Introduction

The ability to evaluate a sensory stimulus and to correctly act upon it is paramount for survival. Abnormal associations of stimuli and/or abnormal actions towards stimuli are hallmark features of psychiatric disorders including anxiety disorders. The basolateral amygdala (BLA) has long been known to support emotional responses [1,2,3], including to both aversive [4,5,6,7,8] and appetitive stimuli [9,10,11]. Affording the BLA with this capacity are both its intrinsic plasticity [12, 13] and its projections into ‘downstream’ structures which can directly influence decisions and behavioral outcomes. For instance, the BLA innervates the central nucleus of the amygdala, and this input is necessary for the expression of learned avoidance [14]. Photostimulating central amygdala projecting BLA neurons evokes avoidance, while photoinhibition of the same neurons reduces fear learning [15]. Other projections of the BLA can influence appetitive responses, including stimulation of BLA neurons that project to the ventral striatum’s nucleus accumbens (NAc) [15]. Thus, it is clear that regionally-separable downstream recipients of BLA input are sufficient to direct emotional responses.

There is also an interplay between BLA projection targets and the cell types which comprise those projections in how specific BLA outputs influence emotion. The genetic identity of BLA neurons is highly diverse and these different cell types appear to be uniquely engaged following emotional responses [16]. A single genetically distinct neuronal population can drive opposing emotional responses if it projects to two brain regions [17] and likewise, two genetically distinct BLA outputs can drive opposing emotional responses if they each project to the same brain region [18]. BLA excitatory neurons are divided into two main genetic classes, which are becoming increasingly understood to have diverse projection targets and functions. These include the Rspo2 expressing neurons (encoding R-spondin 2) which can drive aversive behaviors, and the Ppp1r1b expressing neurons (encoding protein phosphatase 1 regulatory inhibitor subunit 1B) which appear to support appetitive behaviors [19]. BLA neurons distinguished by the expression of the transcription factor RSPO2 are also labeled by Fezf2 (encoding the transcription factor zinc-finger 2) [17] and project to the NAc and also its neighboring ventral striatum subregion, the tubular striatum (TuS, also known as the olfactory tubercle) [20]. Activation of Fezf2 neurons innervating the NAc drives aversive states, and contrastingly, activation of Fezf2 neurons innervating the TuS increases appetitive states. Together, these findings indicate that neither the downstream target nor the genetic identity alone sufficiently explain how the BLA broadcasts emotional information. Instead, where this information goes and who within the BLA sends it are both critical for regulating emotional states. Given that Rspo2/Fezf2 + BLA neurons support both appetitive and aversive states depending upon their downstream targeting, we sought to answer the question of whether the Ppp1r1b + BLA neuron population also contributes to aversive states, and if they do so depending upon their regional innervation within the NAc and TuS.

Ppp1r1b (also known as DARPP-32, dopamine- and cAMP-regulated neuronal phosphoprotein) is a phosphoprotein regulated by both D1 and D2 dopamine receptors [21,22,23], which are encoded for by the Drd1 and Drd2 genes, respectively [24]. Dopamine within the BLA is necessary for fear learning [25]. We know that Drd1+ neurons in the BLA contribute to memory [26]. While the role of Drd2+ neurons in the BLA is not understood, prior pharmacological work has indicated a role for the D2 receptor in emotional responses [27,28,29]. Overall, the respective contributions of Drd1+ and Drd2+ BLA neurons in regulating emotional states are unknown. Moreover, it is unknown if, like the Fezf2 neurons, the influence of Drd1+ and Drd2+ BLA neurons depends upon their projection targets. Here, using a combination of viral tracing, ex vivo brain slice recordings, chemo- and opto-genetics, and behavior, we identified that the Drd1+ and Drd2+ BLA neuron populations form two parallel pathways wherein each innervate both the NAc and the TuS for the modulation of negative emotional states depending upon which ventral striatum subregion they innervate. Overall, these results contribute to a model whereby parallel, genetically-distinct, BLA to ventral striatum circuits inform emotional states in a projection-specific manner and altogether expand our appreciation for how the BLA regulates emotions.

Materials and methods

Ethics approval

All experimental procedures were conducted at the University of Florida or the University of Pennsylvania in accordance with the relevant guidelines from the Guide for the Care and Use of Laboratory Animals. Vertebrate animal procedures were approved and followed the guidelines and regulations set by each institution’s respective Institutional Animal Care and Use Committees (protocol 202300000126 at Univ of Florida and protocol 803920 at Univ of Pennsylvania).

Animals

Adult male and female mice, 2–5 months of age, were housed in a temperature-controlled vivarium on a 12:12 h (hr) light/dark cycle with ad libitum access to food and water, except during behavioral testing. All behavioral testing occurred during the light cycle. Mice that only underwent viral injections were group housed (≤5 mice/cage) and mice with chronic implants were single housed following surgery.

Mouse lines included the following transgenic lines which were maintained on a C57BL/6J background (strain #000664; RRID:IMSR_JAX:000664, The Jackson Laboratory) and were bred in house within a University of Florida vivarium. Drd1-Cre (B6.FVB(Cg)-Tg(Drd1-cre)EY262Gsat/Mmucd, RRID:MMRRC_030989-UCD), Drd2-Cre (B6.FVB(Cg)-Tg(Drd2-cre)ER44Gsat/Mmucd, RRID:MMRRC_032108-UCD), and A2a-Cre (B6.FVB(Cg)-Tg(Adora2a-cre)KG139Gsat/Mmucd, RRID:MMRRC_036158-UCD) mice were obtained from the UC Davis Mutant Mouse Regional Resource Center (MMRRC) [30]. Drd1-tdTomato;Drd2-GFP double transgenic mice originate from that described previously [31]. Ai9 tdTomato Cre reporter mice (B6.Cg-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze/J; RRID:IMSR_JAX:007909, [32]) were obtained from the Jackson Laboratory.

The sample size of animals in each experiment was selected based upon our prior experiences which yielded statistically robust outcomes with similar sample sizes (viz., no a priori statistical power was calculated). Data from animals with implants not appropriately targeted to their region intended (e.g., BLA, NAc, TuS), or with viral expression extending outside of those regions, were eliminated from analyses and are not reported herein. For all experiments, every attempt was made to pseudorandomize the experimental groups so that as many groups as possible would be included in any given replication. While for the behavioral experiments the experimenter was not blind due to the nature of the experiments (e.g., required pre-treatment with drugs, etc), post-processing and handling of data were both performed using semi-automated methods (see below).

Viral vectors

rgAAV.hSyn.HI.eGFP-Cre.WPRE.SV40 (Addgene #105540-AAVrg, 7 × 1012 vg/ml), AAV.Ef1α.DIO.Synaptophysin.mRuby and AAV.Ef1α.FLEX.Synaptophysin.GFP (both generous gifts from Dr Marc Fuccillo, University of Pennsylvania) [33], and AAV.hSyn.FLEx.mGFP-2A-Synaptophysin.mRuby (Addgene #71760-AAV1, 9.8 × 1012 vg/mL) were used for tracing. AAV.Ef1a.DIO.hChR2(E123T/T159C)-EYFP (Addgene #35509-AAV5, 1 × 1012 vg/ml vg/ml) was used for patch-clamp recording and for optogenetic stimulation during the optogenetic real-time place preference/avoidance task. AAV.Ef1a.DIO.EYFP (Addgene #27056-AAV5, 1 × 1012 vg/ml) was used as a control virus for the optogenetic real-time place preference/avoidance task. rgAAV.hSyn.DIO.hM4D(Gi)-mCherry (Addgene #44362-AAVrg, 1.2 × 1013 vg/ml) and rgAAV.hSyn.DIO.mCherry (50459-AAVrg, 1.8 × 1013 vg/ml) were used for chemogenetic inhibition. Both the rgAAV.hSyn.DIO.mCherry and rgAAV.hSyn.DIO.GFP (Addgene #50457-AAVrg, 7 × 1012 vg/ml) were used for retrograde tracing of BLA neurons to assess possible collateralization of inputs to ventral striatum.

Surgical procedures

For all surgical procedures, mice were anesthetized with 2–4% isoflurane (IsoFlo, Patterson Veterinary, Greeley, CO) in 1 L/min (min) O2, and head fixed in a stereotaxic apparatus while their body temperature was maintained using a 38 °C water bath heating pad. The scalp was shaved and cleaned with betadine and 70% ethanol. Following subcutaneous (s.c.) administration of Meloxicam (20 mg/kg) analgesia and local administration of the anesthetic lidocaine (lidocaine, 3 mg/kg, s.c., Patterson Veterinary) to the scalp, a small midline cranial incision was made.

For viral injections, craniotomies were made above the target regions. A pulled glass micropipette containing the AAV was slowly inserted for injection. For TuS injections, 50 nl of viral solution was injected bilaterally at the following coordinates: anteroposterior (AP) + 1.4 mm bregma, mediolateral (ML) ± 1.2 mm lateral midline, dorsoventral (DV) −4.85 mm from the brain surface for DREADD based behaviors, or 50 nl of viral solution was delivered unilaterally for tracing experiments. For NAc injections, 100 nl of viral solution was delivered bilaterally (AP 1.5 mm, ML ± 1.0 mm, DV −3.75 mm) for behavioral experiments, or 50 nl of viral solution was delivered unilaterally for tracing experiments. For BLA injections, 100 nl of viral solution was delivered either unilaterally into the right hemisphere (AP −1.6 mm, ML + 3.25 mm, DV −4.25 mm) for Opto-RTPP/A and brain slice electrophysiology experiments, or bilaterally (AP −1.6 mm, ML ± 3.25 mm, DV −4.25 mm) for tracing experiments. All injections were performed at a rate of 2 nl/s (s), with 20–40 s intervals using a Nanoject III (Drummond Scientific). Following injection, at least 5 min went by before slowly withdrawing the pipette from the brain. Craniotomies were then sealed with dental wax and the incision was closed with wound clips.

For cannula implantation, the skull was scrubbed with 3% H2O2 and covered with a thin layer of cyanoacrylate glue (Vetbond, 3M). Bilateral craniotomies were drilled over the BLA and 26-gauge(G) guide cannulae (#C315GMN/SPC, P1 Technologies) extending 3.5 mm below pedestal were implanted at the coordinates AP −1.3 mm, ML ± 3.2 mm. Cannulae were then lowered into the brain and secured to the skull with a small amount of Vetbond followed by dental cement, and dust caps with a 3.5 mm projection wire (C315DCMN/SPC, P1 Technologies) were inserted.

For optical fiber implantation, following skull preparation for implantation as above, a craniotomy was made and drilled above the ventral striatum on the right hemisphere. Fibers (300 µM core diameter, 0.39NA, cut to 6.0 mm length) for optogenetic stimulation were lowered into the NAc (AP 1.4 mm, ML 1.0 mm, DV −3.85 mm) or the TuS (AP 1.5 mm, ML 1.2 mm, DV −4.9 mm). The fiber was secured with Vetbond followed by dental cement as described for the cannulae implantation.

Following surgery, mice were allowed to recover on a heating pad until ambulatory and were given immediate ad libitum access to food and water. Meloxicam analgesic (20 mg/kg, s.c.) was administered for at least 3 days following surgery. Mice with indwelling cranial implants were single housed and given 7–14 days after surgery to recover before being acclimated to behavioral procedures.

Histology

Immunohistochemistry

Mice were anesthetized with Fatalplus (150 mg/kg; Vortech Pharmaceutical Ltd, Dearborn, MI) and transcardially perfused with cold 0.9% NaCl (Physiological Saline), followed by cold 10% phosphate buffered formalin (#SF100-4, Thermo Fisher Scientific) for fixation. Brains were collected and further fixed and cryoprotected in a 30% sucrose/10% formalin solution for 72 hr at 4 °C. Serial 40 μm thick coronal sections were collected using a sliding microtome (Leica) and stored at 4 °C in a solution of Tris-buffered saline (TBS; 0.242% Tris base, 2.924% sodium chloride, pH = 7.4 ± 0.2) with 0.03% sodium azide.

Sections from Drd1- or Drd2-Cre mice injected with Cre-dependent retrograde mCherry AAV underwent antigen retrieval in citrate buffer (pH 6.0) for 30 mins at 80 °C. Sections from Drd1-Cre;Ai9, Drd2-Cre;Ai9, or A2a-Cre;Ai9 mice were selected to investigate recombination efficiency but did not undergo antigen retrieval. After being rinsed with TBS and diluting buffer (2% bovine serum albumin (Sigma Aldrich), 0.9% sodium chloride (Sigma Aldrich), 0.4% Triton-X 100 (Sigma Aldrich), and 1% normal goat serum (Sigma Aldrich) in TBS), samples were blocked in 20% normal donkey serum solution, then incubated in the primary antibody overnight at 4 °C. Sections were then incubated in the secondary antibody at room temperature and washed with TBS prior to slide-mounting with either DAPI Fluoromount-G® mounting medium (SouthernBiotech, catalog #0100–20). Primary antibodies included rabbit anti-DsRed (Takara Bio, catalog #632496, 1:1000), chicken anti-NeuN/FOX3 (EnCor, catalog #CPCA-FOX3, 1:1000), and mouse anti-Cre Recombinase (Millipore-Sigma, cat#MAB3120, 1:1000). Secondary antibodies included goat anti-chicken Alexa Fluor 488 (Invitrogen, cat#A11039), goat anti-rabbit Alexa Fluor 680 (Invitrogen, cat#A21109), and donkey anti-mouse Alexa-Fluor 488 (Invitrogen, cat#A21202), all at a 1:1000 dilution.

Imaging and quantification

Brain regions were identified using the mouse brain atlas [34]. Images were acquired using a Nikon Eclipse Ti2e fluorescent microscope. For quantification of the number of Drd1+ and Drd2+ NAc and TuS projecting BLA neurons, three mice of each genotype (Drd1-Cre or Drd2-Cre) received an injection in either the NAc or the TuS. From these mice, at least three BLA sections were acquired and imaged from −1.10 mm to −2.10 mm posterior to Bregma. For quantification of the number of Cre+ neurons in the BLA of Drd1-Cre;Ai9, Drd2-Cre;Ai9, or A2a-Cre;Ai9 mice, at least five BLA sections from three mice of each genotype were acquired spanning the same range. Images were acquired at 20x magnification across both hemispheres and Z-stacked every 4 µm. For quantification, regions of interest (ROIs) were drawn around the areas of interest (LA, BA). Images were preprocessed to remove background and to enhance local contrast, a rolling ball algorithm was applied to remove background, and images underwent Gaussian smoothing and Laplace sharpening. A semi-automated thresholding and counting algorithm created within NIS elements (Nikon) software was used to identify cells within selected ROIs, allowing for unbiased estimation of cell numbers. Cells were identified based on fluorescence intensity (via threshold) and diameter.

For quantification of Drd1+ and A2a + BLA to ventral striatum synaptophysin puncta within the ventral striatum, at least three sections from three mice of each genotype were acquired spanning from 1.7–0.6 mm anterior to Bregma. Images were acquired at 20x magnification for the hemisphere ipsilateral to the injection site, and Z-stacked every 0.9 µm. For quantification, ROIs were drawn around the areas of interest (TuS, NAcC, NAcSh, PCx). Images were preprocessed to remove the average background. A semi-automated thresholding and counting algorithm created within NIS elements software was used to identify fluorescent puncta within selected ROIs, allowing for unbiased estimation of the number of fluorescent puncta. Puncta were identified based on fluorescence intensity (via threshold) and diameter.

Brain slice electrophysiology

Whole-cell patch-clamp recordings were performed in ex vivo brain slices from Drd1-Cre;Ai9 or A2a-Cre:Ai9 mice, in which tdTomato expression was directed within cells expressing either DRD1 or DRD2, respectively. A Cre dependent AAV encoding for ChR2 (AAV5-Ef1a-DIO hChR2(E123T/T159C)-EYFP) was injected bilaterally into the BLA of Drd1-Cre:Ai9 or A2a-Cre:Ai9 mice, 2–3months of age. After waiting a minimum of one month to allow for ample AAV expression, acute brain slices were prepared as follows.

Mice were deeply anesthetized with intraperitoneal injection of ketamine-xylazine (200–15 mg/kg body weight) and decapitated. The cranium was dissected and the brain was immediately removed and placed in ice-cold HEPES buffered cutting solution containing (in mM): 92 N-methyl-d-glucamine, 2.5 KCl, 1.2 NaH2PO4, 30 NaHCO3, 20 HEPES, 25 glucose, 5 sodium l-ascorbate, 2 thiourea, 3 sodium pyruvate, 10 MgSO4 and 0.5 CaCl2 (osmolality ~300 mOsm and pH ~7.4, bubbled with 95% O2 and 5% CO2). Coronal brain slice (180–200 µM) containing the OT were cut using a Leica VT 1200S vibratome. Brain slices were incubated in oxygenated artificial cerebrospinal fluid (ACSF) containing (in mM): 126 NaCl, 2.5 KCl, 2.4 CaCl2, 1.2 MgSO4, 1.4 NaH2PO4, 11 glucose, 25 NaHCO3 and 0.6 sodium L-ascorbate (osmolality ~300 mOsm and pH ~7.4, bubbled with 95% O2 and 5% CO2) for 1 hr at 31 °C and at room temperature thereafter. Slices were transferred to the recording chamber for whole-cell patch-clamp recordings and continuously perfused with oxygenated ACSF. 4-Aminopyridine (4-AP; 200 µM) was added to enhance optically evoked synaptic release in ChR2+ axonal terminals. Fluorescent DRD1-/A2A-tdTomato+ cells in the TuS were visualized with a 40X water-immersion objective under an Olympus BX51WI upright microscope equipped with epifluorescence. Electrophysiological recordings were controlled by an EPC-10 amplifier combined with Pulse Software (HEKA Electronic) and analyzed using pulse and Clampfit (Axon instruments). Whole-cell patch-clamp recordings were made in both current and voltage-clamp mode. Patch pipettes were pulled from thin-wall borosilicate glass-capillary tubing (WPI, Sarasota, FL, USA) on a Flaming/Brown puller (P-97; Sutter Instruments Co., Novato, CA, USA). The tip resistance of the electrode was 5–8 MΩ. The pipette solution contained the following (in mM): 120 K-gluconate, 10 NaCl, 1 CaCl2, 10 EGTA, 10 HEPES, 5 Mg-ATP, 0.5 Na-GTP, and 10 phosphocreatine.

To activate ChR2 in the TuS slices, blue light (pE-300ultra, CoolLED, ~25 mW) was delivered through the same 40X objective. Pharmacological reagents including tetrodotoxin (TTX) citrate (Abcam), 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), d,l-2-amino-5-phosphonopentanoic acid (AP5), and 4-Aminopyridine (4-AP) (Sigma-Aldrich) were bath perfused during recording.

in vivo DREADD-based chemogenetic inhibition

For DREADD-based chemogenetic inhibition of Gi coupled inhibitory DREADD receptor (hM4Di) expressing neurons, Drd1+ and Drd2+ mice were injected with rgAAV.hSyn.DIO.hM4D(Gi)-mCherry (1.2 × 1013 vg/ml, 100 nl/hemisphere in NAc, 50 nl/hemisphere in TuS, catalog #44362-AAVrg, Addgene) or rgAAV.hSyn.DIO.mCherry (1.8 × 1013 vg/ml, 100 nl/hemisphere in NAc, 50 nl/hemisphere in TuS, catalog #50459-AAVrg, Addgene) as control. All mice were implanted 1–2 weeks later with bilateral intracranial guide cannulae (Protech International, Inc, catalog #8IC315GMNSPC, 26ga) extending 3.5 mm beyond the pedestal, for direct administration of either the DREADD ligand J60 [35] or vehicle into the BLA. Dust caps without a projection wire (Protech International, Inc, catalog #8IC315DCMNSP) were inserted immediately following surgery, and mice were given 1–2 weeks to recover.

Prior to behavior, mice underwent 2 days of handling in which the dummy cannulae were removed and replaced. On the habituation behavior day, mice received a “mock” infusion, wherein the internal cannulae (Protech International, Inc, catalog #8IC315MNSPC, 5.75 mm projection, 33ga) connected to tubing from a 1 µL Hamilton Syringe (Hamilton, catalog #86211) were inserted into the guide cannulae, and the Harvard Apparatus 22 Syringe Pump (catalog #PY2 55–2222) was turned on for 2 min to simulate the noise of the infusion. This mock infusion occurred 30 min prior to being placed in the plethysmograph for the Pavlovian fear learning behavioral paradigm. On the learning day (Day 2) of the Pavlovian fear learning paradigm, mice were once again tethered to the Hamilton syringe, but this time received an infusion of 100 nL of either 10 nM J60 or vehicle at a rate of 50 nl/min, 30 min prior to the start of the behavioral task.

Behavioral tasks

Odor-shock pavlovian fear learning

We used a whole-body plethysmography chamber (Data Sciences International, St. Paul, MN) that was adapted for the infusion of a neutral odor and the administration of a mild foot shock for an odor-shock Pavlovian fear learning test, as originally developed for use in rats [36]. We constructed an air-dilution olfactometer [37, 38] and used custom code in Synapse (Tucker Davis Technologies) to control the delivery of an otherwise neutral odor, isopentyl acetate (1 torr in liquid state; Sigma Aldrich), at a flow rate of 1 L/min (20 s) which co-terminated with the presentation of a mild foot shock (0.5 mA for 1 s). Respiratory transients were detected using a Data Sciences pressure transducer, gain amplified 100 × (Cygnus Technology Inc), and digitized (0.1–20 Hz) at 300 Hz in Synapse. Positive pressure of clean room air was continuously applied to the chamber using a stable-output air pump (Tetra Whisper). Following each stimulus trial, odor-vaporized air was exhausted from the plethysmograph through an outlet at the chamber’s ceiling.

Mice were acclimated to handling in the behavioral room for two days prior to entering the plethysmograph. Mice were then acclimated to the plethysmograph by undergoing a session in which no odors or shock were delivered, but the associated sounds were present (Supplementary Fig. 5). Twenty-four hr later on the acquisition day, mice were allowed to acclimate to the plethysmograph for a 4 min period and were then presented with 10 trials of 20 s odor delivery co-terminating with an odor-paired 1 s foot shock (0.5 mA) with an inter-trial interval (ITI) of 180 s. For the unpaired fear conditioning task, the foot shock was presented pseudorandomly in the ITI (90 s after the foot shock). For the odor only control mice, the 10 trials consisted of only 20 s odor delivery without the administration of the foot shock. The shocked mice received no odor delivery during the trials but received a foot shock either at the end of the trial (trial shock group) or pseudorandomly in the ITI (ITI-shock group). Mice were then returned to their home cage. Twenty-four hr later on the retrieval day, the odor was presented for 10 trials without the foot shock for all groups receiving odor (paired, unpaired, and odor only groups). Mice who did not previously receive the odor underwent the 10 trials without odor delivery or foot shock. Mobility behavior was recorded throughout the entire fear conditioning task using two digital cameras (Microsoft, 10 Hz frame rate) and was scored in 0.4 s bins during the 19 s of odor presentation prior to shock using ezTrack [39] to identify periods of physical immobility. Respiration digitized from the pressure transducer was imported into MATLAB, and a MATLAB script was used to calculate fast-fourier transform (FFT) power spectra of the respiratory signal during odor (excluding the 1 s when the shock co-occurred) as compared to pre-odor (see Supplementary Fig. 5).

Optogenetic real-time place preference or aversion test (Opto-RTPP/A)

Mice were gently handled and acclimated to the behavior room the day prior to the opto-RTPP/A test. Prior to starting the opto-RTPP/A test, mice were gently scruffed, the dust cap was removed, and the mice were tethered to a 400 µm, 0.57NA fiber (Thorlabs, catalog #M58L01) and placed in a 15.24 × 40.64 × 27.94 cm (length × width × height) apparatus divided into three chambers. This fiber was connected to an LED (Doric, 465 nm) through a rotary joint connected to a 400 µm, 0.39NA patch cable. Mice were placed in the center of a three-chamber apparatus and allowed to explore for 30 min. An infrared video camera was placed above the chamber to record activity of the mouse in each chamber (12 Hz frame rate). When mice entered into one of the three chambers, and subsequently broke the infrared beam path, light stimulation (465 nm, 15 ms pulse width, 40 Hz) was initiated and continuously delivered until mice left the chamber and ceased breaking the infrared beams (controlled by an Arduino). At the end of the 30 min, mice were gently restrained and the tether was removed. Mice were then returned to their home cage. The mice were euthanized and perfused the same day, and brains were collected for histological verification of virus injection and optic fiber placement. Analyses were performed in ezTrack [39] to quantify the time spent in each chamber and to generate maps of physical space for illustration purposes.

Data analysis

Data were analyzed for statistical significance in GraphPad Prism. All data are reported as mean ± SEM unless otherwise noted. Specific tests used can be found in the Results sections or the figure legends. All tests met the assumptions of the test (e.g., were normally distributed, etc.) and variances between groups comparable. All t-tests were paired. When possible, experimenters handling the data were blinded to treatment conditions and all post-processing and analyses of data handled through semi-automated methods when possible.

Results

Drd1+ and Drd2+ BLA neurons innervate the ventral striatum

While we know that the BLA possesses Drd2+ and Drd1+ neurons, with more of the latter [40], we sought to determine if BLA Drd1+ and Drd2+ neurons form a circuit with ventral striatum neurons. We injected a Cre-dependent retrograde (rg) AAV expressing mCherry, rgAAV.hSyn.DIO.mCherry, into either the NAc or the TuS of Drd1-Cre and Drd2-Cre mice [41] (Fig. 1A, B & E, F) and later inspected the BLA for mCherry+ neurons. mCherry+ cells in both groups of mice were found in the BLA in both Drd1- and Drd2-Cre mice (Fig. 1C & G), indicating that these neurons indeed project to the ventral striatum. mCherry+ cells were found throughout the entire anterior-posterior extent of the BLA (Fig. 1C & G). In contrast to the lateral amygdala (LA), which was largely void of mCherry+ cells, the basal nucleus of the amygdala (BA) displayed dense mCherry+ cells (Fig. 1C & G). This organization was observed even when similarly injecting either the NAc or TuS of Ai9.tdTomato Cre-reporter mice [32] with rgAAV.hSyn.HI.eGFP-Cre.WPRE.SV40, suggesting that the BA is a major conduit of ventral striatum input regardless of cell type (Supplementary Fig 1A–C & D–F). No reciprocal connection from ventral striatum to the BLA was found, however (Supplementary Fig. 1G–I).

Fig. 1: Ventral striatum projecting BLA neurons arise from the BA and are comprised of Drd1+ and Drd2+ neurons.
figure 1

A Schematic of approach for identifying BLA Drd1+ and Drd2+ inputs to the NAc. B Example of a NAc injection in Drd1-Cre (top) and Drd2-Cre (bottom) mice (ac = anterior commissure, NAcC & NAcSh = nucleus accumbens core & shell, respectively), and C example images of NAc projecting Drd1+ (top) or Drd2+ (bottom) neurons along the anterior-posterior axis of the BLA. Scale bars = 500 µm. Di Anti-NeuN and anti-DsRed immunofluorescence images to identify the size of the NAc projecting BLA neural population. Scale bars = 40 µm. Dii Quantification of the Drd1 (**p < 0.01, n = 3, 2/1/sex) or Drd2 (*p < 0.05, n = 3, 2/1/sex) expressing BLA cells along the entire AP axis that project to the NAc (Two-way ANOVA, ROI main effect F(1,8) = 26.65, p = 0.001). E Schematic of approach for identifying BLA Drd1+ and Drd2+ inputs to the TuS (TuS = tubular striatum). F Example of a TuS injection in Drd1-Cre (top) and Drd2-Cre (bottom) mice, and G example images of TuS projecting Drd1+ (top) or Drd2+ (bottom) neurons along the anterior-posterior axis of the BLA. Scale bars = 500 µm. Hi Anti-NeuN and anti-DsRed immunofluorescence images to identify the size of the NAc projecting BLA neural population. Scale bars = 40 µm. Hii Quantification of the Drd1 (****p < 0.0001, n = 3, 2/1/sex) or Drd2 (****p < 0.0001, n = 3, 2/1/sex) expressing TuS projecting BLA cells along the entire AP axis (Two-way ANOVA, ROI main effect F(1,8) = 295.3, p < 0.001). Mean ± SEM.

Full size image

To quantify the spatial distribution and to identify the size of the population of Drd1+ and Drd2+ BLA neurons innervating both the NAc and TuS, sections were immunolabeled for both NeuN, to measure the total number of neurons, and the red fluorescent protein DsRed, to amplify the mCherry fluorescent signal (Fig. 1Di & Hi). This revealed that indeed the vast majority of ventral striatum projecting Drd1+ and Drd2+ neurons arise from the BA (Fig 1Dii, NAc: Drd1+ MLSD = |BA-LA| = 18.07, p = 0.002, Drd2+ MLSD = |BA-LA| = 11.95, p = 0.019; Fig. 1Hii, TuS: Drd1+ MLSD = |BA-LA| = 31.0, p < 0.001, Drd2+ MLSD = |BA-LA| = 9.24, p < 0.001; MLSD = mean least square difference). Importantly, we confirmed in Drd1-RFP;Drd2-GFP double transgenic mice [31] that there is minimal co-expression of DRD1 and DRD2 in BLA and ventral striatum cells (Supplementary Fig. 2A & B).

Where throughout the ventral striatum do BLA neurons innervate? To answer this, we injected Drd1-Cre and Adora2a (A2a)-Cre mice into the BLA with an AAV encoding a synaptophysin.mRuby fusion protein (AAV.hSyn.FLEx.mGFP-2A-Synaptophysin.mRuby) [33] (Fig. 2A). A2a-Cre mice were chosen for all anterograde AAV-based experiments to attempt to achieve optimal presynaptic expression in Drd2+ neurons, since both 1) A2A is found downstream of the D2 dopamine receptor in Drd2+ neurons and 2) A2A is enriched in the presynaptic terminals of striatal projection neurons [42]. This resulted in GFP+ neurons in the BLA to confirm the injection-site (Fig. 2B) and mRuby+ puncta, indicative of BLA Drd1+ or A2a+ neuronal synapses, throughout both the NAc and TuS (Fig. 2C). mRuby+ puncta were highly visible in Drd1+ neurons in comparison to A2a+ neurons (Fig. 2C & D). The mRuby+ puncta spanned the entire medial-lateral and anterior-posterior extents of the TuS and were especially prominent in layer 2 (Fig. 2C), which is the densest cell layer. mRuby+ puncta were also observed throughout the medial-lateral and anterior-posterior extents of the NAc, with comparable amounts in both the NAc core and shell (Fig. 2D, Drd1+ MLSD = |NAcC-NAcSh| = 0.035, p > 0.999, A2a + MLSD = |NAcC-NAcSh| = 0.241, p > 0.999). It is notable, given its roles in associative learning including fear learning [43, 44], that the ventral striatum receives more BLA Drd1+ and less A2a+ neuronal input than the neighboring piriform cortex (Fig. 2D, Drd1+ : F(1,16) = 30.8, p < 0.001; A2a + : F(1,14) = 10.6, p = 0.006).

Fig. 2: Drd1+ and Drd2+ BLA neurons innervate the entire span of the ventral striatum.
figure 2

A Schematic of approach for identifying BLA Drd1+ and Drd2+ synaptic innervation of the ventral striatum. B Maps of BLA injection sites for Drd1-Cre (left) and A2a-Cre mice (right), where each color represents individual mice. C Representative images showing direct innervation of BLA neurons into the ventral striatum in both Drd1- and A2a-Cre mice. Inset locations denoted by square white box, inset sale bar = 10 µm. D Quantification of synaptophysin puncta in Drd1-Cre (n = 3, 2/1/sex) and A2a-Cre (n = 3, 2/1/sex) mice. Mean ± SEM. E Approach for defining whether, and if so, BLA neurons send contralateral projections to the ventral striatum using an AAV encoding synaptophysin. GFP in the BLA in one hemisphere and synaptophysin.mRuby in the BLA in the opposite hemisphere. F Representative images from one mouse of ipsilateral and contralateral projections from the BLA to the ventral striatum following injection as in E. These example images show that while some BLA neurons transverse into the contralateral ventral striatum, qualitatively, the bulk of input is ipsilateral. G Diagram of approach for identifying BLA Drd1+ and Drd2+ projecting neurons to the NAc and TuS in the ventral striatum. H Percentage of the Drd1+ and Drd2+ neurons within the BA that project to the NAc (red), TuS (green), or both (yellow) (n = 3 mice/genotype, 2/1/sex, 5–6 sections/mouse). I Representative image of a retrograde injection in a Drd1-Cre mouse and J the anterior to posterior span of VS projecting BLA neurons. Scale bars = 500 µm. PCx piriform cortex, NAcC and NAcSh nucleus accumbens core and shell, respectively, BLA basolateral amygdala, BA basal amygdala, LA lateral amygdala.

Full size image

Next, we asked if the same BLA neurons which innervate the TuS are also those which innervate the NAc and vice versa. We injected rgAAV.hSyn.DIO.GFP into the TuS and rgAAV.hSyn.DIO.mCherry into the NAc of either Drd1-Cre or A2a-Cre mice and then inspected the BLA for GFP and mCherry co-expressing cells (Fig. 2G). As shown in Fig. 2H approximately 50% of the ventral striatum projecting Drd1+ BLA neurons project to the TuS, ~29% project to the NAc, and ~21% project to both (Fig. 2H–J). We found a more even split between the A2a+ ventral striatum projecting BLA neurons, with ~44% projecting to the TuS, 42% projecting to the NAc, and ~14% projecting to both. Thus, a subset of BLA projection neurons innervate both ventral striatum targets (Fig. 2H, right). Taken together, these tracing results establish that both Drd1+ and Drd2+ neurons, largely from the BA, innervate the ventral striatum.

Drd1+ and Drd2+ BLA neurons excite ventral striatum spiny projection neurons

Next, we injected a Cre-dependent AAV expressing channelrhodopsin and EYFP (AAV.Ef1a.DIO.hChR2(E123T/T159C)-EYFP) or EYFP alone as a control (AAV.Ef1a.DIO.EYFP) into the BLA of Drd1-Cre and A2a-Cre mice, which were crossed with the Ai9 tdTomato Cre reporter line. Later we took coronal slices of the ventral striatum for ex vivo recordings to determine which ventral striatum spiny projection neurons (SPNs) the BLA neurons make synapses upon. tdTomato+ neurons were identified and used to identify Drd1+ or Drd2+ SPNs (those expressing tdTomato+ in either Drd1-Cre or A2a-Cre mice respectively; Fig. 3Ai & Aii). In the same slices we also patched onto tdTomato- SPNs to monitor activity of Drd1Ø and Drd2Ø (putative Drd2+ or Drd1+) SPNs, in either Drd1-Cre or A2a-Cre mice respectively. During recordings, blue light pulses were delivered to excite ChR2-expressing BLA terminals in the ventral striatum. As stated above, there is minimal co-expression of Drd1 and Drd2 in the same cells in both the BLA and ventral striatum (Supplementary Fig. 2A & B). Moreover, the BLA to ventral striatum projection is predominantly ipsilateral (Fig. 2E, F). Current injection into ventral striatum neurons confirmed their firing patterns are as expected for TuS SPNs (Supplementary Fig. 3A & B) [45].

Fig. 3: Synaptic properties of Drd1 and Drd2 expressing ventral striatum neurons receiving BLA neuronal projections.
figure 3

Ai Schematic indicating Cre-dependent expression of ChR2 in Drd1+ BLA neurons of Drd1-Cre;Ai9 mice. In these mice, tdTomato + neurons are presumably Drd1+ and tdTomato- neurons are presumably Drd1Ø. During whole-cell patch clamp recordings, ChR2 expressing BLA terminals were activated by 470 nm light. Aii Schematic indicating Cre-dependent expression of ChR2 in Drd2+ BLA neurons of A2a-Cre;Ai9 mice. tdTomato + neurons are presumably Drd2+, and tdTomato- neurons are presumably Drd2Ø. B Example light-evoked monosynaptic EPSCs (top) and light-evoked polysynaptic EPSCs (bottom) from Drd1+ TuS neurons under voltage clamp mode. C Neurons organized by response type upon stimulation of Drd1+ TuS projecting BLA terminals. D Example evoked EPSCs from Drd1Ø TuS neurons. E Neurons organized by response type upon stimulation of Drd1Ø TuS projecting BLA terminals. F Example evoked EPSCs from Drd2+ TuS neurons. G Neurons organized by response type upon stimulation of Drd2+ Tus projecting BLA terminals. H Example evoked EPSCs from Drd2Ø TuS neurons. I Neurons organized by response type upon stimulation of Drd2+ Tus projecting BLA terminals. The holding potential was −70 mV. Drd1-Cre;Ai9 n = 2M/1F, A2a-Cre;Ai9 n = 4F/4M.

Full size image

We found that both BLA cell populations synapse upon both Drd1+ and Drd2+ SPNs, albeit with differing proportions and strengths of responses. The majority of Drd1+ BLA neurons elicited large monosynaptic currents in Drd1+ (mean = 74.46 pAmp, SEM = 7.55 pAmp, n = 23) and Drd1Ø SPNs (mean = 186.1 pAmp, SEM = 100.2, n = 3, Fig. 3B–E). While both ventral striatum cell types were excited by Drd1+ BLA neurons, BLA Drd1+ neurons send stronger input (viz., larger evoked excitatory postsynaptic currents [EPSCs]) and do so more predominantly upon Drd1+ vs Drd1Ø (putative Drd2+) SPNs (Drd1+ vs Drd1Ø monosynaptic: X2(1, N = 91) = 5.45, p = 0.02; Drd1+ vs Drd1Ø polysynaptic: X2(1, N = 91) = 0.096, p = .757; Fig. 3C & E). In a subset of SPNs, monosynaptic glutamatergic connections were verified via pharmacological manipulations (Supplementary Fig 3C & D). Likewise, Drd2+ BLA neurons also synapse upon Drd2+ and Drd2Ø SPNs, but compared to the Drd1+ BLA input, input from Drd2+ BLA neurons was both weaker (mean monosynaptic current in Drd2+ SPN = 10.79 pAmp, SEM = 2.78 pAmp, n = 3; mean monosynaptic current in Drd2Ø SPN = 36.74 pAmp, SEM = 1.98 pAmp, n = 1) and not as predominant (Fig. 3F–I). A far larger percentage of SPNs displayed monosynaptic EPSCs upon Drd1+ BLA neuron terminal stimulation than when stimulating Drd2+ BLA terminals (X2(1, N = 255) = 33.947, p < 0.001). Including polysynaptic EPSCs, 13.1–15.9% of SPNs (Drd2+ and Drd2Ø, respectively) displayed evoked potentials, and these were notably modest in amplitude compared to what was observed when stimulating Drd1+ BLA terminals (e.g., Fig. 3B vs F). Together these results extend the anatomical circuitry (Fig. 1 & 2) by showing that both Drd1+ and Drd2+ BLA neurons excite ventral striatum spiny projection neurons.

Both Drd1+ BLA neurons innervating the NAc and Drd2+ BLA neurons innervating the TuS promote avoidance behavior

Next, we sought to determine a functional role for BLA Drd1+ and Drd2+ input to the ventral striatum. In the first assay, we used an optogenetic approach to excite Drd1+ and Drd2+ BLA neuron terminals innervating either the NAc or TuS to determine if these pathways influence avoidance or approach behaviors. For this we unilaterally injected Drd1-Cre and A2a-Cre mice with AAV.Ef1a.DIO.hChR2(E123T/T159C)-EYFP or AAV.Ef1a.DIO.EYFP as control into their BLA and later implanted optical fibers into the ipsilateral NAc or TuS (Fig. 4A). Given the similar BLA innervation of both the NAc core and shell (Fig. 2C), we targeted both NAc subregions for stimulation (Supplementary Fig. 4A, B). Four weeks post injection, we used a 3-chamber real-time place preference/aversion assay wherein light was delivered to stimulate the Drd1+ or Drd2+ BLA neuron terminals (465 nm, 15 ms pulses, 40 Hz) on only one side of the chamber, with no optical stimulus in either the center or the opposite chambers (Fig. 4B). The location of the mice was tracked with infrared photobeams to trigger optogenetic stimulation, and video was captured for off-line quantification.

Fig. 4: BLA Drd1+ and Drd2+ neurons innervating the ventral striatum promote aversive states depending upon projection target.
figure 4

A Paradigm for optic activation of NAc or TuS projecting Drd1+ or Drd2+ BA neurons and B 3-chamber real-time place preference/aversion assay where optic stimulation occurs in only one side of the chamber (chamber A, blue glow). Ci Optical stimulation of Drd1+ BLA→NAc neurons resulted in less time spent in the light-paired chamber (*p < 0.05, One-way ANOVA, F(2,18) = 5.04, p = 0.018, EYFP controls n = 4M/4F, Drd1+ ChR2 n = 3M/4F, Drd2+ ChR2 n = 3M/3F). Cii Stimulation of Drd1+ BLA→NAc neurons results in avoidance of the light-paired chamber (upper, t(6) = 2.981, *p = 0.025), demonstrated by representative heat map of chamber preference from one mouse (lower). Di Optical stimulation of Drd2+ BLA→TuS neurons resulted in less time spent in the light-paired chamber compared to optical stimulation of EYFP controls (*p < 0.05, Welch’s ANOVA W(2.00,8.31) = 6.02, p = 0.024, EYFP controls n = 6M/2F, Drd1+ ChR2 n = 2M/4F, Drd2+ ChR2 n = 2M/4F). Dii Stimulation of Drd2+ BLA→TuS neurons results in avoidance of the light-paired chamber (upper, t(5) = 2.916, *p = 0.033), demonstrated by representative heat map of chamber preference from one mouse (lower). EYFP control groups were collapsed across genotypes, containing both Drd1-Cre and A2a-Cre mice, since these behavioral results were not different from each other (percent of time spent in non-paired side for Drd1 + vs Drd2+ NAc EYFP controls unpaired t-test, t(6) = 1.85, p = 0.1134; and Drd1+ vs Drd2+ TuS EYFP controls unpaired t-test, t(6) = 0.305, p = 0.771). Behaviors did not differ by sex (all unpaired t-tests: NAc YFP t(6) = 1.29, p = 0.245; NAc drd1+ ChR2 t(5) = 0.933, p = 0.394; NAc Drd2+ ChR2 t(4) = 0.309, p = 0.772; TuS YFP t(6) = 1.10, p = 0.315; TuS Drd1+ ChR2 t(4) = 0.818, p = 0.459; TuS Drd2+ ChR2 t(4) = 0.253, p = 0.813; Mean ± SEM.

Full size image

We found that optical stimulation of Drd1+ BLA→NAc neuron terminals resulted in less time spent in the light-paired chamber compared to optical stimulation of Drd2+ BLA→NAc neuron terminals and EYFP controls (Fig. 4Ci & Cii) (One-way ANOVA F(2,18) = 5.04, p = 0.018). The EYFP control groups were collapsed across genotypes, containing both Drd1-Cre and A2a-Cre mice with the same optic fiber placement, since these behavioral results were not different from each other (percent of time spent in non-paired side for Drd1+ vs A2a + NAc EYFP controls unpaired t-test, t(6) = 1.85, p = 0.1134; and Drd1+ vs A2a + TuS EYFP controls unpaired t-test, t(6) = 0.305, p = 0.771). Indeed, compared to the non-stimulated side, mice spent 49.70 ± 11.10% (mean ± SEM) less time on the chamber paired with Drd1+ BLA→NAc neuron terminal stimulation (t(6) = 2.981, p = 0.025). Similarly, we found that optical stimulation of Drd2+ BLA→TuS neuron terminals resulted in less time spent in the light-paired chamber compared to optical stimulation of EYFP + BLA→TuS controls (Fig. 4Di & Dii) (Welch’s ANOVA W(2.00,8.31) = 6.02, p = 0.024). Compared to the non-stimulated side, mice spent 31.42 ± 6.39% (mean ± SEM) less time on the chamber paired with Drd2+ BLA→TuS neuron terminal stimulation (t(5) = 2.916, p = 0.033). These results show that activation of Drd1+ and Drd2+ BLA input to the NAc and TuS respectively lead to avoidance behavior.

Drd1+ BLA neurons innervating the NAc and Drd2+ BLA neurons innervating the TuS support pavlovian fear learning

Next, we wanted to know the possible influence of these pathways on learned emotional behaviors. To do this we employed an odor-shock Pavlovian fear-learning paradigm [36, 46, 47] wherein an otherwise neutral odor is paired with a mild foot shock (Supplementary Fig. 5Ai). To quantify learning, we monitored both physical immobility and fear-associated respiratory power (4–6 Hz) which increases in power when animals anticipate an aversively-paired stimulus [36, 48]. Mice were placed in a plethysmograph with a custom floor made out of metal connected to a shock stimulus generator. Also connected to the plethysmograph was a tube allowing delivery of clean air or an odor which were both controlled by an odor presentation machine. All behavioral measures and stimuli were controlled by the same computer allowing synchrony in measures and stimulus presentation events. In untreated C57BL/6J mice, we validated that only odors paired with shock were associated with elevations in physical immobility following the conditioning day (Supplementary Fig. 5B). We also validated that fear-associated 4–6 Hz respiratory power is similarly elevated as mice learn to associate an odor with a shock (Supplementary Fig. 5C–H).

We used a chemogenetic approach to suppress BLA→ventral striatum input which included six separate groups of mice to establish the roles of each of the BLA→NAc and BLA→TuS pathways (Fig. 5A). These included Drd1+ and Drd2+ mice injected with rgAAV.hSyn.DIO.hM4D(Gi)-mCherry or rgAAV.hSyn.DIO.mCherry as control. All mice were subsequently implanted with bilateral intracranial cannulae into the BLA for administration of either the DREADD ligand J60 [35] or vehicle (Supplementary Fig. 6A, B). J60 or vehicle were administered 30 min prior to the learning session following a single behavioral session on a prior day to acclimate the mice to the chamber.

Fig. 5: BLA Drd1+ and Drd2+ neurons innervating the ventral striatum support pavlovian fear learning depending upon projection target.
figure 5

A Paradigm for DREADD induced silencing of NAc or TuS projecting Drd1+ or Drd2+ BLA cells. Some aspects created in https://BioRender.com. B Influence of DREADD agonist J60 (100 nL, 10 nM) on fear learning in all NAc injected mice, C left NAc injected mCherry controls (****p < 0.0001, Two-way RM ANOVA, Trial main effect F(1,27) = 86.2, p < 0.001, n = 10M/19F); D middle Drd1+ hM4D(Gi) mice (****p < 0.0001, **p < 0.01, Two-way RM ANOVA, Trial main effect F(1,13) = 48.9, p < 0.001, n = 6M/9F); and E right Drd2+ hM4D(Gi) mice (****p < 0.0001, Two-way RM ANOVA, Trial main effect F(1,13) = 301, p < 0.001, n = 8M/7F). F Influence of DREADD agonist J60 (100 nL, 10 nM) on fear learning in all TuS injected mice, G left mCherry TuS injected controls (****p < 0.0001, Two-way RM ANOVA, Trial main effect F(1,29) = 151, p < 0.001, n = 15M/16F); H middle Drd1+ hM4D(Gi) mice (****p < 0.0001, Two-way RM ANOVA, Trial main effect F(1,13) = 85.6, p < 0.001, n = 7M/8F); and I right Drd2+ hM4D(Gi) mice (***p < 0.001, ****p < 0.0001, Two-way RM ANOVA, F(1,12) = 7.71, p = 0.017, n = 7M/8F). mCherry control groups were collapsed across genotypes, containing both Drd1-Cre and Drd2-Cre mice since these behavioral results were not different from each other (All Two-way RM ANOVAs genotype main effect: vehicle treated Drd1-Cre vs Drd2-Cre NAc mCherry controls, F(1,13) = 0.359, p = 0.559; J60 treated Drd1-Cre vs Drd2-Cre NAc mCherry controls, F(1,12) = 0.323, p = 0.580; vehicle treated Drd1-Cre vs Drd2-Cre TuS mCherry controls, F(1,14) = 0.181, p = 0.677; and J60 treated Drd1-Cre vs Drd2-Cre TuS mCherry controls, F(1,13) = 3.58, p = 0.081). Behaviors differed by sex for Drd1+ hM4D(Gi) mice suggesting sex differences in output to NAc among Drd1+ neurons (all Three-way ANOVAs for sex main effect: NAc mCherry controls F(1,25) = 0.0354, p = 0.852; Drd1+ hM4D(Gi) NAc F(1,11) = 14.0, p = 0.003; Drd2+ hM4D(Gi) NAc F(1,11) = 0.215, p = 0.652; TuS mCherry controls F(1,27) = 2.81, p = 0.102; Drd1+ hM4D(Gi) TuS F(1,11) = 0.617, p = 0.449; Drd2+ hM4D(Gi) TuS F(1,10) = 0.474, p = 0.507. Mean ± SEM.

Full size image

Among both the BLA→NAc and BLA→TuS groups, all control groups displayed elevations in fear-associated respiration by the 10th trial of odor-shock pairings (Fig. 5B, C & F, G; NAc mCherry control: Two-way RM ANOVA, trial main effect F(1,27) = 86.2, p < 0.001; TuS mCherry control: Two-way RM ANOVA, trial main effect F(1,29) = 151, p < 0.001) indicating that they learned to associate an odor with an aversive outcome. Like in the opto-RTPP/A behavioral assay, these control groups were collapsed across genotypes so they contained both Drd1-Cre and Drd2-Cre mice since these behavioral results were not different from each other (All Two-way RM ANOVAs genotype main effect: vehicle treated Drd1-Cre vs Drd2-Cre NAc mCherry controls, F(1,13) = 0.359, p = 0.559; J60 treated Drd1-Cre vs Drd2-Cre NAc mCherry controls, F(1,12) = 0.323, p = 0.580; vehicle treated Drd1-Cre vs Drd2-Cre TuS mCherry controls, F(1,14) = 0.181, p = 0.677; and J60 treated Drd1-Cre vs Drd2-Cre TuS mCherry controls, F(1,13) = 3.58, p = 0.081). As expected, similar elevations in physical immobility were also observed (Supplementary Fig. 7). Importantly, there was no difference in learning between the vehicle and J60 infused groups supporting that there are no off-target effects of this DREADD ligand on odor-shock learning (Fig. 5B, C & F, G; NAc mCherry controls Trial 10: MLSD = |Vehicle-J60| = −0.0443, p = 0.601, TuS mCherry controls Trial 10: MLSD = |Vehicle-J60| = 0.0118, p = 0.888).

While neither inhibition of Drd2+ BLA→NAc nor Drd1+ BLA→TuS pathways impacted fear-learning (Fig. 5E & H, Two-way RM ANOVA, trial main effect: Drd2+ BLA→NAc: F(1,13) = 301, p < 0.001; Drd1+ BLA→TuS: F(1,13) = 85.6, p < 0.001), we found that inhibition of Drd1+ BLA→NAc and Drd2+ BLA→TuS pathways suppressed the magnitude of the learned association. Both Drd1+ BLA→NAc and Drd2+ BLA→TuS pathway inhibition resulted in less fear-related respiration by trial 10 in J60 infused mice compared to those infused with vehicle (Fig. 5D & I; Drd1+ BLA→NAc Trial 10: MLSD = |Vehicle-J60| = 0.319, p = 0.049, Drd2+ BLA→TuS Trial 10: MLSD = |Vehicle-J60| = 0.364, p < 0.001). Fear-related physical immobility was likewise reduced upon Drd2+ BLA→TuS pathway inhibition, yet interestingly not upon Drd1+ BLA→NAc inhibition (Supplementary Fig. 7). These results show that Drd1+ and Drd2+ BLA input to the NAc and TuS respectively are necessary for fear learning in addition to their role in real-time avoidance.

Discussion

It is well established that BLA outputs to specific brain regions influence emotional responses (e.g., [15, 49,50,51,52,53,54]). More recently, several lines of evidence have uncovered divergent valence responding through genetically-distinct neurons within the BLA, including by means of Ppp1r1b and Rspo2 neurons [17,18,19]. Together, both the genetic identity and downstream targets of BLA neurons are necessary to incorporate when understanding the role of BLA cell types in orchestrating the many functions of the BLA.

In the present study we focused on defining the contributions of Drd1+ and Drd2+ BLA neurons to emotional responding. The Drd1 and Drd2 genes encode for the D1 and D2 receptors, respectively [24], which regulate PPP1R1B – a marker for one of the two main classes of BLA excitatory neurons. It has been long known that D1 and D2 receptors are in the BLA (e.g., [55]). We established that both Drd1+ and Drd2+ BLA neurons innervate the NAc and TuS, and we subsequently focused upon these two pathways (BLA→NAc and BLA→TuS) given the recent evidence of regulation of emotional responses through BLA output into these regions [17]. Our findings extend the work of [17] by showing that in addition to the Rspo2/Fezf2 BLA neuron class, both Drd1+ and Drd2+ BLA neurons in the Ppp1r1b neuron class also each innervate the NAc and TuS. These neurons originate from nearly the entire anterior-posterior extent of the BLA, and specifically the vast majority from within the BA (Fig. 1). A small population of Drd1+ and Drd2+ BLA neurons even send collaterals to both the NAc and TuS. Further, our synaptophysin tracing suggests that these BLA neuron populations innervate nearly all TuS and NAc subregions (all layers of TuS and both the NAc core and shell; Fig. 2). While the spatial innervation of Drd1+ and Drd2+ BLA neurons into the ventral striatum is not unlike that reported by [17], it is important to emphasize that Fezf2 + BLA neurons do not co-express PPP1R1B [17], which suggests these three neuron types connecting the BLA with the ventral striatum are distinct.

While the synaptophysin tracing suggests synaptic innervation of the ventral striatum by Drd1+ and Drd2+ BLA neurons, we used brain slice recordings to quantify this. This is interesting given that the primary cell type in the ventral striatum are spiny projection neurons which also express Drd1+ and Drd2+. We focused our recordings on TuS spiny projection neurons given the comparable innervation of both structures (NAc and Tus) by Drd1+ and Drd2+ BLA neurons which allowed us to also perform recordings to identify if there is logic by which ventral striatum neurons these BLA neurons synapse upon. We found that both BLA cell types excite Drd1+ and Drd2+ (identified in this experiment by expression of A2A) TuS neurons in manners which appeared to be predominately glutamatergic, with especially Drd1+ BLA neurons sending a large amount of monosynaptic currents (Supplementary Fig. 3). Further, Drd1+ BLA neurons monosynaptically excited predominately Drd1+ TuS neurons, and Drd2+ BLA neurons non-preferentially excited a small population of both Drd1+ and Drd2+ TuS neurons. While these results were initially surprising given that the Drd1+ BLA→TuS pathway was dispensable for the fear and avoidance behaviors explored in this work, this may indicate a potential role for this pathway in other behaviors, such as those involved in reward. Thus, BLA input to the TuS, and therefore possibly also the NAc, has an organization which allows for recruitment of specific postsynaptic neurons in the TuS which could therefore allow differential output from the ventral striatum in manners supporting specific outputs into the basal ganglia and other brain networks important for behavioral responses.

We found within this circuitry that the parallel pathways generated by the Drd1+ and Drd2+ BLA neuron populations modulate negative emotional states depending upon their ventral striatum projection target (Fig. 6). To show this, we used two distinct behavioral paradigms in combination with either projection specific chemo- or optogenetic manipulations. In all behavioral paradigms, we were able to uncover a role for either Drd1+ and/or Drd2+ neurons, yet, in not all cases did each cell population impact behavior. Instead, the impact on behavior was in most cases also projection target specific. For instance, Drd1+ BLA neurons innervating the NAc increased negative valence states in the real-time place preference/aversion paradigm, whereas the same cell population projecting to the TuS did not (Fig. 4). Likewise, Drd2+ BLA neurons innervating the TuS increased negative valence states in the real-time place preference/aversion paradigm, whereas the same cell population projecting to the NAc did not. Similar differences in how these genetically-distinct BLA cell populations influenced Pavlovian fear learning were also observed to be cell-type and projection target specific. These findings lay the foundation for future work to systematically target Drd1+ or Drd2+ BLA inputs into specific regions within the NAc (core vs shell [56]) or TuS (medial vs lateral [57, 58]) which may provide even more specific behavioral outcomes. This work extends a role for Drd1+ BLA neuron output to the central amygdala, which was found to influence extinction memory [26], into two ventral striatum subregions which are important for valence-based behavioral responses, and allows for comparison with the influence of the neighboring Drd2+ neurons.

Fig. 6: Overview of findings illustrating the behavioral consequences of activating (left) or suppressing (right) Drd1+ or Drd2+ BLA neuron inputs to the NAc and TuS.
figure 6

Solid lines annotate pathways found herein to play a role in real-time avoidance and learned aversion. Dashed lines indicate pathways that were not found to support either of those behaviors/states, at least using the approaches applied in the present study.

Full size image

Interestingly, when comparing changes in fear-associated respiration (well known to be influenced by sympathetic state [36, 48, 59,60,61]) and fear-induced immobility, we saw that manipulation of NAc projecting Drd1+ BLA neurons did not similarly influence both of those fear-associated behaviors (Supplementary Fig. 7). This may be due to either distinct inputs or outputs (collaterals) of the Drd1+ BLA neurons which might differentially guide changes in respiratory behavior versus motor behavior. For instance, differential innervation of the periaqueductal grey might allow for one cell-type to influence respiration over another given the periaqueductal grey’s influence on breathing [62, 63]. While we did not identify the differential pathway, it is interesting to find such instances wherein fear-related behaviors are not simultaneously displayed. Further manipulation of this pathway while similarly taking multiple measures of fear-related behaviors, including even heart rate, skin conductance, and ultrasonic vocalizations for instance, will help refine our understanding of circuitry which specifically supports each to be displayed during emotional contexts.

Given the fact that these BLA→ventral striatum neurons express dopamine receptors, it is tempting to speculate how this pathway may be modulated by dopamine. Dopamine within the BLA is necessary for fear learning [25]. Local antagonism of both D1Rs and D2Rs within the BLA blocks the expression of fear during a potentiated startle paradigm [64, 65]. Antagonism of BLA D1Rs also perturbs the timing of fear behavior [66], and antagonism of BLA D2Rs attenuates freezing during Pavlovian fear conditioning [27, 28, 67]. The role of dopamine receptors is similarly mixed in appetitive behaviors, where antagonizing both D1Rs and D2Rs within the BLA attenuates conditioned reward seeking and taking [29, 68, 69]. Indeed, local application of D1 agonists increases intrinsic excitability and the evoked firing of BLA neurons [70]. D1 receptors have a lower affinity for dopamine than D2 receptors [71, 72]. Further, when dopamine levels are low, D2 receptors are agonized, but when DA levels are elevated, like when receiving an emotionally salient stimulus, both D1 and D2 receptors become agonized [73,74,75]. It is possible these differential roles of D1 and D2 receptors in the BLA might explain our finding that Drd2+ neurons vs Drd1+ neurons contributed differently to the regulation of emotional states.

Several findings throughout this study pose the need to caveat some interpretations. First, while the fiber optic implants in the NAc extended ~1 mm over the TuS, and blue light used in our approach would not penetrate more than ~800 µm through brain, we cannot rule-out that no BLA terminals in the TuS were activated upon NAc stimulation. Related to that, the finding that a subpopulation of Drd1+ and Drd2+ BLA neurons send collaterals to both the NAc and TuS means that the parallel circuitry generated by these genetically-distinct pathways is considerably complicated. This also adds a caveat to the optical stimulation results indicating Drd1+ BLA→NAc neurons influence avoidance since ChR2-mediated activation of those terminals in the NAc might had activated TuS-projecting BLA neurons through antidromic stimulation. Further, while we know that some BLA neurons cross the midline (e.g., Fig. 2F), we did not attempt to disambiguate the role of ipsi- vs contral-lateral Drd1+ and Drd2+ BLA pathways. Although it is notable that during preparation of this paper, a report was published [76] implicating BLA projections to the NAc in valence behaviors which appear to differ depending upon if they transverse through the anterior commissure. Additionally, in experiments wherein AAV injections were performed in the BLA, in some cases, small populations of neurons were labeled outside of the BLA which may impact results. That stated, among amygdaloid nuclei, multiple separate experiments support that it is principally neurons in the BLA which innervate the TuS and NAc (Fig. 1 & Supplementary Fig. 1). Finally, while both our results and that from prior work support that there are more Drd1+ neurons in the BLA than Drd2+ (Fig. 1 and [40]), which was also evident when analyzing levels of A2a+ neurons in the BLA, we found that recombination occurred only in a subpopulation of Cre-expressing neurons (Supplementary Fig. 8). This was especially the case in the Drd2-Cre line. The fact that our chemo- and opto-genetic experiments resulted in behavioral changes (Fig. 4 & 5, Supplementary Fig. 7), despite our approaches not capturing the entire population of Cre-expressing neurons, reflects that these pathways are potent modulators of emotions. Indeed, even chemogenetic manipulation Drd2+ BLA neurons, of which only a spare population showed recombination efficiency, impacted emotional outcomes (Supplementary Fig. 7). These stated, the low recombination efficiency among Drd2+ BLA neurons supports the need for future research to thoroughly understand the role of Drd2+ BLA projections to the NAc since we cannot confidently exclude this pathway from being involved in fear-learning, nor other learned aversive states.

Overall, this work has uncovered that Drd1+ and Drd2+ neurons within the Ppp1r1b BLA neuron class form parallel pathways which influence emotional states when they are activated or suppressed and do so depending upon where they synapse – with unique contributions of Drd1+ and Drd2+ BA→NAc vs BA→TuS circuitry on negative valence states. Given the connectivity to the TuS especially, given its established roles in sensory processing [20], it is interesting to consider if these pathways may contribute to changes in sensory processing in individuals who suffer from anxiety disorders [77]. Overall, our results contribute to a model whereby parallel, genetically-distinct BLA to ventral striatum circuits inform emotional states in a projection-specific manner. This work adds to our understanding of the complex interplay between projection cell types and their projection targets, in how the BLA helps orchestrate emotions.