this system would reduce relapse susceptibility.

Graphical Abstract

eTOC blurb

Behavioral disinhibition is the cardinal feature of substance use disorders, enabling drug use and relapse despite significant consequences. Paniccia, Vollmer, Green and colleagues discover that heroin use induces behavioral disinhibition through maladaptive plasticity in a thalamus-nucleus accumbens circuit, whereas restoration of this circuit prevents heroin seeking in mice.

INTRODUCTION

A lack of behavioral suppression is a hallmark of all substance use disorders, allowing triggers such as environmental cues, drug priming, and stressors to initiate relapse despite negative consequences^1–5. For example, response inhibition deficits and risky decision making are commonly observed in both drug-dependent and drug-abstinent individuals^5–8, effects that are associated with decreased recruitment of cognitive control networks, including cortical, striatal, and thalamic structures^6,7,9. Neuroimaging studies specifically reveal that decreased thalamic activity during response inhibition tasks is associated with more severe substance misuse and is predictive of future relapse suceptibility^9–11. Despite these findings, how neuronal circuits that normally govern behavioral inhibition are affected by drug use remains unclear, and whether restoration of those circuits would prevent relapse is unknown.

Recently, we and others identified a thalamo-accumbal system that pervasively governs the suppression of naturalistic reward-seeking behaviors. Specifically, paraventricular thalamic glutamatergic projection neurons synaptically drive activity in nucleus accumbens parvalbumin interneurons (PVT→NAc^PV-IN), ceasing motivated behaviors in the face of threats, stressors, inhibitory learning, and more^12–15. While these studies demonstrate that PVT→NAc activation can provide a ‘brake’ for natural reward-related behaviors, PVT→NAc stimulation may be incapable of preventing opioid-motivated seeking. For example, chemogenetic activation of PVT→NAc neurons does not prevent heroin seeking under control conditions, but rather has mild inhibitory effects only when heroin seeking is enhanced by food restriction¹⁶. Surprisingly, one study shows that activity in PVT→NAc neurons can enhance, rather than inhibit, heroin seeking during abstinence¹⁷. Considering that behavioral actions across these natural- and opioid-reward seeking paradigms are similar (e.g., lever pressing without reward availability during cue reinstatement), here we test the hypothesis that persistent opioid-mediated neuronal adaptations within the thalamo-accumbal circuit prevents inhibition of opioid seeking.

RESULTS

Inhibitory PVT→NAc neuronal ensemble dynamics predict heroin self-administration.

Activity of PVT→NAc projection neurons is critical for suppression of naturalistic reward-motivated behaviors^18–20, whereas μ-opioid receptor (μ-OR) stimulation can acutely inhibit PVT→NAc neurons and downstream synapses¹⁵. Despite this knowledge, the influence of long-term opioid use on PVT→NAc neurons remains unknown. Here we combined head-fixed heroin self-administration with deep brain two-photon calcium imaging to longitudinally measure the activity dynamics of PVT→NAc projection neurons across heroin use and reinstatement of heroin seeking (Fig. 1A). Head-restrained mice were trained to press an active, but not an inactive, lever resulting in delivery of a tone cue followed by intravenous infusion of heroin and time-out period (Fig. 1B–C). To visualize PVT→NAc neuronal dynamics within this paradigm, mice received injections of a retrogradely-trafficked virus encoding Cre-recombinase bilaterally into the NAc shell (rgAAV2-CAG-Cre) in combination with a Cre-inducible virus encoding a calcium indicator into the posterior PVT (AAVdj-Ef1a-DIO-GCaMP6m; Fig. 1D)²¹. A microendoscopic GRIN lens was implanted dorsal to PVT, allowing visual access to GCaMP6m-expressing PVT→NAc neurons (Fig. 1E).

Figure 1. Inhibitory PVT→NAc neuronal ensemble dynamics predict heroin self-administration.

A,B, Apparatus design (A) and behavioral schematic (B) for intravenous heroin self-administration in head-fixed mice. C, Grouped data for acquisition of heroin self-administration. Mice learn to press the active, but not inactive, lever for heroin (n=27 mice; lever: F_1,52=122.30, P<0.001). D, Surgical strategy allowed two-photon calcium imaging of PVT→NAc neurons in heroin self-administering mice. (E) Example field of view (FOV; top) and extracted signals (bottom) from a two-photon calcium imaging session. F,G, Example waveforms (F) and grouped data (G) depict reductions in PVT→NAc GCaMP6m-mediated fluorescence from the start (first 3-minutes) to the end (last 3-minutes) of each session during early (days 1–2), middle (days 7–8), and late (days 13–14) acquisition, but not during baseline (No-SA) sessions (session length dependent on infusion cap, to normalize we analyzed the first and last 3-minutes; n=3–7 mice; 105–355 neurons/session; interaction: F_3,2194=12.85, P<0.001). H-J, Averaged traces (top) and single-cell heatmaps (bottom) reveal PVT→NAc neuronal encoding of active lever pressing behavior during early (H; n=312 cells; 7 mice), middle (I; n=329 cells; 7 mice), and late (J; n=355 cells; 7 mice) self-administration sessions. K, Spectral clustering revealed three emergent PVT→NAc neuronal ensembles during heroin self-administration: excitatory responders (n=113 cells), somewhat-inhibited responders (n=119 cells), and greatly inhibited responders (n=123 cells). L, Decoding analyses show each neuronal ensemble predicts active lever pressing, with ensemble 3 having the most accuracy (interaction: F_2,704=17.21, P<0.001). Red line indicates the average of shuffled data. M, Example FOVs for PVT→NAc neurons tracked across heroin self-administration (n=6 mice, 66 neurons). N,O, Correlation plots showing mean responses between early vs. late (N; P=0.11) and middle vs. late (O; P<0.01) sessions for all tracked neurons (Pearson-R value displayed in graph). P, auROC z-scores for each ensemble reveal ensemble 3, but not 1 or 2, developed new inhibitory responses across acquisition (interaction: F_2,189=5.09, P<0.001, ensemble 3 post-hoc: P<0.01). Q, Pie charts for each ensemble show a larger proportion of tracked neurons with significant responses in ensemble 3 during late self-administration (ensemble 3: χ²₂ =19.2, P<0.001). See also Figure S1. auROC, area under the receiver operating characteristic; CDF, cumulative distribution frequency; FOV, field of view; Mid, Middle; SA, self-administration; Rein, reinstatement. Group comparisons: *P<0.05, **P<0.01, ****P<0.001.

Using two-photon microscopy, we measured the change in basal fluorescence among PVT→NAc neurons during early (days 1–2), middle (days 7–8), and late (days 13–14) heroin self-administration sessions, as basal fluorescence serves as a proxy for firing rates in tonically active cell populations^22,23 including PVT→NAc projection neurons²⁰. We calculated the change in fluorescence (Δf/f₀) using the average signal from the first 3-minutes as compared to the average of the last 3-minutes of each self-administration session. Previously, we reported that PVT→NAc neurons display modest reductions (~5–10%) in tonic activity during sucrose self-administration¹⁵. Interestingly, these neurons showed profoundly greater reductions in activity during middle and late heroin self-administration sessions (~40%; Fig. 1F–G), effects that were correlated with active lever pressing behavior (Fig. S1A–D).

Next, we measured acute, rather than tonic, PVT→NAc neuronal adaptations within each behavioral session. At a population level, PVT→NAc neurons were acutely inhibited during active lever pressing in middle and late acquisition sessions (Fig. 1H–J). Spectral clustering^22,24 of neuronal responses during late self-administration further revealed three ensembles: activated (ensemble 1), somewhat inhibited (ensemble 2), and greatly inhibited (ensemble 3; Fig. 1K; Fig. S1E–H). Next, we used machine learning-based decoding to predict active lever presses based on the fluorescence of each neuron prior to lever pressing. On average, the decoder could predict a future press at ~6% above chance for each cell, resulting in substantial lever press prediction at the population level (Fig. S1I). Furthermore, while the fluorescence of all ensembles could predict active lever pressing, the acute inhibitory dynamics of ensemble 3 provided superior decoding (Fig. 1L). Finally, by examining tonic fluorescence fluctuations as above, we found that inhibition of only neuronal ensemble 3 correlated with active lever pressing rates (Fig. S1J).

To determine whether PVT→NAc neuronal dynamics developed across learning, we tracked a subset of cells between each phase of heroin self-administration (Fig. 1M). While there was no correlation in the acute lever press dynamics between early and late acquisition sessions, there was a significant correlation between middle and late sessions (Fig. 1N–O). Consistent with these findings, we found that ensemble 3 neurons, but not ensemble 1 or 2 neurons, developed new responses by middle and late sessions that were not present in early acquisition (Fig. 1P–Q). Overall, our data suggest that PVT→NAc neurons develop new, inhibitory dynamics across acquisition of heroin self-administration that subsequently stabilize to predict heroin use.

PVT→NAc neuronal ensemble dynamics stably predict reinstatement of heroin seeking regardless of trigger modality.

To measure PVT→NAc activity dynamics during heroin seeking, mice underwent extinction training through cue and heroin omission (Fig. S1K). While PVT→NAc neurons displayed similar lever press encoding dynamics and tonic inhibition during early extinction sessions (days 1–2) as in late acquisition, these responses were attenuated after extinction was well learned (day 9 or after; Fig. S1L–O). Following extinction, reinstatement or “relapse” to heroin seeking was modeled through cue re-exposure (cue-induced reinstatement), injection of heroin (drug-induced reinstatement), or injection of the pharmacological stressor yohimbine (stress-induced reinstatement), during which active but not inactive lever pressing increased indicative of heroin-seeking behavior (Fig. 2A–C; Fig. S2A–C). Using two-photon calcium imaging, we measured PVT→NAc activity dynamics during and between each reinstatement test. Overall, PVT→NAc neurons showed tonic reductions in activity during each reinstatement test as compared with late extinction (Fig. 2D–F), and this neuronal inhibition was generally correlated with active lever pressing rates (Fig. S2D–I). PVT→NAc activity was also acutely reduced during active lever pressing during reinstatement (Fig. 2G; Fig. S2J–L). Spectral clustering of these responses revealed qualitatively similar neuronal ensembles as compared with acquisition sessions; however, both ensembles 1 (activated) and 3 (greatly inhibited) predicted an active lever press more accurately than ensemble 2 (somewhat inhibited; Fig. 2H–J; Fig. S2M–O). Although the tonic activity of each ensemble decreased across reinstatement, this effect was most pronounced for ensemble 3 (Fig. S2P). Interestingly, while PVT→NAc neurons could be sorted into nearly identical clusters during cue-, drug-, and stress-induced reinstatement, the number of cells within each cluster were distributed differently for each session (Fig. S2Q). Furthermore, behavioral decoding showed that PVT→NAc population dynamics, or discrete neuronal ensemble dynamics, could be used to predict whether an active lever press would occur during grouped or individual reinstatement sessions (Fig. S2R–S). We found that changes in behavioral responding in cue- and drug-induced reinstatement sessions were negatively correlated with changes in fluorescent signal only for ensemble 3, while the changes in active lever pressing and fluorescence were negatively correlated for ensemble 2 during stress-induced relapse (Fig. S2T).

Figure 2. PVT→NAc neuronal ensemble dynamics stably predict reinstatement of heroin seeking regardless of trigger modality.

A-C, Cue- (A), drug- (B), and stress-induced (C) reinstatement tests wherein active lever pressing increased versus previous extinction tests (n=12–13 mice/test; Cue: t₁₂=7.66, P<0.001; Drug: t₁₁=5.04, P<0.001; Stress (yohimbine): t₁₁=5.38, P<0.001). D-F, Grouped data reveal that PVT→NAc GCaMP6m-mediated fluorescence was significantly reduced from the start to end of cue- (D), drug- (E), and stress- (F) induced reinstatement sessions, versus the last day of extinction (averages of first and last 3-minutes of each session; n=84–125 cells, 4–6 mice/session; interaction F_3,840=58.80, P<0.001). G,I, Averaged trace (G) and single-cell heatmap (I) revealing grouped data for PVT→NAc neuronal responses across all three reinstatement sessions (n=4–6 mice/session, 316 neurons). H, Spectral clustering reveals three emergent PVT→NAc neuronal ensembles during reinstatement sessions: excitatory responders (n=106 cells), mildly inhibited responders (n=129 cells), and greatly inhibited responders (n=81 cells). J, Decoding analyses show each neuronal ensemble predicts an active lever press, with ensembles 1 and 3 having the most accurate decoding (interaction: F_2,626=7.84, P<0.001). Red line indicates the average of shuffled data. K, Scatter plots for all reinstatement tests showing correlated mean responses for neurons tracked between tests (see Fig. S2 for each test; n= 20–38 cells; 3–4 mice/group; Pearson-R value on graph; P<0.001). L, auROC z-scores for each ensemble tracked across two reinstatement tests confirm stable responses (F-values<0.57, Ps>0.56). See also Figure S2. Ext, extinction; Rein, reinstatement. Group comparisons: **P<0.01, ****P<0.001.

Considering the similarity in population encoding during reinstatement, but variations in the relative number of cells recruited to each ensemble based on our spectral clustering analysis, we next examined whether individual neurons showed stable or distinct acute lever press responses between each reinstatement session. Overall, we tracked neuronal responses between two tests (cue vs. drug; cue vs. stress; drug vs. stress) and found that neuronal responses during one reinstatement test were highly predictive of responses during the subsequent reinstatement test (Fig. 2K; Fig. S2U). Furthermore, overall response amplitudes did not significantly vary between two tracked reinstatement tests (Fig. 2L). Together, these data show that reinstatement of heroin seeking is associated with discrete neuronal ensembles that are stably maintained across time, regardless of relapse trigger.

PVT→NAc activation prevents reward seeking in drug-naïve but not heroin-experienced mice.

We found that PVT→NAc neuronal inhibition predicts heroin self-administration and reinstatement, but whether stimulation of those neurons would causally prevent heroin seeking remained unclear. Thus, we targeted PVT→NAc neurons for optogenetic manipulation by injecting a retrogradely-trafficked virus encoding Cre-recombinase bilaterally into NAc shell (rgAAV2-CAG-Cre) in combination with viruses encoding Cre-inducible channelrhodopsin (AAV5-DIO-Ef1α-ChR2-EYFP), halorhodopsin (AAV5-DIO-Ef1α-eNpHR3.0-EYFP), or control enhanced yellow fluorescent protein (AAV5-DIO-Ef1α-eYFP) into posterior PVT (Fig. 3A–B). Previously we reported that in mice trained to self-administer sucrose, stimulation but not inhibition of PVT→NAc neurons prevents cue-induced reinstatement sucrose seeking after extinction learning¹⁵. However, in mice trained to self-administer heroin, neither optogenetic stimulation nor inhibition of PVT→NAc neurons prevented cue-, drug-, or stress-induced reinstatement of active lever pressing (Fig. 3C–E). Inactive lever pressing was also unaffected by optogenetic manipulations (Fig. S3). Altogether, these data reveal that PVT→NAc neuronal activation abolishes goal-directed active lever pressing in heroin-naïve but not heroin-experienced mice. However, the circuit mechanism that prevents PVT→NAc neurons from suppressing heroin seeking remains unknown.

Figure 3. PVT→NAc neuronal activation does not prevent heroin seeking.

A,B, Surgical strategy (A) and example viral expression (B) for optogenetic manipulation of PVT→NAc neurons during heroin self-administration. C-E, In heroin-experienced mice, neither optogenetic activation nor inhibition of PVT→NAc neurons prevented cue- (D), drug- (E), or stress-induced (F) reinstatement (n=7–8 mice/group; Cue, day: F_1,19=37.39, P<0.001; group comparisons: eYFP: P<0.05, eNpHR: P<0.01, ChR2: P<0.01; Drug, day: F_1,19=59.52, P<0.001, group comparisons: eYFP: P<0.001, eNpHR: P<0.001, ChR2: P<0.01; Stress (yohimbine), day: F_1,19=29.40, P<0.001, group comparisons: eYFP: P<0.05, eNpHR: P<0.05, ChR2: P<0.01). See also Figure S3. Ext, extinction; Opto, optogenetic manipulation (represented by yellow bars); Rein, reinstatement. Group comparisons: *P<0.05, **P<0.01, ****P<0.001.

PVT→NAc synaptic innervation of parvalbumin interneurons is persistently weakened by heroin self-administration.

PVT→NAc projection neurons synaptically innervate dopamine 1 receptor- and dopamine 2 receptor-expressing medium spiny neurons (PVT→NAc^D1-MSN; PVT→NAc^D2-MSN)^17,25,26, as well as parvalbumin-expressing interneurons (PVT→NAc^PV-IN)^15,27, although the influence of heroin use on the efficacy of these synapses remains unclear. To determine the outcome of heroin self-administration and extinction training on cell-type specific PVT→NAc synapses, we injected a virus encoding the red-shifted excitatory opsin ChrimsonR (AAV5-hSyn-ChrR-tdT)²⁸ into posterior PVT and a Cre-inducible virus encoding eYFP (AAV5-ef1α-DIO-eYFP) into the NAc shell of Cre-driver mouse lines (D2-Cre; PV-Cre; Fig. 4A; Fig. S4A–B). Next, we characterized PVT synaptic innervation of accumbal D1-MSNs (putative), D2-MSNs, and PV-INs 24-hours following the last heroin self-administration or saline control session or following both heroin self-administration and extinction training (Fig. 4B–C). Using patch clamp electrophysiology, we measured glutamatergic excitatory synaptic efficacy through optically evoked excitatory postsynaptic currents (oeEPSCs; primarily mediated by AMPArs), AMPAr/NMDAr current ratios, AMPAr current rectification, and paired-pulse ratios for each cell-type specific NAc neuron. Data revealed PVT→NAc^D1-MSN synaptic efficacy was unchanged by heroin self-administration and/or extinction training, as there was no change in oeEPSC amplitude, AMPAr/NMDAr ratio, or paired-pulse ratio (Fig. 4D, G; Fig. S4C). However, recordings of AMPAr rectification revealed PVT→NAc^D1-MSN synapses were inwardly rectifying following extinction training, suggesting the insertion of calcium-permeable AMPA receptors (CP-AMPArs; Fig. 4J)^29,30. Next, PVT→NAc^D2-MSN synapses displayed an increase in oeEPSC amplitude and AMPAr/NMDAr ratio following heroin self-administration, an effect that was absent after extinction (Fig. 4E, H). Additionally, a decreased paired-pulse ratio was observed at PVT→NAc^D2-MSN synapses after heroin self-administration but not extinction training (Fig. S4D), indicating a transient increase in presynaptic release probability. In contrast, PVT→NAc^D2-MSN synapses did not exhibit a significant change in AMPAr rectification (Fig. 4K). These data are consistent with previous findings showing that PVT synaptic innervation of NAc D2-MSNs is increased during acute opioid withdrawal²⁶, which is likely present in the heroin self-administration group. Finally, we found a robust and persistent decrease in oeEPSC amplitude at PVT→NAc^PV-IN synapses following heroin self-administration with or without extinction training (Fig. 4F). This was accompanied by decreased AMPAr/NMDAr ratio following heroin self-administration and extinction (Fig. 4I). Collectively, these results demonstrate a significant reduction in PVT→NAc^PV-IN synaptic efficacy. In contrast, AMPAr rectification and paired-pulse ratios were unchanged at these synapses (Fig. 4L; Fig. S4E). Altogether, these data reveal that PVT→NAc^PV-IN postsynaptic efficacy is persistently weakened following heroin use, whereas PVT→NAc^D2-MSN synapses undergo a transient increase in excitatory drive consistent with past studies examining opioid withdrawal²⁶.

Figure 4. PVT synaptic innervation of NAc PV-INs is weakened following heroin self-administration and extinction.

A,B, Surgical strategy (A) and experimental timeline (B) for patch-clamp electrophysiology. C, Example DIC images (left), fluorescence expression (middle), and action potential traces (right) for identification of NAc cell types. D, Example oeEPSC waveforms (left) and grouped data (right) show no oeEPSC amplitude changes in NAc D1-MSNs following heroin self-administration or extinction (n=18–21 cells; 8–9 mice/group; F_2,54=0.461, P=0.633). E, Example oeEPSC waveforms (left) and grouped data (right) show elevated oeEPSC amplitudes in NAc D2-MSNs following heroin self-administration, but not extinction (n=19–27 cells; 8–11 mice/group; F_2,64=7.53, P=0.001, group comparisons: D2 control vs D2 heroin: P<0.001, D2 control vs D2 heroin ext: P=0.292). F, Example oeEPSC waveforms (left) and grouped data (right) reveal persistent, decreased oeEPSC amplitudes in NAc PV-INs following heroin self-administration and extinction (n=20–24 cells; 5–9 mice/group; F_2,62=13.53, P<0.001, group comparisons: PV control vs PV heroin: P<0.001, PV control vs PV heroin ext: P<0.001). G, Example AMPAr and NMDAr currents (left) and grouped data (right) for D1-MSNs reveal no change in AMPAr/NMDAr ratio (n=11 cells; 5–6 mice/group; F_2,30=0.714, P=0.498). H, Example AMPAr and NMDAr currents (left) and grouped data (right) for D2-MSNs reveal a trending, transient change in AMPAr/NMDAr ratio (n=10–11 cells; 5 mice/group; F_2,29=2.98, P=0.066, group comparisons: D2 control vs D2 heroin: P<0.05, D2 control vs D2 heroin ext: P=0.716). I, Example AMPAr and NMDAr currents (left) and grouped data (right) reveal no change in PV-IN AMPAr/NMDAr ratio following heroin self-administration but exhibit a significant decrease following extinction (n=11–17 cells; 5–8 mice/group; F_2,42=6.12, P<0.01, group comparisons: PV control vs PV heroin: P=0.063, PV control vs PV heroin ext: P<0.01). J, Example AMPAr rectification waveforms (left) and grouped data (right) show significantly increased rectification in D1-MSNs following heroin self-administration plus extinction, but not self-administration alone (I-70/I+50; n=11–12 cells; 5–6 mice/group; F_2,31=4.12; P<0.05, group comparisons: D1 control vs D1 heroin: P=0.992, D1 control vs D1 heroin ext: P<0.05). K, Example AMPAr rectification waveforms (left) and grouped data (right) show no significant changes in rectification in D2-MSNs following heroin self-administration or extinction (n=10–11 cells; 5 mice/group; F_2,29=2.48; P=0.101). L, Example AMPAr rectification waveforms (left) and grouped data (right) reveal no significant changes in rectification index in PV-INs following heroin self-administration or extinction (n=13–18 cells; 5–7 mice/group; F_2,43=0.054; P=0.948). See also Figure S4. Acq, acquisition; ChR, ChrimsonR; DIC, diffusion interference contrast; Ext, extinction; Habit, habituation; oeEPSC, optically evoked excitatory postsynaptic current. Group comparisons: *P<0.05, **P<0.01, ****P<0.001; ns, not significant.

Rescuing activity in the PVT→NAc^PV-IN circuit prevents heroin seeking.

Based on electrophysiological findings showing persistently reduced PVT→NAc^PV-IN synaptic efficacy following heroin self-administration (Fig. 4), and the requirement of PVT→NAc^PV-IN synapses for behavioral inhibition¹⁵, we predicted that restoration of activity at these synapses would restore behavioral inhibition in heroin-experienced mice. Thus, we used a chemogenetic approach to elevate PV-IN excitability, with and without simultaneous optogenetic stimulation of PVT→NAc neurons, during the reinstatement of heroin seeking. Bilateral cannulae were implanted dorsal to the NAc shell and a Cre-inducible virus encoding an excitatory DREADD (AAV5-hSyn-DIO-hM4D(Gq)-mCherry) was injected into NAc shell of PV-Cre or wild-type control mice (Fig. 5A). Patch-clamp electrophysiology revealed that clozapine-N-oxide (CNO) reliably increased the excitability of Gq-DREADD-expressing NAc^PV-INs (Fig. 5B; Fig. S5A). Next, following heroin self-administration and extinction, mice were given intra-NAc microinfusions of CNO prior to cue-, drug-, and stress-induced reinstatement tests. Chemogenetic enhancement of PV-IN excitability was insufficient to decrease active lever pressing during these tests (Fig. 5C–E). Thus, we next combined the chemogenetic manipulation with simultaneous optogenetic stimulation of PVT→NAc neurons. PV-Cre or wild-type control mice received bilateral microinjections of a retrogradely-trafficked virus encoding ChR2 and a Cre-inducible virus encoding the excitatory DREADD into NAc shell, with bilateral cannula implanted dorsal to the NAc shell (Fig. S5B–C). Additionally, an optical fiber was implanted dorsal to the PVT for activation of PVT→NAc projection neurons (Fig. 5F–G; Fig. S5D–E). As shown above, optogenetic stimulation of PVT→NAc neurons following heroin self-administration and extinction training in wild-type mice did not prevent heroin seeking (see Fig. 3). However, when optogenetic activation of PVT→NAc neurons was paired with chemogenetic enhancement of NAc PV-IN excitability in PV-Cre mice, heroin seeking was abolished during cue-, drug-, and stress-induced reinstatement (Fig. 5H–J). These manipulations did not alter inactive lever pressing (Fig. S5F–K). Altogether, these data reveal that selective restoration of PVT→NAc^PV-IN synaptic activity can prevent heroin-seeking behavior, likely by reestablishing a feedforward inhibitory brake for reward seeking. The mechanism whereby heroin self-administration leads to persistent dysregulation of PVT→NAc^PV-IN synapses, however, remains unclear.

Figure 5. Rescuing activity in the PVT→NAc^PV-IN circuit prevents reinstatement of heroin seeking.

A, Surgical strategy for chemogenetic activation of NAc PV-INs. B, Electrophysiological recordings confirmed CNO wash-on increased NAc PV-IN neuronal activity. C-E, Chemogenetic activation of PV-INs through intra-NAc CNO infusions did not prevent cue- (C), drug- (D), or stress-induced (E) reinstatement (n=6–8 mice/group; Cue, day: F_1,14=37.98, P<0.001; Drug, day: F_1,11=14.39, P<0.01; Stress, day: F_1,12=14.37, P<0.01). F,G, Surgical strategy (F) and example viral expression (G) for simultaneous optogenetic stimulation of PVT→NAc neurons and chemogenetic activation of PV-INs. H-J, Optogenetic activation of PVT→NAc neurons combined with chemogenetic activation of PV-INs prevented cue- (H), drug- (I), and stress-induced (J) reinstatement in PV-Cre, but not WT, mice (n=6–8 mice/group; Cue, interaction: F_1,14=13.86, P<0.01, Ext: P>0.999, Rein: P<0.001; Drug: interaction: F_1,12=8.90, P<0.05, Ext: P=0.998, Rein: P<0.001; Stress: interaction: F_1,11=11.31, P<0.01, Ext: P=0.986, Rein: P<0.001. See also Figure S5. WT, wild-type; Ext, extinction; Rein, reinstatement; Opto, optogenetic manipulation (represented by yellow bars). Group comparisons: ****P<0.001; ns, not significant.

Cell-type specific paraventricular thalamo-accumbal projection neurons display enrichment of Oprm1 expression.

There is significant transcriptional heterogeneity within the PVT, with unique molecular cell subpopulations responsible for regulating distinct behavioral phenotypes^31,32. Thus, to gain mechanistic insight into the heroin-mediated adaptations within the PVT→NAc^PV-IN circuit, we first aimed to characterize the circuit- and cell type-specific expression of presynaptic Oprm1. We focused on four transcriptional markers of distinct posterior PVT neuronal cell types: Drd2, Drd3, Col12a1, and Esr1³². Using Oprm1^fl/fl mice, we injected Red Retrobeads^™ into the NAc, and conducted RNAscope^® on posterior PVT slices to simultaneously visualize Oprm1, and either Drd2, Drd3, Col12a1, or Esr1 (Fig. 6A–B; Fig. S6A–D). We detected a high percentage of Retrobead^™-positive cells in the PVT (Fig. 6C) and found that both Retrobead^™-positive and -negative cells express Oprm1 (Fig. 6D–E). Interestingly, higher levels of Oprm1 expression significantly correlated with higher levels of Retrobead^™ fluorescence (Fig. 6F), indicating densest Oprm1 expression in PVT projection neurons with the strongest innervation of NAc.

Figure 6. Quantification of circuit- and cell type-specific expression of Oprm1 in posterior PVT.

A, Surgical strategy for Red Retrobead^™ experiments. B, Representative heatmap of NAc Retrobead^™ placement. C, Percentage of Retrobead^™-positive neurons detected in PVT, i.e., percentage of PVT→NAc projection neurons, independent of gene expression profiles (n=6 mice, 10–11 optical datasets/RNAscope^® staining batch: e.g., Oprm1 + Col12a1, Oprm1 + Drd3, Oprm1 + Drd2 or Oprm1 + Esr1). D, Representative 10x confocal z-stack of PVT Retrobead^™ fluorescence (red) and Oprm1 gene expression (green). Inserts (right) illustrate high colocalization of Oprm1 in Retrobead^™-positive PVT→NAc neurons. E, Percentage of Oprm1-positive Retrobead^™-positive (+) or -negative (−) cells detected in PVT. (n=6 mice, 10–11 optical datasets per RNAscope^® batch: Oprm1 + Col12a1, Drd3, Drd2 or Esr1). F, Correlation of PVT→NAc Retrobead^™ and PVT Oprm1 3D volume. Expression of PVT Oprm1 was significantly correlated with density of Retrobead^™ signal in PVT→NAc neurons (n=6 mice, 30–80 cells/optical dataset; Pearson-R value displayed on graph; P<0.001). G, Representative 10x confocal z-stacks of Oprm1 fluorescent signal (green) and Drd2, Drd3, Col12a1, or Esr1 fluorescence (blue). Inserts (right) depict colocalization of PVT Oprm1 fluorescence with expression of each of the four genes. H, Cell type-specific expression of Oprm1 signal volume in Drd2-, Col12a1-, Drd3-, and Esr1-positive PVT cells. Oprm1 expression is highest in Drd2-positive cells, followed by Col12a1-positive cells (Kruskal-Wallis H_3,2354=352.9, P<0.001, group comparisons shown on graph). I, Representative images of Retrobead^™-positive (top rows) or Retrobead^™-negative (bottom rows) defined cells (dashed circles), expressing Oprm1 and Drd2, Drd3, Col12a1 or Esr1. Scale bar represents 10 μm. J, Levels of Oprm1 volume were significantly higher in Retrobead^™-positive versus Retrobead^™-negative Drd2-positive neurons, but not Drd3-, Col12a1-, and Esr1-expressing neurons (interaction: F_3,2303=7.438, P<0.001, group comparisons: Drd2: P<0.001; Col12a1: P=0.554; Drd3: P=0.066; Esr1: P=0.606). See also Figure S6. Group comparisons **** P<0.001; ns, not significant.

Next, we analyzed colocalization of Oprm1 with Drd2, Drd3, Col12a1, and Esr1 in posterior PVT (Fig. 6G; Fig. S6E–G). We detected the highest Oprm1 expression in Drd2-positive cells, followed next by Col12a1-positive cells, and comparable Oprm1 levels in Drd3- and Esr1-positive cells (Fig. 6H). These data suggest that within the posterior PVT, Oprm1 expression is enriched in dopamine receptor type 2 (D2R)-positive neurons. Finally, we assessed thalamo-accumbal circuit-specific expression patterns of Oprm1 in Drd2-, Drd3-, Col12a1-, or Esr1-positive cells within the PVT (Fig. 6I). We observed higher levels of Oprm1 expression in Retrobead^™-positive PVT→NAc neurons than in Retrobead^™-negative neurons only in Drd2-positive cells (Fig. 6J). Altogether, these data indicate there is a high density of μ-ORs on thalamo-accumbal projection neurons, and these μ-ORs are preferentially expressed on PVT→NAc cells that co-express dopamine D2 receptors. However, whether presynaptic μ-ORs mediate heroin-induced adaptations at PVT→NAc^PV-IN synapses to drive behavioral disinhibition remains unknown.

Thalamic μ-ORs are required for heroin-mediated dysregulation of PVT→NAc^PV-IN synapses and behavioral disinhibition.

Due to the high functional density of μ-ORs on PVT→NAc projection neurons¹⁵, we predicted that thalamic μ-ORs are required for heroin-induced dysregulation of PVT→NAc^PV-IN synaptic connectivity and PVT→NAc-dependent behavioral inhibition. Thus, we replicated the above optogenetic (Fig. 3) and electrophysiological (Fig. 4) experiments in Oprm1^fl/fl mice, enabling Cre-dependent knockout (KO) of μ-ORs³³. We validated this approach for knockout of μ-ORs in PVT→NAc neurons through RNAscope^® and slice electrophysiology (Fig. 7A–B)¹⁵.

Figure 7. PVT μ-ORs knockout attenuates heroin-induced synaptic alterations and enables PVT→NAc-dependent suppression of motivated behavior in heroin-experienced mice.

A, Example confocal images depicting Oprm1 (blue) and Oprk1 (red) gene expression in eYFP-expressing PVT→NAc neurons (green). Data reveal selective μ-OR knockout in PVT of Oprm1^fl/fl, but not wild-type control, mice. B, Example waveforms (left) and grouped data (right) of PVT somatic recordings in Oprm1^fl/fl mice before and after application of μ-OR agonist DAMGO for control (top) and μ-OR KO (bottom) mice. DAMGO significantly reduced spiking in control but not μ-OR KO neurons (n=7–17 cells; 5–9 mice/group; F_1,22=42.27, P<0.001, group comparisons: Control heroin P<0.001; μ-OR KO: P=0.724). C, Surgical strategy for electrophysiology experiments. D, Example oeEPSC waveforms (left) and grouped data (right) for PV-INs confirmed significant decreases in oeEPSC amplitude following heroin self-administration and extinction, an effect that was blocked by PVT μ-OR knockout (n=10–17 cells; 5–6 mice/group; F_2,37=4.61, P<0.05, group comparisons: Control saline vs Control heroin: P<0.05, Control saline vs μ-OR KO heroin: P=0.989). E, Example waveforms (left) and grouped data (right) confirmed significant decreases in AMPAr/NMDAr ratios following heroin self-administration and extinction, an effect that was blocked by PVT μ-OR knockout (n=10–17 cells; 5–6 mice/group; F_2,37=4.13, P<0.05, group comparisons: Control saline vs Control heroin: P<0.05, Control saline vs μ-OR KO heroin: P=0.922). F, Surgical strategy for behavioral experiments. G, H, In mice with a history of heroin self-administration and extinction, knockout of PVT μ-ORs enabled optogenetic activation of PVT→NAc neurons to suppress active lever pressing in both cue- (G) and heroin-primed (H) reinstatement tests relative to control heroin mice (n=5–7 mice/group; Cue Rein: interaction: F_1,10=8.17, P<0.05, group comparisons: Ext: P>0.999, Rein: P<0.01; Drug Rein: interaction: F_1,9=14.87, P<0.01, group comparisons: Ext: P=0.994, Rein: P<0.001). I, In mice with a history of heroin self-administration or saline control sessions, there were no significant differences in active lever pressing for sucrose during the laser off test. Optogenetic activation of PVT→NAc neurons significantly reduced active lever presses for sucrose in control saline mice compared to the control heroin group. Knockout of thalamic μ-ORs in heroin-experienced mice enabled PVT→NAc photoactivation to suppress active lever pressing for sucrose to levels comparable to control saline mice (interaction: F_2,15=5.00, P<0.05; group comparisons: Laser Off: Control saline vs Control heroin: P=0.904, Control saline vs μ-OR KO heroin: P=0.918; Laser On: Control saline vs Control heroin: P<0.01, Control saline vs μ-OR KO heroin: P=0.944). See also Figure S7. Base, baseline; KO, knockout; Ext, extinction; Opto, optogenetic manipulation (represented by yellow bars); OR, opioid receptor; Rein, reinstatement. Group comparisons *P<0.05, **P<0.01, ****P<0.001; ns, not significant.

To determine the necessity of PVT μ-ORs for heroin-induced PVT→NAc^PV-IN synaptic adaptations, we injected a cocktail of viruses allowing expression of the excitatory opsin ChR2 (AAV5-hSyn-hChR2-eYFP) and either control eYFP (AAV5-hSyn-eYFP) or Cre-recombinase (AAV5-hSyn-GFP-Cre) in posterior PVT (Fig. 7C). Additionally, we injected a PV-specific virus to label PV-INs in NAc (AAV1-s5e2-GCaMP6f). In separate groups of heroin naïve mice, we found that rheobase was elevated in PVT neurons in μ-OR KO mice as compared to controls, with no differences in synaptic drive onto downstream PV-INs (Fig. S7A–F). Thus, following recovery from surgery, Oprm1^fl/fl mice were split into one of three groups for electrophysiological experiments: (1) Control saline (PVT injections of a virus encoding eYFP; PVT∷eYFP), (2) Control heroin (PVT injections of a virus encoding eYFP; PVT∷eYFP), or (3) μ-OR KO heroin (PVT injections of a virus encoding Cre-recombinase; PVT∷Cre). Control saline mice underwent 14-days of saline control sessions followed by 10-days of extinction training, whereas heroin groups underwent 14-days of heroin self-administration followed by 10-days of extinction training. Recordings from NAc PV-INs confirmed reduced synaptic efficacy at PVT→NAc^PV-IN synapses in control heroin mice, measured through oeEPSCs and AMPAr/NMDAr ratios, an effect that was prevented by μ-OR KO in heroin-trained mice (Fig. 7D–E). These results establish presynaptic PVT μ-ORs as a mechanism whereby heroin weakens PVT→NAc^PV-IN synapses.

Next, we determined if PVT μ-OR KO would also prevent heroin-induced behavioral disinhibition. A virus encoding Cre-recombinase or eYFP was injected into posterior PVT of Oprm1^fl/fl mice, a retrogradely-trafficked virus encoding ChR2 was injected into NAc, and an optical fiber was implanted dorsal to PVT (Fig. 7F). There were no observed differences in acquisition of heroin self-administration or extinction between groups (Fig. S7G–I). Consistent with our earlier results (see Fig. 3), optogenetic activation of PVT→NAc neurons was insufficient to inhibit active lever pressing in either cue- or drug-induced reinstatement tests in control heroin mice. However, knockout of thalamic μ-ORs enabled optogenetic activation of PVT→NAc neurons to suppress heroin seeking primed by heroin-conditioned cues or heroin itself (Fig. 7G–H). In the absence of optogenetic activation, both control heroin and μ-OR KO heroin mice reinstated active lever pressing (Fig. S7J–K). Furthermore, optogenetic stimulation did not alter inactive lever pressing in either control or μ-OR KO heroin mice (Fig. S7L–M), indicating this manipulation was specific to the heroin-paired lever and goal-directed behavior. Thus, presynaptic PVT μ-ORs serve as a mechanism for heroin-induced behavioral disinhibition.

Our previous work established that acute opioid exposure rapidly disengages the thalamo-accumbal behavioral suppression system, functionally removing a physiological ‘brake’ on sucrose-seeking behavior in a μ-OR-dependent manner¹⁵. However, whether chronic heroin exposure functionally disinhibits sucrose seeking through a similar opioidergic mechanism remains unclear. To address this issue, we used the same viral strategy as in Fig. 7F, and Oprm1^fl/fl mice were assigned to one of three groups for heroin self-administration or saline control sessions: (1) Control Saline (PVT∷eYFP); (2) Control Heroin (PVT∷eYFP); or (3) μ-OR KO heroin (PVT∷Cre). Following acquisition of heroin self-administration or saline control sessions, all mice underwent extinction training. Next, mice were trained to self-administer sucrose and acquired sucrose-taking behavior. We found that in the absence of optogenetic PVT→NAc activation, all groups self-administered sucrose reliably (Fig. S7N–O). However, PVT→NAc activation suppressed sucrose self-administration in mice with a history of control saline, but not heroin self-administration, behavioral training (Fig. 7I; Fig. S7P). These data reveal that chronic heroin use causes behavioral disinhibition of natural reward-motivated behaviors. Critically, knockout of thalamic μ-ORs restored the ability of PVT→NAc activation to suppress sucrose self-administration in heroin-experienced mice (Fig. 7I; Fig. S7P). PVT→NAc photoactivation did not alter inactive lever pressing across groups (Fig. S7P). These results suggest that heroin use functionally disinhibits both natural- and drug-reward seeking through thalamic μ-OR activation. In a subset of mice, we further tested whether heroin self-administration would also disinhibit sucrose seeking following stress. We show that while predator odor (2,5-dihydro-2,4,5-trimethylthiazoline; TMT) exposure suppressed sucrose self-administration in saline control mice, it did not in heroin-experienced mice (Fig. S7Q). Altogether, our findings reveal that thalamic μ-ORs provide a mechanism underlying heroin-induced weakening of PVT→NAc^PV-IN synapses and heroin-induced behavioral disinhibition.

DISCUSSION

Here we characterize a neuronal circuit mechanism for opioid-induced behavioral disinhibition. We identify PVT→NAc neuronal ensembles that develop acute and tonic inhibitory dynamics predictive of heroin self-administration and relapse to heroin seeking. These dynamics are stably maintained after acquisition in single neurons and are consistent across reinstatement modalities. Additionally, we find that heroin use causes long-lasting attenuation of PVT→NAc^PV-IN synapses, which are known to govern the suppression of reward seeking in drug-naïve rodents¹⁵. We reveal that restoration of activity at PVT→NAc^PV-IN synapses, but not PVT→NAc neurons or NAc PV-INs alone, abolishes cue-, drug-, and stress-induced reinstatement of heroin seeking. Next, we find a high density of Oprm1 expression in PVT→NAc cells, primarily on Drd2-positive neurons. Finally, we establish PVT μ-ORs as the functional mechanism underlying heroin-mediated weakening of PVT→NAc^PV-IN synapses and disinhibition of natural- and drug-motivated behaviors. Overall, we find that a keystone neuronal system for behavioral suppression becomes dysfunctional following heroin use but can be re-established to prevent heroin-seeking behaviors.

We and others have found that posterior PVT→NAc neurons are inhibited by naturalistic reward-predictive cues, and further inhibition of this circuit can enhance reward seeking when competing behavioral suppressors are present or when an expected reward is omitted^12,14,15. Similarly, here we identify PVT→NAc neuronal ensembles with inhibitory dynamics that are predictive of heroin-seeking behavior. These data indicate that decreased PVT→NAc activity encodes both natural- and drug-motivated behavioral actions. However, whereas optogenetic activation of PVT→NAc neurons is sufficient to suppress sucrose seeking in drug-naïve mice¹⁵, heroin use functionally weakens PVT→NAc^PV-IN synaptic efficacy and prevents PVT→NAc photoactivation from suppressing heroin- or sucrose-seeking behavior. Specifically, PVT→NAc^PV-IN synaptic efficacy is reduced following heroin self-administration, and this effect is not transient or reversible through extinction training. It is likely that dysregulation of PVT→NAc^PV-IN synapses is a result of continued heroin-driven stimulation of presynaptic thalamic μ-ORs, which can reduce the activity of presynaptic neurons^34–36 including those in PVT³⁷. The resulting chronic, heroin-induced reductions of presynaptic thalamic firing could lead to long-term depression postsynaptically at downstream PV-IN neurons. This idea is supported by our evidence that (1) it is necessary to restore activity in both PVT→NAc projection neurons and NAc PV-INs to rescue behavioral inhibition, (2) knockout of presynaptic thalamic μ-ORs prevents heroin-induced postsynaptic adaptations within the PVT→NAc^PV-IN circuit, and (3) knockout of thalamic μ-ORs preserves the ability of PVT→NAc circuit excitation to suppress drug- and sucrose-seeking behavior. Interestingly, optogenetic inhibition of PVT→NAc neurons had no effect on active or inactive lever pressing in heroin-seeking mice. There may be two explanations for this. First, in heroin-naïve, sucrose-seeking mice, optogenetic inhibition of PVT→NAc neurons unleashes reward-seeking behavior only when a competing behavioral suppressor is present¹⁵. However, we did not include a behavioral suppressor in our reinstatement of heroin seeking experiments. Second, PVT→NAc neurons are profoundly inhibited during reinstatement of heroin seeking, such that further inhibition of these neurons may have little effect on overall neuronal activity.

Overall, our two-photon calcium imaging data reveal that PVT→NAc inhibitory dynamics are predictive of heroin self-administration and reinstatement, though significant heterogeneity was observed (e.g., active, somewhat inhibited, and greatly inhibited ensembles). This may be due to PVT containing classes of neurons that can be distinguished by anatomical location^{31,32,38–40}, function^41–43, gene expression^31,32,43, and input/output connectivity^20,44. As a result, a mechanistic understanding of how the PVT orchestrates opioid-related behaviors requires identifying heroin-induced adaptations within thalamo-accumbal neuronal subpopulations and downstream projection targets. While our data suggest that heroin-mediated depression of PVT→NAc^PV-IN synapses is vital for behavioral disinhibition, we also identify changes in synaptic physiology within PVT→NAc^D1-MSN and PVT→NAc^D2-MSN synapses. Thalamo-accumbal innervation of D1- or D2-MSNs is critical for maintenance of context-opioid associations and drive opioid-conditioned appetitive²⁵ and aversive behaviors¹⁷. In the current study, we find selective, transient increases in PVT→NAc^D2-MSN excitatory drive and presynaptic release probability 24-hours after heroin self-administration. While we did not specifically examine how PVT→NAc neurons regulate acute or protracted withdrawal, our findings are in line with work demonstrating acute morphine withdrawal relies on the potentiation of PVT→NAc^D2-MSN synapses²⁶. Conversely, while there were no changes in PVT→NAc^D1-MSN synaptic strength following heroin use and extinction training, PVT→NAc^D1-MSN synapses became more inwardly rectifying after extinction. This suggests insertion of CP-AMPArs at PVT→NAc^D1-MSN synapses, increasing the potential for synaptic plasticity^26,30. Interestingly, PVT→NAc^D1-MSN synapses have been shown to differentially regulate heroin seeking following either an abstinence period or extinction training¹⁷. Following abstinence from heroin, PVT→NAc photoactivation can promote drug seeking. However, extinction significantly decreases PVT→NAc^D1-MSN synaptic plasticity and blocks the ability of PVT→NAc activation to promote heroin-seeking behavior¹⁷. The conflicting behavioral phenotypes observed following PVT→NAc activation may be a result of the divergent neurobiological mechanisms involved in abstinence or extinction training^17,30,45,46 combined with the heterogeneous activity profiles of thalamic projection neurons and downstream synaptic targets^14,31,47. Collectively, our findings are consistent with the idea that the opioid-mediated plasticity observed in PVT→NAc^D1-MSN and PVT→NAc^D2-MSN synapses is either transient or dependent on extinction training. In contrast, we observed persistent heroin-induced reductions in PVT→NAc^PV-IN postsynaptic, but not presynaptic, efficacy. Consequently, we targeted PVT→NAc^PV-IN synapses to determine whether restoring activity in this circuit would restore behavioral inhibition. Our findings demonstrate that neither optogenetic activation of PVT→NAc neurons nor chemogenetic excitation of NAc PV-INs alone can restore behavioral inhibition. However, when these manipulations are combined with the purpose of restoring PVT→NAc^PV-IN synaptic coupling, heroin seeking is abolished. These findings lend support to the idea that PVT→NAc^PV-IN synaptic depression underlies opioid-induced behavioral disinhibition.

Here we determined that PVT μ-ORs serve as a mechanism for behavioral disinhibition resulting from chronic heroin use. Consistent with previous findings showing high μ-OR expression in PVT^15,35,36, we find that Oprm1 is particularly enriched in Drd2-expressing PVT→NAc neurons. Drd2-positive neurons have been shown to densely populate posterior PVT^31,48,49, and their activity can drive aversion³¹ and encode goal-directed behaviors⁴⁹. Considering our findings that μ-ORs are most densely expressed in Drd2-positive PVT→NAc neurons, it is possible that heroin-induced behavioral disinhibition relies specifically on μ-OR stimulation within this neuronal subpopulation. However, whether genetically defined, projection-specific PVT cell types differentially influence opioid-driven behaviors has not yet been determined.

Recent technical advances now allow for deep-brain holographic optogenetic modulation of single neurons in awake, behaving animals^50–52. Future studies can therefore employ holographic optogenetics to determine the function of unique PVT neuronal ensemble dynamics for opioid-mediated behaviors. Further, by combining calcium imaging studies with spatial transcriptomics^53,54, the molecular profile of these unique neuronal ensembles could be defined. Overall, our discovery that PVT→NAc^PV-IN circuit adaptations underlie opioid-induced behavioral disinhibition, a hallmark of opioid use disorder, suggests that further study of this circuit could aid in the development of therapeutics aimed to restore behavioral inhibition for relapse prevention.

STAR Methods

RESOURCE AVAILABILITY

Lead Contact

Further information and request for resources and reagents should be directed to and will be fulfilled by the lead contact, James M. Otis (otis@musc.edu).

Materials Availability

This study did not generate new unique reagents.

Data and Code Availability

Behavioral and electrophysiological data generated in this study, and all original code, have been deposited in the Otis Lab GitHub database and are publicly available as of date of publication. Accession links are listed in the key resources table. Two-photon imaging datasets will be made available upon request but are not immediately available for download due to file size. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies/Probes
Mouse monocolonal antibody - NeuN	Millipore	RRID AB_2298772
RNAscope® Probe - Mm-Oprm1-C1	Advanced Cell Diagnostics	Cat#315841
RNAscope® Probe- Mm-Col12a1-C3	Advanced Cell Diagnostics	Cat#312631-C3
RNAscope® Probe- Mm-Drd3-C3	Advanced Cell Diagnostics	Cat#447721-C3
RNAscope® Probe- Mm-Esr1-C2	Advanced Cell Diagnostics	Cat#478201-C2
RNAscope® Probe - Mm-Drd2-C3 -	Advanced Cell Diagnostics	Cat#406501-C3
RNAscope® Probe- Mm-Oprk1-C2	Advanced Cell Diagnostics	Cat#316111-C2
RNAscope® Probe Diluent	Advanced Cell Diagnostics	Cat#300041
TSA Plus Fluorescein 50-150 slides - 1X Plus Amplification Diluent (15 mL); Fluorescein Plus Amplification Reagent (2 tubes, add 150 μL DMSO/tube)	Akoya Biosciences	Cat#NEL741001KT
TSA Plus Cyanine 3 50-150 slides - 1X Plus Amplification Diluent (15 mL); Cyanine 3 Plus Amplification Reagent (2 tubes, add 150 μL DMSO/tube)	Akoya Biosciences	Cat#NEL744001KT
TSA Plus Cyanine 5 50-150 slides - 1X Plus Amplification Diluent (15 mL); Cyanine 5 Plus Amplification Reagent (2 tubes, add 150 μL DMSO/tube)	Akoya Biosciences	Cat#NEL745001KT
Bacterial and virus strains
rgAAV2-CAG-Cre	UNC Vector Core	RRID:Addgene 196140; Lot#AV7703F
AAV5-DIO-Ef1α-ChR2-EYFP	Addgene	RRID:Addgene 35509
AAV5- DIO-Ef1α-eNpHR3.0-EYFP	Addgene	RRID:Addgene 26966; Lot#AV4806H
AAV5-hSyn-ChrR-tdT	Addgene	RRID:Addgene 59171
AAV5-hSyn-hChR2-eYFP	Addgene	RRID:Addgene 26973
AAV5-hSyn-eYFP	Addgene	RRID:Addgene 117382
AAV5-hSyn-GFP-Cre	UNC Vector Core	RRID:Addgene 175381; Lot#AVV6446C
AAV1-s5e2-GCaMP6f	Addgene	RRID:Addgene 135632
AAV5-hSyn-DIO-hM4D(Gq)-mCherry	Addgene	RRID:Addgene 44361
AAVdj-ef1a-DIO-GCaMP6m	UNC Vector Core	Lot#AV6505
AAV5-ef1a-DIO-eYFP	Addgene	RRID:Addgene 27056
rgAAV2-hSyn-ChR2-eYFP	Addgene	RRID:Addgene 26973
Red Retrobeads™	Lumafluor, Inc.	N/A
AAV5-hSyn-ChrimsonR-tdTomato	Addgene	RRID:Addgene 59171
Chemicals, peptides, and recombinant proteins
DNQX	Tocris	Cat#0189
Picrotoxin	Tocris	Cat#1128
DAMGO	Tocris	Cat#1171
Clozapine-N-oxide	Tocris	Cat#4936
Dimethyl Sulfoxide	Sigma-Aldrich	Cat#276855
AP5	Tocris	Cat#0106
IEM1460	Tocris	Cat#1636
TMT	SRQ	Cat#1G-TMT-97
Critical commercial assays
RNAscope® Multiplex Fluorescent Detection Kit v2	Advanced Cell Diagnostics	Cat#323110
RNAscope® 50x Wash Buffer Reagents	Advanced Cell Diagnostics	Cat#310091
RNAscope® Hydrogen Peroxide	Advanced Cell Diagnostics	Cat#322335
RNAscope® Multiplex TSA Buffer	Advanced Cell Diagnostics	Cat#322809
Deposited data
Behavioral and electrophysiological data	This paper	https://github.com/jimotis/Paniccia-Vollmer-Green-et-al
Experimental models: Organisms/strains
Mouse: C57BL6/J (Wild-Type)	Jackson Laboratory	RRID:IMSR_JAX:000664
Mouse: B6.Cg-Pvalbtm1.1(cre)Aibs/J (PV-Cre)	Jackson Laboratory	RRID:IMSR_JAX:012358
Mouse: Drd2, line ER44 (D2-Cre)	Jackson Laboratory	RRID:MMRRC 017263-UCD
Mouse: B6.129-Oprm1tm1.1Cgrf/KffJ	Jackson Laboratory	RRID:IMSR_JAX:030074
Software and algorithms
Python (v2.7)	Anaconda	https://conda.io
FIJI - ImageJ (v2)	Schneider et al., 2012	RRID:SCR_003070
GraphPad PRISM (v8)	GraphPad	RRID:SCR_002798
Adobe Illustrator (v26)	Adobe	RRID:SCR_010279
SIMA (v1.3.2)	Kaifosh et al., 2014	RRID:SCR_024466
Clampfit (pClamp v11.2)	Molecular Devices	RRID:SCR_011323
Ocular (v2.0)	Teledyne Photometrics	RRID:SCR_024467
MatLab	MathWorks	RRID:SCR 001622
Arduino	Arduino	https://www.arduino.cc/
Other
GRIN lens; 8mm long, 0.5mm diameter	Inscopix	Cat#1050-004600
ImmEdge® Hydrophotic Barrier Pen	Advanced Cell Diagnostics	Cat#310018
ACD HybEZ™ II Hybridization System (110v)	Advanced Cell Diagnostics	Cat#321711
ACD HybEZ™ Humidity Control Tray (with lid)	Advanced Cell Diagnostics	Cat#310012
RNAscope® EZ-Batch™ Slide Rack (20 slide capacity) (with Wash Tray)	Advanced Cell Diagnostics	Cat#310007
Fisherbrand™ Superfrost™ Plus Microscope Slides	Fisher Scientific	Cat#12-550-15
473nm Blue DPSS Laser (T8) and Adjustable Power Supply	Shanghai Laser & Optics Century Co., Ltd.	Model BL473T8- 150FC + ADR-800A
0.39 NA, Ø200 μm Core Multimode Optical Fiber	ThorLabs	Part#FT200EMT
PFP Elite LC 1.25mm OD Multimode Ceramic Zirconia Ferrules, Ferrule ID Bore - 270 um	Precision Fiber Products, Inc.	Cat#MM-FER2007C-2700-E
Miller CS-30-W Carbide Scribe-Wedge	Precision Fiber Products, Inc.	Cat#M1-46124
Fiber Stripping Tool	ThorLabs	Cat#T12S21
Fiber Cable, MM, 200 μm, 0.5NA FC/PC to 1.25mm Ferrule, 1m	ThorLabs	Cat#M73L01
PFP 353ND Epoxy Bottles	Precision Fiber Products, Inc.	Cat#PFP-353ND1
CoolLED (pE-300white)	CoolLED	RRID:SCR 021073
Borosilicate glass capillaries	World Precision Instruments	Cat#50-821-819

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Animals

All experiments were approved by the Institutional Animal Care and Use Committee (IACUC) at the Medical University of South Carolina in accordance with the NIH-adopted Guide for the Care and Use of Laboratory Animals. Adult male and female C57BL6/J wild-type, PV-Cre (B6.Cg-Pvalbtm1.1(cre)Aibs/J, Strain #012358)⁵⁵, D2-Cre (Drd2, line ER44, RRID:MMRRC_017263-UCD)⁵⁶, and Oprm1^fl/fl (B6.129-Oprm1^tm1.1Cgrf/KffJ, Jax Strain #030074) mice were group-housed pre-operatively and single-housed post-operatively, with access to standard chow and water ad libitum throughout all experiments (mice were at least 8-weeks of age and 20g prior to study onset). Male and female mice were randomly assigned to experimental groups. Mice were housed under a reverse 12:12-hour light cycle (lights off at 8:00am), with experiments performed during the dark phase.

METHOD DETAILS

Surgery

For intracranial surgeries, mice were anesthetized with isoflurane (0.8–1.5% in oxygen; 1L/minute) and placed within a stereotactic frame (Kopf Instruments). Ophthalmic ointment (Akorn), topical anesthetic (2% Lidocaine; Akorn), analgesic (Ketorolac, 2 mg/kg, ip), and subcutaneous sterile saline (0.9% NaCl in water) were given pre- and intra-operatively for health and pain management. An antibiotic (Cefazolin, 200 mg/kg, sc) was given post-operatively to reduce the possibility of infection.

Two-photon calcium imaging surgeries:

To target PVT→NAc projection neurons for two-photon calcium imaging, a single microinjection of a Cre-inducible virus encoding the calcium indicator GCaMP6m (AAVdj-ef1a-DIO-GCaMP6m; 300nL) was infused into the posterior PVT (AP: −1.55mm; ML: −1.13mm; DV: −3.30mm; 20° angle), along with bilateral microinjections of a retrogradely-trafficked virus encoding Cre-recombinase (rgAAV2-CAG-Cre; 500nL/side) aimed for the anterior/medial NAc shell (AP: +1.45mm; ML: ± 0.65mm; DV: −4.65mm) in wild-type mice. A microendoscopic gradient refractive index lens (GRIN lens; 8mm long, 0.5mm diameter, Inscopix) was then implanted dorsal to the PVT injection site (AP: −1.55mm; ML: −1.13mm; DV: −3.00mm; 20° angle), allowing chronic visual access to PVT→NAc projection neurons^20,57. Next, a stainless-steel head ring was cemented around the GRIN lens using dental cement and skull screws for later head fixation²³. GRIN lens placement and GCaMP6m fluorescence of posterior PVT→NAc neurons was confirmed post-mortem (Fig. S5D).

Behavioral optogenetics and chemogenetics surgeries:

We targeted PVT→NAc projection neurons for optogenetic manipulations using injection coordinates and volumes that were akin to two-photon calcium imaging experiments described above. For optogenetics experiments (see Fig. 3), we gave a single microinjection of a Cre-inducible virus encoding one of the three constructs (AAV5-ef1a-DIO-ChR2-eYFP; AAV5-ef1a-DIO-eNpHR3.0-eYFP; AAV5-ef1a-DIO-eYFP) into the posterior PVT, along with bilateral microinjections of the retrogradely-trafficked virus encoding Cre-recombinase into the NAc shell (rgAAV2-CAG-Cre) of wild-type mice. In later experiments, when combining optogenetics with chemogenetics (see Fig. 5), we instead injected a retrogradely trafficked virus encoding channelrhodopsin (rgAAV2-hSyn-ChR2-eYFP) bilaterally into NAc shell, along with bilateral microinjections of a Cre-inducible virus encoding an excitatory DREADD^58,59 into the NAc shell (AAV5-hSyn-DIO-hM4D(Gq)-mCherry) of PV-Cre or wild-type control mice. Finally, a bilateral double-barrel guide cannula (Plastics One: 26-gauge, 5mm length, 1.2mm barrel separation) was implanted dorsal to the anterior/medial NAc shell allowing local CNO microinfusion (AP: +1.45mm; ML: ± 0.60mm; DV: −4.15mm). For μ-OR knockout optogenetic experiments (see Fig. 7), a retrogradely trafficked virus encoding channelrhodopsin (rgAAV2-hSyn-ChR2-eYFP) was bilaterally infused into NAc shell and a virus encoding either eYFP control (AAV5-hSyn-eYFP) or Cre-recombinase (AAV5-hSyn-GFP-Cre) was infused into the posterior PVT of Oprm1^fl/fl mice. For all optogenetics experiments, a custom-made optical fiber⁶⁰ was implanted dorsal to the PVT injection site (AP: −1.55mm; ML: −1.13mm; DV: −3.00mm; 20° angle), allowing laser-evoked perturbation of activity in PVT→NAc projection neurons. For all optogenetics and chemogenetics surgeries, a stainless-steel head ring was cemented around the optical fiber and/or guide cannula using dental cement and skull screws. Optical fiber, cannula, and viral placements were confirmed post-mortem via histology (Fig. S5B–E).

Slice electrophysiology surgeries:

PVT→NAc projection neurons were targeted for slice electrophysiological studies using injection coordinates that were identical to two-photon calcium imaging experiments described above. A virus encoding a red-shifted excitatory opsin (AAV5-hSyn-ChrimsonR-tdTomato; 400nL)²⁸ was injected into the posterior PVT, along with bilateral microinjections of a Cre-inducible virus encoding an eYFP into the NAc shell (AAV5-ef1a-DIO-eYFP; 400nL/side) of D2-Cre or PV-Cre mice. For PVT μ-OR knockout electrophysiology experiments, we combined AAV-eYFP (AAV5-hSyn-eYFP) or AAV-Cre (AAV5-hSyn-Cre-eGFP) with AAV-Channelrhodopsin (AAV5-hSyn-ChR2-eYFP)²⁸ at a 1:10 ratio (400nL), and microinjections were targeted to the posterior PVT of Oprm1^fl/f mice. For identification of PV-INs, bilateral NAc shell microinjections of a PV-interneuron targeted calcium indicator (AAV1-s5e2-GCaMP6f; 400nL)⁶¹, which expresses GCaMP6f under control of the E2 regulatory element, were administered.

Retrobead surgeries:

PVT→NAc projection neurons were targeted for RNAscope^® experiments examining the colocalization of Oprm1 with known genes expressed in posterior PVT neurons (Col12a1, Drd3, Esr1, or Drd2). Microinfusions of Red Retrobeads^™ (Lumafluor Inc.) were directed at the NAc shell (AP +1.5mm; ML ±1.75mm; DV −4.75mm; 10° angle; 300 nL/side). The incision site was closed using Vetbond^™ tissue adhesive (cat#1469SB; 3M). Mice were given 48-hours for postoperative care for Retrobead^™ expression, prior to euthanization.

IV Catheter Surgeries:

Most drug self-administration mice were allowed at least 7-days of recovery from intracranial surgery before catheterization occurred, with a subset of mice receiving IV catheters within the same intracranial surgery. All mice were anesthetized as described above and were implanted with custom-made intravenous catheters, using a method previously described⁶². Catheters were implanted subcutaneously with the tubing inserted into the external jugular vein. All mice received analgesic, ophthalmic, and antibiotic treatments as described above, as well as topical antibiotic ointment and lidocaine (2%) jelly around incision cites. Following a minimum of 7-days of recovery, mice began behavioral experiments wherein catheters were flushed daily with heparinized saline (60 units/mL, 0.02mL) to maintain patency. Mice with non-patent catheters were to be excluded from the study, however all mice remained patent throughout acquisition and catheter patency was no longer monitored once mice entered extinction. If necessary, patency was determined by giving mice an i.v. infusion of brevital (2 mg/mL, 0,02 mL).

Head-fixed Behavior

Heroin Self-administration:

Experiments involving heroin self-administration or saline control sessions were performed based on a previous study wherein we developed a model of natural- and drug-reward seeking in head-restrained mice, enabling simultaneous two-photon calcium imaging⁶². After recovery from surgery, mice were habituated for 3-days to head fixation during 30-minute sessions wherein levers were not presented.

Acquisition:

Mice next underwent heroin self-administration or saline control sessions through 14 daily 2-hour sessions, during which two levers were placed within forelimb reach. A press on the active lever, but not inactive lever, resulted in the presentation of a tone cue (8 kHz, 1.6s) followed by the intravenous infusion of heroin or saline (administered over a 2s epoch). A timeout period (20s) was given after each cue- and heroin-reinforced active lever press, wherein active lever pressing had no effect. Heroin self-administering and saline control mice were trained on a fixed ratio 1 (FR1) schedule of reinforcement using a decreasing dose design (Day 1–2: 0.1 mg/kg/12.5 μL heroin, 10 infusion maximum; Day 3–4: 0.05 mg/kg/12.5 μL heroin, 20 infusion maximum; Day 5–14: 0.025 mg/kg/12.5 μL heroin, 40 infusion maximum), for a maximum of 1 mg/kg of heroin per session (as previously described⁶²; a 0.9% saline/12.5 μL infusion was used on all days for saline control mice). Due to quick responding on the active lever, mice were capped to receiving 1 mg/kg per session to prevent overdose. Self-administration sessions were a maximum of 2-hours. A subset of mice only underwent 13-days of heroin self-administration or saline control sessions, as these mice were euthanized 24-hours later for patch clamp electrophysiology. Additionally, we confirmed that electrophysiological results in saline control mice were equivalent to results in naïve mice, and thus those datasets were merged.

Extinction:

Following acquisition, heroin self-administering mice underwent 2-hour extinction training sessions, wherein active lever presses resulted in neither cue nor drug delivery until extinction criteria were reached. Extinction criteria were determined a priori, as (1) ≥ 9-days of extinction training and (2) the last 2-days of extinction training resulting in ≤20% of the average active lever pressing observed during the last 2-days of acquisition. Heroin self-administering mice that did not reach extinction criteria following 20-days of consecutive extinction training sessions were excluded from the study (n=2). In electrophysiology experiments used in Fig. 4, saline control mice did not undergo extinction training as they were euthanized 24-hours following their last saline control session. For μ-OR KO studies that employed saline control mice, these animals received comparable 2-hour sessions to the heroin-experienced group, in which they were given access to the levers in the head-fixed set-up. Following a minimum of 10-days of extinction training, control saline Oprm^fl/fl mice were either euthanized for electrophysiological studies or underwent sucrose self-administration training.

Reinstatement:

After extinction criteria was reached, heroin self-administering mice underwent cue-, drug-, or yohimbine-induced reinstatement tests in a pseudorandom order. For cue-induced reinstatement testing, active lever presses resulted in cue presentation as in acquisition, however drug infusions were excluded. A timeout period (20s) was given after the onset of each cue, wherein active lever pressing did not result in cue delivery. For drug-induced reinstatement, mice received an acute injection of heroin (1 mg/kg, ip) immediately before the session, and active lever presses resulted in neither cue nor drug delivery. For yohimbine-induced reinstatement, mice received an acute injection of the pharmacological stressor, yohimbine (0.625 mg/kg, ip; Sigma Chemical)⁶³ 15-minutes before the session; active lever presses did not result in cue nor drug delivery. Mice were given at least 2-days of extinction between reinstatement tests to ensure that lever pressing was ‘re-extinguished’ before each test.

Sucrose self-administration and testing:

To assess whether heroin self-administration caused behavioral disinhibition of sucrose-seeking behavior, Oprm1^fl/fl mice underwent sucrose self-administration training following either the heroin-self administration or saline control sessions described above. Mice received at minimum 6, 1-hour sucrose self-administration sessions in which animals were head-restrained and with access to an active lever and inactive lever. A press on the active lever would result in a 3 kHz tone (1.6-seconds) and a 12.5 μL droplet of sucrose from a metal lick spout in front of the animal, while a press on the inactive lever resulted in no cue or reward delivery. A 20-second time out period followed an active lever press, in which an animal could press the levers without resulting in cue or reward. During acquisition mice were capped at 40 sucrose rewards maximum per training session. Following acquisition of sucrose-taking behavior, animals underwent 45-minute, uncapped sucrose seeking tests in the presence or absence of 473 nm laser excitation of ChR2-expressing PVT→NAc neurons (order counterbalanced). As during previous acquisition sessions, active lever presses resulted in a tone cue and sucrose delivery followed by the 20-second time-out period, and an inactive lever press yielded no cue or reward. Following optogenetic testing, a subset of animals (control heroin and control saline mice) underwent 45-minute, uncapped sucrose seeking tests that were preceded by a 15-minute exposure to the fox feces derivative 2,5-dihydro-2,4,5-trimethylthiazoline (TMT; 30 μL; 1% v/v ddH2O) or vehicle (0.9% sterile saline; order counterbalanced). Mice were given at minimum two sessions in between each uncapped sucrose test to re-establish sucrose-seeking behavior.

Two-photon Calcium Imaging

Data collection and processing:

We visualized GCaMP6m-expressing PVT→NAc projection neurons using a two-photon microscope (Bruker Nano Inc) equipped with a tunable InSight DeepSee laser (Spectra Physics, laser set to 920nm, ~100fs pulse width), resonant scanning mirrors (~30Hz framerate), a 20X air objective (Olympus, LCPLN20XIR, 0.45NA, 8.3mm working distance), and GaAsP photodetectors. In some cases, two fields of view (FOVs) were visible through the GRIN lens (separated by >75μm in the Z-axis to avoid signal contamination from chromatic aberration), in which case we recorded from each FOV during separate imaging sessions. Data were acquired without averaging using PrarieView software, converted into hdf5 format, and motion corrected using SIMA⁶⁴. Following motion correction, a motion-corrected video and averaged time-series frame were used to draw regions of interest (ROIs) around dynamic and visually distinct cells using the polygon selection tool in FIJI⁶⁵. Fluorescent traces for each ROI were then extracted using SIMA, and all subsequent analyses were performed using custom Python codes in Jupyter Notebook^20,22. Two-photon imaging was performed during select acquisition sessions (early: days 1–2; middle: days 7–8; late: days 13–14) and extinction sessions (early: days 1–2; late: last 2-days) to simplify data analysis, and during all reinstatement tests.

Behavioral Optogenetics and Chemogenetics

Optogenetics:

We used optogenetics to stimulate or inhibit the activity of PVT→NAc neurons during cue-, drug-, or yohimbine-induced reinstatement tests, as well as during sucrose seeking tests. For photoactivation experiments in ChR2 or control eYFP mice, the laser (473nm; ~10mW) was pulsed (5ms; 20Hz) for 30-second intervals once/minute throughout the session. For photoinhibition experiments in eNpHR3.0 or control eYFP mice, the laser (532nm; ~10mW) was displayed (pure light, not pulsed) for 30-second intervals once/minute throughout the session. Because each laser manipulation had no effect in control eYFP mice, those data were collapsed across groups for each experiment.

Chemogenetics:

We used site-specific chemogenetics to determine the function of NAc PV interneurons for the expression of heroin-seeking behaviors. Microinfusions of CNO (0.1μg in 0.3μL) were administered into NAc 5-minutes before each behavioral session. Sessions wherein mice received chemogenetics alone or chemogenetics combined with optogenetics were counterbalanced.

Patch-clamp Electrophysiology

Mice were anesthetized with isoflurane (1.5% in oxygen; 1L/minute) before transcardial perfusion with oxygenated, ice-cold sucrose-based cutting solution containing the following (in mM): 225 sucrose, 119.0 NaCl, 1.0 NaH₂PO₄, 4.9 MgCl₂, 0.1 CaCl₂, 26.2 NaHCO₃, 1.25 glucose (305–310 mOsm). Brains were then rapidly removed and bathed in cutting solution, while coronal sections 300μm thick were taken using a vibratome (Leica VT1200S). Sections were incubated in warm aCSF (32° C) containing the following (in mM): 119 NaCl, 2.5 KCl, 1.0 NaH₂PO₄, 1.3 MgCl, 2.5 CaCl₂, 26.2 NaHCO₃, 15 glucose (305–310 mOsm). After at least 1-hour of recovery, slices were constantly perfused with oxygenated aCSF and visualized using differential interference contrast (DIC) through a 40X water-immersion objective mounted on an upright microscope (Olympus BX51), and a microscope-mounted camera (Scientifica, SciCam Pro, Oculus). Whole-cell patch-clamp recordings were obtained using borosilicate pipettes (~3–5MΩ) back-filled with a potassium gluconate-based internal solution composed of the following (in mM): 130 K-gluconate, 10 KCl, 10 HEPES, 10 EGTA, 2 MgCl₂, 2 ATP, 0.2 GTP (pH 7.35, mOsm 280) to characterize intrinsic excitability, sEPSCs, or oeEPSCs. Alternatively, recordings were obtained using a cesium methylsulfonate-based internal solution composed of the following (in mM): 117 Cs methanesulfonic acid, 20 HEPES, 2.8 NaCl, 5 TEA, 2 ATP, 0.2 GTP (pH 7.35, mOsm 280) to measure the amplitude of synaptic currents (mediated by AMPA and NMDA receptors) and AMPA receptor rectification. Electrophysiological data acquisition occurred at a 10-kHz sampling rate through a MultiClamp 700B amplifier connected to a Digidata 1550B digitizer (Molecular Devices) and were analyzed using Clampfit 11.2 (Molecular Devices).

NAc Electrophysiology:

Patch-clamp recordings were obtained from eYFP⁺ (PV-INs in PV-Cre mice, D2-MSNs in D2-Cre mice) and eYFP⁻ (putative D1-MSNs in D2-Cre mice) neurons surrounding the virus injection site in anterior/medial NAc shell. eYFP was visualized using a blue LED (490nm; <1mW) integrated into the light path (CoolLED: pE-300white) and a GFP epifluorescence filter set. In a subset of neurons, depolarizing current pulses (800ms; 50pA steps) were applied in current-clamp mode to confirm that recordings were from fluorescence-identified cell types. Specifically, PV-INs were confirmed in a subset of neurons based on their fast-spiking properties, whereas D1- and D2-MSNs were confirmed by their relatively limited spike frequency, ramping depolarization, and/or late spiking features. We also evaluated functional synaptic innervation from PVT to each of the NAc cell types. Visually identified NAc neurons were held at −70mV in voltage-clamp mode, and presynaptic ChrimsonR-tdT⁺ axons from PVT were activated using a green LED (545nm; single 10ms pulse, 1mW) applied every 10–15-seconds, resulting in optically evoked excitatory postsynaptic currents (oeEPSCs). To investigate presynaptic release probability, two LED pulses in quick succession were applied to induce two oeEPSCs (545nm; 10ms pulses, 1mW, 50ms apart). The paired-pulse ratio was calculated by dividing the amplitude of the second peak by the amplitude of the first peak. In a subset of recordings, we confirmed that oeEPSCs were blocked by 10-minute bath application of the glutamatergic AMPA receptor antagonist DNQX (10μM). AMPAr/NMDAr ratio and AMPAr rectification were measured using the cesium-based internal solution in voltage-clamp mode, with a GABA receptor antagonist picrotoxin (100μM). For AMPAr/NMDAr ratios, we optogenetically evoked EPSCs at holding potentials of −80mV and +50mV to isolate fast AMPAr- and slow NMDAr-mediated oeEPSCs, respectively. AMPAr/NMDAr ratios were calculated by dividing the amplitude of the AMPAr current (peak current, at −80mV) by the amplitude of the NMDAr current (current at 50ms following start of pulse, at +50mV). AMPAr rectification was recorded in the presence of both picrotoxin (100 μM) and the NMDA receptor antagonist APV (50μM)²⁶. Rectification was measured by evoking oeEPSCs as above, with 5 sweeps averaged at a range of voltages (−80, −70, −50, −30, −10, +10, +30, +50 mV). AMPAr rectification data were normalized to the peak oeEPSC at −80mV, and an AMPAr rectification index was calculated for each neuron as I−70/I+50 where I is the peak oeEPSC amplitude at each voltage.

DIO-Gq-DREADD validation:

For Gq-DREADD validation, patch-clamp recordings were obtained from mCherry+ PV-INs in the anterior/medial NAc shell of PV-Cre mice. In current-clamp, neurons were stimulated using direct current injection (800ms; 200pA) every 10-seconds. PV-IN spiking was recorded for a 2.5-minute baseline, 5-minute CNO wash-on period, and 5-minute post-wash period. In the control group, aCSF was perfused across all three recording periods, whereas the CNO group received aCSF combined with CNO (1μM)⁵⁹ during the wash-on period.

NAc electrophysiology with PVT μ-OR knockout:

In Oprm1^fl/fl mice that received an infusion of Cre or eYFP in the posterior PVT, patch-clamp recordings were obtained from GCaMP6f⁺ (PV-INs) neurons surrounding the virus injection site in anterior/medial NAc shell. In a subset of GCaMP6f⁺ neurons (visible with the blue LED described above), depolarizing current pulses (800ms; 50pA steps) were applied in current-clamp mode to confirm PV-IN-like waveform properties. Next, the GCaMP6f⁺ neurons were held at −70mV in voltage-clamp mode, and presynaptic Channelrhodopsin-eYFP⁺ axons from PVT were activated using a blue LED (10ms pulse, 1mW) pulsed every 10–15-seconds. oeEPSCs and AMPAr/NMDAr ratios were collected using the methods described above.

PVT electrophysiology with μ-OR knockout:

Patch-clamp recordings were obtained from eGFP⁺ or eYFP⁺ neurons in the posterior PVT of eGFP-Cre or eYFP-control infused Oprm1^fl/fl mice. Current-clamp recordings were obtained to identify the intrinsic adaptations in PVT caused by μ-OR knockout in drug-naïve mice. First, action potential firing was examined by applying a series of long depolarizing sweeps (800ms) at +50pA steps (−100–400pA). Next, rheobase (the minimum amount of current required for an action potential to fire) was measured by applying a series of short depolarizing sweeps (50ms) at +5pA steps (starting at 0pA) until the neuron fired an action potential. Finally, in voltage-clamp mode, synaptic changes were examined through collection of spontaneous EPSCs (sEPSCs) over a 2-minute period while neurons were held at −70 mV.

μ-OR knockout validation:

In current-clamp mode, cells were held at −70mV and depolarizing current pulses (800ms; 10–300pA) were applied across 15 10-second sweeps in an attempt to evoke 4–10 action potentials per sweep. Baseline recordings were taken directly before drug application, whereas ‘DAMGO’ recordings were taken immediately following 25-minute bath application of DAMGO (3μM)⁶⁶.

RNAscope^®

Animals were deeply anesthetized using isoflurane and tissue was collected via rapid decapitation. Brain tissue was extracted and rinsed with 0.1M phosphate buffered saline (PBS) and immediately submerged into −40°C 2-methyl-butane for 45-seconds. Tissue was stored at −20°C until sectioning where brains were cut into 20μm coronal sections via Leica cryostat and immediately mounted onto Superfrost Plus slides (FisherScientific). Slides were stored at −20°C until assay 1-day later. Three to four adjacent posterior PVT coronal slices (AP −1.2 mm to AP −1.7mm) per mouse were used in the μ-OR knockout validation study or in the μ-OR colocalization experiment. Slides were processed with RNAscope^® probes and reagents according to the RNAscope^® Multiplex Fluorescent Reagent Kit v2 manufacturer’s instructions (Advanced Cell Diagnostics) and as previously published¹⁵. Briefly, tissue was fixed in 4% paraformaldehyde for 15-minutes, followed by two washes in PBS. Tissue sections were dehydrated in sequential washes of 50%, 70%, and 100% ethanol and a hydrophobic barrier was drawn around sections to prevent the spread of solutions on each slide. For validation of Cre-induced knockout of Oprm1 in the Oprm1^fl/fl mouse line, a RNAscope Reagent kit was used, and the slides were incubated in RNAscope^® Hydrogen Peroxide for 10-minutes, rinsed twice in dH2O, and then treated with Mm-Oprm1 and Mm-Oprk1 for 2-hours in an ACD hybridization oven at 40°C. For tissue used in the PVT→NAc μ-OR colocalization experiment, a RNAscope^® Reagent kit was used, and all tissue was incubated in hydrogen peroxide and treated with Mm-Oprm1 and either Mm-Col12a1, Mm-Drd3, Mm-Esr1, or Mm-Drd2 in the hybridization oven for 2-hours. For all slices, sections were incubated with amplification (AMP) solution1 for 30-minutes at 40°C, solution2 for 15-minutes at 40°C, and solution3 for 30-minutes at 4°C, with a double rinse in 1xRNAscope^® Wash Buffer for 2-minutes between each amplification step. Next, slides were incubated with fluorescently labeled probes (Akoya Biosciences) to distinguish each channel. For the validation experiment, TSA + Cyanine 3 (1:1500 dilution) and TSA + Cyanine 5 (1:900 dilution) were used. For the colocalization experiment, TSA + Fluorescein (1:1500 dilution) was used for visualization of Oprm1 expression and TSA + Cyanine 5 (1:900 dilution) was used to visualize Col12a1, Drd3, Esr1, or Drd2. The tissue was incubated in a drop of DAPI for 30-seconds, then coverslipped with Prolong Gold fluorescent mounting medium (Invitrogen).

Immunohistochemistry

Free-floating 80μm coronal sections containing the PVT were blocked in 0.1M PBS with 2% Triton X-100 (PBST) with 2% normal goat serum (NGS, Jackson Immuno Research, Westgrove, PA) for 2-hours at room temperature with agitation. Sections were then incubated overnight at 4°C with agitation in NeuN primary antisera diluted in 2% PBST with 2% NGS, washed 3 times for 5-minutes in PBST, then incubated in the appropriate secondary antisera diluted in PBST with 2% NGS for 4-hours at room temperature with agitation. Secondary antisera were raised in goat, conjugated to Alexa fluorophores, were used at a concentration of 1:1000, and were purchased from Invitrogen (Carlsbad, CA). Sections were then washed 3 times for 5-minutes in PBST, mounted on SuperFrost+ slides, and cover slipped with ProLong^™ Gold Antifade. Slides were stored in a dark area. Brain sections were imaged using a Leica SP8 laser-scanning confocal microscope. Care was taken to only acquire images within dense fields of virally transduced neurons (eYFP+) in the PVT. Simultaneously, we also imaged immunohistochemically (IHC) labeled markers for NeuN. For detection of eYFP+ cells, an OPSL 488nm laser line was used. For NeuN, a Diode 638nm laser line was used for detection. During all imaging experimentation, a frame size of 1024×1024 was used and pinhole size, laser power, gain, and Z-step thickness were held constant throughout.

Confocal Microscopy and Imaris Image Analysis

For projection- and cell type-specific quantification of μ-OR gene expression, RNAscope^® was performed for Oprm1 in combination with probes for Col12a1, Drd3, Drd2 or Esr1³² representing 4 separate RNAscope^® staining batches. Confocal images of RNAscope^® stained slides were imaged where Oprm1 signal was acquired in 488 nm laser line, NAc-derived retrobead signal was acquired in 552 nm laser line, and Col12a1, Drd3, Drd2 or Esr1 signal was acquired in 638 nm laser line. Individual optical datasets were comprised of Z-series digital image sets through each section, which were acquired on a Leica SP8 laser-scanning confocal microscope with a 10X air objective using an optical section thickness of 1μm. All imaging parameters (i.e., optical zoom, field size, laser power, gain and pinhole) were kept consistent between each Z-series and each animal for each round of imaging. Following acquisition, datasets were imported into Imaris 9.0 (Bitplane). Within Imaris, the “cells” function was used to mark each cell, and then to calculate the associated total 3D volume of Oprm1, Retrobead^™, and Col12a1, Drd3, Drd2 or Esr1 signal in μm³ (Fig. S6). Approximately 30–80 cells were identified within each optical dataset. 10–11 optical datasets were collected for each staining run (Oprm1 + Col12a1, Drd3, Drd2 or Esr1), across 6 Oprm1^fl/fl mice, with ~2 datasets collected for each mouse.

QUANTIFICATION AND STATISTICAL ANALYSIS

Two-photon Calcium Imaging

We quantified the average or ‘basal’ fluorescence of each neuron across time, as basal fluorescence can serve as a proxy for firing rates in tonically active cell populations^22,23, including PVT→NAc neurons²⁰. Fluorescent traces were averaged across 3-minute bins and normalized to the first 3-minutes of each session. Because heroin self-administration sessions were of varying length depending on the day and speed of intake, we compared the first and last 3-minutes of each session. These data were compared across time and sessions using a two-way ANOVA, followed by Sidak’s post-hoc analyses for between-session comparisons. In addition to these analyses, we include time course data showing fluorescence adaptations across each session, capped at 30-minutes for acquisition of self-administration, but uncapped for extinction or reinstatement (which were 2-hours in length), as most mice reached heroin infusion criteria after about 30-minutes and thus the sessions ended. Finally, we correlated the change in fluorescence for each neuron with active lever pressing responses on a minute-by-minute basis, by taking the average amplitude of fluorescence and number of active lever presses in any 1-minute epoch and subtracting the same measures from the previous 1-minute bin. These change scores for each minute for all neurons were then used to inform a Pearson-R analysis to determine the correlation between fluorescence adaptations and behavior. Pearson-R coefficients were calculated with two-tailed P-values at a 95% confidence interval, which are reported in Fig. S1 and S2.

In addition to fluorescence adaptations within sessions, we aligned fluorescent traces of each neuron to active lever presses, including the 4.5-seconds beforehand, 3-seconds between the lever press and heroin delivery, and 4.5-seconds after heroin delivery. The 12-second fluorescent trace was averaged across trials and plotted as a peri-stimulus time heatmap across neurons. Due to the robust active lever pressing rates during late acquisition sessions, early extinction sessions, and each reinstatement test, the resulting 2-dimensional arrays from those sessions were used to inform separate principal components analyses. The number of principal components were determined using the inflection point of a scree plot, which graphs the peristimulus time histogram of variance explained versus an increasing number of principal components. The remaining principal components were then plotted into a subspace and used to inform the Scikit-learn function sklearn.cluster.SpectralClustering, a spectral clustering algorithm that uses a k-nearest neighbor connectivity matrix to identify unique cell clusters. Spectral clustering was chosen due to its improved performance for separating dynamic neuronal datasets as compared with other clustering algorithms^22,24. Finally, a decoding analysis was used to determine how the activity of each neuron could predict future active lever pressing behavior. A binary decoder was used through the Scikit-learn functions sklearn.discriminant_analysis, sklearn.smv, and sklearn.decomposition, informed by the fluorescence of each neuron during 2 epochs: 1-second before each active lever press vs a random 1-second baseline. As a control, these 2 epochs were randomly shuffled, and the decoding analysis was repeated. The decoding scores for each neuron were subtracted from the average of shuffled data for the corresponding neuronal ensemble, and the data were plotted against the shuffled data for all neurons. These data were compared across ensembles and corresponding shuffled data using a two-way ANOVA, followed by Sidak’s post-hoc comparisons.

Subsets of neurons were reliably identified across days based on shape and relative position within each FOV, allowing us to visualize the adaptation and maintenance of responses from individual neurons across specific timepoints. Single-cell tracking was performed across early, middle, and late heroin self-administration, as well as from cue to drug, cue to stress, and stress to drug-induced reinstatement tests. To compare neuronal responses across these tracked sessions, we calculated the auROC which compared the fluorescence of a 1-second baseline to the fluorescence during and surrounding the lever press, cue, and/or infusion interval (from 0.5-seconds before the lever press to 3.5-seconds after the lever press). Due to the variance in the number of lever presses per session, these scores were then z-scored. Data were compared across tests using Pearson-R correlation tests and a two-way ANOVA, followed by Sidak’s post-hoc comparisons if required. Note that for reinstatement tracking, we display ‘grouped’ reinstatement tests in the main figure (Fig. 2K, L), which includes data for cue to drug, cue to stress, and drug to stress tracking (individual figures shown in Fig. S2).

Behavioral Data

Acquisition and Extinction:

Lever pressing was analyzed across levers and days using two-way ANOVAs followed by Sidak’s post-hoc comparisons when applicable. Drug infusions (heroin/saline) or sucrose deliveries were analyzed using repeated measures one-way ANOVAs followed by Sidak’s post-hoc comparisons when applicable. Reinstatement: Reinstatement tests wherein no additional manipulations were compared with extinction using two-tailed paired t-tests. Reinstatement tests with optogenetic manipulations were analyzed across groups and behavioral sessions using two-way ANOVAs, followed by Sidak’s post-hoc comparisons when applicable. Reinstatement tests with chemogenetics, or a combination of optogenetics and chemogenetics, were analyzed across groups and behavioral sessions using a two-way ANOVAs, followed by Sidak’s post-hoc comparisons when applicable. Sucrose seeking tests: Sucrose-seeking behavior (active lever responses, sucrose rewards, inactive lever responses) was analyzed using two-way ANOVAs followed by Sidak’s post-hoc comparisons when applicable. All statistical analyses were performed using Prism (GraphPad Prism) statistical software. Behavioral data is represented as mean ± standard error of the mean.

Electrophysiology

General:

oeEPSC amplitudes, AMPAr/NMDAr ratios, AMPAr rectification indices, and paired-pulse ratios measured for each neuron were compared across experimental groups using a one-way ANOVA followed by Sidak’s post-hoc tests for between group comparisons (Fig. 4D–L; Fig. 7D–E; Fig. S4C–E). For all electrophysiology experiments, sample sizes (n) of cells/animals per group and statistical information are stated in the corresponding figure legends. Data points outside of 2 standard deviations above or below the mean were excluded from analyses. All statistical analyses were performed using Prism (GraphPad Prism) statistical software. Data is represented as mean ± standard error of the mean.

Gq-DREADD Validation:

The number of spikes evoked before and after aCSF control versus CNO wash were averaged for each epoch and compared across groups and time using a two-way ANOVA, followed by Sidak’s post-hoc tests for between group comparisons (Fig. S5A). μ-OR Knockout: To validate μ-OR knockout, the number of spikes from baseline recordings and post-DAMGO wash groups were normalized to the average number of spikes from the baseline recording and compared across groups and time using a two-way ANOVA followed by Sidak’s post-hoc test for between group comparisons (Fig. 7B). For characterization of μ-OR knockout effects on PVT→NAc neurophysiology, electrically evoked action potentials analyzed using a two-way ANOVA, followed by Sidak’s post-hoc test for between group comparisons for each current step (Fig. S7A). Other measures of intrinsic excitability and synaptic efficacy, including maximum spikes, rheobase, sEPSC amplitude and frequency, oeEPSC amplitude, and AMPAr/NMDAr ratio, were compared between groups using a two-tailed unpaired t-test (Fig. S7B–F).

Confocal Microscopy and Imaris Image Analysis

Correlations of Retrobead^™ signal and μ-OR expression were carried out using a Pearson correlation coefficient with two-tailed P-values at a 95% confidence interval. Calculations of μ-OR expression were analyzed across cell types using a Kruskal-Wallis ANOVA, with Dunn’s multiple comparison test. Calculations of μ-OR projection- and cell type-specific μ-OR expression were calculated using a two-way ANOVA, followed by a Sidak’s post-hoc comparison. Graphical representations of these data as scatter plots or bar graphs refer to the mean and standard error of the mean.

Highlights.

1. Heroin use and seeking are predicted by inhibition of PVT→NAc neuronal ensembles

2. Heroin also weakens PVT→NAc^PV-IN synapses, collectively inducing behavioral disinhibition

3. Behavioral inhibition is restored by rescuing activity at PVT→NAc^PV-IN synapses

4. Heroin-induced synaptic adaptations and behavioral disinhibition rely on PVT μ-ORs