A master regulator of opioid reward in ventral prefrontal cortex

Abstract

In addition to their intrinsic rewarding properties, opioids can also evoke aversive reactions that protect against misuse. Cellular mechanisms that govern the interplay between opioid reward and aversion are poorly understood. We used whole-brain activity mapping to show that neurons in the dorsal peduncular nucleus (DPn) are highly responsive to the opioid oxycodone. Connectomic profiling revealed that DPn neurons innervate the parabrachial nucleus (PBn). Spatial and single-nuclei transcriptomics resolved a population of PBn-projecting pyramidal neurons in the DPn that express μ-opioid receptors (μORs). Disrupting μOR signaling in the DPn switched oxycodone from rewarding to aversive and exacerbated the severity of opioid withdrawal. These findings identify the DPn as a key substrate for the abuse liability of opioids.

One Sentence Summary

Pyramidal neurons in ventral prefrontal cortex that express μ-opioid receptors and project to the parabrachial nucleus regulate hedonic responses to opioids.

Opioids elicit their rewarding effects in large part by stimulating inhibitory μORs located on GABAergic inputs to dopamine neurons in the ventral tegmental area (VTA), thereby disinhibiting dopamine transmission in the nucleus accumbens (NAc) and other components of the mesolimbic system (1-5). Accumulating evidence suggests that dopamine-independent are also involved (6-11). Furthermore, excitotoxic lesions of the NAc that markedly decrease cocaine self-administration only modestly attenuate responding for opioids (12, 13). This highlights the importance of non-mesolimbic brain circuitries in the motivational properties of opioids (14), but little is currently known about their identity (15-21). Paradoxically, the same doses of opioids that elicit reward-related behaviors can also provoke aversive reactions (22, 23). Aversion to opioids and other drugs of abuse after initial exposure decreases the likelihood of developing habitual patterns of use (24, 25). Brain mechanisms that determine whether opioids elicit reward or aversion are mostly unexplored.

Ventral PFC contains opioid-responsive neurons

Unbiased whole-brain mapping of c-Fos was used to identify brain regions in which neural activity was modified by a rewarding dose of the μOR agonist oxycodone (OxyContin^®) (26). Mice were injected with saline or a dose of oxycodone (5 mg kg⁻¹) that established conditioned place preference (CPP) behavior (Fig. 1a). Brains were collected from all mice 120 min after injection, cleared using the iDISCO+ procedure, and immunostained for c-Fos (27, 28) (Fig. 1b). c-Fos-immunopositive (Fos+) cells were mapped onto the Allen Brain Atlas (Fig. 1c). Statistically significant increases in Fos+ cell densities were detected in 28 brain regions in oxycodone-treated mice relative to control mice (Fig. 1d and Table S1). Included were regions known to densely express μORs (29). Robust induction of c-Fos also occurred in the DPn, an almost entirely unexplored area of vPFC (30) (Fig. 1d and 1e). We focused our attention on this region. K-nearest neighbor (KNN) analysis was performed on the Fos+ data (27) (Fig. 1f). Closer proximity of brain regions in tSNE space reflected greater likelihood that neural activities in those regions was functionally connected (27). The DPn clustered with other cortical regions in control and oxycodone-treated mice, including the infralimbic (IL) and prelimbic (PrL) areas of medial prefrontal cortex (mPFC) (Fig. 1f). The DPn also clustered with regions known to modulate behavioral responses to painful and stressful stimuli in oxycodone-treated but not control mice, including the PBn, interpeduncular nucleus (IPn), and locus coeruleus (LC) (Fig. 1f).

Fig. 1. Opioid-regulated DPn neurons encode aversion.

(a) Graphical depiction of oxycodone CPP procedure (left). Proportion of time (± SEM) male mice (n=7) spent in saline and oxycodone-paired sides of CPP apparatus (right). *P<0.05 compared to saline-paired side, two-tailed paired t-test. (b) Illustration of DISCO+ brain clearing and whole-brain c-Fos immunostaining procedures. (c) Representative iDISCO+ cleared brain from oxycodone-injected mouse showing c-Fos+ cells (red) and white matter tracts (green). (d) Mean (± SEM) numbers of Fos+ cells in brain regions of saline- and oxycodone-injected male mice (n=4 per group). Regions sorted according to p value. ***P<0.001, **p<0.01, *p<0.05, Fisher’s exact test. (e) Summary heat-maps showing higher c-Fos+ cell densities in DPn of oxycodone-injected relative to saline-injected mice. (f) tSNE plots of KNN analyses of c-Fos data from saline-injected (left) and oxycodone-injected (right) mice. Brain regions in which c-Fos expression was correlated across mice in each group are shown in the different colored clusters. (g) Graphical representation of AAV-ChR2-eYFP or AAV-eYFP injection into DPn of male C57BL/6J mice (left) and an example of fluorescence expression in DPn of AAV-ChR2-eYFP-injected mouse (right). (h) Design of RTPP experiment to investigate the effect of DPn photo-stimulation on reward/aversion behavior. (i) Heat-map of RTPP behavior in a ChR2-expressing mouse. (j) Mean (± SEM) time (s) spent in sides of RTPP apparatus in which the LED delivering DPn photo-stimulation was activated (LED-on) or inactivated (LED-on) in ChR2-expressing mice injected with saline or oxycodone (5 mg kg⁻¹) before testing (n=4 per group) and eYFP-expressing control mice injected with saline before testing (n=4). Treatment x Session interaction in two-way repeated-measures ANOVA (F_{(2, 10)}= 7.677, p=0.0095); **p<0.01, Šídák’s multiple comparisons test. (k) Design of the RTPP experiment in which oxycodone-responsive DPn neurons were photo-stimulated in male FosTRAP2 mice (upper). Depiction of 4-OHT-mediated ChR2-eYFP expression in oxycodone-responsive neurons in DPn of FosTRAP2 mice (lower). (l) iDISCO+ cleared brain showing ChR2-eYFP expression in DPn of FosTRAP2 mouse. (m) Mean (± SEM) time (s) spent in LED-on and LED-off sides of RTPP apparatus in ChR2-expressing (n=5) and eYFP-expressing FosTRAP2 mice (n=4). TRAP x Session interaction effect (F_{(1, 14)}= 7.397, p=0.0166); **p<0.01, Šídák’s multiple comparisons test.

Opioid-regulated DPn neurons encode states of aversion

Next, we investigated the role of the DPn in regulating reward and aversion-related behaviors. AAV5-hSyn-ChR2-eYFP or control AAV5-hSyn-eYFP was injected into the DPn of wild-type mice and a fiber optic implanted 200 μm above the injection site (Fig. 1g and Fig. S1). Three weeks later, all mice were tested in a real-time place-preference (RTPP) procedure (Fig 1h). ChR2-expressing but not control mice avoided the side of the RTPP apparatus paired with DPn photo-stimulation (Fig. 1i and 1j). This aversion-related behavior was abolished in ChR2-expressing mice injected with oxycodone (5 mg kg⁻¹) prior to testing (Fig. 1j). Next, Fos-2A-iCre^ERT2 (FosTRAP2) mice (31) were used to target only those opioid-responsive DPn neurons in which oxycodone induced c-Fos expression (Fig. 1k). AAV8-hSyn-DIO-ChR2-eYFP or control AAV8-hSyn-DIO-eYFP were injected unilaterally into the DPn of FosTRAP2 mice and 3 weeks later all mice were injected with oxycodone (5 mg kg⁻¹) and 4-OHT (30 mg kg⁻¹) to TRAP opioid-responsive neurons (Fig. 1k and Fig. S1). iDISCO+ brain clearing and eYFP immunostaining confirmed that ChR2-expressing neurons were concentrated in the DPn of oxycodone-treated FosTRAP2 mice (Fig. 1l). ChR2-expressing but not eYFP-expressing (eYFP+) control mice avoided the side of the RTPP apparatus paired with DPn photo-stimulation (Fig. 1m). Finally, AAV2-Ef1a-DIO-eNpHR 3.0-eYFP or control AAV2-Ef1a-DIO-eYFP were injected into the DPn of FosTRAP2 mice, and oxycodone and 4-OHT injected to TRAP opioid-responsive neurons (Fig. S2). DPn photo-inhibition had no effect on RTPP behavior in the eNpHR-expressing mice (Fig. S2).

Connectomic profiling identifies discrete networks of DPn neurons

We sought to identify the circuit-based mechanisms by which opioid-regulated DPn neurons control aversion. Pyramidal neurons in the DPn and adjacent tenia tecta (TT) project to the dorsomedial hypothalamus (DMH), and activation of these vPFC→DMH neurons facilitates autonomic responses to stress (30). Besides the DMH, little is known about the brain regions innervated by DPn neurons. To characterize the projection profiles of opioid-regulated DPn neurons, FosTRAP2 mice were injected into the DPn with AAV8-hSyn-DIO-eYFP and 3 weeks they were later injected with 4-OHT and oxycodone (5 mg kg⁻¹) or saline (control) (Fig. 2a). Brains were collected >3 weeks later, cleared, and immunostained for eYFP (Fig. 2a). In addition to the DPn, 26 other brain regions contained higher levels DPn-derived eYFP+ axons in oxycodone-treated vs. saline-treated FosTRAP2 mice (Fig. 2b and 2c; Table S2).

Fig. 2. DPn aversion neurons project to the parabrachial nucleus.

(a) iDISCO+ cleared brain showing eYFP+ axons from oxycodone-responsive neurons (green) in a FosTRAP2 mouse injected into the DPn with AAV-hSyn-DIO-eYFP. (b) Bar graph summarizing brain regions with higher concentrations of eYFP+ DPn axons in brains of oxycodone-treated (n=9) versus saline-treated (n=8) male FosTRAP2 mice. Regions sorted according to −log P value (Fisher’s exact test). (c) Renderings of coronal brain images from Allen Mouse Brain Atlas highlighting regions containing DPn-derived eYFP+ axons (green). (d) Graphical summary of MAP-seq procedure. (e) Fluorescence micrograph showing DPn neurons infected with MAP-seq barcoding virus. cc, Corpus callosum; ac; anterior commissure. (f) Bar graph summarizing numbers of unique DPn-derived barcodes in sequenced brain regions (n=6 male C57BL/6J mice). (g) Barcode matrix showing the distributions of each unique DPn barcode across sequenced brain regions. (h) PCA was used to cluster sequenced regions based on their barcode content. (i) Radial graphs summarizing the distribution of DPn-derived barcodes in each brain region and the relative proportion of the same barcodes in the other regions. (j) Numbers of unique barcode reads (± SEM) detected in DPn neurons in saline-injected (n=3) and oxycodone-injected (n=3) mice. **P<0.01, unpaired two-sided t-test. (k) Relative numbers (± SEM) of each barcode (normalized to total barcode reads) across sequenced brain regions in saline and oxycodone-injected mice. *P<0.05, ***p<0.001, post-hoc test after significant interaction effect in two-way repeated-measures ANOVA (F_{(5, 1695)}= 5.915, p<0.0001). (l) eYFP+ axons were detected in PBn of C57BL/6J mice (n=3) after injection of AAV5-hSyn-eYFP into DPn. scp, Superior cerebellar peduncle. (m) CTb-488+ cell bodies were detected in DPn of C57BL/6J mice (n=3) injected with CTb-488 into PBn. (n) Design of RTPP experiment assessing the effect of photo-stimulating the terminals of oxycodone-responsive DPn neurons in PBn of male FosTRAP2 mice (upper). Graphical representation of FosTRAP2 mice in which the terminals of oxycodone-responsive DPn→PBn neurons were photo-stimulated in PBn (lower). (o) Mean (± SEM) time (s) spent in the LED-on and LEF-off sides of RTPP apparatus by ChR2-expressing (n=5) and eYFP-expressing (n=5) FosTRAP2 mice. DPn photo-stimulation was delivered only in the LED-on side. TRAP x Session interaction effect in two-way repeated-measures ANOVA (F_{(1, 8)}=12.81, p=0.0072); *p<0.05, Šídák’s multiple comparisons test.

Next, we used Multiplexed Analysis of Projections by Sequencing (MAP-seq) to generate higher resolution connectomic maps of opioid-responsive DPn neurons (32-35) (Fig. 2d). We confirmed that the MAP-seq barcoding virus efficiently infected DPn neurons (Fig. 2e). Mice were injected daily with saline or oxycodone (5 mg kg⁻¹) for 10 days, the MAP-seq virus was then injected into the DPn, and brains collected from all animals 40 h later. Repeated oxycodone injections were used to facilitate identification of opioid-responsive DPn neurons based on altered barcode densities in efferent brain regions. The DPn was dissected for barcode sequencing, along with the DMH and regions of interest identified above by eYFP mapping. A total of 343 unique barcodes were detected in the DPn and one or more of the other sequenced regions from control and oxycodone-treated mice (Fig. 2f and 2g; Table S3). The VTA and PBn contained the highest densities of DPn-derived barcodes (Fig. 2f and 2g). Principal component analysis (PCA) was used to cluster barcode distributions and thereby resolve connectomic motifs of DPn neurons (Fig. 2h). DMH, ventromedial hypothalamus (VMH), and periaqueductal gray (PAG) clustered together in PCA space (Fig. 2h), suggesting the same DMH-projecting DPn neurons that control autonomic stress responses (30) provide concurrent input to the VMH and PAG (Fig. 2i). The entorhinal cortex (ENT) clustered independently in PCA space (Fig. 2h), suggesting that ENT-projecting DPn neurons provide minimal input to the other sequenced regions (Fig. 2i). Separately, the PBn and VTA clustered in PCA space (Fig. 2h), suggesting these areas are innervated by a prominent population of DPn neurons separate from those that project to the DMH or ENT (Fig. 2i). Finally, we assessed oxycodone-induced changes in barcode densities in these regions, which is thought to reflect altered strength of connectivity (36, 37). Numbers of barcodes expressed per neuron in the DPn were much lower in oxycodone-injected than control mice (Fig. 2j). Inspection of the data showed that relative barcode densities were unaltered in the DMH and ENT and were modestly increased in the VMH and PAG of the oxycodone-treated mice (Fig. 2k). By contrast, barcode densities were markedly decreased in the PBn and VTA of the oxycodone-treated mice (Fig. 2k). This suggests that DPn neurons that project to the PBn and VTA are highly responsive to oxycodone and are likely to undergo opioid-induced functional remodeling.

Opioid-regulated DPn aversion neurons project to parabrachial nucleus

The PBn contributes to respiratory depression and other physiological responses to opioids (38-40). The PBn also regulates aversive reactions to acutely administered opioids (41, 42). Thus, we investigated the role of the opioid-regulated DPn neurons that project to the PBn (DPn→PBn neurons) in reward-related behaviors. First, we injected AAV5-hSyn-eYFP into the DPn of a new cohort of mice and confirmed that eYFP+ axons were detected in the PBn (Fig. 2l). Then, we injected the retrograde tracer cholera toxin subunit B conjugated to a green fluorescent tag (CTb-488) into the PBn of mice and detected CTb-488-labelled neurons in the DPn (Fig. 2m). Next, we injected AAV8-hSyn-DIO-ChR2-eYFP into the DPn of FosTRAP2 mice and 3 weeks later injected them with 4-OHT and oxycodone or saline (control) (Fig. 2n). A fiber optic was implanted above the PBn of all animals (Fig. 2n). The ChR2-expressing but not control mice avoided the side of a RTPP apparatus paired with photo-stimulation of DPn terminals in the PBn (Fig. 2o).

DPn contains a unique population of PBn-projecting pyramidal neurons

Next, we sought to identify the DPn→PBn neurons that encode aversion-related behavioral states in an opioid-regulated manner. We began by using fluorescence in situ hybridization (FISH)-based spatial transcriptomics (Spatial-seq) to characterize the expression of a library of 120 marker genes used to define cellular phenotypes with single-cell resolution (Table S4). Coronal slices containing the PFC were prepared for Spatial-seq from C57BL/6J mice and a 1 mm² area encompassing the DPn was profiled (Fig. 3a and Fig. S3). Gene transcripts considered markers of discrete cortical cell types exhibited spatially variable expression corresponding to the established cellular anatomy of the surveyed area (Fig. 3b). A total of 31975 cells segregated into discrete clusters in Uniform Manifold Approximation and Projection (UMAP) space based on their transcriptional profiles (Fig. 3c and Fig. S3). Clusters were enriched in canonical marker genes of discrete cell types (Fig. 3d; Fig. S3 and S4). Each profiled cell was mapped to its precise location in the PFC (Fig. 3e and Fig. S5). Cells expressing markers of cortical GABAergic or glutamatergic (Glut) neurons were concentrated in two major clusters in UMAP space (Fig. 3c and 3d). GABAergic neurons were ~3 times less abundant than Glut neurons in the surveyed area (3577 vs. 9766 cells) (Fig. 3c and Fig. S3). GABAergic neurons could be further segregated into 5 subpopulations based on their transcriptional profiles (Fig. S6 and Table S5). GABAergic subtypes were distributed throughout the PFC, with no apparent enrichment of any subtype in the DPn (Fig. S7). Glut neurons segregated into 6 subpopulations (Fig. 3f and Fig. S6; Table S5). Glut subtypes 3 and 4, which accounted for 2% and 1%, respectively, of all profiled Glut neurons, were concentrated in the DPn relative to surrounding regions (Fig. 3f and 3g; Fig. S7). These DPn-enriched cells were distinguished from other Glut neurons by their expression of Slc17a6 transcripts, which encode vesicular glutamate transporter 2 (vGlut2) (Fig. 3h and Table S5). Inspection of the Allen Mouse Brain Expression Atlas confirmed that the DPn contains a prominent population of vGlut2+ neurons (43) (Fig. 3i). vGlut2+ neurons were recently implicated in reward and aversion-related behavioral responses to opioids (15, 44).

Fig. 3. DPn contains a unique population of pyramidal neurons.

(a) Slc17a7 (vGlut1) expression in DPn-containing coronal brain slice from Allen Mouse Brain Atlas. cc, Corpus callosum; ac; anterior commissure. (b) Spatial-seq was performed on a 1 mm² area of PFC containing the DPn from male C57BL/6J mice (n=5). Shown are representative examples of genes with spatially variable expression. (c) UMAP plot of PFC cell clusters identified by Spatial-seq. (d) Dot plot of canonical marker gene expression showing enrichment in discrete UMAP clusters. (e) Representative PFC slice showing location of cells profiled by Spatial-seq and clustered in UMAP space. (f) Spatial distribution of glutamatergic (Glut) neurons from the same PFC slice subclustered into 6 discrete subtypes. (g) Glut 3 and Glut 4 subtypes were concentrated in the DPn. (h) Volcano plot showing that Glut 3 and Glut 4 subtypes were distinguished from the other Glut subtypes based on Slc17a6 (vGlut2) expression (−log P value; Fisher’s exact test). (i) Slc17a6 expression in a DPn-containing brain slice from Allen Mouse Brain Atlas. (j) Representation of vGlut2-Cre mice (n=3) injected into PBn with rgAAV-DIO-GFP to label DPnv^Glut2→PBn neurons. (k) Fluorescence micrograph showing site of rgAAV-DIO-GFP injection in PBn of a vGlut2+ mouse. (l) Higher magnification images of injection area identified by a white box. (m) Fluorescence micrograph showing GFP+ neurons in DPn-containing coronal brain slice from same animal. (n, o and p) Higher magnification images of areas identified by white boxes and labeled (i), (ii), and (ii) in panel m. Ac, anterior cingulate cortex; Ins, insular cortex; MO, motor cortex; Orb, orbitofrontal cortex. (q) Graphical representation of vGlut2-Cre mice injected into the DPn with AAV-DIO-ChR2-eYFP or AAV-DIO-eYFP to express ChR2 (or eYFP) in vGlut2+ DPn neurons. (r) Mean (± SEM) time (s) spent in the LED-on and LED-off sides of the RTPP apparatus by ChR2-expressing mice injected with saline (n=5), ChR2-expressing mice injected with oxycodone (5 mg kg⁻¹) (n=5), and eYFP-expressing mice injected with saline (n=5) prior to testing. ***P<0.001, unpaired two-tailed t test. (s) Graphical summary showing that the vGlut2+ DPn neurons that project to the PBn and regulate aversion in an opioid-regulated manner.

The DPn and TT neurons that project to DMH to regulate autonomic stress responses express vGlut1 but not vGlut2 (30). Thus, vGlut2 expression appears to define a unique population of DPn glutamatergic neurons (DPn^vGlut2 neurons). We investigated whether DPn^vGlut2 neurons were the same population identified by MAP-seq that project to the PBn. A retrograde-traveling AAV that expressed GFP in a Cre-dependent manner (rgAAV-DIO-GFP) was injected into the PBn of vGlut2-Cre mice (Fig. 3j-3l). GFP+ neurons were detected throughout the DPn (Fig. 3m and 3n). More sparsely distributed GFP+ cells were also detected in the IL area of mPFC but not in surrounding cortical regions (Fig. 3m and 3o). Next, we investigated the role of DPn^vGlut2 neurons in regulating reward-related behaviors. AAV8-hSyn-DIO-ChR2-eYFP or control AAV8-hSyn-DIO-eYFP was injected into the DPn of vGlut2-Cre mice and a fiber optic implanted above the injection site (Fig. 3q and Fig. S1). ChR2-expressing but not control mice avoided the side of a RTPP apparatus paired with DPn photo-stimulation (Fig. 3r). This aversion-related behavior was abolished by oxycodone injection (5 mg kg⁻¹) (Fig. 3r). Hence, DPn^vGlut2 neurons that project to the PBn (DPn^vGlut2→PBn neurons) encode states of aversion in an opioid-regulated manner (Fig. 3s). MAP-seq showed that ~60% of DPn→PBn neurons also project to the VTA (Fig. 2i), suggesting that DPn→VTA neurons may function similarly.

DPn pyramidal neurons express μORs

Why are DPn^vGlut2 neurons so highly responsive to oxycodone? We used single nuclei RNA-sequencing (snRNA-seq) to characterize the transcriptional profiles of these cells more thoroughly. Mice were injected with saline or oxycodone (5 mg kg⁻¹), brains collected 30 min later, and DPn tissue punches prepared for snRNA-seq. Nuclei from a total of 20763 cells were included in the analyses after filtering for quality control (Fig. 4a). Cells segregated into 14 discrete clusters in UMAP space (Fig. 4a). Patterns and relative densities of clusters were similar between oxycodone-treated and control mice (Fig. 4b). Gene transcripts considered hallmarks of GABAergic interneurons were concentrated in clusters 3, 5, and 6 (Fig. 4c and 4d; Table S6). Cells expressing vGlut1 and other markers of cortical Glut neurons were concentrated in clusters 0, 1, 4, and 7 (Fig. 4c and 4e; Table S6). Glut neurons in cluster 7 expressed vGlut2 (Fig. 4f), and contained transcripts considered hallmarks of subcortical-projecting pyramidal neurons (45) (Fig. 4e and Table S6). High concentrations of Oprm1 gene transcripts, which encode μORs, were also detected in cluster 7 cells (Fig. 4f and Table S6). Gene ontology analysis of differentially expressed genes (DEGs) showed that oxycodone engaged pathways involved in opioid signaling in these cells (Fig. 4g; Tables S7 and S8). These DEGs were enriched in genes regulated by c-Fos, CREB, and other transcriptional regulators known to participate in μOR signaling (46-48) (Fig. 4h).

Fig. 4. DPn pyramidal neurons express μORs.

(a) UMAP plot of DPn cell clusters identified by snRNA-seq. (b) UMAP cell clusters (upper) and cell densities across clusters (lower) from saline- and oxycodone-treated male C57BL/6J mice (n=6 per group). (c) Violin plot of canonical marker genes across clusters. (d) Dot plot of gene transcripts identifying GABAergic interneurons. Red boxes indicate clusters enriched in GABAergic marker genes. (e) Dot plot of clusters expressing genes identifying glutamatergic neurons (red boxes). (f) Dot plot showing Slc17a6 and Oprm1 expression in glutamatergic clusters. (g) Gene ontology (GO) analysis of DEGs in cluster 7 cells (−log P value; Fisher’s exact test). (h) Upstream regulator analysis of DEGs in cluster 7 (−log P value; Fisher’s exact test). (i) DPn-containing brain slice from male C57BL/6J mice (n=3) used for FISH (left). Higher magnification images from the same slice (middle). Higher magnification images of areas identified by white boxes in DPn (i) and IF region of mPFC (ii). (j) Summary of the gene expression profiles of μOR-expressing DPn pyramidal neurons resolved by FISH. (k) Mean (± SEM) numbers of μOR+ neurons in the DPn and IL detected by FISH. *P<0.05, two-tailed unpaired t test. (l) Mean (± SEM) numbers of μOR+/vGlut2+ co-expressing neurons in DPn and IL. *P<0.05, two-tailed unpaired t test. (m) Representation male vGlut2-Cre mice (n=5) injected into PBn with rgAAV-DIO-GFP. (n) Fluorescence micrograph showing site of rgAAV-DIO-GFP injection in PBn. (o) Fluorescence micrograph of GFP-labelled neurons in DPn. (p) Higher magnification image showing current-clamped DPn^vGlut2→PBn neuron. (q) Representative voltage traces in DPn^vGlut2→PBn neuron depolarized by current injections (200 and 300 pA) before and after DAMGO (1 μM) application. (r) Mean (± SEM) numbers of spikes emitted by DPn^vGlut2→PBn neurons before and after DAMGO. Current intensity x DAMGO interaction in two-way repeated-measures ANOVA (F_{(1, 5)}=7.770, p=0.0386). ###P<0.001, 200 vs. 300 pA current injections; ***p<0.001, before vs. after DAMGO; Tukey post-hoc test. (s) Illustration of male vGlut2-Cre mice (n=3) injected into DPn with AAV-DIO-ChR2-eYFP. (t and u) ChR2-eYFP expression in DPn and PBn, respectively. (v) Representative oEPSC traces in PBn neurons triggered by photo-stimulating the terminals of DPn^vGlut2→PBn neurons before (control) and after DAMGO (1 μM) application. Mean (± SEM) peak current amplitudes of oEPSCs detected in PBn before (control) and after DAMGO. *P<0.05, unpaired two-tailed t test.

RNAscope-based FISH confirmed that the DPn contains a dense population of vGlut2+ neurons that co-express vGlut1 transcripts but not transcripts encoding the vesicular GABA transporter (vGat), a marker of GABAergic neurons (Fig. 4i-4l and Fig. S8). The majority of DPn^vGlut2 pyramidal neurons expressed μOR transcripts (Fig. 4i-4l and Fig. S8). Real-time PCR verified that μORs were expressed at higher levels in the DPn than the adjacent mPFC (Fig. S9). iDISCO+ brain clearing of μOR-mCherry reporter mice was used to characterize the distribution of cortical μOR+ neurons (49) (Fig. S9). μOR+ neurons were detected in the DPn (Fig. S9). mCherry-expressing μOR+ neurons in DPn were largely segregated from cells that expressed GAD1, a marker of GABAergic neurons (Fig. S9). This is consistent with μORs in the DPn being expressed primarily by local pyramidal neurons.

Next, whole-cell recordings in brain slices from μOR-mCherry mice were used to compare the electrophysiological properties of μOR+ DPn neurons with those of established subtypes of cortical pyramidal neurons (50, 51) (Fig. S10). Membrane resistance and cell capacitance did not differ between μOR+ and μOR− DPn neurons (Fig. S10). μOR+ neurons had relatively depolarized resting membrane potentials (RMPs) compared with μOR− neurons (~60 vs. ~70 mV, respectively) (Fig. S10). Nevertheless, the minimal current amplitudes necessary to evoke an action potential (rheobase) were higher in μOR+ than μOR− neurons (Fig. S10). μOR+ neurons also had higher action potential amplitudes and larger hyperpolarization-activated inward (Ih) currents than μOR− neurons (Fig. S10). μOR− neurons had near linear input-output (I-O) relationships, whereas μOR+ neurons had flattened I-O curves that plateaued at lower currents injections than μOR− neurons (~175 pA vs. ~325 pA, respectively) (Fig. S10). These features of μOR+ DPn neurons are consistent with the regular spiking (RS) class of pyramidal neurons (50, 51). When the synthetic opioid peptide DAMGO was applied to brain slices, μOR+ but not μOR− DPn neurons demonstrated hyperpolarization of RMPs, elevated rheobase, flattening of I-O curves, but no change in Ih currents (Fig. S10). This suggests that μOR+ DPn neurons are inhibited by opioids.

Finally, rgAAV-DIO-GFP was injected into the PBn of vGlut2-Cre mice to label DPn^vGlut2→PBn neurons (Fig. 4m-4p). DAMGO but not aCSF decreased the numbers of action potentials evoked by current injections (200 or 300 pA) in these cells (Fig. 4q and 4r; Fig. S11). AAV8-hSyn-DIO-ChR2-eYFP was injected into the DPn of a new cohort of vGlut2-Cre mice (Fig. 4s-4u). Optically stimulating the terminals of DPn^vGlut2 neurons evoked excitatory post-synaptic currents (oEPSCs) in PBn neurons that were abolished by the AMPA receptor antagonist NBQX (Fig. 4v). DAMGO decreased the amplitude of these oEPSCs by >50% (Fig. 4v and 4w). This suggests that opioids inhibit DPn^vGlut2→PBn neurons.

DPn neurons regulate the hedonic valence of opioids

Next, we investigated the role of μOR signaling in the DPn in regulating behavioral responses to oxycodone. AAV-EF1a-mCherry-IRES-Cre virus was injected into the DPn of Oprm1^fl/fl mice or Oprm1^WT/WT (wild-type) littermates (Fig. 5a and Fig. S1). In addition, a non-Cre-expressing AAV-mCherry virus was injected into the DPn of Oprm1^fl/fl mice (Fig. S1). As μOR expression did not differ between Cre-injected wild-type and mCherry-injected Oprm1^fl/fl mice, their data were collapsed into a single group (control mice). μOR expression was reduced in the DPn of Cre-injected Oprm1^fl/fl (cKO) mice relative to the control mice (Fig. 5b), while μOR expression was unaltered in the adjacent mPFC (Fig. 5b). Locomotor activity after saline injection was similar in cKO and control mice (Fig. 5c and 5e). Oxycodone (5 mg kg⁻¹) injection increased locomotor activity to a similar extent in both groups across three daily test sessions (Fig. 5d and 5e). This suggests that the stimulant effect of opioids, which is mediated by mesolimbic dopamine transmission (19, 52), is unaltered by disruption of μOR signaling in the DPn. Oxycodone reward was assessed in the cKO and control mice using the same CPP procedure shown in Fig. 1 (Fig. 5f). Control mice spent a greater proportion of time in the oxycodone-paired side (Fig. 5g). By contrast, cKO mice avoided the oxycodone-paired side (Fig. 5g). This suggests that disrupting μOR signaling in the DPn enhances aversive reactions to oxycodone. Thus, we investigated the effect of oxycodone on food responding in control and cKO mice. AAV8-hSyn-Cre-IRES-mCherry or control AAV8-hSyn-mCherry was injected into the DPn of Oprm1^fl/fl mice and >3 weeks later they were permitted to lever-press for food pellets (25 mg) (Fig. 5h). Both groups responded at similar rates after saline injection (Fig. 5i). However, food responding was suppressed in cKO mice relative to control mice after oxycodone (5 mg kg⁻¹) injection (Fig. 5i), consistent with enhanced opioid aversion in the cKO mice.

Fig. 5. DPn neurons regulate opioid aversion.

(a) Conditional deletion of μORs from DPn of Oprm1^Fl/Fl mice. (b) Mean (± SEM) μOR expression in DPn and mPFC of Cre-injected Oprm1^Fl/Fl (cKO) mice (n=8) and control (n=4) male mice. Main effect of Genotype (F_{(1, 11)}=9.318, p=0.011) in two-way repeated-measures ANOVA.**P<0.01, Bonferroni's multiple comparisons test. (c and d) No difference in mean (± SEM) distance travelled between control (n=17) and cKO (n=16) male and female mice after saline or oxycodone injection. (e) Mean (± SEM) total distance travelled after saline or oxycodone injection across daily sessions. Main effect of Treatment (F_{(1, 31)}=138.1, ***p=0.0001). (f) CPP procedure to assess opioid reward in μOR cKO mice. (g) Mean (± SEM) time (s) spent in oxycodone-paired side by control (n=17) and cKO (n=16) male and female mice. Genotype x Session interaction effect (F_{(1, 31)}=29.84, p<0.0001); ***P<0.001, Fisher’s LSD test. (h) Operant procedure to assess food responding. (i) Mean (± SEM) number of food pellets earned by control (n=7) and cKO (n=6) male mice after saline or oxycodone injection. Main effect of Treatment (F_{(1, 11)}=62.93, p<0.0001); *P<0.05, Fisher’s LSD test. (j) Illustrations of AAV-Cre injection into DPn (upper); and whole-body tracking of mice during opioid withdrawal (lower). (k) Freeze-frame image of oxycodone-dependent mouse during naloxone-precipitated withdrawal. (l) Representative 2D traces of tracked body parts in mouse before (left) and during (right) naloxone (1 mg kg⁻¹)-precipitated oxycodone withdrawal. (m) Mean (± SEM) % time spent in locomotion (10 min epochs) after daily oxycodone injection, then after naloxone injections in control (n=4) and cKO (n=4) male mice. Genotype x Injection interaction (F_{(2, 18)}=18.52, p<0.0001); ***P<0.001, Bonferroni’s test. (n) Mean (± SEM) % time spent in withdrawal behaviors. Genotype x Injection interaction (F_{(2, 18)}=11.86, p=0.0005); ***P<0.001, Bonferroni's test. (o) Mean (± SEM) % time spent in other distress-related behaviors. Genotype x Injection interaction (F_{(2, 18)}=38.89, p<0.0001); ***P<0.001, **p<0.01, Bonferroni's test. (p) Summary of DPn regulation of opioid withdrawal. (q) Representative AAV-hM4Di-mCherry expression in DPn. (r and s) Mean (± SEM) number of food pellets earned by hM4Di-expressing (n=13) and mCherry-expressing (n=14) male mice after CNO (3 mg kg⁻¹) and/or naloxone (1 mg kg⁻¹) injection before and after induction of oxycodone dependence. Drug treatment x DREADD interaction in oxycodone-dependent mice (F_{(2, 50)}=4.224, p=0.0202); ***P<0.001, **p<0.01, *p<0.05, Bonferroni's test. (t) Procedure to photo-stimulate opioid-responsive DPn neurons. (u) Mean (± SEM) number of food pellets earned by oxycodone-dependent FosTRAP2 male mice expressing ChR2 (n=5) or eYFP (n=4). ChR2 x Session interaction (F_{(1, 7)}=8.12, p=0.0247); ***P<0.01, Fisher’s LSD test.

DPn neurons regulate opioid withdrawal

Opioids are consumed by dependent individuals not only for their rewarding effects but also to escape the aversive state experienced during withdrawal (53, 54). To investigate DPn involvement in opioid withdrawal, we injected AAV8-hSyn-Cre-IRES-mCherry or control AAV8-hSyn-mCherry into the DPn of Oprm1^fl/fl mice and 3 weeks later they were chronically treated with oxycodone to induce opioid dependence (55, 56) (see Methods). Withdrawal-related behaviors were precipitated by naloxone injection (0.1 and 1 mg kg⁻¹) and characterized using DeepLabCut/VAME (Fig. 5j-5l). Prior to naloxone injection, opioid-dependent control and cKO mice behaved similarly in the test arena (Fig. 5l and 5m). Naloxone (1 mg kg⁻¹) reduced locomotor activity (Fig. 5l and 5m) and precipitated physical withdrawal signs in the control mice (Fig. 5n). Other distress-related behaviors, including grooming and immobility, were also precipitated by naloxone (1 mg kg⁻¹) in control mice (Fig. 5o). The intensity of these behaviors was greater in the cKO mice relative to control mice (Fig. 5m-5o). The lower naloxone dose (0.1 mg kg⁻¹) had no effect on behavior in control mice but precipitated withdrawal responses in cKO mice (Fig. 5m-5o). This suggests that deficits in μOR signaling in the DPn exacerbates the severity of opioid withdrawal (Fig. 5p). Next, we prepared a group of FosTRAP2 mice in which ChR2 or eYFP (control) was conditionally expressed in oxycodone-responsive DPn neurons, and both groups chronically treated with oxycodone (Fig. S12). Photo-stimulating the DPn in the absence of naloxone injection had no effect on locomotor activity or withdrawal signs in either group. However, DPn photo-stimulation exacerbated locomotor suppression and physical withdrawal signs precipitated by naloxone (1 mg kg⁻¹) in ChR2-expressing but not control mice (Fig. S12).

Negative affective aspects of withdrawal play a more prominent role than physical symptoms in maintaining opioid dependence (53, 54, 57, 58). Disruption of appetitive responding for food rewards is thought to reflect negative affective components of opioid withdrawal in rodents and non-human primates (59-61). AAV8-hSyn-hM4Di-mCherry or control AAV8-hSyn-mCherry was injected into the DPn of wild-type mice and >3 weeks later they were trained to lever-press for food pellets (Fig. 5q and Fig. S1). Clozapine-N-oxide (CNO; 3 mg kg⁻¹) injected alone or in combination with naloxone (0.1 mg kg⁻¹) had no effect on food responding in hM4Di-expressing and control mice (Fig. 5r). After the induction of oxycodone dependence, CNO injection again had no effect on food responding (Fig. 5s). However, CNO blocked the decrease in food responding that was precipitated by naloxone in hM4Di-expressing but not control mice (Fig. 5s). Finally, we investigated the effect of photo-stimulating the DPn on food responding in FosTRAP2 mice expressing ChR2 or YFP in opioid-regulated DPn neurons (Fig. 5t). DPn photo-stimulation precipitated a withdrawal-like decrease in food responding ChR2-expressing but not control mice that were oxycodone-dependent even in the absence of naloxone injection (Fig. 5u). By contrast, photo-stimulating the DPn failed to suppress food responding in ChR2-expressing mice that were oxycodone-naïve and instead increased responding (Fig. S13).

Discussion

Opioids that engage μOR signaling in the brain are highly addictive (62). The VTA and NAc have long been recognized as important substrates for the rewarding actions of opioids that drive the development of opioid use disorder (OUD) (1-4, 63, 64). Less is known about non-mesolimbic mechanisms of their abuse potential (15-19, 65, 66). In addition to their rewarding effects, opioids can also be highly aversive (22, 40, 67). How the rewarding and aversive effects of opioids interact in the brain to control addiction-related behaviors is poorly understood. Here, we identify a population of glutamatergic neurons in the DPn that express vGlut2 and μORs, which are not typically expressed by cortical pyramidal neurons (16, 68-70). These unusual DPn neurons project to the PBn, encode states of aversion, and are inhibited by opioids. Crucially, when μOR signaling in the DPn was disrupted, the hedonic valence of oxycodone was switched from rewarding to aversive. This suggests that aversive reactions to opioids are regulated by the DPn, and that μOR-mediated inhibition of DPn neurons is necessary for the rewarding effects of opioids to dominate over their aversive effects that protect against OUD. Disrupting μOR signaling in the DPn also exacerbated the severity of oxycodone withdrawal, suggesting that cellular adaptations in the DPn contribute to the development of opioid dependence. That a population of cortical pyramidal neurons should directly regulate reward and aversion-related responses to oxycodone may seem surprising in light of classical reports showing that opioids infused into mPFC failed to elicit hedonic reactions (65, 66). Indeed, it is generally believed that PFC neurons are not direct substrates for the addiction-related actions of opioids, but instead serve to coordinate opioid-seeking behaviors during abstinence (71, 72). However, little is known about the vPFC and its involvement in OUD, as few studies have investigated the function of this cortical territory (30). Our data suggest that DPn neurons in the vPFC control hedonic reactions to opioids and serve as key regulators of their abuse potential.