phosphoprotein-32, or DARPP-32; Fig. 1a-b) among other canonical MSN markers (Fig. 4d). MSN populations in the rat NAc included previously characterized D1-MSNs and D2MSNs, as well as a novel Ppp1r1b+ population termed visualize the spatial distribution of targeted gene transcripts. oThugh

Oprm1 and Grm8 expression can be observed across the broader NAc, Chst9 expression was restricted to a much smaller group of cells, again highlighting the selectivity of this marker across NAc subpopulations (Fig. 2c). Curiously, this Chst9+ population appears to form discrete cellular clusters along the medial and ventral borders of the NAc shell subregion (Fig. 2c). To confirm this observation across all animals, region of interest (ROI) coordinates from quantified cell types were combined into an integrated administration tasks16,18,19. µOR antagonists decrease sucrose intake and feeding behavior15. Additionally, opioid receptor antagonists delivered into the

NAc block opioid reinforcement and social reward16,17, and direct infusion of µOR receptor agonists into the NAc is reinforcing in selfhT

e principal cells of the NAc are GABAergic medium spiny neurons (MSNs) that directly target the output nuclei of the basal ganglia (ventral tegmental area and substantia nigra) or indirectly target these

Drd1 distinct and

Drd2 dopamine receptors, which form populations termed

D1-MSNs and D2-MSNs23,24. understanding striatal physiology25, circuitry26, and molecular responses to drugs of abuse27–29, both cell populations express Oprm130 and contribute to opioidinduced behaviors31. Further, recent single nucleus RNA sequencing (snRNA-seq) studies have revealed unexpected heterogeneity in MSNs within the NAc and broader striatum, with multiple subtypes of Drd1 and Drd2 expressing MSNs present in the mouse32, rat33–35, monkey36,37, and human NAc37,38. Strikingly, NAc MSN subtypes exhibit highly variable expression of Oprm134, highlighting the need for further investigation into the roles of distinct MSN subclasses in opioid action. Here, to we leveraged available snRNA-seq characterize a subpopulation of

NAc datasets

MSNs that exhibit the highest level of Oprm1 expression across all defined NAc cell types. Using data-driven comparisons, we identified a marker gene (Chst9, which encodes a carbohydrate sulfotransferase) that defines this neuronal population. Next, we employed single molecule RNA uflorescence in situ hybridization to validate this marker and map the spatial location of this transcriptionally defined cell type in the rat NAc. Finally, comparison of transcriptional profiles of MSN subtypes from the rat, monkey, and human NAc revealed that this cell population is conserved across species. Together, these findings highlight a unique population of NAc neurons with abundant Oprm1 expression and identify a specific molecular marker that may be useful in future targeting of this cell type.

marker gene for the

Grm8-MSN population, as it exhibits selectively high expression for that population only (Fig. 1f). eTh evaluation of

Chst9 expression, identifying this as a unique marker gene (Fig. 1g).

Validation of Chst9 as an Oprm1+ subpopulation marker gene

To validate snRNA-seq findings demonstrating that Chst9 is in a distinct population

of Grm8+/ Oprm1+ MSNs in the NAc, we employed RNAscope39, a

widely ulforescence implemented single molecule

RNA in situ

hybridization (smRNA-FISH) platform for multiplexed transcript detection. Tissue sections containing the NAc from naïve rats were used for multiplexed hybridization using distinct probes for Grm8, Oprm1, and Chst9 (Fig. 2a-c). Consistent with rat NAc snRNA-seq data, Grm8, Oprm1, and Chst9 mRNA colocalized in a subset of NAc neurons (Fig.

Chat-Int.

Sst-Int.

GABAergic

Drd1 MSN-2

NAc snRNA-seq Astrocyte Polydendrocyte

Pvalb-Int.

Microglia Grm8-MSN

Mural DMrSd3N- GlutamatOerlgigic-1

Drd1-MSN-1 spatial map, forming a spatially co-registered map of cell type location of the NAc (Fig. 2d). eThse maps further revealed that although Oprm1+ cells are distributed across the medial/lateral and dorsal/ventral axes of the NAc, Chst9+ cells were relatively confined along the NAc shell’s medial and lateral border (Fig. 2d). Additionally, by reflecting the degree of Oprm1 expression in the area of each point, this map suggested that nearly all Chst9+ cells also abundantly expressed Oprm1 (Fig. 2d).

To confirm the relative expression of Oprm1 within diefrent NAc cell types, we quantified the intensity and transcript-level puncta of Oprm1 signal within all Oprm1+ cells. Oprm1+ cells were then further identified as Chst9+ or Chst9-. eThse analyses verified significantly elevated expression of Oprm1 in Chst9+ cells as compared to Chst9- cells (Fig. 2e). We additionally evaluated the percentage of diefrent NAc cell types, including Drd1+ and Ppp1r1b+ neurons, that express Oprm1. Oprm1 was detected in in ~40% of Drd1+ neurons, and in ~30% of total NAc cells (Fig. 2f). However, Oprm1 is expressed in over 90% of Chst9+ NAc neurons (Fig. 2f). In combination with snRNA-seq data, these observations confirm Chst9 as a selective marker for NAc neurons that abundantly express Oprm1 mRNA.

Conservation of Chst9+ neurons across mammalian species

While these studies identify Chst9+ MSNs in Grm8 Oprm1 Chst9 DAPI

b M

L D V

Grm8 Oprm1 Chst9 DAPI d 30 a t cn10 u p 1 m r p O 3 1

All Oprm1+ NAc core

AC

NAc shell *

* a

AC NAc shell Pdyn Drd2 Penk Adora2a Grm8 Oprm1 Chst9 Foxp2 D1−SDh2e−lS/ODheT1l−/NOUTDD1AD/DP12−−ICHDjy1b−rDMDid21a−−trMSiDxta2rti−oriSsxotrmioesome

D1-NUDAP .D1N_.DS1.D .D .D .D1.DF _.DA2_.D2N.D

A _B1_C1_D1_E _ 2 B _C2_D

NSNSNSNSNSNSNS MSNMSM M M M M M M M

MSN.D1_F DRD1 EBF1 PDYN DRD2 PENK ADORA2A GRM8 OPRM1 CHST9 FOXP2 PPP1R1B

PPP1R1B FOXP2 BCL11B GAD1 DRD1 PDYN EBF1 HTR4 DRD2 ADORA2A PENK GRM8 OPRM1 CHST9 c f llvee 4 n o iss0.6 e r p xE 4 2 3 2 3 4 3 5 3

DRD1 EBF1 PDYN DRD2 PENK ADORA2A GRM8 OPRM1 CHST9 FOXP2 PPP1R1B

PPP1R1B FOXP2 BCL11B GAD1 DRD1 PDYN EBF1 HTR4 DRD2 ADORA2A PENK GRM8 OPRM1 CHST9

Rat MSNs Phillips, et al., 2023 (Mol. Cell. Neurosci.)

NHP MSNs He, et al., 2021 (Current Biology)

Human MSNs Tran, et al., 2021

(Neuron) 11BPPR 2FPXO 11LBBC 1ADG 1RDD PYND 1FEB 4THR 2RDD 2AARO EKPN 8RGM 1PROM 9TSHC P AD coexp%reosfscinegllttwypoegenes 11BPPR 2FPXO 11LBBC 1ADG 1RDD PYND 1FEB 4THR 2RDD 2AARO EKPN 8RGM 1PROM 9TSHC

P AD 0 20 40 60 80 100 Figure 3. Identification of conserved populations of Chst9+ MSNs in non-human primates and humans. a-c, Violin plots of MSN marker genes in transcriptionally defined cell populations of the rat (a), non-human primate (b), and human (c) NAc. d-f, Coexpression heatmaps showing the percent of cells within the rat Grm8-MSN cluster (d), non-human primate D1-NUDAP cluster (e), or human MSN.D1_F cluster (f) coexpressing identified genes. the rat brain, we next investigated conservation of expression of other marker genes as well. To examine this NAc cell population across species, with a focus whether this was the case, we generated co-expression on higher-order mammals. To do this, we leveraged heatmaps of selected transcripts within Grm8publicly available snRNA-seq atlases of the non- MSNs, D1-NUDAP cells, and MSN.D1_F cells (Fig. human primate (NHP)36 and human38 NAc that have 3d-f). For this analysis, we prioritized previously previously been used to compare gene expression characterized markers of MSNs or MSN subtypes, profiles and MSN subtypes across species 34. As in addition to Grm8, Chst9, and Oprm1. Consistent previously described, both NHP and human datasets with a potentially conserved role across rodent and identified unique MSN clusters that exhibit similar primate NAc samples, we observed similar patterns marker gene expression patterns to those observed in of expression for additional marker genes in these the rat NAc (Fig. 3a-c). eTh direct comparison of Grm8, populations. For example, each cluster included a Oprm1, and Chst9 expression between these species high percentage of cells expressing the genes Foxp2 revealed that one cell type in each respective dataset and Bcl11b, commonly cited markers of MSNs40,41. exhibited colocalization and abundant expression Furthermore, an appreciable subset of cells in each of both Grm8/GRM8 and Oprm1/OPRM1. Notably, cluster expressed Drd1/DRD1 and Pdyn/PDYN, both the same population was marked by cluster-selective established markers of D1-MSNs42,43. Additionally, Chst9/CHST9 expression. While the rat literature had feature expression plots from each species confirmed termed these populations Grm8-MSNs, NHP datasets a similar expression of Drd1/DRD1, among other termed them D1-NUDAP cells, and human datasets genes, across these three clusters (Fig. S1). Taken referred to these as MSN.D1_F cells (Fig. 3a-c). together, this finding demonstrates that Grm8-MSNs,

Transcriptionally conserved cell types should and specifically Chst9+ neurons, are conserved across be expected to display a similar enrichment and co- species and exhibit similar gene expression signatures

investigation into the circuit and behavioral role of single or repeated cocaine administration34. Future hT

e anatomical localization of Chst9+ MSNs at electrophysiological and transcriptional responses of to previously described cell types in higher-order mammals.

family member Fosb) in Grm8-MSNs following either studies will be needed to comprehensively define the Chst9+ cells to opioids and other psychoactive drugs.

Chst9 as the most specific marker for Grm8-MSNs, we also detected elevated levels of other potential marker genes in this population (Fig. 1e). eThse included zinc finger homeobox 1) and

Tac3

Tshz1 (Teashirt (Tachykinin precursor 3; the rat ortholog of mouse Tac2). Each of these genes have recently been discovered by other groups to mark specific populations of MSNs in the dorsal and ventral striatum, and cells expressing these genes have been shown to be important for aversive aspects of motivated behavior43,51. Tshz1 is expressed by a subgroup of striosomal D1-MSNs located in the dorsal striatum, but is also found in the ventral striatum.

Tshz1+ neurons in the dorsal striatum are activated by aversive stimuli, and optogenetic stimulation of these cells causes place aversive and suppresses locomotion43. Similarly, inhibiting Tshz1+ MSNs impairs learning related to punishment, but does not impact reward learning. However, while this study characterized Tshz1+

nothing is known about the behavioral role of Tshz1+ neurons in the ventral striatum or NAc, and we did not observe significant Chst9 expression in the dorsal striatum. Similarly, another recent study identified Tac2 mRNA to mark a subpopulation of D1-MSNs in the mouse NAc51. This study found that acute cocaine decreased activity of Tac2+ neurons, and stimulation of this population suppressed cocaine place preference brains were then wrapped in aluminum foil and placed and self-administration. Given the enrichment of Tac3 on dry ice. Tissue was stored at -80°C until the day of in rat Grm8-MSNs, the cells targeted in this mouse sectioning. Aeftr a 30-minute equilibration at -20°C, study likely overlap to some extent with the Chst9+ 10 μm coronal sections were obtained using a Leica cells investigated in the present manuscript. However, CM 1850 cryostat (Deer Park, IL, USA) set to -20°C. further work is needed to characterize the circuit and Sections containing the NAc, ranging from 1.68behavioral role of this population, including synaptic 1.92 mm anterior-posterior (A/P) to Bregma, were inputs and downstream targets. mounted onto room temperature frosted microscope

In summary, we have identified a slides, air dried for 60 minutes at -20°C, and stored at transcriptionally and anatomically unique population -80°C until staining. of NAc MSNs marked by high expression of Oprm1 and selective expression of Chst9. Given that this population Tissue preparation and imaging for RNAscope. eTh expresses other markers of D1-MSNs, it is likely that RNAscope Multiplex Fluorescent v2 assay kit (ACD prior labeling studies or genetic manipulations may Bio, 323110) and RNAscope 4-plex ancillary kit (ACD have targeted this cell population along with more Bio, 323120) were used to stain sections, following the classically defined D1-MSNs that do not express Chst9 manufacturer’s recommended protocol. Probe sets and exhibit lower expression of Oprm1. For example, used include DAPI (320858), Rn-Oprm1-C1 mRNA use of Drd1 or prodynorphin (Pdyn) gene promoter (410691-C1), Rn-Grm8-C2 mRNA (1149671-C2), sequences is a common strategy for cell-restricted Rn-Chst9-C4 mRNA (1149681-C4), Rn-Ppp1r1b-C2 expression of transgenes in D1-MSNs2,21,52–54. Using this mRNA (1048941-C2), and Rn-Drd1-C3 mRNA strategy, prior work has revealed that Oprm1 expression (317031-C3). Three diefrent groups of sections were in Pdyn+ cells (using a Pdyn BAC transgenic mouse) probed, with varying marker probe combinations— is suficient to restore opioid reward in an Oprm1-/- group 1 (Chst9 population validation experiment) mouse10. However, based on the present results (as well probes included Oprm1, Grm8, and Chst9; group as recent work demonstrating that Pdyn regulatory 2 (MSN marker experiment) probes included elements are accessible in Grm8-MSNs55), it is likely Oprm1, Ppp1r1b, and Drd1; group 3 (MSN marker that this manipulation restores Oprm1 expression in experiment) probes included Oprm1, Ppp1r1b, and multiple MSN subtypes including Chst9+ cells, which Chst9. Channels were matched with genes according may each contribute to opioid reward in distinct ways. to relative expression level and channel background Our results highlight the previously underappreciated in diefrent experiments: the green channel (520 nm diversity within broad striatal MSN subclasses, and dye, recommended for high expressors) was assigned underline the need for continued exploration of the to Grm8, Ppp1r1b, and Oprm1, the red channel distinct roles of unique cell types. (570 nm dye, recommended for low expressors) was assigned to Oprm1, and the white channel (690 nm METHODS dye, recommended for low expressors) was assigned to Chst9 and Drd1. 5 total sections (1 each from 3M/2F) were used for the Chst9 population validation experiment, and 6 total sections (1 each from 3M/3F) were used for each of the MSN marker experiment.

Aeftr staining, sections were imaged using the BZX800 Keyence Microscope, and then stitched using the BZ-X800 Keyence Analyzer software. High sensitivity images were taken using OP-87767 filter cubes for GFP (Channel 1, green), Texas Red (PE) (Channel 2, red), DAPI (Channel 3, blue) and Cy5 (Channel 4, white) at 4x (PlanApo NA 0.20), 20x (Plan Fluor NA 0.45 Ph1), and 100x (PlanApo 1.45/0.113 mm Oil) magnifications.

Animals. Adult male and female Sprague-Dawley rats were obtained from Charles River Laboratories (Wilmington, MA, USA). Rats were cohoused in pairs in plastic filtered cages with nesting enrichment in an Association for Assessment and Accreditation of Laboratory Animal Care–approved animal care facility maintained between 22° and 24°C on a 12hour light/12-hour dark cycle with ad libitum food (Lab Diet Irradiated rat chow) and water. Bedding and enrichment were changed weekly by animal resources program staf. A total of 6 naïve rats (n = 3M/3F) were used in initial Chst9 population and MSN characterization experiments. Animals were randomly assigned to experimental groups. All experiments were approved by the University of Alabama at Birmingham Institutional Animal Care and Use Committee (IACUC).

Brain tissue collection. Rats were euthanized by live decapitation using a guillotine (World Precision Instruments, Sarasota, FL, USA). Brains were rapidly removed and submerged in 2-methylbutane, chilled on dry ice, for approximately 30 seconds. Flash-frozen Image analysis. Approximate A/P coordinates of sections were determined by comparing anatomical landmarks within 4x whole section images to those in coordinate and region-labeled diagrams of the Paxinos and Watson rat brain atlas56. eThse A/P distinctions were used to process 20x ROI images of the NAc. This protocol was used for the Chst9 population validation and MSN marker experiments.

To isolate the NAc for quantification, brain atlas templates at the corresponding A/P coordinate to that of a given section were overlaid on 20x images of

Statistical analysis. All image data were analyzed statistically with Prism software using the nested t-test (Chst9 population validation experiment) or Welch’s t-test (fentanyl experiment). A 95% confidence level was set for every test, with a definition of statistical significance at p < 0.05. All graphs, diagrams, and image edits were made using Prism, Adobe Illustrator, R Studio, and FIJI.

Published snRNA-seq data analysis. Data objects for NHP and human NAc snRNA-seq datasets were obtained from data repositories outlined in each respective publication. eTh data object for the rat NAc snRNA-seq dataset was provided directly by Robert Phillips at the University of Alabama at Birmingham (UAB). All data objects were analyzed with R software. that section in Adobe Illustrator (San Jose, CA, USA) using anatomical landmarks as a guide. This overlay image was then imported into FIJI57 to create NAcspecific ROIs for that 20x image. eThse ROIs were then applied to each individual channel of the original 20x image, and the NAc was cropped from the rest of each channel’s image using the Clear Outside plugin. DAPI channel nuclei-specific ROIs were generated with the StarDist-2D plugin58, using the “Versatile (uflorescent nuclei)” model with a probability/score threshold of 0.20 and an overlap threshold of 0.45. Nuclei-specific ROIs were then applied to images containing Oprm1, Grm8, Chst9, Ppp1r1b, or Drd1 probe signal, and raw signal intensity within each ROI was measured.

Cell types were categorized based on positive marker colocalization within a single ROI, and intensity values were determined within diefrent cell types, using R Studio. Nuclei-specific ROI positions were used to plot locations of diefrent cell types, using R Studio.

100x magnification images were taken using the z-stack setting. One image for each channel within a stack was chosen for analysis, for a total of 10-14 images per animal within the Chst9 population validation experiment. Selected channel images for one total z-stack image were merged, and the resulting overlay was converted to Nikon (.nd2) format using NIS-Elements software. Nuclei-specific ROIs were generated around DAPI staining using QuPath59 software version 0.3.2. Cell detection parameters were set to background radius 0 px, median filter radius CONFLICTS OF INTEREST 0 px, minimum area 100 px^2, maximum area 3000 ehT authors declare no competing interests. px^2, threshold 35, and cell expansion 3 px. For cell detection, “split by shape”, “include cell nucleus”, “smooth boundaries”, and “make measurements” boxes were checked. Marker puncta-specific ROIs were then generated, and counts were measured within each DAPI ROI for an image. Subcellular detection parameters were set to expected spot size 12 px^2, minimum spot size 0.5 px^2, and maximum spot size 25 px^2. Cell types were categorized based on positive marker colocalization within a single ROI, and puncta counts were determined within diefrent cell types, using R Studio.