All animal procedures were conducted in accordance with the guidelines of the French Agriculture and Forestry Ministry for handling animals (authorization number/license B34-172-41) and approved by the relevant local and national ethics committees (authorization APAFIS#38912). Female and male mice were used for all experiments. Animals were housed in groups of 2 to 5 per cage under standardized conditions with a 12-h light/dark cycle, ad libitum food and water, stable temperature (22 ± 2°C) and controlled humidity (55 ± 10%). The mouse models used in this study were derived from the following original strains: Pvalb-IRES-Cre/+ (B6.129P2-Pvalb^{tm1(cre)Arbr/}J, RRID: IMSR_JAX:017320), Ribotag (B6J.129(Cg)-Rpl22^{tm1.1Psam/SjJ}, RRID: IMSR_JAX:029977), Ai14 (B6;129S6-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze}/J), Ai32 (B6;129S-Gt(ROSA)26Sor^{tm32(CAG-COP4*H134R/EYFP)Hze}/J). To generate Pvalb-reporter mice, Pvalb-IRES-Cre/+ mouse line was crossed with the Ribotag line to generate Pvalb-Ribotag mice, with the Ai32 line (which express an improved channelrhodopsin-2/EYFP fusion protein) to generate Pvalb-ChR2 mice, with the Ai14 line to generate Pvalb-Ai14 mice, and with the Ai32; Ribotag line to generate Pvalb-ChR2-Ribotag mice.

Mouse strains used for immunofluorescence included Pvalb-ChR2-Ribotag mice (n = 2), Pvalb-ChR2 mice (n = 3), Pvalb-Ribotag mice (n = 4) and Pvalb-Ai14 mice (n = 3). Pvalb-Ribotag mice were used for low-input RNA-sequencing (RNA-seq). The experimental groups included: saline (n = 7), D-amphetamine (n = 8), highly palatable food master (n = 6), highly palatable food yoked (n = 6), standard food master (n = 6), and standard food yoked (n = 6). For each RNAseq sample, striatal tissue from two to three mice were pooled.

(+)-α-Methylphenethylamine [D-amphetamine (D-amph)] sulfate salt (5 mg/kg) from Tocris was dissolved in 0.9% (w/v) NaCl (saline) and injected intraperitoneally (i.p) in a volume of 10 mL/kg. Mice were administered with d-amphetamine (5 mg/kg) during 5 days and euthanized 3 days after the last injection.

Free-floating sections (30 μm) of the dorsal striatum were prepared as previously described (Cutando et al., 2021). On day 1, slices were washed for 10 min in PBS (3x), incubated 15 min in 0.2% Triton X-100 in PBS before incubation overnight at 4°C with primary antibodies including chicken anti-GFP (1:500, Invitrogen, #A10262), rat anti-HA (1:500, Roche, #11867431001), goat anti-PV (1:500, Swant, #PVG-213), rabbit anti-PV (1:1000, Swant, #PV25), mouse anti-RFP (1:1,000, MBL, #M155-3), rabbit anti-TTF1 (1:500, Santa-Cruz, #sc-13040) and N-acetylgalactosamine-binding Wisteria floribunda agglutinin (WFA, 1:1000, Sigma, #L1516). Non-specific binding was blocked with 10% Normal Donkey Serum (NDS) in PBS for 2 h for WFA staining. On day 2, slices were rinsed in PBS and incubated 45 min with various combination of the following secondary antibodies: goat Alexa Fluor 488-coupled anti-chicken (1:500, Jackson ImmunoResearch, #103-545-155), donkey Alexa Fluor 488-coupled anti-goat (1:500, Jackson ImmunoResearch, #103-545-003), donkey Alexa Fluor 647-coupled anti-goat (1:500, Jackson ImmunoResearch, #103-605-003), donkey Alexa Fluor 594-coupled anti-mouse (1:500, Jackson ImmunoResearch, #103-585-150), goat Alexa Fluor 594-coupled anti-mouse (1:500, Jackson ImmunoResearch, #103-585-003), goat Alexa Fluor 488-coupled anti-rabbit (1:500, Invitrogen, #A11034), goat Cy3-coupled anti-rabbit (1:500, Jackson ImmunoResearch, #111-165-003), donkey Alexa Fluor 647-coupled anti-rat (1:500, Jackson ImmunoResearch, #712-605-150), goat Alexa Fluor 647-coupled anti-rat (1:500, Thermo Fisher Cat#112-005-003) and Streptavidin A488 (1:500, Invitrogen, #S11223) or Streptavidin Atto 550 (1500, Sigma, #96404). Sections were then rinsed twice in PBS before mounting on SuperFrost® slides (VWR #631-0108) and coverslipped with a xylene-based mounting medium (Leica Micromount #3801731). Slices were stored at 4°C until imaging. Confocal miscroscopy/imaging were carried out at the Montpellier RIO Imaging Facility. Fluorescent images of labeled cells in the region of interest were captured using sequential laser scanning confocal microscopy (Leica SP8). Three to eight images per hemisphere from four to six sections for a given marker were used for quantifications. Adjacent serial sections were never counted for the same marker to avoid any potential double counting of hemisected neurons. Images have been analyzed using Fiji software.

Whole striata were extracted from adult Pvalb-Ribotag mice, as previously described (Puighermanal et al., 2020). Mice chronically treated with d-amph (5 mg/kg) or saline were sacrificed 3 days after the last administration. Mice that underwent the operant behavior task were euthanized 1 day after the last FR5 training session. Striatal tissue from two to three mice was pooled to generate a single sample for polyribosome immunoprecipitation and subsequent RNAseq.

HA-tagged ribosome cell-type-specific mRNA purification by Translating Ribosome Affinity Purification (TRAP) from striatal samples of Pvalb-Ribotag mice was performed as previously described (Cutando et al., 2022), using 5 μL of anti-HA antibody (Clone 16B12; BioLegend, 901513) and magnetic beads (Fisher Scientific, #88803). Total RNA from the pellet fractions was extracted using the RNeasy Micro Kit (Qiagen, #73934), and from the input fractions using the RNeasy Mini Kit (Qiagen, #74104), following the manufacturer’s instructions. An on-column DNase treatment was included to eliminate genomic DNA contamination. RNA quality and quantity were assessed using 1 μL of sample on a Nanodrop One spectrophotometer (Thermo Scientific). For the translatome profile experiment, three biological replicates were used for the RNAseq analysis, each consisting of pooled tissue from 2 to 3 mice. For the d-amphetamine experiment, three biological replicates were used, each consisting of pooled tissue from 2 to 3 mice. For the operant behavior experiment, two biological replicates were used per group, each consisting of pooled striata from 2 to 3 mice.

RNA sequencing libraries from Pvalb-Ribotag mice striatal tissue were prepared following the SMART-seq2 protocol (Picelli et al., 2014), with some modifications. Briefly, total RNA samples were quantified using the Qubit® RNA BR Assay Kit (Thermo Fisher Scientific), and RNA integrity was assessed with the Agilent DNF-471 RNA (15 nt) Kit on the Fragment Analyzer 5,200 system (Agilent). Reverse transcription was performed on 1.8 μL of total RNA input (6–11 ng, depending on sample availability) using SuperScript II (Invitrogen) in the presence of oligo-dT30VN primers (1 μM; 5′-AAGCAGTGGTATCAACGCAGAGTACT30VN-3′), template-switching oligonucleotides (1 μM), and betaine (1 M). The resulting cDNA was amplified using KAPA HiFi HotStart ReadyMix (2×) (Roche) and 100 nM IS PCR primer (5′-AAGCAGTGGTATCAACGCAGAGT-3′), with 8 cycles of PCR. After purification with Agencourt AMPure XP beads (1:1 ratio; Beckman Coulter), product size distribution and concentration were evaluated using the Bioanalyzer High Sensitivity DNA Kit (Agilent). A total of 200 ng of amplified cDNA was fragmented for 10 min at 55°C using the Nextera XT Kit (Illumina), followed by 12 cycles of amplification with indexed Nextera PCR primers. The resulting libraries were purified twice using Agencourt AMPure XP beads (0.8:1 ratio) and quantified again using the Bioanalyzer High Sensitivity DNA Kit. Sequencing was performed on an Illumina NovaSeq 6000 system in paired-end mode with a read length of 2 × 51 bp, according to the manufacturer’s protocol for dual indexing. Image analysis, base calling, and quality scoring were carried out using Real Time Analysis (RTA) software version 3.4.4, followed by the generation of FASTQ files.

For the operant behavior experiment (Figure 1), RNA-seq reads were mapped against the Mus musculus genome (GRCm38) using STAR 2.5.3a (Dobin et al., 2013) with ENCODE parameters. Gene-level quantification was performed with RSEM 1.3.0 (Li and Dewey, 2011) using the gencode. M21 annotation. For the d-amphetamine experiment (Figure 1), RNAseq reads were aligned against the Mus musculus genome (GRCm39) using STAR 2.7.8a (Dobin et al., 2013), also with ENCODE parameters and gene quantification was performed with RSEM 1.3.0 (Dobin et al., 2013) using the gencode. M34 annotation.

Genes with at least 1 count-per-million reads (cpm) in at least 3 samples were retained. Counts were normalized using the trimmed mean of M-values (TMM) method and transformed into log2-counts per million (logCPM). Differential gene expression analysis was performed using the limma R package v3.54.2 (Ritchie et al., 2015). The voom function (Law et al., 2014) was used to estimate the mean–variance relationship and compute observation-level weights. These voom-transformed counts were used to fit linear models. Functional enrichment analysis was conducted using g: Profiler via the gprofiler2 R package v0.1.8 (Kolberg et al., 2023) and ShinyGO (Ge et al., 2020). Multidimensional scaling plots were generated with the limma function plotMDS using the top 500 most variable genes.

Statistical analyses were performed with GraphPad Prism v10.4.0. Behaviors were analyzed with two-way repeated measure ANOVA or three-way ANOVA as detailed in Supplementary Table 4.

We first used Pvalb-ChR2-Ribotag mice to visualize PV interneurons and to confirm Ribotag recombination as co-expression of the HA tagged ribosomal subunit in PV + neurons. Triple immunofluorescence analyses revealed that endogenous PV interneurons identified using PV antibody outnumbered by far the ChR2-and/or HA-positive interneurons in the DS (Figures 1a,b). Indeed, no recombination was found in ~68% and ~57% of the PV interneurons analyzed in the dorsomedial (DMS) and dorsolateral striatum (DLS), respectively (Figure 1b). We also found a fraction of PV cells expressing either only ChR2 (DMS: ~10% and DLS: ~8%) or to a lesser extent expressing only HA (DMS: ~1% and DLS: ~1%) (Figures 1a,b). Facing this low rate of recombination, we decided to reevaluate the degree of recombination in two additional mouse lines, the Pvalb-Ribotag and the Pvalb-Ai14 mice expressing the red fluorescent protein tdTomato (Figure 1c). In Pvalb-Ribotag mice, double immunofluorescence analyses revealed a higher percentage of PV interneurons expressing HA (Figures 1b,d). We also found that the percentage of PV/HA-expressing interneurons was higher in the DLS (~75%) compared to the DMS (~49%). Similar results were found when analyses were performed in Pvalb-Ai14 mice in which ~67% and ~47% of the PV interneurons expressed RFP in the DLS and DMS, respectively (Figures 1e,f). Importantly, in none of the three mouse lines tested, did we detect cells expressing ChR2, HA or tdTomato in the ventral striatum.

The translatome of DS PV interneurons was established by identifying the relative enrichment of genes in the pellet fraction containing tagged Ribosomes-bound mRNAs compared to the input fraction where mRNAs from all cell types were present (Figure 2a). We first validated the selectivity of the approach by demonstrating the enrichment in the pellet fraction of well-established PV markers including Pvalb, Cox6a2, Kcnc1, Vwc2, Clstn2, Nrip3, Nxph1, Pthlh, Ubash3b and Kcnip1 (Figure 2b; Supplementary Table 1) (Muñoz-Manchado et al., 2018; Saunders et al., 2018). Conversely, transcripts identifying spiny projection neurons (SPNs: Gpr88, Ppp1r1b, Arpp19, Pde10a) from both the direct (dSPNs: Drd1, Pdyn, Tac1, Eya1) and the indirect pathway (iSPNs: Gpr6, Adora2a, Penk, Necab1) (Gokce et al., 2016; Märtin et al., 2019) as well as other classes of striatal GABAergic interneurons (INs: Npy, Sst, Htr3a, Calb2, Chodl) and large cholinergic interneurons (CINs: Chat, Slc17a8, Slc10a4, Ntrk1) (Muñoz-Manchado et al., 2018) were all decreased (Figure 2c; Supplementary Table 1). Similar de-enrichment was found for genes used to classify astrocytes (Gfap, Aldh1l1, S100b), microglia (Aif1, Trem2, Tmem119) and oligodendrocytes (Olig2, Cnp, Mog) (Figure 2d). Among the 14,185 protein-coding genes detected in our RNAseq, 3,484 were identified as enriched in DS PV interneurons as compared to the DS inputs (adjusted p value of < 0.05) (Figure 2e; Supplementary Table 1). This number dropped down to 2,750 for those displaying a fold-change > 1.5 (Figure 2f; Supplementary Table 1). We also found a significant enrichment of several non-coding transcripts including 663 long non-coding RNAs (LncRNA), 963 To be Experimentally Confirmed RNAs (TEC) and 172 pseudogenes mainly composed of processed pseudogenes (147 out of 172) (Figure 2g; Supplementary Table 1).

Cross-analyses with previous single-cell RNAseq studies revealed that PV-enriched transcripts found in Pvalb-Th-Pnoc (Saunders et al., 2018) and Pvalb-Pthlh (Muñoz-Manchado et al., 2018) populations match at 55 and 62% with those identified in the present study (Figure 2h; Supplementary Table 1). We also compared the overlap between the genes we found to be enriched in the DS PV interneurons with those enriched in the PV interneurons of the nucleus accumbens (Acb) unveiled using the Pvalb-Sun1-GFP mouse line (Gallegos et al., 2023). Pvalb-Sun1-GFP mice express the nuclear envelope protein Sun1 fused to GFP specifically in PV-expressing neurons, enabling the isolation of nuclei and analysis of the nuclear transcriptome. In contrast, Pvalb-Ribotag mice express the ribosomal protein Rpl22 tagged with HA in PV neurons, allowing for the isolation of ribosome-associated, actively translated transcripts. Despite targeting distinct RNA populations, we identified 1,015 shared transcripts enriched between DS and Acb PV interneurons (Figure 2i; Supplementary Table 1). Among these common genes, 18 were found in the Pvalb-Th-Pnoc population and 30 in the Pvalb-Pthlh population (Figure 2j; Supplementary Table 1). All in all, only 4 transcripts (Pvalb, Pthlh, Nxph1, Nrip3) were found systematically enriched in PV interneurons regardless their location (DS vs. Acb) and the RNAseq approaches used (scRNAseq vs. TRAP) (Figure 2j; Supplementary Table 1). Finally, in situ hybridization from the Allen Brain Atlas dataset for Nxph1, Nrip3, Laptm4b, Hmga1b and Setd7 transcripts presented a sparse distribution typical of DS PV interneurons (Supplementary Figure 1a).

Gene Ontology (GO) enrichment analysis using the ShinyGO v0.8 web tool (Ge et al., 2020) revealed that synapse organization represented one of the most significant GO terms associated to Biological Process and Cellular Component (Supplementary Figure 2). Because synaptic stabilization is tightly associated with extracellular matrix (ECM) structures (Christensen et al., 2021; Wingert and Sorg, 2021), we examined whether ECM-related genes were enriched in DS PV interneurons using the MatrisomeDB database. Among 274 annotated core matrisome genes, we found transcripts encoding ECM glycoproteins (19), collagens (6) and proteoglycans (5) enriched in PV interneurons (Figure 3a; Supplementary Figure 3; Supplementary Table 2). We also identified several matrisome-associated genes among which some encode ECM regulators (22), ECM-affiliated proteins (12) and secreted factors (31) (Figure 3a; Supplementary Figure 3; Supplementary Table 2). Interestingly we noticed that 44 ECM-related genes enriched in DS PV interneurons were also found in Acb PV interneurons (labeled in magenta) (Figure 3a; Supplementary Table 2). Closed inspection of the identity of proteoglycan-enriched genes revealed the presence of core constituents of perineuronal nets (PNNs) including Acan and Vcan transcripts encoding for two chondroitin sulfate proteoglycans (Aggrecan and Versican) and Hapln1 and Hapln4 encoding for hyaluronan and proteoglycan link protein 1 and 4 (Figure 3a; Supplementary Table 2). To examine the presence of PNNs around DS PV interneurons, we performed Wisteria floribunda agglutinin (WFA) staining (Härtig et al., 2022) on striatal slices from Pvalb-Ribotag mice (Figure 3b), Pvalb-ChR2 and Pvalb-Ai14 mice (Supplementary Figure 4) mice in which PV interneurons were identified indirectly with GFP, tdTomato or HA and directly using PV antibody. Immunofluorescence analysis revealed that regardless the mouse line and PV detection method used, PNNs ensheathe the majority of the DS PV interneurons, with WFA staining observed in more than 94% of them (Figure 3b; Supplementary Figure 4).

Cell adhesion molecules are also core constituents of synaptic organization and stabilization. We therefore analyzed the distribution of PV-enriched genes among cadherins, protocadherins, integrins as well as several different categories of cell adhesion molecules (Figure 3c; Supplementary Table 2). We found that genes encoding alpha isoforms of clustered protocadherins (9 out of 13: Pcdhac1, Pcdhac2, Pcdha5, Pcdha6, Pcdha1, Pcdha2, Pcdha3, Pcdha7, Pcdha8), non-clustered protocadherins (5 out of 12: Pcdh7, Pcdh9, Pcdh10, Pcdh19, Pcdh15), contactin-associated proteins (5 out 7: Cntnap1, Cntnap4, Cntnap5c, Cntnap2, Cntnap5a) as well as amigo (2 out of 3: Amigo1, Amigo3) or neurexin (2 out of 3: Nrx2, Nrx3) were particularly enriched in DS PV interneurons (Figure 3c; Supplementary Table 2). In contrast to ECM-related genes, only 17 cell adhesion-related genes were common to DS and Acb PV interneurons (labeled in magenta). These findings allow the identification of a unique ECM and adhesion molecule signature for DS PV interneurons (Figure 3c; Supplementary Table 2).

Cellular Component GO terms associated with synapses functions were also highly represented in our analysis (Supplementary Figure 2). We therefore took advantage of a system classification previously implemented allowing the rapid visualization of genes encoding receptors, transporters, and enzymes involved in the turnover of a given neurotransmitter system including serotonergic, catecholaminergic, GABAergic, cholinergic and glutamatergic (Puighermanal et al., 2020) (Figures 4a–e; Supplementary Table 3). These analyses revealed the enrichment of a restricted set of serotonergic (Htr2b, Htr7) (Figure 4a; Supplementary Table 3) and catecholaminergic (Drd4, Adra1a) receptors (Figure 4b). Several components of the GABAergic system were also found to be enriched including genes encoding GABAa and GABAb receptor subunits (Gabra1, Gabrg3, Gabrd, Gabbr2) as well as transcripts encoding proteins involved in GABA turnover (Gad1, Gad2) and transport (Slc6a1, Slc6a8) (Figure 4c; Supplementary Table 3). Chrnb2 was the only cholinergic gene found to be enriched in DS PV interneurons (Figure 4d). Regarding the glutamatergic system, 10 out of the 18 genes encoding ionotropic glutamate receptors were enriched in PV interneurons (Grik1, Grik3, Grik5, Gria3, Gria4, Grin1, Grin2a, Grin2b, Grin2d, Grid2) while none of the transcripts encoding metabotropic glutamate receptors (Grm) were found to be enriched (Figure 4e; Supplementary Table 3). Finally, we found that DS and Acb PV interneurons only shared 5 neurotransmitter system-related genes, 3 associated to the GABAergic system (Gabra1, Gad1, Slc6a8) and 2 with the glutamatergic one (Gria4, Grin2d) (labeled in magenta) (Figure 4).

We also performed a complete classification of voltage-gated ion channels (Figure 5; Supplementary Table 3), including calcium channels (Figure 5a), sodium channels (Figure 5b), and potassium channels comprising voltage-gated potassium channels (Figure 5c), inwardly rectifying potassium channels (Figure 5d), two-P potassium channels (Figure 5e), calcium-activated potassium channels (Figure 5f), and accessory subunits (Figure 5g). The main pore-forming alpha 1 subunits of voltage-gated calcium channels were particularly enriched compared to the other subunits (7 out of 10 constituents) (Figure 5a). Voltage-gated potassium channels distributed among the different families were also highly enriched in DS PV interneurons, several of them displaying a fold-enrichment > 5 (Kcna2, Kcnc1, Kcnc3, Kcng4, Kcnh2) (Figure 5c). Our analysis also unveiled a biased expression in DS PV interneurons of G-protein activated inward rectifying potassium channels (Kcnj3, Kcnj6, Kcnj9, Kcnj11, Kcnj12) (Figure 5d). In contrast to neurotransmitter system-related genes, the expression of voltage-gated ion channels appears to be more conserved between DS and Acb PV interneurons as suggested by the important shared enriched expression of genes encoding voltage-gated potassium channels (9 out of 13 genes including Kcna1, Kcna2, Kcnc1, Kcnc2, Kcnc3, Kcnb2, Kcng4, Kcns3, Kcnh2) (Figure 5c), G-protein gated potassium channels (3 out of 5 genes: Kcnj3, Kcnj9, Kcnj12) (Figure 5d), Two-P potassium channels (Kcnk3, Kcnk12) and accessory subunits (3 out 3 genes: Kcnab2, Kcnab3, Kcnmb2) (Figure 5g).

Finally, to gain insights into the specification of DS PV interneurons, we classified transcription factors by family using the AnimalTFDB v4.0 database (Shen et al., 2023) (Supplementary Figure 5). We analyzed the distribution of DS PV interneurons enriched genes among 6 core families of transcription (Supplementary Figure 5). The classification was further refined by analyzing the distribution DS PV neurons enriched genes within each transcription factors subfamilies. The filtering of enriched genes for a FC > 4 allowed the description of a putative transcription factors landscape of DS PV interneurons comprising 5 members of the basic domain groups (Tfap2b, Hes3, Bach2, Maf, Mafb) (Supplementary Figure 5a), 1 member of beta scaffold factors (Lin28b) (Supplementary Figure 5b), 8 transcripts from the helix-turn helix group (Onecut2, Satb1, Arid3a, Arx, Barx2, Lhx1, Six4, Pou5f2) (Supplementary Figure 5c), 4 genes from other alpha helix groups (Sox12, Sox6, Tox2, Tox3) (Supplementary Figure 5d), 5 transcripts from the unclassified structure (Hmga1b, Pura, Prug, Zfp318, Zfp750) (Supplementary Figure 5e) and 11 members of the zinc coordinating group (Esrrb, Essrg, Nr6a1, Zfp618, Zim1, Gm14308, Gm14408, Gm14443, Prdm11, Scrt1, Scrt2) (Supplementary Figure 5f). Among all the genes encoding transcription factors enriched in the DS PV neurons, 75 transcripts were also found in the Acb PV neurons dataset (colored transcripts) with percentages of overlap ranging from 40% for the basic domain groups (8 out of 20 genes in common) and 35% for the helix-turn helix group (20 out of 57 common transcripts) to 12.5% for the beta scaffold factors (1 gene out of 8 genes) (Supplementary Figures 5a–c).

Previous works showed that striatal PV interneurons were required for the action of psychostimulant drugs (Wang et al., 2018; Wiltschko et al., 2010). We therefore sought to determine the impact of repeated exposure to d-amphetamine on the DS PV interneurons translatome. Pvalb-Ribotag mice used for RNAseq were administered either saline or d-amphetamine (5 mg/kg, i.p.) for 5 consecutive days during which psychomotor responsiveness was measured in a circular corridor (Supplementary Figure 6). As expected, d-amphetamine-treated mice exhibited a robust increase in both horizontal (locomotion) and vertical (rearings) locomotor activity as compared to saline-treated mice (Supplementary Figures 6a,b, p = 0.003 (locomotion) and p = 0.007 (rearings) at day1, p < 0.0001 (locomotion) and p = 0.0238 (rearings) at day 5). Although repeated d-amphetamine administration triggered a robust conditioned locomotor response to the context, mice failed to develop locomotor sensitization and strongly decreased their rearing behaviors at the expense of the development of stereotyped behaviors (Supplementary Figures 6b–d). To detect stable modifications, tagged ribosomes-bound mRNAs from DS PV interneurons were isolated 3 days after the last administration of d-amphetamine and compared to DS PV interneurons isolated mRNAs from saline-treated Pvalb-Ribotag mice (Supplementary Figure 6e). Comparison between PV neurons saline and d-amphetamine samples identified only one differentially translated transcripts (Gm20683) between these groups (Supplementary Figure 6e). In contrast, when input fractions from saline and d-amphetamine samples were compared, we identified 1 downregulated and 96 upregulated transcripts in d-amphetamine group most of them being microtubule cytoskeleton-related genes (Supplementary Figure 6f).

To evaluate whether reward type and contingency impacted gene expression in DS PV neurons, food-restricted Pvalb-Ribotag mice were divided into two groups based on pellet type: a high-palatable diet or a standard chow (Castell et al., 2024). Within each dietary condition, animals were subsequently allocated to two groups: master mice, which earned food rewards through lever pressing, and yoked mice, which passively received a pellet each time their paired master obtained one. During the initial FR1 schedule (days 1 to 5), behavioral performance began to diverge after the third training session (Figure 6a, upper panel). Compared to master mice given regular chow, master mice trained for high-palatable rewards exhibited a significantly higher number of active lever presses (p < 0.0001) (Figure 6a, upper panel; Supplementary Table 4). Accordingly, both high-palatable masters and their yoked counterparts received a greater number of pellets than the standard food groups (p < 0.0001) (Figure 6a, middle panel and Supplementary Table 4). As expected, lever pressing remained low and stable across days in yoked animals, reflecting the absence of action-outcome contingency (Figure 6a, upper panel; Supplementary Table 4). Upon transition to an FR5 schedule (days 6–9), high-palatable masters showed a progressive increase in lever pressing across sessions. In contrast, standard food masters exhibited a marked increase in responding on the first day of FR5, followed by performance stabilization, as reflected in both lever press frequency and pellet delivery (Figure 6a, upper and middle panels; Supplementary Table 4). From day 10 to 15, animals were returned to ad libitum feeding. Under these conditions, both groups of master mice reduced operant responding, consistent with reduced motivational drive as their physiological needs were now met. High-palatable masters continued to lever press at a higher rate than standard chow masters, although both groups demonstrated a progressive decline in responding over time (Figure 6a, upper panel; Supplementary Table 4). To further highlight the behavioral dissociation between master and yoked groups, we quantified food magazine visits (Figure 6a, lower panel; Supplementary Table 4). During the food-restricted phase, yoked mice exhibited a higher frequency of magazine checks compared to their corresponding master mice consistent with the non-contingent nature of reward delivery in yoked animals. This pattern was abolished under ad libitum feeding, during which all groups displayed comparable low levels of food magazine inspection (Figure 6a, lower panel; Supplementary Table 4).

One day after the last session, immunoprecipitation of tagged ribosomes was performed and isolated DS PV interneurons mRNAs from the 4 groups were analyzed by high-throughput RNAseq. Paired comparisons between master groups (standard vs. high palatable) revealed the highest number of the differentially translated transcripts (Figure 6b; Supplementary Table 2). Indeed, 1,092 genes were found to be enriched in DS PV interneurons in standard food masters whereas 143 genes were more expressed in high palatable food masters (Figure 6b). In contrast, analysis between yoked groups (standard vs. high palatable food) identified 44 genes displaying differential expression DS PV interneurons (Figure 6b). However, these latter changes were unlikely directly linked to food palatability as 37 genes out of 44 were similarly modified between standard and high palatable master mice, suggesting that most of the changes observed in DS PV interneurons were a snapshot of the interaction between food-seeking behavior and the palatability (Figure 6b). Comparison of upregulated genes in master and yoked mice fed a high-palatable diet revealed 32 genes commonly upregulated in both groups, indicating that these genes may be associated with the sucrose content of the diet rather than operant behavioral contingency (Figure 6c). In contrast, only 5 genes were commonly upregulated between master and yoked mice on the standard diet, suggesting a more limited shared transcriptional response under these conditions (Figure 6c). To get insights into the biological functions of the changes observed we conducted GO enrichment analysis. We found that most of the genes upregulated in high palatable food master mice were related to actin cytoskeleton organization (Figure 6d). Although transcripts related to tubulin, actin, cell adhesion and ECM regulation were also increased in standard master mice, GO terms associated to transcription were the most represented in this group (Figures 6d,e) (Gallegos et al., 2023; Manz et al., 2022; Wang et al., 2018). Among the 1,092 upregulated genes, 155 were identified as transcription factors distributed in the 6 core families.