The Perineuronal Net Protein Brevican Acts in Nucleus Accumbens Parvalbumin-expressing Interneurons of Adult Mice to Regulate Excitatory Synaptic Inputs and Motivated Behaviors

Abstract

Background:

Experience-dependent functional adaptation of nucleus accumbens (NAc) circuitry underlies the development and expression of reward-motivated behaviors. Parvalbumin-expressing GABAergic interneurons (PVINs) within the NAc are required for this process. Perineuronal nets (PNNs) are extracellular matrix structures enriched around PVINs that arise during development and are proposed to mediate brain circuit stability. However, their function in adult NAc is largely unknown. Here we study the developmental emergence and adult regulation of PNNs in the NAc of male and female mice and test the cellular and behavioral consequences of reducing the PNN component Brevican in NAc PVINs.

Methods:

We characterized the expression of PNN components in mouse NAc using immunofluorescence and RNA in situ hybridization. We lowered Brevican in NAc PVINs of adult mice using an intersectional viral and genetic method and quantified the effects on synaptic inputs to NAc PVINs and reward-motivated learning.

Results:

PNNs around NAc PVINs are developmentally regulated and appear during adolescence. In the adult NAc, PVIN PNNs are also dynamically regulated by cocaine. Transcription of the gene encoding Brevican is regulated in a cell-type and isoform-specific manner in NAc, with the membrane-tethered form of Brevican highly enriched in PVINs. Lowering Brevican in NAc PVINs of adult mice decreased their excitatory inputs and enhanced both short-term novel object recognition and cocaine conditioned place preference.

Conclusions:

Regulation of Brevican in NAc PVINs of the adult mouse modulates their excitatory synaptic drive and sets experience thresholds for the development of motivated behaviors driven by rewarding stimuli.

Introduction

The Nucleus Accumbens (NAc) drives reward-motivated behaviors by integrating information about environmental context, internal state, and motivation with past outcomes (1). NAc Striatal Projection Neuron (SPN) activity mediates the expression of these behaviors, and SPN experience-dependent transcriptional plasticity has been widely studied as a mechanism of reward learning (2). However, SPN firing is coordinated by strong, feed-forward inhibition from a sparse population of local Parvalbumin-expressing GABAergic interneurons (PVINs) that also shape reward-relevant NAc output (3–9). NAc PVINs are functionally required for the expression of conditioned place preference (CPP) following appetitive rewards (4,6). Furthermore, following repeated exposure to the psychostimulant cocaine, NAc PVINs display increased excitability and selectively strengthened excitatory synaptic inputs (5,10), suggesting that cocaine-dependent functional adaptations in NAc PVINs could contribute to the development of addictive-like behaviors. The synaptic connectivity of PVINs in many brain areas is regulated by a specialized extracellular matrix (ECM) structure called the perineuronal net (PNN) (11). We recently identified synaptic and ECM-related genes that were selectively regulated in NAc PVINs after acute psychostimulant exposure (12). Here, we examine the role of one class of these ECM gene products in NAc PVIN synaptic connectivity and reward-motivated behaviors.

PNNs have been highly studied in developing cortex, where their activity-dependent emergence (13–15) correlates with PVIN physiological maturation (16–18), formation of PVIN synaptic inputs (11,19–21), and closure of developmentally regulated critical periods of plasticity (22–24). These data have led to a model in which PNNs promote circuit stability (11), and an emerging idea that PNN remodeling contributes to adult learning (11,25) including addiction-related behaviors (26). In adult mice, enzymatically disrupting PNNs and other ECM structures (27) generally impairs memory, with the nature of the deficit dependent on brain region (28–35). Two studies have manipulated PNNs in the adult striatum (36,37), but none, to our knowledge, have targeted NAc.

Major PNN components include the chondroitin sulfate proteoglycans (CSPGs) aggrecan (ACAN), brevican (BCAN), neurocan (NCAN), and versican (VCAN) (27). The gene encoding BCAN is distinct among the CSPGs in that it produces two transcripts (Bcan1 and Bcan2), one of which (Bcan2) encodes a glycosylphosphatidylinositol (GPI)-linked isoform that remains tethered to the cell surface (38). GPI-linked BCAN2 cell-autonomously promotes the development of excitatory inputs onto hippocampal PVINs (21). In adult mice, one study found that knocking down BCAN in hippocampal PVINs perturbed PNNs and impaired memory (33). These studies raise the exciting possibility that BCAN2 may cell-autonomously control synaptic inputs and/or PNNs on PVINs during learning. However, whether BCAN is required to maintain NAc PVIN synaptic inputs and reward-motivated behaviors in adulthood is unknown.

Here we first characterize the expression of PNN components including BCAN1 and BCAN2 in mouse NA PVINs through postnatal development and following cocaine exposure in adulthood. Next, using cell-type specific viral knockdown methods in NAc PVINs of adult mice, we demonstrate that BCAN is required to maintain PVIN excitatory inputs. Interestingly, we find that loss of BCAN in NAc PVINs accelerates learning in response to both natural and drug rewards. These data provide new insights into mechanisms by which PNN components of NAc PVINs regulate their plasticity and shape learning thresholds for reward-motivated behaviors in adulthood.

Methods and Materials

Animals

Homozygous Pvalb-T2A-Cre mice (RRID:IMSR_JAX:012358) were crossed with C57BL/6J mice (RRID:IMSR_JAX:000664). Except where indicated, experiments used adult (p60–235) male and female mice heterozygous for the Pvalb-T2A-Cre transgene. All procedures were performed under a protocol approved by the Duke University Institutional Animal Care and Use Committee.

PNN and Bcan labeling

PNNs were detected with WFA-FITC and the signal was amplified by incubation with rabbit anti-FITC. For RNA visualization, multiplex RNAscope FISH (Advanced Cell Diagnostics) was performed on fresh-frozen tissue. All images within an experiment were obtained under identical acquisition settings.

Stereotactic surgery

Adeno-associated viruses (AAVs) containing DIO-mCherry-shLacZ, DIO-mCherry-shBcan, or DIO-Bcan2-HA (21) were generated by the Duke University viral vector core facility in serotype AAV1/5. For electrophysiology, viruses were mixed 1:1 with AAV1-DIO-eGFP (4). To bilaterally deliver viruses to NAc, mice were anesthetized using isoflurane and virus (0.5μL/side) was stereotactically delivered at coordinates from Bregma AP+1.8mm, ML+/−1.5mm, and DV-4.4mm from dura. Experiments occurred at least 2 weeks later.

Electrophysiology

Voltage-clamp recordings of miniature EPSCs were performed using 250μm parasagittal acute brain slices containing NAc. mEPSCs were obtained at −70mV in the presence of artificial cerebrospinal fluid (124mM NaCl, 2.5mM KCl, 1mM MgCl₂, 2mM CaCl₂, 26mM NaHCO₃, 1.2mM NaH₂PO₄, and 10mM glucose, saturated with 95% O₂ and 5% CO₂, pH 7.4, 300mOsm/l), 1μM TTX, and 50μM picrotoxin, using a potassium methanesulfonate-based internal solution (130mM KMeSO₃, 7.5mM NaCl, 0.2mM EGTA, 10mM HEPES, 4.2mM ATP-Mg, 0.3mM GTP-Na, pH 7.3, 295mOsm/liter).

Behavior

Behavioral tests were performed in the light portion of the light/dark cycle. Novel object recognition sessions lasted 5min, and tests occurred 30min and 24hr after training. Cocaine conditioned place preference (CPP) used a 3-chamber apparatus. All sessions lasted 30min, except for the interleaved 10min test sessions in the Bcan knockdown study. For PNN and Bcan regulation studies, experimental mice received saline on days 2, 4, and 6, and cocaine (15mg/kg, i.p.) on days 3, 5, and 7. Control mice received saline on days 2–7 in alternating chambers. ‘Acute’ mice were sacrificed 3hr or 24hr after the injection on day 3. ‘CPP’ mice were tested on day 8, and sacrificed 24hr later. For the Fos study, mice were perfused 2hr after the injection on day 3. For the BCAN knockdown study, experimental mice received cocaine (10mg/kg, i.p.) on days 2, 5, and 8, and saline on days 3, 6, and 9, with tests on days 4, 7, and 10. Preference was quantified as (cocaine/novel time – saline/familiar time)÷(cocaine/novel time + saline/familiar time).

Statistical Analyses

All experiments were performed or quantified by an investigator blind to experimental condition. All data were tested for normality and analyzed using indicated tests in Prism 9. p<0.05 was considered significant. For further details, refer to Supplemental Information and the Key Resources Table.

Key Resources Table.

Resource Type	Specific Reagent or Resource	Source or Reference	Identifiers	Additional Information
Mouse	B6.Cg-Pvalbtm1.1(cre)Aibs/J	Jackson Laboratory	RRID:IMSR_JAX:012358	Pvalb-T2A-Cre
Mouse	C57BL/6J	Jackson Laboratory	RRID:IMSR_JAX:000664
Antibody	anti-PV - guinea pig	Synaptic Systems	195004; RRID:AB_2156476	IHC, 1:1000
Antibody	anti-FITC - rabbit	Invitrogen	A889; RRID:AB_221561	IHC, 1:1000
Antibody	anti-HA - mouse	EMD Milipore	05-904; RRID:AB_417380	IHC, WB 1:1000
Antibody	anti-MCHERRY - chicken	EnCor Biotechnology	CPCA-mCherry; RRID:AB_2572308	IHC, 1:1000
Antibody	anti-c-FOS - rabbit	Synaptic Systems	226008; RRID:AB_2891278	IHC, 1:1000
Antibody	anti-BCAN - rabbit	Protein Tech	19017-1-AP; RRID:AB_10643526	IHC, 1:1000
Antibody	anti-VGLUT1 - guinea pig	Synaptic Systems	135304; RRID:AB_887878	IHC, 1:500
Antibody	anti-VGLUT2 - guinea pig	Synaptic Systems	135404; RRID:AB_887884	IHC, 1:500
Antibody	anti-VGAT - guinea pig	Synaptic Systems	131004; RRID:AB_887873	IHC, 1:500
Antibody	anti-PSD95 - rabbit	Invitrogen	516900; RRID:AB_2533914	IHC, 1:300
Antibody	anti-GEPHYRIN - rabbit	Synaptic Systems	147008; RRID:AB_2619834	IHC, 1:500
Antibody	anti-FLAG - mouse	Sigma-Aldrich	F3165; RRID:AB_259529	WB, 1:1000
Antibody	anti-ACTIN - mouse	Milipore	MAB1501; RRID:AB_2223041	WB, 1:20,000
Lectin	WFA-FITC	Invitrogen	L32481	IHC, 1:1000
Commercial Assay Or Kit	RNAscope Multiplex Fluorescent Assay	Advanced Cell Diagnostics		version 1
RNAscope Probe	Pvalb	Advanced Cell Diagnostics	421931	NM_013645.3, nucleotides 2-884
RNAscope Probe	Bcan	Advanced Cell Diagnostics	516911-C2	NM_007529.2, nucleotides 1548-2549
RNAscope Probe	Bcan1	Advanced Cell Diagnostics	531771-C3	NM_007529.2, nucleotides 2246-3231
RNAscope Probe	Bcan2	Advanced Cell Diagnostics	531761-C2	NM_001109758.1, nucleotides 2147-2643
Cell line	Neuro2a	ATCC	ATCC #CCL-131; RRID:CVCL_0470
Plasmid	ADAMTS15	Gift from Dylan R. Edwards (https://doi.org/10.1002/ijc.29129)		FLAG-tagged
Plasmid	ADAMTS15E362A	Gift from Dylan R. Edwards (https://doi.org/10.1002/ijc.29129)		FLAG-tagged
Plasmid	DIO-shLacZ-mCherry	Gift from Beatriz Rico (https://doi.org/10.1016/j.neuron.2017.06.028)
Plasmid	DIO-shBcan-mCherry	Gift from Beatriz Rico (https://doi.org/10.1016/j.neuron.2017.06.028)
Plasmid	DIO-BCAN1-HA	Gift from Beatriz Rico (https://doi.org/10.1016/j.neuron.2017.06.028)
Plasmid	DIO-BCAN2-HA	Gift from Beatriz Rico (https://doi.org/10.1016/j.neuron.2017.06.028)
Virus	AAV1-DIO-BCAN2-HA	Duke Viral Vector Core; https://doi.org/10.1016/j.neuron.2017.06.028		3.0×10^12 vg/mL
Virus	AAV1-DIO-shLacZ-mCherry	Duke Viral Vector Core; https://doi.org/10.1016/j.neuron.2017.06.028		6.8×10^12 vg/mL
Virus	AAV1-DIO-shBcan-mCherry	Duke Viral Vector Core; https://doi.org/10.1016/j.neuron.2017.06.028		3.5×10^12 vg/mL
Virus	AAV1-DIO-eGFP	Duke Viral Vector Core; https://doi.org/10.1038/npp.2017.178		4.0×10^12 vg/mL
Equipment	Three compartment CPP apparatus	Med Associates	MED-CPP-3013	Duke Behavioral and Neuroendocrine Core
Software	EthoVision XT	Noldus		Duke Behavioral and Neuroendocrine Core
Software	GraphPad Prism 9	GraphPad Softwares

Results

PNNs emerge around NAc PVINs during adolescence and are locally remodeled in cocaine CPP

Wisteria floribunda agglutinin (WFA) binds to the chondroitin sulfate (CS) side-chains of proteoglycans enriched in PNNs (39,40). In adult NAc, WFA signal was present both in a diffuse pattern reflecting ubiquitous ‘loose’ ECM (27) and in brighter perisomatic PNNs (Fig. 1A). We defined PV+ cells as having somatic PV signal ≥25% above surrounding tissue, and PV+/WFA+ cells as also having perisomatic WFA signal ≥25% above surrounding tissue. Similar to medial prefrontal cortex (mPFC) and other brain regions (41), NAc PNNs were primarily found around PV+ cells, though some WFA+ cells were not detectably PV+ (Fig. S1A). Although adult NAc showed weaker PV and WFA signal compared to mPFC in both PV+ cells and loose ECM, similar proportions (NAc, 52.8±7.4%; mPFC, 65.9±4.9%) of NAc and mPFC PV+ cells were also WFA+ (Fig. S1B–E). These proportions are consistent with a recent report sampling PV and WFA staining across the entire mouse brain (42).

Figure 1.

NAc PVINs upregulate PV and acquire PNNs during the transition to adulthood. A) Representative immunostaining showing WFA-positive PNNs (cyan) surrounding the soma of PVINs (red) in the P90 NAc. ac, anterior commissure; PNN, perineuronal net; ECM, loose extracellular matrix; boxes indicate areas magnified in panels below. Scale bar = 40μm. B) The number of PV+ cells detected in NAc increases from P15 to P50 and P90 (ANOVA F(3,24)=3.163, p=0.043; *p<0.05 by Dunnett’s post-hoc test; n=4–10 mice/group). C) The median relative PV intensity of all PV+ cells increases from P15 to P30, P50, and P90 (Kruskal-Wallis statistic=15.75, p<0.01; *p<0.05, **p<0.01, ***p<0.001 by Dunn’s post-hoc test; n=52–393 cells from 4–10 mice/group). D) The number of PV+/WFA+ cells detected in NAc increases from P15 to P50 and P90 (ANOVA F(3,24)=3.726, p=0.025; *p<0.05 by Dunnett’s post-hoc test; n=4–10 mice/group).

In cortex, PVINs undergo an activity-dependent period of postnatal maturation during which they upregulate PV expression, develop mature fast-spiking properties, and acquire PNNs (14,15,17,43,44). To determine whether NAc PNNs show similar developmental regulation, we assessed WFA signal at P15, P30, and P50 to compare against P90 (adult) mice (45). We observed an increase in the number and relative PV intensity of NAc PV+ cells during adolescence (Fig. 1B–C). The number of PV+/WFA+ cells in the NAc also increased over the course of adolescence (P15-P50) but did not increase further as mice aged (P50-P90) in adulthood (Fig. 1D). WFA labeling of loose ECM increased from P15 to P50 (Fig. S1F), consistent with changes in ECM composition over this period (46). However, we saw no accompanying differences in WFA intensity in PNNs (Fig. S1G), showing these two types of ECM can be regulated independently. These results reveal that similar to cortical PVINs, PNNs around NAc PVINs are developmentally regulated during adolescence.

Adult cortical and hippocampal PNNs are dynamically remodeled during associative learning (21,33,34) and addiction-related paradigms (47). To determine whether PNNs around adult NAc PVINs are also regulated by experience, we examined PNN integrity in NAc PVINs at distinct stages of learning an association between spatial context and drug reward. PV and WFA staining was measured 24hr after a single cocaine conditioning session (Acute) or 24hr after the CPP test (CPP) (Fig. 2A). As expected, acute cocaine increased locomotion (Fig. S2A) and repeated pairing of cocaine induced a significant place preference compared with saline (Fig. 2B). Individual mice showed variable preference magnitude after CPP training (Fig. 2B), and when analyzed with individual immunostaining data, CPP score was positively correlated with the number of both PV+ cells and PV+/WFA+ cells within NAc (Fig. 2C,D). These data suggest that NAc PVINs with PNNs may set thresholds for reward-motivated behaviors in adult mice.

Figure 2.

Adult NAc PNNs are remodeled following cocaine experience. A) Representative PV (red) and WFA (cyan) immunostaining. B) CPP cocaine mice showed a significant preference for the drug-paired chamber (rmANOVA Time Point × Treatment interaction, F(1, 11)=11.89, p<0.01; **p<0.01 by Bonferroni’s post-hoc test; n=6–7 mice/group). The magnitude of CPP score was positively correlated with the density of PV+ cells per NAc region of interest (ROI), bounded by anatomical landmarks (C; Pearson’s correlation saline R²=0.016, p=0.79; cocaine R²=0.695, p=0.04; n=6–7 mice/group), and PV+/WFA+ cells per NAc ROI (D; Pearson’s correlation saline R²=0.062, p=0.59; cocaine R²=0.788, p=0.02; n=6–7 mice/group). PV+ cells from CPP cocaine mice had increased PV intensity (E) and decreased WFA signal (F) compared to Acute cocaine or CPP saline mice (Kruskal-Wallis statistic>18.3, p<0.001; **p<0.01, ***p<0.001, ****p<0.0001 by Dunn’s post-hoc test; n=423–603 cells from 6–7 mice/group). G) Mean loose ECM WFA intensity was increased in CPP mice (2-way ANOVA main effect of Time Point, F(1, 21)=10.33, p<0.01), but there was no effect of cocaine treatment (2-way ANOVA Treatment main or interaction effect, F(1, 21)<0.6, p>0.46; n=6–7 mice/group).

The correlation between PV+ or PV+/WFA+ neurons and CPP score was not due to a cocaine- or training-dependent change in the number of these cells, because the average number of PV+ and PV+/WFA+ neurons in the NAc did not differ between treatment groups (Fig. S2B,C). However, there were significant differences in the intensity of immunostaining in mice that underwent CPP. PV intensity was increased in the cocaine CPP group (Fig. 2E), which is notable in light of prior studies showing that changes in hippocampal PVIN PV expression during spatial learning correlates with changes in their synaptic inputs (48). By contrast, WFA staining in PVIN PNNs was weaker in the cocaine CPP mice (Fig. 2F). This was not due to a general loss of NAc ECM, because WFA signal in loose ECM was actually elevated in mice that had been through CPP testing (Fig. 2G). These data show a dynamic regulation of CSPG ECM in the NAc over the course of CPP and further indicate that ECM-regulatory mechanisms local to NAc PVINs are associated with CPP.

Cell-autonomous synthesis and cleavage of GPI-linked Brevican in PVINs of NAc

PNN integrity depends on the balance between expression and degradation of PNN components. The local reduction of WFA intensity around PVINs in cocaine CPP led us to investigate genes encoding PNN proteins and remodeling enzymes that are selectively expressed in PVINs and regulated by psychostimulant exposure.

Among the CSPGs, PVINs of the adult rodent NAc are enriched for expression of both Acan and Bcan relative to other NAc cells; between these, Bcan is more highly expressed (12,49). However, little is known about the expression of Bcan isoforms in NAc. To assess the expression of secreted and GPI-linked BCAN within adult NAc PVINs, we designed custom RNAscope probes targeting either secreted Bcan1-exclusive exons or the alternative 3’UTR in GPI-linked Bcan2 (Fig. 3A). In adult NAc, similar to hippocampus (21), Bcan transcripts are found in virtually all Pvalb+ cells and more sparsely in surrounding Pvalb− cells (Fig. 3B). Compared to surrounding Bcan+/Pvalb− cells, which are probably astrocytes and polydendrocytes (49), Pvalb+ cells express greater than 2-fold more Bcan2 (Fig. 3C) and a small fraction of Bcan1 (Fig. 3D), demonstrating that transcripts for GPI-linked BCAN2 are markedly enriched in PVINs in the adult mouse NAc.

Figure 3.

Expression of Bcan isoforms in NAc Pvalb+ cells. A) Schematic of Bcan1 and Bcan2 isoform transcripts, and target regions of the Bcan1 (cyan) and Bcan2 (magenta) isoform-specific RNAscope probes. B) Representative RNAscope signal for Pvalb (green), Bcan2 (magenta), and Bcan1 (cyan) probes in adult NAc tissue. Nuclei are labeled with DAPI (blue). Example Pvalb+ (left) and Pvalb−/Bcan1+ (right) cells are shown. Scale bar = 5μm. C) Pvalb+ nuclei have significantly more Bcan2 transcripts/cell (C) and significantly fewer Bcan1 transcripts/cell (D) than Pvalb−/Bcan1+ nuclei (p<0.0001 by Mann Whitney U test, n=49–68 cells/group from 4 mice). Data graphed are from the Acute Saline group in Fig. 5.

Next, we considered the two major PNN-degrading enzyme families: the matrix metalloproteinases (MMPs) and the disintegrin and metalloproteinase with thrombospondin motifs (ADAMTS) family (50–55) (Fig. 4A). Adult NAc PVINs are enriched for transcripts encoding eight ADAMTSs and two MMPs (Table S1). Of these, only Adamts15 is significantly upregulated following psychostimulant exposure (12). ADAMTS15 cleaves ACAN (56) and VCAN (57), but whether it can also cleave BCAN was unknown. Furthermore, though ADAMTSs are major in vivo regulators of BCAN (55,58), prior assays of bulk brain tissue could neither distinguish fragments originating from BCAN1 or BCAN2, nor BCAN fragments originating from loose ECM or PNNs.

Figure 4.

Cleavage of BCAN by an ADAMTS family member. A) Schematic of HA-tagged BCAN1 and BCAN2 constructs. Consensus MMP and ADAMTS cleavage sites are indicated by arrows, and amino acid coordinates of HA-tagged constructs, with the predicted molecular weight of the HA-tagged cleavage product for each site. B) Representative blot of BCAN1-HA conditioned media incubated with Neuro2a cells expressing empty pcDNA, ADAMTS15 wild-type (WT), or ADAMTS15^E362A, probed with anti-HA. C) BCAN1-HA conditioned medium showed a significantly greater proportion of HA signal in the 55kD band compared to the 145kD band in the ADAMTS15 condition relative to the pcDNA control (ANOVA F(2, 6)=9.218, p=0.015; **p<0.01 by Dunnett’s post-hoc test; n=3 wells/condition). D) Representative HA blot of cell lysates (Ly) and conditioned medium (CM) from Neuro2a cells expressing BCAN2-HA and either pcDNA or ADAMTS15. Consistent proportions of CM and Ly were loaded for each sample. E) A significantly greater proportion of BCAN2-HA signal was seen in the CM 55kD band compared to the Ly 100–125kD bands in the ADAMTS15 condition compared to pcDNA control (Student’s t-test, *p=0.0199; n=7 wells/condition from 2 independent experiments). F) Schematic of viral strategy for detecting cleavage fragments of BCAN2-HA expressed specifically in adult NAc PVINs. G) Representative immunostaining showing immunoreactivity of HA (magenta) specifically in the membrane of PV+ (green) cells of infected NAc in PV-Cre mice. Nuclei are labeled with Hoechst dye (blue). Scale bar = 25μm. H) HA western blot of NAc protein isolated from 2 mice expressing BCAN2-HA in NAc PVINs, BCAN1-HA/ADAMTS15 CM, BCAN2-HA/ADAMTS15 CM, and NAc protein isolated from a control mouse. * indicates non-specific bands.

To determine if ADAMTS15 can cleave BCAN, we incubated conditioned media from Neuro2a cells expressing HA-tagged, secreted BCAN (BCAN1-HA; Fig. 4A) with Neuro2A cells transfected with empty pcDNA, ADAMTS15, or the enzymatically dead mutant ADAMTS15^E362A (56). Wild-type or mutant ADAMTS15 was detected in both zymogen and mature forms (59) (Fig. S3A,B). In BCAN1-HA conditioned media, two specific, HA-immunoreactive bands were seen at ~145kD and 55kD, as expected for full-length and ADAMTS-cleaved BCAN1-HA (Fig. 4A,B), and we quantified the proportion of 145kD and 55kD signal to measure brevicanase activity. Overexpression of wild-type, but not mutant, ADAMTS15 increased brevicanase activity above endogenous levels (Fig. 4C). To determine whether ADAMTS15 can also cleave GPI-linked BCAN2, we co-transfected Neuro2a cells with HA-tagged BCAN2 (BCAN2-HA; Fig. 4A) and either pcDNA or ADAMTS15. We observed two HA-immunoreactive bands at ~100–125kD in the cell lysate, and one band at ~55kD in the conditioned medium, consistent with full-length and ADAMTS-cleaved BCAN2-HA (Fig. 4D; Fig. S3C–E). Cells overexpressing ADAMTS15 again showed greater brevicanase activity (Fig. 4E).

To determine whether BCAN2 expressed by NAc PVINs is cleaved by ADAMTSs in vivo, we injected a Cre-inducible BCAN2-HA-expressing AAV (21) into the NAc of PV-Cre mice (Fig. 4F–G), and assessed HA-immunoreactive protein fragments by western blot. NAc samples from AAV-DIO-BCAN2-HA mice showed HA-immunoreactive bands at 145kD and 55kD that were absent in protein from uninfected NAc, consistent with the fragment produced when BCAN2-HA is co-expressed with ADAMTS15 (Fig. 4H). These data indicate that ADAMTS family members, including the psychostimulant-inducible and PVIN-enriched Adamts15, can remodel PNNs containing GPI-linked BCAN2 on NAc PVINs in vivo.

To further determine how cocaine experience affects expression of Bcan isoforms, we performed in situ hybridization on NAc tissue harvested 3hr after a single cocaine dose, or 24hr after the CPP test. 3hr after acute cocaine, Pvalb+ cells have more Pvalb transcripts per cell compared to saline controls (Fig. 5B). This is driven by a subset of Pvalb+ cells consistent with the proportion in which psychostimulant exposure increases action potential firing (60), and is consistent with previous reports that PV expression is regulated by increased PVIN firing (61). Within Pvalb+ cells, there is also an increase in Bcan2 transcripts 3hr after acute cocaine (Fig. 5C). The distribution of Bcan1 transcripts per cell was slightly right-shifted, but this effect was not significant when comparing mean transcripts/cell/mouse between treatment groups (Fig. 5D). Transcription of Pvalb and Bcan2 within individual cells appears to be co-regulated, as there is no drug-induced change in the positive correlation between Pvalb and Bcan2 transcripts in single cells (Fig. 5E). In Pvalb− cells 3hr after cocaine, no changes were seen in Bcan2 expression (Fig. 5F), but Bcan1 transcripts were increased (Fig. 5G). 24hr after the CPP test, no cocaine-dependent change was observed for Pvalb, Bcan2, or Bcan1 in either cell type (Fig. 5H–M). These data show that Bcan transcription is upregulated acutely after cocaine exposure in an isoform-specific and cell-type specific manner, where Bcan1 increases in putative glia, and Bcan2 increases in NAc PVINs.

Figure 5.

Bcan transcription is regulated after acute cocaine treatment in an isoform- and cell-type-specific manner. A) Representative RNAscope in situ signal for Pvalb+ (left panels) and Bcan1+/Pvalb− cells (right panels) from mice collected 3hr after an acute dose of saline or cocaine, or 24hr after the CPP test. B) 3hr after acute cocaine treatment, the distribution of Pvalb transcripts per Pvalb+ cell becomes right-shifted (****p<0.0001 by Mann Whitney U test, n=73–87 cells from 4 mice/group), and the average number of Pvalb transcripts/cell/mouse is significantly increased (inset, **p=0.003 by Student’s t test, n=4 mice/group), compared to saline controls. These data are from the same animals and experiment, and graphed to illustrate both the distribution of cells within the overall population of PVINs, as well as the average values per mouse. C,D) as in (B), but graphing the number of Bcan2 transcripts (C; *p=0.025 by Mann Whitney U test; inset, **p=0.008 by Student’s t test) or Bcan1 transcripts (D; **p=0.003 by Mann Whitney U test; inset, p=0.118 by Student’s t test) within the same Pvalb+ cells graphed in (B). E) Cell-by-cell counts of Pvalb and Bcan2 transcripts are positively correlated (Pearson’s correlation R²=0.103, p<0.0001, n=160 cells from 8 mice) across both acute saline and acute cocaine conditions, while there is no difference in the linear regressions between treatment groups (ANCOVA slopes F(1, 156)=0.271, p=0.603; intercepts F(1, 157)=0.588, p=0.444). Within Pvalb−/Bcan1+ cells, there are no changes seen in the number of Bcan2 transcripts (F), but there is a significant increase in Bcan1 transcripts/cell (G; *p=0.012 by Mann Whitney U test, n=73–95 cells from 4 mice/group; inset, *p=0.025 by Student’s t test, n=4 mice/group) 3hr after acute cocaine treatment. 24hr after the CPP test, there are no differences detected in the number of Pvalb (H), Bcan2 (I), or Bcan1 (J) transcripts within Pvalb+ cells (n=67–90 cells from 4–5 mice/group). K) Pvalb and Bcan2 transcripts are positively correlated at this time point (Pearson’s correlation R²=0.0836, p<0.001, n=157 cells), with no significant effect of treatment (ANCOVA slopes F(1, 153)=0.05208, p=0.820; intercepts F(1, 154)=0.7044, p=0.403). No differences in Bcan2 (L) or Bcan1 (M) transcripts were detected in Pvalb−/Bcan1+ cells 24hr after the CPP test (n=70–95 cells from 4–5 mice/group).

BCAN expression in PVINs of the adult NAc maintains excitatory synaptic drive

In developing hippocampus, BCAN2 cell-autonomously promotes the formation of excitatory inputs on PVINs (21). We observe that Bcan2 is selectively upregulated in NAc PVINs following acute cocaine (Fig 5) and that repeated cocaine persistently increases PV protein expression (Fig 2), which in hippocampus is associated with increased PVIN excitatory inputs (48). To determine whether BCAN expression in NAc PVINs has a role in maintaining excitatory synapses in the adult, we used Cre-dependent AAV shRNA constructs (21) to knockdown BCAN mRNA in NAc PVINs of adult PV-Cre mice (Fig. 6A–G). Knocking down BCAN decreased WFA signal around virally transduced (mCherry+) neurons, supporting the hypothesis that cell-autonomous synthesis of CSPG proteins by PVINs is essential for PNN integrity (Fig. 6H,I). Next, we stained pre- (VGlut1, VGlut2, VGAT) and post- (PSD95, Gephyrin) synaptic markers of major excitatory and inhibitory input populations (Fig. 7A) and quantified the density of colocalized puncta around mCherry+ PVIN somata (Fig. 7B–D). shBcan cells had decreased densities of both VGlut1+/PSD95+ (Fig. 7B) and VGlut2+/PSD95+ puncta (Fig. 7C) compared with shLacZ controls. By contrast there was no effect on VGAT+/Gephyrin+ puncta (Fig. 7D), indicating that this loss was selective for excitatory synapses.

Figure 6.

Validation of BCAN knock-down in adult NAc PVINs. A) Schematic of viral strategy for BCAN knock-down. B) Representative immunostaining showing colocalization of PV (green) and viral mCherry (magenta). Nuclei are labeled with Hoechst dye (blue). Scale bar = 10μm. C) Schematic of the Bcan1 and Bcan2 isoform transcripts, and target regions of the Bcan probe (black). This probe labels both Bcan isoforms to different degrees; as a result, transcript numbers cannot be consistently estimated from Bcan RNAscope signal intensity, and we reverted to relative signal intensity measurements. D) Representative RNAscope labeling of Bcan (green), nuclei (blue), and overlaid immunostaining for mCherry (white). Arrowheads indicate mCherry+ cells. Scale bar = 5μm. E) Bcan signal was significantly lower in mCherry+ cells from shBcan mice compared to shLacZ controls (**p=0.009 by Mann Whitney U test, n=44–59 cells/condition). F) Representative immunostaining of BCAN (green) and viral mCherry (white) in NAc of viral mice. Nuclei are labeled with Hoechst dye (blue). Scale bar = 10μm. G) mCherry+ cells from shBcan mice have significantly less BCAN signal compared to shLacZ controls (****p<0.0001 by Student’s t test, n=16–28 cells from 2–4 mice/group). H) Representative WFA (green) and mCherry (white) immunostaining in NAc of shBcan and shLacZ mice. Nuclei are labeled with Hoechst dye (blue). Scale bar = 10μm. I) mCherry+ cells from shBcan mice have significantly less WFA signal compared to shLacZ controls (***p=0.0008 by Mann Whitney U test, n=6–12 cells from 4 mice/group).

Figure 7.

BCAN knock-down decreases adult NAc PVIN excitatory drive. A) Representative immunostaining (left) and processed synaptic marker masks (right) of VGlut1 or VGlut2 (magenta), PSD95 (green), VGAT (cyan), Gephyrin (red), and mCherry (gray). Arrowheads indicate colocalized pre- and post-synaptic puncta. Scale bar = 2μm. shBcan mice have lower average densities of of VGlut1+/PSD95+ synaptic puncta (B; **p=0.0075 by Student’s t test, n=5 mice/group) and VGlut2+/PSD95+ synaptic puncta (C; **p=0.0022 by Student’s t test, n=5 mice/group), but not VGAT+/Gephyrin+ puncta (D; p=0.8401 by Student’s t test, n=4–6 mice/group) compared to shLacZ controls. E) Representative mEPSC current traces from shLacZ (gray) and shBcan (salmon) cells. BCAN knock-down significantly reduced mEPSC frequency (F; t=4.032, F(1,11)=16.25, **p=0.002 by nested t test), but not amplitude (G; t=1.623, F(1,11)=2.635, p=0.1328 by nested t test). E-G, n=8–13 cells from 4–9 mice/group. H) Representative Fos (green) and mCherry (magenta) immunostaining in NAc 2hr after an acute dose of saline or cocaine in shLacZ or shBcan mice. Scale bar = 50μm. I) Cocaine treatment significantly increased the fraction of mCherry+ cells that were also Fos+ in shLacZ, but not shBcan, mice (2-way ANOVA Virus × Treatment interaction F(1, 24)=17.29, p=0.0004; ****p<0.0001 by Bonferroni’s post-hoc test). J) Cocaine increased the number of Fos+ cells in NAc compared to saline treatment (2-way ANOVA main effect of Treatment F(1, 24)=5.147, p=0.0326), but there was no effect of viral condition (2-way ANOVA Virus main or interaction effect F(1, 24)<0.227, p>0.638). H-J, n=6–8 mice/group.

To determine whether structural differences in synapse number corresponded to functional changes in excitatory input onto NAc PVINs, we recorded mEPSCs from PVINs expressing shLacZ or shBcan. BCAN knockdown markedly decreased the frequency (Fig. 7E,F) but not the amplitude (Fig. 7G) of mEPSCs, consistent with the observed decrease in physical synaptic contacts. There were no changes in other properties of recorded mEPSCs (Table S3), though there was a trending increase in PVIN input resistance (Table S3), which is a predictable consequence of decreased excitatory synaptic input (62).

Finally, we asked whether BCAN knockdown impaired NAc PVINs’ in vivo response to cocaine, as indicated by expression of the neural activity-inducible protein Fos (63). Two hours after treatment, cocaine increased the fraction of mCherry+ cells that were Fos+ in shLacZ mice, but this effect was absent in shBcan mice (Fig. 7I). Reduced Fos induction was specific to PVINs because cocaine treatment increased the total number of Fos+ cells throughout the NAc in both control and PVIN-specific BCAN knockdown mice (Fig. 7J). Together, these data demonstrate a key role for BCAN2 in adult NAc PVINs in regulating their excitatory synaptic inputs and their likelihood for activation in response to cocaine in vivo.

BCAN knock-down in adult NAc PVINs accelerates reward learning

To assess the behavioral consequences of regulating BCAN in adult NAc PVINs, we tested NAc shLacZ control and shBcan knockdown mice for accumbens-related phenotypes. At baseline, relative to control, shBcan mice showed no difference in locomotion, anxiety-like behavior, or depression-like behavior in open field, elevated zero maze, or tail suspension tests (Fig. S4A–H). In novel object exploration, both shLacZ mice and shBcan mice preferred a novel object 30min and 24hr after the original exposure to a comparison familiar object. However, while their preferences were similar after 24hr, shBcan mice showed enhanced novel object preference after 30min (Fig. 8B). There were no differences in locomotion between groups, suggesting that novelty exploration did not merely reflect greater overall arousal in shBcan mice (Fig. S4I) and might instead indicate accelerated reward-based learning.

Figure 8.

BCAN knock-down accelerates reward learning in response to both physiological and drug rewards. A) Schematic of Novel Object Recognition paradigm. B) While both shLacZ and shBcan mice showed a significant preference for the novel object at both tests (Mixed-effects model main effect of Phase F(1.697, 46.66)=31.29, p<0.0001; shLacZ 30min, 24hr vs Training p<0.001; shBcan 30min, 24hr vs Training p<0.01 by Bonferroni’s post-hoc tests), shBcan mice showed a stronger preference at 30 min compared to shLacZ controls (Mixed-effects model Phase x Virus interaction F(2, 55)=3.959, p=0.0248; *p=0.0382 by Bonferroni’s post-hoc test; n=12–18 mice/group/time point). C) Schematic of CPP paradigm with test trials (10min) interleaved between conditioning sessions (30min). Exp, experimental animals; Ctrl, saline-only control animals. D) Cocaine-conditioned shLacZ mice display a significant preference after 3 conditioning sessions, while shBcan mice display a significant preference after a single conditioning session, which is maintained for the remainder of the paradigm (rmANOVA Day × Group interaction F(9, 102)=4.134, p=0.0001; #p=0.0149 shLacZ Test vs Acclimation; *p=0.0116, **p=0.0033, ***p=0.0008 shBcan Test vs Acclimation by Bonferroni’s post-hoc tests; n=9–11 mice/group). E) Cocaine treatment significantly increased the average number of infrared beambreaks on each treatment day (Mixed-effects model Day x Group interaction F(15, 160)=10.55, p<0.001; *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 vs saline controls; within-virus comparisons of Coc 1, 2, 3 p>0.999 by Bonferroni’s post-hoc tests; n=8–10 mice/group/time point).

To directly test this hypothesis, we asked whether loss of BCAN in NAc PVINs would alter the development of cocaine CPP. To monitor the development of CPP we interleaved 10min test sessions after each round of conditioning, and we used a moderate dose of cocaine (10mg/kg) to avoid ceiling effects. As expected, saline-conditioned mice showed no chamber preference over the course of the paradigm (Fig. 8D). Cocaine-conditioned shLacZ mice displayed a significant preference for the drug-paired chamber after 3 conditioning sessions (Fig. 8D). However, cocaine-conditioned shBcan mice preferred the drug-paired chamber after just one round of conditioning, and this preference was maintained for the remainder of the paradigm (Fig. 8D). There was no effect of viral group on cocaine-induced locomotion (Fig. 8E), demonstrating that loss of BCAN in NAc PVINs does not enhance the acute pharmacological response to cocaine, but rather impacts how cocaine experience is encoded into place associations over the course of learning. Thus, BCAN knockdown in NAc PVINs accelerates motivated learning in two behavioral settings, novel object recognition and cocaine CPP.

Discussion

Here, we report that PNNs around NAc PVINs emerge during adolescence and are dynamically and locally remodeled over the course of cocaine CPP in adult mice via the acutely induced transcription of Bcan2 and its regulatory enzyme Adamts15 in NAc PVINs. Using cell-autonomous manipulations, we identified a specific role for BCAN in regulating NAc PVIN excitatory synaptic inputs and reward-motivated learning in adult mice. These data suggest that stimulus-dependent regulation of PNNs in the adult NAc tunes PVIN synaptic connectivity and regulates thresholds for reward learning.

The developmental emergence of NAc PNNs mimics the experience-driven maturation of PVINs and PNNs in cortex that closes critical periods for sensory plasticity (18,64). The size of the PV+/WFA+ population further correlates with the magnitude of cocaine CPP in adult mice, suggesting that variability in NAc PVIN maturation establishes the potential range of adult reward-seeking behavior. Direct perturbation of NAc PNNs through BCAN knockdown is sufficient to accelerate learning in response to both novelty and cocaine. These data further raise the possibility that developmental exposures that compromise the maturation of NAc PVINs and their PNNs may persistently adjust thresholds for addictive behaviors later in life.

Our data showing regulation of adult NAc PV and WFA labeling during cocaine CPP complement studies demonstrating psychostimulant-induced changes in WFA and PV labeling in prelimbic PFC (47) or WFA labeling in cerebellum (65). In contrast to our behavioral results, enzymatic manipulation of CSPGs in these areas impairs cocaine seeking (32,35), demonstrating that the role of PNNs varies according to the role of PVINs in specific circuits. To our knowledge, only one other study has demonstrated enhanced acquisition of associative learning after PNN perturbation, specifically CSPG degradation in dorsal striatum (37). A difference between those studies and ours is that we used intersectional genetic methods to knock out single PNN components in specific cell populations of the adult mouse brain. This is important because regulation of distinct CSPGs may have differential consequences for PVINs and circuit function. For example, a recent study augmented adult fear memory extinction by briefly manipulating Acan expression in mPFC PVINs (66). Future studies that address functions of the PNN components at different times and in distinct brain regions will advance understanding of the cellular and circuit consequences of these structures.

Psychostimulants acutely drive transcription of both Bcan2 and its regulatory enzyme Adamts15 in NAc PVINs, revealing a mechanism by which cocaine exposure can lead to local remodeling of PNNs early in cocaine association learning. Knockdown of BCAN in adult NAc PVINs accelerated reward learning, but did not affect the final learned behavior, further supporting the idea that changes in NAc PVIN physiology and synaptic drive may primarily tune the early stages of reward association learning. Loss of BCAN expression in adulthood destabilizes both VGlut1+ and VGlut2+ synaptic inputs to PVIN somata, potentially via mislocalization of AMPA receptor subunits (21). Together, these data suggest that BCAN expression in NAc PVINs may act as a brake on processes which selectively affect specific NAc PVIN inputs following drug experience (5). Changes in the feed-forward inhibition motif in NAc—such as alterations to the relative excitatory drive to PVINs and SPNs from specific inputs (5,8)—likely affect the coordinated firing of different SPN ensembles that underlie the expression of reward-motivated behaviors. We predict that interfering with NAc PVIN PNN remodeling, such as by overexpressing a non-cleavable BCAN2 in PVINs, modeled on prior studies of BCAN1 (55), may be able to slow this process, and thus slow the acquisition of addiction-like behaviors.

Although Bcan2 transcription is acutely induced in a subset of PVINs after acute cocaine, overall PNN staining is decreased after repeated exposure, indicating that the balance of PNN component expression and degradation is eventually shifted toward degradation. This may be due in part to more persistent changes in NAc PVIN physiology (5,10) and the decay of psychostimulant-induced transcription after repeated exposure (67). Additionally, non-cell autonomous mechanisms may contribute to changes in ECM, including expression of PNN components, MMPs, and ADAMTSs by surrounding glia. Indeed, MMP activity has been shown to locally regulate SPN synapses during later stages of addiction-like paradigms (68–72). While our data show a primary role for NAc PVIN PNNs early in reward learning, consequences for persistent changes in these structures remain to be explored.

The data presented here support an emerging model in the field where PNNs, rather than statically restricting circuit plasticity, are dynamic structures that make important contributions to PVIN synaptic inputs and circuit output. In NAc, this process is controlled at least in part by cell-autonomous mechanisms to allow experience-dependent regulation of adult PVIN synaptic connectivity.

Limitations

BCAN knockdown is reported to decrease PV expression (22), which could contribute to the learning phenotypes described here. We did not determine whether the expression of other PNN component CSPGs such as ACAN were disrupted in our BCAN knockdown neurons. Although our studies included mice of both sexes (Table S4), they were not powered to detect potential sex differences in the outcome measures. Finally, our analysis of motivated behaviors did not extend to operant conditioning paradigms such as cocaine self-administration.