Male Sprague-Dawley rats (N = 82) were used for all experiments. All animal use protocols were approved by the Medical University of South Carolina and were performed according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals (8th ed., 2011). Viral vector information can be found in Table 1. Detailed methods can be found in the Supplementary Information.

Cocaine hydrochloride (NIDA, Research Triangle Park NC; 200 μg/50 μl bolus) was dissolved in sterile saline for SA experiments. Clozapine-N-Oxide (CNO) was obtained through the National Institute of Mental Health Chemical Synthesis Program. CNO was dissolved in 5% DMSO in sterile saline and was administered at a dose of 5 mg/kg (i.p.) 30 min before testing.

Cocaine and sucrose SA (SA) were performed as previously described (Giannotti et al., 2018; Siemsen et al., 2018a,b, 2020). Briefly, rats were mildly food-deprived (∼15 g rat chow/day) to increase exploration prior to beginning SA and were then maintained at 20 g/day for the remainder of the experiment. Rats were weighed daily, and catheters were flushed with sterile saline prior to daily SA sessions. Rats then underwent SA [Fixed ratio (FR) 1 schedule of reinforcement] in standard MedPC operant chambers (Med Associates, St. Albans, VT, United States) fixed with two retractable levers, a house light, tone generator, sucrose pellet dispenser, and two lights above each lever for 2 h/day (12–14 days). Presses on the active lever elicited a light and tone cue-complex followed by a single infusion of cocaine or a single 45 mg flavored sucrose pellet (Bioserv, Flemington, NJ #F0025, United States) followed by a 20 s timeout period. Presses on the inactive lever had no programmed consequence. The day after the final SA session, rats either entered extinction training or received an intra-NAcore cannula implantation for electrochemical experiments. Active lever presses (ALP), inactive lever presses (ILP) and infusions earned are shown within each figure for all SA experiments.

During daily extinction sessions (2 h/day) ALPs had no programmed consequence. Animals underwent extinction sessions until they met criterion (average ≤ 25 ALPs over the last 2 days). Rats then underwent a cue-induced reinstatement test whereby ALPs elicited the light and tone cue-complex, but no cocaine or sucrose delivery. Yoked saline animals were sacrificed 24 h after the final extinction session without undergoing reinstatement. At the beginning of each reinstatement session, a single non-contingent cue was presented. Following the 2-h reinstatement session, a subset of rats were perfused 15 min after the session for Fos immunohistochemistry (see below). Virus expression in both the PrL cortex and NAcore were mapped according to the atlas of Paxinos and Watson.

Antisera information can be found in the key resource table. Detailed methods can be found in the Supplementary Information.

Rats were anesthetized with isoflurane and rapidly decapitated for brain extraction. Brains were sectioned (300 μm) using a vibratome (Leica VT 1200) in ice-cold, oxygenated sucrose cutting solution containing (in mM): 225 sucrose, 119 NaCl, 1.0 NaH2PO4, 4.9 MgCl2, 0.1 CaCl2, 26.2 NaHCO3, 1.25 glucose, ~305 mOsm. Brain slices recovered for 30 min in warm (32°C) aCSF containing (in mM): 119 NaCl, 2.5 KCl, 1.0 NaH₂PO₄, 1.3 MgCl, 2.5 CaCl₂, 26.2 NaHCO₃, 15 glucose, ∼305 mOsm. During recording, slices were perfused with room-temperature aCSF (1 mL/min). NAcore cells were visualized using differential interference contrast (DIC) through a 40x liquid-immersion objective mounted on an upright light microscope (Olympus, BX-RFA). NAcore MSNs expressing mCherry were visualized using an integrated green LED (545 nm; <1 mW). Whole-cell recordings were obtained using borosilicate pipettes (3–7 MΩ) back-filled with Cesium methanesulfonate, which contained (in mM): 117 Cs methanesulfonic acid, 20 HEPES, 10 EGTA, 2 MgCl₂, 2 ATP, 0.2 GTP (pH 7.35, mOsm 280).

Voltage-clamp recordings were obtained from mCherry⁺ and neighboring mCherry^– NAcore MSNs to characterize inputs from PrL. During recordings, NAcore MSNs were held at –70 mV and PrL axons containing AAV2-EF1a-DIO-hChR2-EYFP were activated through a 10 ms pulse of a blue LED (470 nm; 1 mW) delivered every 5 s. Importantly, the exact same stimulation parameters were kept throughout all recordings. The resulting optogenetically evoked EPSCs were recorded and peak amplitudes were analyzed using Clampfit (v10).

In vivo electrochemical detection and quantification of glutamate levels were performed as previously described (Siemsen et al., 2020). Following SA, surgery, abstinence, and extinction, Glutamate oxidase (GluOx)-coated electrochemical electrodes were implanted (Pinnacle Technology, Lawrence, KS, United States) and connected to a wireless potentiostat housed within a 3D-printed enclosure; data was transmitted from the potentiostat via Bluetooth. Electrodes were calibrated and implanted as described previously (Siemsen et al., 2020). Following implantation animals were returned to the colony overnight. The next morning, a ∼1-h baseline recording began; animals undergoing cocaine seeking were injected with CNO (5 mg/kg, i.p.) 30 min prior to the reinstatement session. Animals undergoing sucrose seeking did not receive CNO. One Hz measurements of glutamate-dependent currents were expressed relative to a 15-min pre-session baseline, then converted to glutamate concentrations (in nM) based off the in vitro calibration. While behavioral testing lasted two hours, recordings were stopped after the first hour. Animals were perfused following testing. Virus expression and probe placement were mapped according to the atlas of Paxinos and Watson.

Detailed methods regarding virus vector mapping, cell counting, and Fos imaging and analyses can be found in the Supplementary Information.

All statistical analyses were performed with GraphPad Prism (v9) software. When comparing groups across time or bins a mixed-effects model (for missing values) or mixed-model two-way ANOVA with virus or treatment as a between subject variable and session/test/bin as the within subject variable followed by Bonferroni-corrected multiple comparison test when an interaction or main effect was significant. Fos data were analyzed with a two-way ANOVA with virus (mCherry vs. hM4Di) and reward (cocaine vs. sucrose) as between subject variables followed by Tukey’s multiple comparison tests when an interaction was significant. Correlation analyses were performed using a two-tailed Pearson’s correlation. Electrochemical calibration data was analyzed with linear regression and changes in glutamate concentrations during reinstatement were analyzed using area under the curve (AUC) analyses. AUC’s were then compared across groups using a one-way ANOVA with a Dunnett’s multiple comparison test. Paired and unpaired observations were analyzed with a paired or un-paired, with the appropriate correction if needed, t-test, respectively. Significance was set at p < 0.05 and data is expressed as the mean ± the standard error of the mean. All non-significant behavioral data is presented in Supplementary Table S1.

To test whether PrL neurons projecting to the NAcore (PrL^NAcore) are required for cue-induced reinstatement and increased glutamate release in the NAcore, we expressed inhibitory hM4Di-DREADDs in PrL^NAcore neurons using an intersectional chemogenetic approach (Figure 1A). Animals underwent cocaine or sucrose SA, followed by one week of home cage abstinence, then extinction training (Figure 1B). Following cocaine or sucrose SA, 7-day home-cage abstinence, and extinction training (Figure 1C and Supplementary Figure S2). Within Figure 1C, both ALPs and infusions earned share the same axis. Animals were then administered CNO 30 min prior to the presentation of cocaine-paired cues. Cocaine seeking was then measured as the number of non-reinforced presses of the lever formerly paired with cocaine (i.e., “active” lever presses or ALPs). Comparing the last day of extinction, cue presentation to DIO-mCherry control animals produced a significant increase in drug seeking on the active lever [two-way mixed-model ANOVA, significant group by test interaction, F(1,21) = 11.73, p = 0.003, Figure 1D], which was suppressed in the DIO-hM4Di rats compared to mCherry controls (p < 0.0001). When ALPs during reinstatement were summated in 15 min bins, there was a main effect of group [F(1,21) = 12.60, p = 0.002, Figure 1E], indicating that ALPs were lower in hM4Di compared to mCherry rats regardless of the time bin. Representative PrL^NAcore hM4Di expression is shown in Figure 1F and a compiled histology map for hM4Di rats is shown in Figure 1G.

In a subset of the cocaine SA animals, we also examined the influence of PrL^NAcore neurons on cue-induced glutamate release in the NAcore. Extracellular glutamate levels were detected using a calibrated GluOx-coated electrode implanted in the NAcore (Figures 1A,H) and cued glutamate release was measured in cocaine (mCherry and hM4Di) and sucrose reinstating rats (Figure 1I). When analyzing the area under the curve (AUC) of 1 Hz measurements of glutamate release in the three groups, we detected a main effect of cue on NAcore glutamate levels [one-way ANOVA, main effect of treatment, F(2,12) = 4.55, p = 0.034, Figure 1J]. Compared to PrL^NAcore DIO-mCherry controls, the PrL^NAcore DIO-hM4Di animals and animals reinstating to sucrose-paired cues both showed a significant reduction in cue-induced glutamate release in the NAcore (p = 0.04).

Like other AAV serotypes, AAV1 viral particles transduce neurons in the primary injection site. However, one unique property of AAV1 vectors is that a portion of viral particles are trafficked anterogradely down axons then released via vesicle fusion at axon terminals. This allows for transfer of viral particles and transgene expression in downstream neurons receiving synaptic contacts (Zingg et al., 2017, 2020). Importantly, AAV1-mediated transsynaptic transduction is monosynaptic (Oh et al., 2020) and strongest and strongest in neurons that receive numerous functional inputs from cells at the primary injection site, with the original characterization of these vectors and their anterograde transsynaptic transfer demonstrating a significant association between postsynaptic AAV transduction and presynaptic connectivity (Zingg et al., 2017, 2020). As such, we took advantage of this characteristic of AAV1 vectors to investigate NAcore neurons receiving dense and functional PrL inputs (NAcore^PrL). First, we determined what percentage of NAcore neurons are transduced via AAV1 transsynaptic transduction by infusing AAV1-CamKIIα-Cre in the PrL cortex and AAV1-CAG-Flex-Ruby2sm-Flag (Flex-smFP) or AAV2-hSyn-DIO-hM4Di-mCherry in the NAcore (Figures 2A–C). Within the field of viral transduction in the NAcore, ∼10% of the neurons were labeled (Figure 2D). Further analysis of NAcore^PrL neurons revealed that a large fraction of NAcore^PrL neurons exhibited reactivity for pre-pro Enkephalin (ppENK), a marker for D2 MSNs (Lu et al., 1998), whereas a smaller subset exhibited reactivity for various interneuron markers (Figure 2E). As CamKIIa expression has been demonstrated in both MSNs (Klug et al., 2012) and interneurons (Keaveney et al., 2020), the transsynaptic transduction in these downstream targets observed here is consistent with these observations. Moreover, the CamKIIα promoter was selected to avoid transduction of long range PrL to NAcore GABAergic projections (Lee et al., 2014). Interestingly, even though a minority of NAcore^PrL neurons were identified as parvalbumin interneurons (PV), we found ∼60% of total NAcore PV neurons were transduced via their inputs from the PrL (Figure 2F). Representative images for each immunohistochemical cell type identification experiment are shown in Figure 2G.

In a separate cohort of drug-naïve rats, we co-injected AAV1-CamKIIα-Cre and AAV2-EF1a-DIO-hChR2-EYFP into the PrL cortex and AAV2-hSyn-DIO-mCherry in the NAcore and used a blue light-evoked EPSC protocol (Otis et al., 2019) to record EPSC amplitude in mCherry⁺ and neighboring mCherry^– neurons in the NAcore (Figures 2H,I). Fifty-eight percent of mCherry^– neurons either showed no response (33%), or a response < 10 pA (25%), whereas 100% of mCherry⁺ cells responded (see Table 2). When removing mCherry^– cells that showed no response, mCherry⁺ cells showed a greater EPSC amplitude compared to mCherry^– cells [Welch’s-corrected t-test, t(16.34) = 2.259, p = 0.038, Figure 2J].

To examine the contribution of Nacore^PrL neurons in reward-seeking behavior, we microinjected AAV1-CamKIIα-Cre into the PrL cortex and AAV2-hSyn-DIO-hM4Di-mCherry or AAV2-hSyn-DIO-mCherry into the NAcore (Figure 3A). Animals underwent cocaine or sucrose SA, extinction, and cue-induced reinstatement (Figure 3B). AAV1-CamKIIα-Cre injections were confined to the PrL cortex, and AAV2-DIO-hM4Di transduced a subpopulation of cells in the NAcore (Figures 3C,D). Importantly, throughout all experimentation, expression of DIO-mCherry or DIO-hM4Di was confirmed to be restricted to the core subregion of the nucleus accumbens (see Figure 3D inset panel), given that a small population of PrL neurons do exist that send input the shell subcompartment of the nucleus accumbens (Wayman and Woodward, 2018a). Animals that were infused with either DIO-mCherry or DIO-hM4Di were taken through cocaine self-administration and extinction (Figure 3E) and both reinstated to cocaine-paired cues [two-way mixed-model ANOVA, group by test interaction, F(1,21) = 13.94, p = 0.001; mCherry (p < 0.0001), hM4Di (p = 0.006), Figure 3F]. Compared to mCherry controls, the chemogenetic inhibition of NAcore^PrL neurons reduced cue-induced reinstatement (p < 0.0001), which was most pronounced during the first 30 min of the cue test when ALP data was binned into 15-min time blocks [two-way RM ANOVA, virus by time point interaction, F(7,147) = 3.743, p = 0.0009, Figure 3G] and significant differences between mCherry and hM4Di were observed at 15 (p < 0.0001) and 30 min (p < 0.05) minutes. Interestingly, for DIO-mCherry and DIO-hM4Di rats undergoing sucrose self-administration and extinction (Figure 3H), there was a main effect of cued sucrose seeking [two-way mixed model ANOVA, F(1,14) = 46.03, p < 0.0001, Figure 3I], yet NAcore^PrL inhibition did not affect cue-induced reinstatement. There was also no difference in cued sucrose seeking with NAcore^PrL inhibition when ALP data was binned [two-way RM ANOVA, F(7,98) = 0.82, p = 0.573, Figure 3J], indicating that NAcore^PrL neurons are required for cued seeking for cocaine, but not sucrose rewards.

Next, we used Fos immunoreactivity to examine neuronal activation in the NAcore following cued reinstatement of cocaine or sucrose seeking. Compared to yoked-saline controls sacrificed 24 h after the last extinction session, cued seeking in virus control animals (mCherry-Cocaine and mCherry-sucrose) produced a significant increase in Fos staining in the NAcore, which included NAcore^PrL neurons (Supplementary Figure S3). When also including cue-induced Fos staining in the hM4Di-Cocaine and hM4Di-sucrose groups, a significant interaction between reward and virus was observed [two-way ANOVA, F(1,35) = 5.178, p = 0.029, Figure 3K]. Compared to mCherry-Cocaine animals, only hM4Di-Cocaine animals showed reduced Fos in the NAcore overall (p = 0.0007). When limiting the analysis to NAcore^PrL neurons, a two-way ANOVA revealed a significant reward by virus interaction [F(1,35) = 4.746, p = 0.036, Figure 3L]. Compared to mCherry-Cocaine rats, all other groups showed reduced Fos expression in NAcore^PrL neurons (p’s < 0.05). As expected, hM4Di-Sucrose rats showed reduced Fos expression in NAcore^PrL neurons compared to mCherry-Sucrose (p = 0.005). Interestingly, ALPs during reinstatement positively correlated with Fos activation in NAcore^PrL neurons in cocaine (r = 0.447, p = 0.033, Figure 3M), but not sucrose animals (r = 0.002, p = 0.992, Figure 3N). Representative images for Fos quantification analyses are shown in Figure 3O.

Given NAcore^PrL neurons are necessary for cued drug seeking, we next asked whether NAcore^PrL dendritic spine heads and surrounding perisynaptic astrocyte processes (PAPs) undergo morphological plasticity during cued cocaine seeking. To label NAcore^PrL neurons and surrounding astrocytes, we microinjected AAV1-CamKIIα-Cre in the PrL cortex and a mix of AAV1-CAG-Flex-Ruby2sm-Flag to label dendritic spines and AAV5-GfaABC1D-Lck-GFP to label astrocyte PAPs in the NAcore (Figure 4A). Following cocaine SA and extinction training, animals were sacrificed 15 min into cue-induced reinstatement (cocaine), and these were compared to yoked-saline control animals sacrificed 24 h after extinction (Figure 4B). There was no difference in the percent of transduced neurons in the NAcore between cocaine and yoked saline rats [t(21) = 0.794, p = 0.436, Figure 2D], or in the percentage of NAcore^PrL neurons that were ppENK⁺ (Supplementary Figure S4). Akin to prior experiments, CamKIIα-driven Cre expression in the primary injection site was confined to the PrL cortex and Flex-Ruby2sm-Flag provided sparse, fully labeled NAcore neurons, including MSNs (identified via their morphological signature) for spine analyses (Figures 4C,D). Following cocaine SA and extinction training (Figure 4E), we observed that cued reinstatement significantly increased dendritic spine head diameter (d_H) of NAcore^PrL spines [paired t-test, t(115) = 5.780, p < 0.0001, Figure 4F]. This increase in d_H was associated with a rightward shift in dendritic spine d_H when spine d_H was binned to generate a frequency distribution. While the area under the curve for the spine head data presented in Figure 4G is similar for control and cocaine seeking animals our data indicate a rightward shift in the frequency distribution of dendritic spine heads during cue-induced cocaine seeking, indicating an increase in the frequency of large spine heads and a decrease in frequency of smaller spine heads [two-way mixed model ANOVA, group by bin interaction, F(5,575) = 13.22, p < 0.0001, Figure 4G), with no change in spine density [t(115) = 0.905, p = 0.367, Figures 4H,I].

Next, we evaluated whether astrocyte PAPs undergo morphological plasticity at dendritic spine heads of NAcore^PrL MSNs during seeking and whether this occurs at specific subsets of spines (Figure 4J and Supplementary Figure S1A). Interestingly, compared to yoked-saline controls, cue-induced cocaine seeking animals had a significantly lower percent of spines without any astroglial association [Welch’s-corrected two-tailed t-test, t(80.07) = 3.933, p = 0.0002, Figure 4K], and the average astroglial association with dendritic spine heads on NAcore^PrL neurons was significantly increased during cocaine seeking (Mann–Whitney test, U = 781, p = 0.014, Figure 4L), which positively correlated with the dendrites’ average spine d_H (r = 0.216, p = 0.037, Figure 4M). We also examined which NAcore^PrL spines showed the increased astroglial association. When analyzing the astroglial association across binned spine head sizes (Figures 4N–P), we observed a main effect of cue [F(1,92) = 7.08, p = 0.009] and an increase in the percent astrocyte association with the largest NAcore^PrL spines (Bonferroni post hoc, 0.5–0.6 μm–p = 0.049, >0.6 μm–p = 0.023). Importantly, these effects were not due to differences in the intensity of GFP in the astrocyte processes (Supplementary Figure S1B).

We also asked whether these adaptations occur 24 h after the last SA session given that repeated cocaine increases basal dopamine concentrations in the NAc 24 h after the final injection (Weiss et al., 1992) without altering dendritic spine density on distal dendrites of NAcore MSNs (Dumitriu et al., 2012), and astrocytes respond to synaptically released dopamine by increasing Ca²⁺ mobility to drive gliotransmission (Corkrum et al., 2020), a phenomenon often paired with increased PAPs association at synapses (Bernardinelli et al., 2014). Akin to previous findings regarding spine density 24 h after repeated daily cocaine injections (Dumitriu et al., 2012), dendritic spine density or head diameter of NAcore^PrL neurons was unaltered 24 h after cocaine SA and no changes in astrocyte-dendritic spine association were observed (Supplementary Figures S5A–G).

Previous findings have indicated a clear role for PrL^NAcore neurons in driving cued cocaine seeking, but not food seeking, via glutamate release in the NAcore (Gipson et al., 2013a; Scofield et al., 2016a; Stefanik et al., 2016). Alternatively, activation of infralimbic to nucleus accumbens shell neurons has been shown to reduce cue-induced cocaine seeking (Peters et al., 2008; Augur et al., 2016). While glutamate release in the nucleus accumbens is thought to drive these effects, evidence exists that recruitment of the PrL neurons that supply excitatory input to the NAcore requires DA release in the PrL (McGlinchey et al., 2016; James et al., 2018). Moreover, while our manipulations were restricted to the PrL and its input to the NAcore, it is also important to note that glutamate release from neurons located in the adjacent infralimbic cortex can also influence extracellular dopamine levels in the shell subcompartment of the nucleus accumbens through infralimbic-mediated activation of dopamine neurons in the ventral tegmental area (Quiroz et al., 2016). Despite what is currently known, the contribution of PrL-dependent glutamate release in the NAcore has yet to be directly compared between cocaine and sucrose reinstating animals. Our study is the first to demonstrate in real time that chemogenetic inhibition of PrL^NAcore neurons suppresses both cue-induced seeking and the associated elevations in NAcore glutamate release. We find that chemogenetic inhibition of PrL^NAcore neurons normalized cue-induced glutamate release in the NAcore to levels that were nearly identical to those observed in animals undergoing sucrose seeking. Thus, PrL^NAcore neurons represent a primary source of cue-induced glutamate release in the NAcore to drive cocaine seeking.

Conversely, it has been demonstrated that inhibition of the infralimbic cortex prevents appetitive eating and fearful behaviors precipitated by disruption of glutamatergic signaling in the nucleus accumbens shell (Richard and Berridge, 2013). Speaking to a potential PrL and infralimbic dichotomy, following exposure to the inhalant Toluene and a single day of withdrawal, decreased intrinsic excitability in layer 5/6 PrL neurons that project to the NAcore and increased intrinsic excitability in infralimbic neurons (in either Layer 2/3 or 5) that project to the NAcore was observed (Wayman and Woodward, 2018b). Additionally, decreased intrinsic excitability was observed in the subset of layer 5/6 infralimbic neurons that project to the nucleus accumbens shell (Wayman and Woodward, 2018b). A subsequent study quantified the unique characteristics of Toluene-induced adaptations in excitability in PrL and infralimbic projection neurons and demonstrated that chemogenetic activation of accumbens projecting infralimbic cortex neurons prevents Toluene-induced conditioned place preference (Wayman and Woodward, 2018a). Taken together, these studies support the conclusion that hypoactivity in infralimbic to nucleus accumbens shell neurons promotes Toluene reward-associated behaviors. Finally, such a dichotomy is clearly drug-dependent as cue-induced heroin seeking (Rogers et al., 2008), morphine-induced conditioned place preference (Hearing et al., 2016), and cue-induced methamphetamine seeking (Kearns et al., 2022) all require activity in either the infralimbic cortex and/or its specific projection to the nucleus accumbens shell.

Recent advances in AAV1-mediated transsynaptic transduction allow for the manipulation of neurons by virtue of their inputs (Zingg et al., 2017, 2020). Importantly, transsynaptic transgene expression is highly dependent on the number of active connections made between the upstream and downstream neuron (Zingg et al., 2017). Consistent with this property of AAV1 transsynaptic transduction, we found that 100% of virally transduced NAcore^PrL neurons responded when optically stimulating PrL terminals expressing DIO-hChR2, yet 58% of non-transduced cells either did not respond (33%) or showed a negligible response (≤10 pA, 25%), and EPSC amplitude was significantly greater in NAcore^PrL neurons when compared to neighboring unlabeled cells. We observed that NAcore^PrL neurons represent ∼10% of total NeuN⁺ neurons in the NAcore, the majority of which were MSNs. While this estimate is likely conservative given that this relies upon 100% transduction in the PrL, 1:1 transsynaptic transfer and 100% transduction in the NAcore, our data do support the existence of a subpopulation of NAcore neurons receiving the high density, functionally relevant PrL input.

The NAcore is a critical region mediating cue-induced seeking for all drugs of abuse (Kalivas and Volkow, 2011; Roberts-Wolfe and Kalivas, 2015; Spencer et al., 2018), whereby drug-induced cellular adaptations distinct from those induced by natural rewards set the stage for relapse vulnerability. Given the importance of the corticostriatal pathway in drug seeking (Figure 1) and the heterogeneity of PrL input density and functional innervation of NAcore neurons (O’Donnell and Grace, 1995; Finch, 1996), we used AAV1 vectors to isolate and manipulate the NAcore MSNs most heavily innervated by the PrL cortex. We found that chemogenetic inhibition of NAcore^PrL neurons suppressed cue-induced cocaine, but not sucrose, seeking. One limitation of our approach is that NAcore^PrL neurons were transduced by virtue of their input from the PrL and not by virtue of their input and cell type. This is important given that we observed NAcore^PrL neurons constitute a mix of D1 and D2-expressing MSNs as well as interneurons. We suspect that chemogenetic inhibition of D1-NAcore^PrL and D2-NAcore^PrL neurons would have dichotomous effects on cued seeking, consistent with previous reports (Lobo et al., 2010; Heinsbroek et al., 2017). Our AAV1-derived estimates indicate that only 20% of total D2-MSNs receive dense PrL input, with NAcore^PrL neurons being approximately 50% D2-MSNs and by estimation 36% D1-MSNs. While just 7% of NAcore^PrL neurons were PV interneurons, 60% of total PV interneurons were labeled with our AAV1 transsynaptic vector, indicating that a large portion of total PV interneurons receive a functionally relevant PrL cortical input. As such, future experimentation will need to address whether chemogenetic inhibition of the population of PrL input receiving PV interneurons is necessary to suppress cue-induced reinstatement. Indeed, suppressing PV interneuron activity in the NAc reduces amphetamine sensitization in mice (Wang et al., 2018) and BLA-mediated activation of nucleus accumbens shell interneurons enhances acquisition of cocaine self-administration through interneuron-mediated inhibition of MSNs (Yu et al., 2017). A recent study also demonstrates that activation of NAcore PV interneurons in the NAcore produces conditioned place preference and reduces LiCl-induced conditioned place aversion (Chen et al., 2019). The authors find increased activation of D1 MSNs following optogenetic stimulation of NAcore PV interneurons and speculate that activation of PV interneurons in the NAcore promotes conditioned place preference through preferential targeting of MSNs in the indirect pathway or through polysynaptic disinhibition of direct pathway MSNs (Chen et al., 2019). Overall, the interaction between NAcore^PrL subtypes in regulating cocaine-seeking behavior is undoubtedly complex, and future studies will need to be performed to delineate the involvement of each specific subtype.

Regardless of the cell types involved, activity in NAcore^PrL neurons appears to be critical for overall NAcore activation during cocaine seeking given that chemogenetic inhibition of NAcore^PrL neurons led to an overall suppression of Fos⁺ neurons in the NAcore. Consistent with recent reports (Bobadilla et al., 2020), we observed that overall levels of Fos induction in the NAcore per μm³ were similar in cocaine and sucrose reinstating mCherry control animals. However, Fos induction was significantly higher in NAcore^PrL neurons in mCherry cocaine animals compared to mCherry sucrose. Further, the magnitude of Fos activation in NAcore^PrL neurons positively correlated with the amount of seeking in cocaine, but not sucrose, animals. Collectively, these data demonstrate that NAcore^PrL neuronal activity is necessary for cue-induced cocaine, but not sucrose, seeking. Given that inhibition of PrL^NAcore neurons during cocaine seeking normalized NAcore glutamate levels to that of sucrose seeking animals (Figure 1), we speculate that both potentiated PrL-dependent excitatory transmission in the NAcore combined with cellular adaptations in NAcore^PrL neurons following extinction from cocaine, but not sucrose, SA sets the stage for relapse vulnerability. Additionally, when comparing the results from Figures 1, 3, the binned lever pressing data during cocaine seeking were not identical (Figure 1E vs. 3G). However, it is important to note that we did observe parity across these two distinct viral manipulations at different neurons, with an overall reduction in cue-induced cocaine seeking observed when inhibiting either PrL neurons the project to the NAcore (Figure 1), or NAcore neurons that receive PrL input (Figure 3).

Cue-induced seeking for multiple drugs of abuse elicits time-dependent increases in synaptic strength within the NAcore MSNs, including increased AMPA/NMDA ratio and enlarged dendritic spine d_H (Gipson et al., 2014), the degree of which positively correlates with drug-seeking behavior (Gipson et al., 2013a). As a recent addition to established cue-induced neuronal structural plasticity, heroin seeking has been linked to increased astrocyte association with NAcore synapses. This cue-induced astrocytic structural plasticity limits drug seeking (i.e., promotes within-session extinction), likely by increasing glutamate clearance (Kruyer et al., 2019). However, previous studies did not provide evidence that increased astrocyte association occurred at relevant synapses. Our data are the first to demonstrate that NAcore^PrL MSNs undergo increased dendritic spine d_H during the peak of reinstatement, as well as increased association of perisynaptic astrocyte processes with the most potentiated spines on NAcore^PrL neurons, an effect that was not observed 24 h following cocaine SA. These data suggest that cue-induced motility of astroglial processes and increased interaction with structurally potentiated synapses at NAcore^PrL MSNs serves to limit glutamate spillover during cocaine seeking.