A molecularly defined basalo-prefrontal-thalamic circuit regulates sensory and affective dimensions of pain

Abstract

Both the medial prefrontal cortex (mPFC) and thalamus have been implicated in pain regulation. However, the roles of the mPFC-thalamus connection in pain and how the mPFC modulates nociceptive processing within the brain remain unclear. Here, we show that the mPFC neurons that project to thalamus are marked by Foxp2 expression and deactivated in both acute and chronic pain. Persistent inactivation of the mPFC Foxp2⁺ neurons enhances nociceptive sensitivity, while their activation alleviates multiple aspects of pain. Circuit-specific manipulations revealed that the projections to parataenial nucleus, mediodorsal and ventromedial thalamus differentially modulate sensory and affective pain. Additionally, the mPFC Foxp2⁺ neurons receive cholinergic input from the basal forebrain, particularly the horizonal diagonal band (HDB). Notably, activation of the α4β2-containing nicotinic acetylcholine receptor in mPFC exerts antinociceptive effects in Foxp2⁺ neuron-dependent manner. Together, our study defines an HDB→mPFC^Foxp2→thalamus circuit essential for sensory and affective pain modulation and underscores the therapeutic potential of targeting mPFC cholinergic signaling in chronic pain management.

Introduction

Pain is an unpleasant sensory and emotional experience associated with actual or potential tissue damage and is essential to protect the body from potentially harmful stimuli^{1, 2}. Chronic pain, however, is not just a temporal continuum of acute pain but one of the major causes of human suffering, affecting about 20% of the global population with immense clinical and economic consequences^{3, 4}. Unfortunately, current pain management strategies are largely unsatisfactory with inadequate pain relief and multiple adverse effects^{5, 6}. Thus, understanding the mechanisms underlying pain processing remains to be an urgent need for developing effective treatments.

Accumulating evidence supports the crucial roles of medial prefrontal cortex (mPFC) in regulating multiple aspects of pain, including the sensory, affective and cognitive dimensions^7–11. Neuroimaging studies in humans have showed functional deactivation and altered connectivity of mPFC in patients with chronic pain^12–14. Consistently, preclinical studies have also revealed the diminished activities of mPFC in several forms of chronic pain and the critical roles of mPFC in pain relief^15–18. The mPFC is widely connected with other brain regions, including the periaqueductal gray (PAG), thalamus, amygdala and basal nuclei¹⁹. As part of the descending modulation pathway that governs the noxious inputs in the spinal cord, the mPFC to PAG circuit in pain processing has been well documented^{5, 16, 17}. However, it is much less clear about the roles of the mPFC outputs to other brain regions in pain regulation as well as how mPFC modulates the processing of noxious information in the brain.

Thalamus, a major target of the mPFC output, serves as a key hub for integrating and processing nociceptive information ascending from the spinal cord and modulates both sensory and affective dimensions of pain through the lateral and medial pathways^{5, 20}. Imaging studies have linked chronic pain to thalamic alterations, including reduced volume, activity and imbalanced transmitters^{12, 14, 21}. Recent findings also highlight the critical functions of the connections between thalamus and other cortices, such as motor cortex and ACC, in regulating affective pain^{22, 23}. Despite the strong anatomical connections between mPFC and thalamus and their critical roles in pain and other behaviors such as attention and cognition^{24, 25}, the specific functions of the mPFC to thalamus circuit in pain processing remain largely unexplored.

The mPFC receives inputs from multiple brain regions¹⁹. Among these, the strengthened glutamatergic input from basolateral amygdala (BLA), which synapses onto the inhibitory neurons in mPFC and sends feedforward inhibition to the pyramidal neurons, has been reported as a key mechanism underlying the deactivation of mPFC in chronic pain.^{17, 18}. However, the roles of other mPFC inputs in chronic pain remain largely unknown. Notably, the mPFC receives dense cholinergic innervation from the basal forebrain, a pathway involved in various functions such as learning, attention and arousal²⁶. Although the clinical importance of cholinergic signaling in pain has been well recognized²⁷, the roles and the circuit mechanism of cholinergic modulation of mPFC in pain are not fully understood. Moreover, the mPFC is highly heterogeneous, with differential structural and functional dysregulation of mPFC in chronic pain⁸. The variable functional responses, either pain relief or exacerbation following mPFC activation^{8, 28, 29}, might result from the cell heterogeneity across subregions and layers of mPFC, which requires higher cell-type and projection-specific resolution to accurately delineate.

In this study, we identify Foxp2 as a marker of the mPFC neurons that project to the thalamus and show that these neurons are deactivated under acute and chronic pain. Using cell type-specific and circuit-specific manipulations, we demonstrate that activation of the mPFC Foxp2⁺ neurons alleviates both the sensory and affective components of pain by targeting distinct thalamic subregions. Additionally, we show that the mPFC Foxp2⁺ neurons receive cholinergic innervation from the basal forebrain, particularly the horizonal diagonal band of Broca (HDB), and that activation of the cholinergic projection is antinociceptive. Notably, we found that activation of the α4β2-containing nicotinic acetylcholine receptor (nAChR) in mPFC relieves pain, which requires the Foxp2⁺ neuronal activity. Collectively, our study reveals a novel mechanism by which the mPFC regulates pain through projection to thalamus, and highlights mPFC cholinergic signaling as a potential therapeutic target for chronic pain treatment.

Results

Foxp2 specifically marks the mPFC outputs to thalamus

The cerebral cortex is organized into layers with diverse neuron subtypes that have distinct projection patterns distributed in laminar pattern³⁰. In the mPFC, the projections to thalamus, a hub for pain processing, primarily originate from the layer (L) 6 corticothalamic (CT) neurons^{31, 32}. Surprisingly, single molecule fluorescence in situ hybridization (smFISH) revealed that Ntsr1, a widely used L6 CT neuron marker in the cortex^{22, 33, 34}, is broadly expressed from L2 to L6 in the mPFC (Extended Data Fig. 1a), indicating the different organization of the mPFC compared to other cortex. Integrative analysis of our previous single-cell RNA sequencing and spatial transcriptomic data indicated Foxp2 and Oprk1 respectively mark the L6 CT and L6 intratelencephalic (IT) neurons^{35, 36}, the two major excitatory neuron types in L6 (Extended Data Fig. 1b). This is consistent with previous studies suggesting exclusive expression of Foxp2 in CT neurons of the sensory cortex^{33, 37–40}.

To determine whether Foxp2 can serve as a marker for L6 CT neurons in the mPFC of adult mice, we injected cholera toxin subunit B (CTB) 555, a retrograde tracer, into the medial dorsal thalamus (MD) and performed smFISH for Foxp2 in the mPFC (Fig. 1a). We found that the CTB⁺ cells are mainly located in L6 with over 90% of them express Foxp2 (Fig. 1a), indicating that the Foxp2⁺ neurons account for the major outputs of mPFC to thalamus. Around 60% of the Foxp2⁺ neurons are labeled by CTB, which could be due to the limited area of CTB diffusion and that the mPFC Foxp2⁺ neurons may project to other thalamic regions. To examine whether the mPFC Foxp2⁺ neurons project to other major mPFC targets, CTB was also injected into the PAG (Fig. 1b), contralateral mPFC (Extended Data Fig. 1c), amygdala (Extended Data Fig. 1d) or nucleus accumbens (NAc) (Extended Data Fig. 1e). Minimal colocalization of CTB and Foxp2 was observed in these conditions (Fig. 1b and Extended Data Fig. 1c–e). These findings confirm that Foxp2 specifically marks the major outputs of mPFC to thalamus.

Figure 1. Foxp2 specifically marks the outputs of mPFC to thalamus.

a, Diagram of the injections of CTB into the thalamus of wildtype mice (left). Representative images of the mPFC in mice with CTB injected into the thalamic area (middle, left); the enlarged images of the indicated regions (middle, top); the injection site of CTB (middle, bottom). Percentage of the Foxp2⁺ or Foxp2⁻ cells of all CTB⁺ neurons (top right); Percentage of the CTB⁺ or CTB⁻ cells of all Foxp2⁺ neurons (bottom right). Scale bars: middle left, 200 μm; middle top, 20 μm; middle bottom, 500 μm. b, Same as a with CTB injected into the PAG. Scale bars: middle left, 200 μm; middle top, 20 μm; middle bottom, 500 μm. c, Representative image of the smFISH for Foxp2 (red), Slc17a7 (green) and Gad1 (white) in the mPFC (left) and the enlarged image of the indicated region (top right), as well as the percentages of the Foxp2⁺ neurons co-expressing Slc17a7 or Gad1 (bottom right). Scale bars: left, 200 μm; top right, 20 μm. d, Representative image of smFISH for Foxp2 (red), Etv1 (white) and Ctgf (green) in the mPFC (left) and the percentages of the Foxp2⁺ neurons co-expressing Etv1 or Ctgf (right). Scale bar, 200 μm. e-g, Representative image of the smFISH for Foxp2 (red) and Oprk1 (green) (e), Pou3f1 (green) (f), and Tshz2 (green (g) in the mPFC (left) and the enlarged image of the indicated region (top right), as well as the percentages of the Foxp2-only and Foxp2-Oprk1/Pou3f1/Tshz2 double positive cells in the mPFC (bottom right). Scale bars: left, 200 μm; top right, 20 μm. h, Diagram of the axon labeling strategy (left), and representative images showing the mGFP and mRuby signals in the thalamic subregions (middle and right). Enlarged images of the indicated regions in PTN, MD and VM are shown in the inserted boxes. Scale bars: 100 μm; inserted boxes, 10 μm. i, Quantification of the mean mRuby fluorescence intensity per area in different brain regions. j, Diagram of the projection pattern of mPFC L6 neurons.

Data are represented as mean ± SEM. 3V, third ventricle; AD, anterodorsal thalamus; Amy, amygdala; CTB: cholera toxin subunit B; DLPAG, dorsal lateral periaqueductal grey; NAc, nucleus accumbens; Hb, habenula nucleus; HPC, hippocampus; LD, laterodorsal thalamus; LPAG, lateral periaqueductal grey; MD, mediodorsal thalamus; PAG: periaqueductal gray; PTN, parataenial thalamic nucleus; PVT, paraventricular thalamus; Re, reuniens nucleus; Str, striatum; VM, ventromedial thalamus.

Given the outputs of mPFC to thalamus mainly originate from the L6 CT neurons, we then asked whether Foxp2 specifically marks this distinct neuronal subtype. To this end, we performed smFISH for Foxp2, Slc17a7 (a marker for glutamatergic neurons) and Gad1 (a marker for inhibitory neurons), which revealed that over 95% of Foxp2⁺ cells express Slc17a7 (Fig. 1c), indicating that Foxp2⁺ cells are predominantly glutamatergic neurons. To further characterize the distribution of Foxp2⁺ cells in the mPFC, smFISH for Foxp2, Etv1 (a marker for L5), and Ctgf (a marker for L6b) was performed with the mPFC sections along the anterior-posterior axis (Fig. 1d and Extended Data Fig. 2). Quantification of the fluorescence signals confirmed the expression of Foxp2 in L6 including L6b, with about 10% of Foxp2⁺ cells also expressing Etv1 (Fig. 1d and Extended Data Fig. 2a–c). Further quantification of the fluorescence signals in the mPFC from pia to white matter revealed that the Foxp2⁺ cells are mainly located in L6 and to a much less extent in deep L5, especially in the anterior part of mPFC (Extended Data Fig. 2d). Since Foxp2 and Oprk1 are indicated to respectively mark the L6 CT and IT neurons (Extended Data Fig. 1b), we examined their expression in the mPFC and found that they label two distinct neuronal populations in L6 with minimal overlap (Fig. 1e). Because there are also a small number of Foxp2⁺ neurons in L5, where resides the L5 IT neurons, as well as the extratelencephalic (ET) neurons labeled by Pou3f1³⁵ and near-projecting (NP) neurons labeled by Tshz2⁴¹, we then performed smFISH and found little overlap between Foxp2 and Pou3f1 (Fig. 1f) or Tshz2 (Fig. 1g). Collectively, these results indicate that Foxp2 labels a distinct neuronal subtype in L6 different from the IT, ET and NP neurons in the mPFC, consistent with the tracing results.

To further confirm that the mPFC Foxp2⁺ neurons specifically project to the thalamus, Cre recombinase (Cre)-dependent adeno-associated virus (AAV) encoding mGFP and synaptophysin-fused mRuby that labels axon terminals was injected into the mPFC of Foxp2-Cre mice (Fig. 1h). The virus expression was confirmed by post hoc histological examination (Extended Data Fig. 1f). Consistent with the fact that Foxp2 labels the L6 CT neurons in the mPFC, whole brain mapping revealed that the mPFC Foxp2⁺ neurons primarily project to thalamus, with the highest intensity of mRuby signals in parataenial thalamic nucleus (PTN), mediodorsal (MD) and ventromedial (VM) thalamus (Fig. 1h). Weaker signals were observed in the nucleus reuniens (Re), paraventricular nucleus of thalamus (PVT) and the medial striatum (Str), while no signal was detected in other downstream targets of mPFC such as NAc, amygdala, PAG, hippocampus, hypothalamus and ventral tegmental area (VTA) (Fig. 1h and Extended Data Fig. 1g–m). Quantification of the average axon-labeling mRuby signals indicated that the medial thalamic nuclei, including PTN, MD and VM, are the main downstream targets of the mPFC Foxp2⁺ neurons (Fig. 1i). Collectively, these results demonstrate that Foxp2 specifically labels the mPFC outputs to thalamus (Fig. 1j).

The mPFC Foxp2⁺ neurons are deactivated in acute and chronic pain

To determine whether the mPFC Foxp2⁺ neurons are involved in pain processing, in vivo miniscopic calcium (Ca²⁺) imaging was performed to monitor the activity of the individual mPFC Foxp2⁺ neurons under various pain states. To this end, Cre-dependent AAV encoding the Ca²⁺ indicator GCaMP7s was injected into the prelimbic cortex (PL), a subregion of mPFC critical for pain regulation⁸, of Foxp2-Cre mice, followed by the implantation of a gradient refractive index (GRIN) lens above the injection site two weeks later (Fig. 2a). Virus expression and proper GRIN lens implantation were confirmed by post hoc histological examination (Fig. 2a). Four weeks after the virus injection, fluorescence signals of the individual Foxp2⁺ neurons were recorded through the GRIN lens using a head-mounted miniaturized microscope, and the Ca²⁺ signal traces of individual neurons were identified and extracted (Fig. 2b). The mice were subjected to acute noxious stimuli and complete Freund’s adjuvant (CFA) treatment to induce chronic inflammatory pain, with the Ca²⁺ signals recorded at the indicated time points (Fig. 2c).

Figure 2. The mPFC Foxp2⁺ neurons are deactivated in response to acute and chronic pain.

a, Diagram of the in vivo miniscopic calcium imaging (left) and a representative image of virus expression and GRIN lens implantation (right), Scale bar: 200 μm. b, An example field of view (left) and identified contours of neuronal regions of interest (right). c, Timeline of treatments and calcium recordings. d, Diagram of calcium recording for noxious heat stimulation. e, Left, heatmap of the calcium event frequency averaged to 1 min of bin of all neurons in response to and after the tail pinch stimulation (n = 270). The white dotted lines indicate the start and end of tail pinch, respectively. Right, representative traces showed fluorescence intensity changes (ΔF/F) from the identified individual neurons in response to and after the tail pinch stimulation. The red and the blue background indicate the time of during and after the tail pinch stimulation, respectively. f, Cumulative distribution of calcium event frequencies of baseline (BL) and during the tail pinch stimulation (left). Percentages of neurons that show no change (grey), decrease (red), and increase (yellow) in frequency in response to noxious heat during the last minute compared to the baseline (right). n = 248. g, Average calcium event frequencies of all identified neurons in baseline and during the last minute of heating. n = 248. h-k, Same as d-g but for formalin treatment. n = 302. l, Diagram of calcium recording before and after CFA injection. m, Averaged frequency of spontaneous calcium events of all the identified neurons in the baseline (n = 297), day 1 (n = 297) and day 7 (n = 344) after CFA injection. n, Heatmap of the calcium transient frequency averaged to 3 min of bin of the registered same neurons across different days (n = 124). o, Representative calcium traces of the same individual neurons that show decrease (red), no change (black), and increase (yellow) in the frequency in the baseline, day 1 and day 7 after CFA injection. p, Cumulative distribution of the calcium event frequency of the registered same neurons in baseline, day 1 and day 7 after CFA injection. q, Comparison of baseline and day 1 events (left); the percentages of the neurons that show no change (grey), decrease (red), and increase (yellow) in frequency in 1 day after CFA injection (right). r, Same as q except it is for comparison of day 7 after CFA injection.

Data are represented as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. Each dot indicates one individual neuron.

To examine the activities of the Foxp2⁺ neurons in response to noxious mechanical stimulation, we recorded the baseline activity for 2 minutes, then subjected the mice to tail pinch for 1 minute, and monitored an additional 9 minutes for recovery (Fig. 2d). We found that most of the Foxp2⁺ neurons exhibited a decreased frequency of Ca²⁺ events during pinch stimulation, followed by a gradual recovery (Fig. 2e). A small subset of the Foxp2⁺ neurons displayed the opposite response (Fig. 2e). Quantitative analysis revealed that around 80% of the Foxp2⁺ neurons exhibited a significant decrease in the Ca²⁺ event frequency during pinch stimulation compared to the baseline (Fig. 2f and Extended Data Fig. 3a). A significant decrease in the overall Ca²⁺ event frequency across all Foxp2⁺ neurons was also observed (Fig. 2g), while the amplitude was unaltered (Extended Data Fig. 3b). These results indicate that the overall Foxp2⁺ neuronal activity is decreased in response to noxious mechanical stimulation.

The mice were also subjected to formalin treatment (Fig. 2h), which induces biphasic pain responses as previously described⁴². Enhanced coping behaviors were observed during the first 5 minutes (phase 1) and between 15 to 30 minutes (phase 2) following the formalin injection (Extended Data Fig. 3c). Similar to the responses to noxious mechanical stimulation, a large subset of the Foxp2⁺ neurons exhibited a decrease in the frequency of Ca²⁺ events during phase 1, while a small subset showed increased activity (Fig. 2i). Quantitative analysis revealed a significant decrease in the Ca²⁺ event frequency during both phases, with approximately 30% of Foxp2⁺ neurons significantly deactivated in the phase 1 (Fig. 2j and Extended Data Fig. 3d). The overall Ca²⁺ event frequency but not amplitude across all Foxp2⁺ neurons was also significantly decreased in both phases (Fig. 2k and Extended Data Fig. 3e), indicating a decreased Foxp2⁺ neuronal activity in response to formalin treatment. Collectively, these results indicate that the mPFC Foxp2⁺ neurons are deactivated in acute nociception.

Finally, we asked whether the mPFC Foxp2⁺ neurons are involved in chronic pain. To this end, we subjected the mice to CFA treatment to induce chronic inflammatory pain (Fig. 2l). The overall frequency of Ca²⁺ events in the recorded mPFC Foxp2⁺ neurons significantly decreased in both day 1 and day 7 after the CFA injection compared to the baseline (Fig. 2m), indicating a decrease in the Foxp2⁺ neuronal activity in chronic inflammatory pain. To better compare the Ca²⁺ events of the individual Foxp2⁺ neurons among different days, we identified and analyzed the same neurons recorded in different days (Extended Data Fig. 3f). We found that a large proportion of the Foxp2⁺ neurons exhibited decreased Ca²⁺ activities on day 1 and day 7 after CFA injection (Fig. 2n, o). Quantitative analysis revealed the decreased Ca²⁺ events frequency of the registered Foxp2⁺ neurons in both day 1 and day 7 (Fig. 2p), and that 25.8% and 43.5% of Foxp2⁺ neurons were significantly deactivated on day 1 and day 7, respectively (Fig. 2q, r). Together, these results demonstrate that the mPFC Foxp2⁺ neurons are deactivated in CFA-induced chronic inflammatory pain.

Persistent inactivation of the mPFC Foxp2⁺ neurons enhances nociceptive sensitivity, while activation of these neurons alleviates different aspects of pain

Given the critical roles of mPFC and thalamus in pain processing and the observed deactivation of the mPFC Foxp2⁺ neurons in both acute and chronic pain (Fig. 2), we asked whether these neurons play a causal role in pain regulation. To this end, Cre-dependent AAVs expressing EYFP or tetanus neurotoxin (TeNT) were bilaterally injected into the PL mPFC of Foxp2-Cre mice to persistently block the synaptic output of the Foxp2⁺ neurons (Fig. 3a). Behavioral tests were conducted five weeks later, with virus expression confirmed by post hoc histological examination (Fig. 3a). We first performed von Frey test (VFT) to measure the mechanical sensitivity of the mice. The TeNT group exhibited a significantly decreased paw withdrawal threshold compared to the control group, indicating that persistent inactivation of the mPFC Foxp2⁺ neurons induces mechanical hypersensitivity of the mice (Fig. 3b). Similarly, a significant decrease in paw withdrawal latency in the hot plate test (HPT) was also observed in the TeNT group, indicating thermal hypersensitivity in these mice (Fig. 3c). Together, these results indicate that persistent inactivation of the mPFC Foxp2⁺ neurons enhances nociceptive sensitivity in mice.

Figure 3. The mPFC Foxp2⁺ neurons regulate both sensory and affective aspects of pain.

a, Diagram of virus injection (left) and a representative image of TeNT-EYFP expression in the PL mPFC of Foxp2-Cre mice (right). Scale bar: 100 μm. b, Diagram of the von Frey test (VFT) (left). Paw withdrawal threshold of the mice assessed in VFT (right). c, Diagram of the hot plate test (HPT) (left). Paw withdrawal latency of the mice assessed in HPT (right). d, Diagram of the virus injection into the PL mPFC of the Foxp2-Cre mice (left) and a representative image of hM3Dq-mCherry expression in the mPFC (right). Scale bar: 100 μm. e, Paw withdrawal threshold (left) and latency (right) of the mice with mCherry or hM3Dq-mCherry expression assessed in VFT and HPT, respectively. f, Diagram of formalin test (left). Licking duration of the mice in response to formalin injection after CNO treatment and summarized licking duration in the phase 1 and phase 2 (right 3 panels). g, Diagram of CFA injection (left) and paw withdrawal threshold/latency assessed by VFT on post CFA day 2 (middle) and HPT on post CFA day 3 (right). h, Diagram of the conditioned place preference (CPP) test starting from post CFA day 5 (left). The time that mice spent in the chamber before and after the conditioning (middle) and CPP scores (right) of the mice in CPP test. The CPP score is calculated by subtracting the time that the mice spent in chamber coupled with CNO treatment in the pre-test from the one in the post-test.

Data are represented as mean ± SEM. Each dot indicated one mouse. *p < 0.05, **p < 0.01, ***p < 0.001; ns, no significant difference.

We next asked whether short term manipulation of the mPFC Foxp2⁺ neurons affects the nociception. To this end, we bilaterally injected Cre-dependent AAVs encoding mCherry or hM3Dq, an activating DREADD (designer receptors exclusively activated by designer drug), into the PL mPFC of the Foxp2-Cre mice (Fig. 3d). Virus expression was confirmed by post hoc histological examination (Fig. 3d). Three weeks later, behavioral assays were performed 20 min after intraperitoneal (i.p.) injection of clozapine-N-oxide (CNO). In both the VFT and HPT, the hM3Dq group showed a significant increase in paw withdrawal threshold and latency (Fig. 3e), indicating the antinociceptive effects induced by the mPFC Foxp2⁺ neuron activation. To examine the roles of the mPFC Foxp2⁺ neurons in coping behaviors, formalin test was performed which revealed a significant decrease in licking duration in both phase 1 and phase 2 in the hM3Dq group (Fig. 3f), further supporting the antinociceptive effects of the mPFC Foxp2⁺ neuron activation.

We next assessed the roles of the mPFC Foxp2⁺ neurons in chronic pain by treating the mice with CFA to induce chronic inflammatory pain. VFT and HPT were performed 2 or 3 days after CFA injection which revealed significant increases in paw withdrawal threshold and latency in the hM3Dq group (Fig. 3g), indicating activation of the mPFC Foxp2⁺ neurons alleviates the mechanical and thermal hypersensitivity in chronic inflammatory pain. To determine the contribution of these neurons to the affective components of pain, we subjected the mice to a conditioned place preference (CPP) test. The mice were conditioned with saline in one chamber and CNO in another (Fig. 3h). After three days of training, the hM3Dq group displayed a preference for the chamber paired with CNO treatment, as indicated by increased time spent in the chamber and a higher CPP score (Fig. 3h), suggesting that activation of the mPFC Foxp2⁺ neurons could alleviate affective aspects of chronic inflammatory pain. Collectively, these results indicate that activation of the mPFC Foxp2⁺ neurons relieves both nociceptive hypersensitivity and negative affects of chronic inflammatory pain.

To evaluate the effects of short-term inactivation, we bilaterally injected Cre-dependent AAVs expressing mCherry or hM4Di, an inhibitory DREADD, into the PL mPFC of the Foxp2-Cre mice (Extended Data Fig. 4a). Surprisingly, no significant effects were observed in the nociceptive sensitivity, coping behaviors or negative affects of pain as indicated by VFT, HPT, formalin test and CPP test (Extended Data Fig. 4b–d). The efficiency of chemogenetic manipulation was confirmed as the expression of an immediate-early gene cFos was respectively decreased or increased after CNO treatment in the hM4Di or hM3Dq group (Extended Data Fig. 4e).

Collectively, these results indicate that persistent inactivation of the mPFC Foxp2⁺ neurons enhances nociceptive sensitivity, while activation of these neurons exerts robust antinociceptive effects, alleviating both the sensory and affective components of inflammatory pain.

Projection-specific roles of the mPFC Foxp2⁺ neurons in regulating acute and chronic pain

Having demonstrated the crucial roles of the mPFC Foxp2⁺ neurons in regulating pain, we next investigated the underlying circuit mechanisms. We have revealed that the PTN, MD and VM thalamic subregions are the major downstream targets of the mPFC Foxp2⁺ neurons (Fig. 1h, i and Extended Data Fig. 1f–m). Previous studies have revealed the roles of MD and VM in regulating affective, but not sensory aspects of pain^{7, 22, 23}, while the PTN is much less studied but implicated in regulating pain sensitivity in mice⁴³. To determine the roles of the mPFC Foxp2⁺ projections in regulating pain, Cre-dependent AAVs encoding EYFP or channelrhodopsin-2 (ChR2) were injected into the PL mPFC of Foxp2-Cre mice, and optical fibers were implanted above the PTN (Fig. 4a), MD (Fig. 4b) or VM (Fig. 4c), respectively for projection-specific activation. Virus expression and fiber implantation were confirmed by post hoc histological examination (Fig. 4a–c). Three weeks later, VFT and HPT were conducted. We found that optogenetic activation of the mPFC Foxp2⁺ projection to PTN significantly increased the paw withdrawal threshold in VFT and latency in HPT compared to the baseline or the control group (Fig. 4d), indicating reduced mechanical and thermal sensitivity. In contrast, activation of the projections to MD or VM had no significant effect on nociceptive sensitivity (Fig. 4e, f). These results suggest that mPFC Foxp2⁺ neurons reduce mechanical and thermal sensitivity through projecting to PTN, but not the MD or VM of thalamus.

Figure 4. Projection-specific effects of the mPFC Foxp2⁺ neurons on acute nociception.

a, Diagram of virus injection and optical fiber implantation in the Foxp2-Cre mice (left). Representative images showing the injection site in PL mPFC (middle) and the cannula implantation above the projection to PTN (right). Scale bars: 100 μm. b-c, Diagram of surgery in the Foxp2-Cre mice (left) and representative images (right) showing the implantation of the optical fiber above the projection to MD (b) or VM (c). Scale bars: 100 μm. d, Paw withdrawal threshold/latency assessed by VFT (left) and HPT (right) of the GFP-expressing and ChR2-expressing mice with (on) or without (off) light stimulation targeting the projection to PTN. e, f, Same as the d but targeting the projection to MD (e) or VM (f). g, Diagram of formalin injection (left). Licking duration of the mice in response to formalin injection with light stimulation (second panel), and summarized licking duration in phase 1 and phase 2 (right two panels) targeting the projection to PTN. h, i, Same as g but targeting the projection to MD (h) or VM (i).

Data are represented as mean ± SEM. Each dot in bar graph indicated one mouse. *p < 0.05, ***p < 0.001; ns, no significant difference.

To examine the roles of the projections in coping behaviors under tonic pain, we subjected the mice to formalin test with optogenetic activation of the specific projections. We found that activation of the projections to PTN and VM, but not MD, significantly reduced the licking duration in both phase 1 and phase 2 of the formalin test compared to the control group (Fig. 4g–i). These results indicate that the projections of the mPFC Foxp2⁺ neurons to the PTN and VM reduce the coping behaviors of the mice in response to tonic pain. Collectively, these results demonstrate that the mPFC^Foxp2→PTN circuit modulates both the nociceptive sensitivity and coping behavior and the mPFC^Foxp2→VM circuit specifically modulate the coping behavior, while the mPFC^Foxp2→MD circuit is not involved in either.

To assess the roles of the projections in chronic pain, we treated the mice with CFA to induce chronic inflammatory pain and performed the pain-related behavioral assays with optogenetic stimulation (Fig. 5a). Two and three days after the CFA injection, VFT and HPT were performed, respectively, to measure mechanical allodynia and thermal hyperalgesia in chronic inflammatory pain. We found that activation of the PTN projection significantly increased the paw withdrawal threshold and latency (Fig. 5b), indicating reduced mechanical and thermal hypersensitivity of the mice. In contrast, no significant change in the VFT and HPT was observed by activating the MD or VM projections (Fig. 5c, d). These results demonstrate that the PTN projection is primarily responsible for the roles of the mPFC Foxp2⁺ neurons in alleviating nociceptive hypersensitivity in chronic pain.

Figure 5. Projection-specific effects of the mPFC Foxp2⁺ neurons on chronic inflammatory pain.

a, Schematic diagram of the behavioral tests. b, Paw withdrawal threshold/latency assessed by VFT (left) and HPT (right) of the EYFP- or ChR2-expressing mice with (On) or without (Off) light stimulation targeting the projection to PTN. c, d, Same as b but targeting the projection to MD (c) or VM (d). e, The time spent in the chamber paired with light stimulation (left) and the CPP scores (right) of the mice targeting the projection to PTN. f, g, Same as e but targeting the projection to MD (f) or VM (g).

Data are represented as mean ± SEM. Each dot indicated one mouse. *p < 0.05, **p < 0.01, ***p < 0.001; ns, no significant difference.

To investigate the roles of the projections in regulating affective aspects of chronic pain, we subjected the mice to the CPP test, pairing one chamber with optogenetic stimulation (Fig. 5a). The mice with activation of the PTN or MD projections exhibited a significant increase in the time spent in the light-paired chamber as well as a higher CPP score (Fig. 5e, f), indicating the alleviated negative affects of pain in these mice. Activation of the VM projection, however, had no significant effects on the chamber preference (Fig. 5g). These results indicate that both the mPFC^Foxp2→PTN and mPFC^Foxp2→MD circuits alleviate the affective aspects of chronic inflammatory pain, while the mPFC^Foxp2→VM circuit is not involved.

Collectively, these results demonstrate that while the mPFC^Foxp2→PTN circuit is involved in regulating nociceptive sensitivity, coping behaviors and negative affects of pain, the mPFC^Foxp2→MD circuit modulates the negative pain affects and the mPFC^Foxp2→VM circuit regulates coping behaviors, respectively.

The mPFC Foxp2⁺ neurons receive antinociceptive cholinergic inputs from HDB

Having demonstrated the functional deactivation of the mPFC Foxp2⁺ neurons in pain and their pain-relieving effects through projections to distinct thalamic nuclei, we next explored the circuit mechanism regulating the activity of the mPFC Foxp2⁺ neurons. Previous studies have shown that mPFC receives intense cholinergic innervation from the basal forebrain and is strongly modulated by acetylcholine (ACh) ^{26, 44–46}. Notably, cholinergic modulation of the mPFC is severely impaired in neuropathic pain, while activating the cholinergic projection from the nucleus basalis magnocellularis to the mPFC induces antinociceptive effects^{47, 48}. These observations prompted us to investigate whether the mPFC Foxp2⁺ neurons are modulated by cholinergic innervation.

To determine whether the mPFC Foxp2⁺ neurons receive direct monosynaptic cholinergic inputs, rabies virus (RV)-based monosynaptic retrograde tracing was performed (Fig. 6a). Cre-dependent AAV helpers encoding TVA-mCherry and rabies G protein were injected into the PL mPFC of Foxp2-Cre mice, followed by EnvA-ΔG rabies virus expressing EGFP injected into the same region 4 weeks later. One week after the RV injection, the virus expression and starter cells were confirmed in the mPFC (Fig. 6a). Brain-wide analysis revealed RV-EGFP labeled neurons in several brain regions, including the basal forebrain, insular cortex, thalamus, ventral hippocampus and basolateral amygdala, while fewer labeled neurons were also observed in hypothalamus, dorsal raphe nucleus and parabrachial nucleus (Fig. 6b and Extended Data Fig. 5a, b). Notably, a substantial number of neurons were labeled with RV-EGFP in basal forebrain, particularly the HDB, with the RV-EGFP labeled neurons distributed from anterior to posterior part of HDB (Fig. 6b). Further immunostaining showed that most RV-EGFP labeled neurons HDB also express choline acetyltransferase (ChAT), a marker for cholinergic neurons (Fig. 6c), indicating the direct monosynaptic cholinergic inputs from HDB to the mPFC Foxp2⁺ neurons. These results suggest that the mPFC Foxp2⁺ neurons receive monosynaptic inputs from the HDB cholinergic neurons.

Figure 6. The mPFC Foxp2⁺ neurons receive cholinergic innervation from HDB which is antinociceptive.

a, Monosynaptic retrograde rabies tracing of the PL mPFC Foxp2⁺ neurons. The timeline and diagram of virus injection (left). Representative image showing the virus expression in the injection site and magnified image of the indicated region (right). Scale bars: 200 μm; inserted box, 20 μm. b, Representative images showing the RV-EGFP labeled cells from the anterior to posterior HDB (left), and quantification of the labeled cells in HDB (right) along the Anterior-Posterior axis. Scale bars: 200 μm. c, Immunostaining of ChAT in HDB showing colocalization with RV-EGFP labeling (left) and an enlarged image of the indicated region (middle). Quantification of the ChAT⁺ cells among the GFP⁺ cells (right). Scale bars: left, 200 μm; right, 20 μm. d, Diagram of the surgery of ChAT-Cre mice (left), and representative images showing the expression of the virus in HDB and fiber implantation in the PL mPFC (middle) as well as a magnified image of the indicated region (right). Scale bars, middle, 200 μm; right, 20 μm. e, Paw withdrawal threshold (left) and latency (right) of the mice expressing EYFP or ChR2 with (on) or without (off) light stimulation assessed in VFT and HPT, respectively. f, Diagram of CFA injection to induce chronic pain (left), and the paw withdrawal threshold (middle) and latency (right) of the mice with (on) or without (off) light stimulation after CFA treatment.

Data are represented as mean ± SEM. Each dot indicated one mouse. *p < 0.05, **p < 0.01, ***p < 0.001; ns, no significant difference.

To examine the functions of the HDB inputs to the mPFC in pain, Cre-dependent AAVs encoding EYFP or ChR2 were injected into the HDB of the ChAT-Cre mice and an optical fiber was implanted above the PL mPFC region to enable specific activation of the HDB^ChAT→mPFC circuit (Fig. 6d). The virus expression as well as the fiber implantation were confirmed by post hoc histological examination (Fig. 6d). Three weeks after the surgery, VFT and HPT were performed and a significant increase in the VFT was observed in response to optogenetic activation of the circuit (Fig. 6e), indicating reduced mechanical sensitivity of the mice. However, no significant effect was observed in the HPT (Fig. 6e), indicating that the baseline thermal sensitivity is unaffected. To further evaluate the roles of this circuit in chronic pain, the mice were treated with CFA to induce chronic inflammatory pain. VFT and HPT analyses indicated that optogenetic activation of the HDB cholinergic terminals in mPFC significantly alleviated both the mechanical and thermal hypersensitivity in chronic pain, as indicated by increased paw withdrawal threshold and latency (Fig. 6f). Together, these results demonstrate that the mPFC Foxp2⁺ neurons receive direct monosynaptic cholinergic innervation from the HDB, which regulates the nociceptive sensitivity in mice.

The mPFC Foxp2⁺ neurons highly express α4β2 nAChR which facilitates pain relief

Given the critical roles of the cholinergic signaling in pain processing and the innervation of mPFC Foxp2⁺ neurons by cholinergic inputs, we next sought to identify the specific acetylcholine receptor (AChR) responsible for modulating the mPFC Foxp2⁺ neurons. In the cerebral cortex, there are mainly two types of nicotinic acetylcholine receptors (nAChR), the α7 and the α4β2 nAChRs, and two types of muscarinic acetylcholine receptors (mAChR), the M1 and M2 receptors⁴⁶. To this end, we performed smFISH analyses for these receptors as well as Foxp2 and Oprk1. The results showed that over 95% of the Foxp2⁺ cells express Chrna4, the gene encoding the α4 subunit of α4β2 nAChR, while very few express Chrna7 (Fig. 7a). Further smFISH analysis of Chrna4 and Chrnb2, which encodes the β2 subunit, revealed that around 93% of the Foxp2⁺ cells are positive for both (Extended Data Fig. 6a), while less than 10% of Oprk1⁺ cells express both (Fig. 7b and Extended Data Fig. 6b), indicating the predominant expression of α4β2 nAChR in the Foxp2⁺ neurons. smFISH analysis of mAChRs showed that Chrm1, encoding the M1 receptor, was widely expressed across the mPFC, with expression in nearly all Foxp2⁺ and Oprk1⁺ cells in L6 (Fig. 7c and Extended Data Fig. 6c), while minimal expression of Chrm2 in Foxp2⁺ cells were observed (Fig. 7d and Extended Data Fig. 6d). Notably, quantitative analysis revealed the selective expression of the α4β2 nAChRs in Foxp2⁺ cells in mPFC L6, with nearly 80% of α4β2 nAChR-expressing cells being Foxp2⁺ and only a minority is Oprk1⁺ (Fig. 7e). Collectively, these results indicate that while α7 nAChR and M2 expressions are relatively low and M1 is broadly expressed in mPFC, the α4β2 nAChR is highly expressed and enriched in the L6 Foxp2⁺ neurons of mPFC (Fig. 7f).

Figure 7. Cell-type specific expression pattern of acetylcholine receptors in mPFC, and the α4β2 nAChR relieves pain through the Foxp2⁺ neurons in mPFC.

a, Left: smFISH for Foxp2 (red), Chrna4 (green) and Chrna7 (white) in the mPFC and the enlarged image of the indicated region. Right: quantification of Chrna4⁺ and Chrna7⁺ cells among the Foxp2⁺ cells. Scale bars: 200 μm; inserted box, 20 μm. b, Same as a but for Oprk1 (green), Chrna4 (red) and Chrna7 (white). c, Same as a but for Foxp2 (red) and Chrm1 (green). d, Same as a but for Foxp2 (red) and Chrm2 (green). e, Percentages of the Foxp2⁺ cells (left) and the Oprk1⁺ cells (right) among the cells expressing different types of acetylcholine receptor. f, Summary of the expression pattern of acetylcholine receptors in the mPFC. g, Diagram (left) and validation (right) of the cannula implantation into the PL mPFC. Scale bar: 200 μm. h, Paw withdrawal threshold (left) and latency (right) of the mice assessed in VFT and HPT before and after the intracranial injection of ABT-594 (5 pmol/site) into the mPFC. i, Diagram of CFA injection (left) and paw withdrawal threshold (middle) and latency (right) of the mice assessed in VFT and HPT in response to intracranial injection of ABT-594. j, Diagram (left) and validation (right) of the virus injection and cannula implantation. Scale bar: 200 μm. k, Paw withdrawal threshold in VFT (left) and latency in HPT (right) of the mice in response to ABT-594 treatment and CNO treatment.

Data are represented as mean ± SEM. Each dot indicated one mouse. *p < 0.05, **p < 0.01, ***p < 0.001; ns, no significant difference.

Given the critical roles of the mPFC Foxp2⁺ neurons in pain processing and the strong effects of ACh in modulating neuronal activity in mPFC⁴⁶, we next asked whether the α4β2 nAChR in mPFC regulates pain. To this end, we bilaterally implanted guided cannula for intracranial drug infusion into the mPFC of WT mice, confirmed by post hoc histological examination (Fig. 7g). We first infused artificial cerebrospinal fluid (ACSF) or ABT-594, a specific α4β2 nAChR angonist²⁷, at different dosages (2.5, 5, and 25 pmol/site) into the mPFC and performed VFT to measure the mechanical sensitivity of the mice. We found that ABT-594 at 5 or 25 pmol per site significantly increase the paw withdrawal threshold in VFT compared to both baseline and the ACSF control (Fig. 7h and Extended Data Fig. 7a), indicating reduced mechanical sensitivity by activating the α4β2 nAChR. Since the antinociceptive effect is sufficiently elicited by the treatment of ABT-594 at the dosage of 5 pmol per site, this dosage is used in all the subsequent experiments. Similar antinociceptive effects were observed in the hot plate test (HPT), with increased paw withdrawal latency following ABT-594 treatment (Fig. 7h). To test whether the α4β2 nAChR is also involved in regulating chronic pain, we treated the mice with CFA to induce chronic inflammatory pain. We found that ABT-594 treatment significantly increased the paw withdrawal threshold in VFT and latency in HPT (Fig. 7i), indicating the treatment alleviated mechanical and thermal hypersensitivity in chronic inflammatory pain. However, CPP test revealed no significant effect of ABT-594 on the affective components of pain, although a trend of increased CPP score was observed (Extended Data Fig. 7b), indicating that ABT-594 treatment is not sufficient to relieve the affective components of pain.

The high levels of α4β2 nAChR expression in the mPFC Foxp2⁺ neurons prompted us to ask whether the antinociceptive effects of ABT-594 depend on activation of the mPFC Foxp2⁺ neurons. To this end, Cre-dependent AAVs encoding mCherry or hM4Di were bilaterally injected into the PL mPFC of the Foxp2-Cre mice, followed by guided cannula implantation for intracranial drug delivery (Fig. 7j). The virus expression and the cannula implantation were confirmed by post hoc histological examination (Fig. 7j). Three weeks after the surgery, VFT and HPT were performed. In both the mCherry and hM4Di groups, intracranial ABT-594 delivery significantly increased the paw withdrawal threshold in VFT and latency in HPT (Fig. 7k). Notably, systemic administration of CNO, which selectively inhibits hM4Di-expressing mPFC Foxp2⁺ neurons, abolished these effects in the hM4Di group, while no significant changes were observed in the mCherry group (Fig. 7k). These results indicate that the antinociceptive effects of α4β2 nAChR activation are mediated by the mPFC Foxp2⁺ neuron activity.

Collectively, these results demonstrate that the α4β2 nAChR is highly expressed in the mPFC Foxp2⁺ neurons and activation of this receptor is antinociceptive, which depends on the activation of the mPFC Foxp2⁺ neurons.

Discussion

The present study reveals how the mPFC modulates different dimensions of pain through different thalamic projections and uncovers the critical roles of the cholinergic signaling in mPFC in pain regulation (Fig. 8). We identified Foxp2 as a marker for the thalamus-projecting neurons in mPFC by neural tracing and smFISH (Fig. 1). In vivo recordings showed that these neurons are deactivated in both acute and chronic pain (Fig. 2). Combined cell-type specific and circuit-specific manipulation studies revealed the critical roles of the mPFC Foxp2⁺ neurons in regulating sensory and affective pain through discrete thalamic projections (Fig. 3–5). We also showed the cholinergic innervation from HDB and the expression of α4β2 nAChR, which is antinociceptive, in the mPFC Foxp2⁺ neurons, underscoring the roles of the HDB^ChAT→mPFC^Foxp2→thalamus circuit in pain regulation (Fig. 6, 7).

Figure 8. Graphic summary of the HDB^ChAT→mPFC^Foxp2→thalamus connections in regulating different aspects of pain.

In the mPFC, the major outputs to thalamus originate from the corticothalamic neurons, which are specifically marked by the expression of Foxp2. These neurons are drastically deactivated in response to various acute noxious stimulation and in chronic inflammatory pain, and activation of these neurons regulates different aspects of pain through segregated thalamic projections. The mPFC Foxp2⁺ neurons project to PTN are involved in relieving both the sensory and affective aspects of pain, while the projection to MD relieves affective pain and the one to VM regulates coping behaviors, respectively. The mPFC Foxp2⁺ neurons receive cholinergic innervation from the HDB and highly express the α4β2 nAChR, which relieves pain by activating the Foxp2⁺ neurons in mPFC.

The cerebral cortex is organized into layers with intermingled distribution of projection neurons that target diverse brain regions³⁷. While our understanding of cortical circuitry has predominantly emerged from sensory cortices^{19, 49}, recent studies have revealed some distinct features of cellular organization and connections of the mPFC ^{19, 35, 36, 50}. Notably, we found that Ntsr1, a widely used marker for L6 CT neurons in sensory cortex and motor cortex^{22, 33, 34}, is broadly expressed across multiple layers in mPFC. The expression of Foxp2 has been well described in developing brain and the sensory cortex of adult brain^37–40, but it is not clear whether Foxp2 is specifically expressed in the L6 CT neurons in mPFC of adult mice. Our smFISH results showed that the Foxp2 expression is restricted in the glutamatergic neurons of the deep layers in mPFC, mostly in L6, and has few overlaps with the markers for other projection neurons such as IT, ET and NP neurons (Fig. 1c–g). Retrograde labeling further confirmed that Foxp2 specifically labels the mPFC neurons projecting to thalamus, but not those to cPFC, PAG, NAc or BLA (Fig. 1a, b and Extended Data Fig. 1c–e). Consistently, axon terminal labeling revealed that the mPFC Foxp2⁺ neurons almost exclusively project to thalamus with much weaker signals in medial striatum, and no detectable signal in NAc, BLA and PAG (Fig. 1h, i and Extended Data Fig. 1f–m). Interestingly, it is reported that a subset of Foxp2⁺ neurons in PFC project to VTA³⁹, but we did not observe axon-labeling signals in VTA (Extended Data Fig. 1m). One possible explanation is that the reported VTA-projecting Foxp2⁺ neurons are mainly located in the ventral part of PFC, especially the dorsal peduncular nucleus (DPn) which show different projection patterns from that of the mPFC⁵¹. Moreover, we observed limited number of neurons in L5 that express Foxp2 and found that these neurons do not express markers of IT, ET and NP neurons in mPFC (Fig. 1e, g), consistent with tracing results. Our study therefore provides multiple lines of evidence supporting Foxp2 as a specific marker for the thalamus-projecting neurons in mPFC.

Substantial evidence supports the involvement and critical roles of mPFC in pain regulation^{8, 10} and previous studies have uncovered the roles of the mPFC L5 ET neurons projecting to PAG, a key node for the descending modulatory pathway that dictates the processing of noxious information in the spinal cord^{16, 17}. However, although the functions of the mPFC IT neurons that project to BLA and striatum have also been indicated^{8, 9}, it is much less clear how the mPFC modulates the processing of noxious information in the brain, particularly for the roles of the output to thalamus³⁰. Thalamus is a key hub for integrating noxious information and regulates multiple aspects of pain⁵. We showed that the thalamus-projecting Foxp2-expressing neurons in mPFC are deactivated in response to various noxious stimulation and recover gradually after the removal of the stimuli (Fig. 2), indicating the involvement of these neurons in both acute and chronic pain. In contrast to the motor cortex to thalamus circuit that mainly involved in affective pain²², we found that persistent inactivation of the mPFC Foxp2⁺ neurons sensitizes nociception (Fig. 3a–c), while activation relieves both sensory and affective components of pain (Fig. 3d–h), demonstrating the crucial functions of the mPFC CT neurons in the endogenous pain relief pathway. Interestingly, short-term chemogenetic inhibition of the mPFC Foxp2⁺ neurons has few effects on pain-related behaviors (Extended Data Fig. 4a–d). Indeed, the organization of the mPFC-thalamus connection is consistent with the canonical model, in which the L6 CT neurons as “modulators” project to thalamus with low initial release probability, but are able to evoke robust firing in thalamus by strongly facilitating synapses^{52, 53}. Our results indicate that short-term inhibition of the mPFC Foxp2⁺ neurons may not induce adaptive alterations in the mPFC-thalamus connection and long-term adaptations in this pathway are responsible for the sensitized nociceptive responses induced by persistent inhibition of the mPFC Foxp2⁺ neurons.

Our study also delineates the circuit mechanism underlying the roles of the mPFC Foxp2⁺ neurons in regulating pain. The thalamus receives nociceptive information directly from the spinothalamic tract and is widely connected to other brain regions⁵⁴. It has been proposed that the lateral thalamic system mediates discriminative features of pain, while the medial system mediates the emotional aspects of pain⁵. Our results show PTN, MD and VM of thalamus as the major targets of the mPFC Foxp2⁺ neurons (Fig. 1h, i) and reveal the critical roles of mPFC in modulating the nociceptive processing by these thalamic projections (Fig. 4, 5). We found that while the projection of the mPFC Foxp2⁺ neurons to PTN regulates both sensory and affective pain, the projection to MD relieves affective pain and the one to VM regulates coping behaviors specifically (Fig. 4, 5). These findings support the idea that segregated circuit mechanisms are responsible for different dimensions of pain^{5, 20}. Our results reveal PTN, which is a relatively less studied region with the functions and circuits remained to be fully characterized, as a master hub in the medial thalamus for regulating both the sensory and affective aspects of pain. Future study is needed to further clarify the circuit mechanism underlying the critical roles of PTN in pain regulation. In summary, our study reveals the critical roles and the segregated circuit mechanisms of mPFC in governing the nociceptive processing in the thalamus.

The mPFC is broadly connected to a wide range of brain regions¹⁹. While the strengthened input from BLA to the inhibitory neurons in mPFC has been proposed to be a key mechanism underlying the functional deactivation of mPFC in chronic pain¹⁷, the contributions of other inputs are far from fully understood. The mPFC receives dense cholinergic innervation from the basal forebrain and expresses high levels of cholinergic receptors^{26, 46}. It is reported that the ACh receptors are expressed in a layer-specific manner, with some expression of α4β2 nAChR in the inhibitory neurons of L2/3 and much higher expression in the excitatory neurons in L5/6 of mPFC⁴⁴. Consistently, our results showed that the expression of α4β2 nAChR is enriched in deep layers of mPFC and more specifically in the L6 CT neurons, while the α7 nAChR is expressed in IT neurons (Fig. 7a–f and Extended Data Fig. 6). We further revealed that the cholinergic inputs to the Foxp2⁺ neurons are from the basal forebrain, particularly the HDB, by monosynaptic retrograde tracing (Fig. 6a–c and Extended Data Fig. 5). Our study thus revealed the cell-type specific expression patterns of the AChRs in mPFC as well as the cholinergic inputs to the mPFC L6 CT neurons.

Cholinergic signaling has been extensively studied in the context of pain and analgesia, with significant insights into its roles in the peripheral, spinal cord and descending modulation pathway^55–58. However, the functions and circuit mechanisms of cholinergic system in the brain in nociceptive processing are not fully understood⁴⁷. While cholinergic input is known to be essential for the mPFC functions such as attention and cognition, its role in pain regulation is less clear^{26, 45, 46}. Notably, cholinergic signaling in the mPFC is markedly impaired in neuropathic pain models⁴⁸, and activation of cholinergic projections from the nucleus basalis magnocellularis to the mPFC has been shown to produce antinociceptive effects⁴⁷. Our study revealed that the mPFC Foxp2⁺ neurons receive cholinergic inputs from the HDB, which regulates the sensory aspects of pain (Fig. 6d–f). We also found that targeting the α4β2 nAChR in mPFC with a specific agonist ABT-594, which is antinociceptive through systemic administration or brainstem injection²⁷, alleviates pain (Fig. 7g–i). Importantly, this antinociceptive effect requires the activity of the mPFC Foxp2⁺ neurons (Fig. 7j, k). Notably, desensitization is an important property of the nAChRs, and the receptors are desensitized in response to high concentration or affinity of agonist⁵⁹. In our study, we used a relatively low concentration of agonist that has been found to activate the neurons in brainstem²⁷. On the other hand, the L6 pyramidal neurons in cortex express Chrna5, which encodes an accessory subunit of the α5-containing α4β2 nAChR⁶⁰. This receptor complex could increase conductance and prolong inward currents in response to persistent nicotine application and is important for the tonic effects of ACh^{46, 61}. Thus, the α4β2 nAChR in Foxp2⁺ neurons could play an important role in maintaining the baseline neuronal activity and activating the endogenous pain relief pathway. While further studies are still needed to fully elucidate how the cholinergic inputs from HDB modulate the mPFC activity in chronic pain, our results support the crucial role of the cholinergic signaling in the mPFC Foxp2⁺ neurons in pain processing.

Besides the cholinergic inputs, we also revealed other monosynaptic inputs to the mPFC L6 CT neurons, including the IC and BLA (Extended Data Fig. 5), which are important for pain processing^{5, 62}. It is well-known that in chronic pain, BLA sends strengthened feedforward inhibition to the L5 ET neurons of mPFC that further project to the PAG to regulate pain¹⁷. However, the effects of the BLA inputs to the L6 CT neurons are unknown. Future studies will reveal the respective contributions of these specific inputs.

Collectively, our study uncovers the crucial roles of the mPFC in modulating sensory and affective aspects of pain through the discrete HDB^ChAT→mPFC^Foxp2→thalamic nuclei circuits and raises the possibility of managing chronic pain by targeting the mPFC L6 CT neurons.

Materials and Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Animals

All experiments were conducted in accordance with the National Institutes of Health Guide for Care and Use of Laboratory animals and approved by the Institutional Animal Care and Use Committee (IACUC) of Boston Children’s Hospital and Harvard Medical School. The wild-type C57BL/6 mice (Jax, 000664), B6.Cg-Foxp2^{tm1.1(cre)Rpa}/J mice (Jax, 030541) and B6.129S6-Chat^tm2(cre)lowl/J mice (Jax, 006410) were purchased from Jackson Laboratory. Mice were housed with littermates (3–5 mice/cage) in 12-hr light/dark cycle (light on from 7:00 to 19:00) with food and water ad libitum unless otherwise specified. Ambient temperature (23–25°C) and humidity (55–62%) were automatically controlled. For the present study, 2–4-month-old male mice were used and all mice were randomly assigned into different groups.

METHODS DETAILS

Single molecule Fluorescence in situ hybridization (smFISH) and immunofluorescence (IF) staining

Mice were euthanized by inhalation of CO₂ and transcardially perfused with PBS and 4% paraformaldehyde (PFA) before the brains were collected and fixed in 4% PFA at 4 °C overnight, followed by dehydration in 30% sucrose solution for 3 days. The brains were then frozen in Optimal Cutting Temperature embedding media (Sakura, #4583) and coronal sections (14 μm for smFISH and 30 μm for IF) were cut using a cryostat (Leica, #CM3050S). The smFISH was performed using a RNAscope Fluorescent Multiplex Assay kit (Advanced Cell Diagnostics, ACD) following the manufacturer’s instruction. Briefly, the brain slices were fixed with fixed with 4% PFA and washed with PBS, then dehydrated with ethanol and treated with hydrogen peroxide. The samples were then treated with the target retrieval reagents and protease for 10 minutes, after which the samples were hybridized with the probes in 40°C for 2 hours. Then the signals were amplified using the reagents included in the kit, followed by the labeling of relative channels with the TSA fluorophores. The samples were treated with the included blocking buffer before proceeding to DAPI staining and imaging. For the combined smFISH with CTB retrograde tracing, the brain samples were prepared from the mice with CTB injected into the relative regions. The samples were treated following the manufacturer’s instruction as described above till the labeling with TSA fluorophore without being treated with the blocking buffer. The samples were then directing stained with DAPI and the images were captured immediately. Probes (ACD) used were as followed: Foxp2 (Cat. No. 428791-C1; Cat. No. 428791-C2); Slc17a7 (Cat. No. 416631-C3); Gad1 (Cat. No. 400951-C2); Ctgf (Cat. No. 314541-C1); Etv1 (Cat. No. 557891-C3); Oprk1 (Cat. No. 316111-C1; Cat. No. 316111-C2); Tshz2 (Cat. No. 431061-C3); Pou3f1 (Cat. No. 436421-C2); Chrna4 (Cat. No. 429871-C3); Chrna7 (Cat. No. 465161-C2); Chrnb2 (Cat. No. 449231-C2); Chrm1 (Cat. No. 495291-C2); Chrm2 (Cat. No. 495311-C1).

For IF staining, the brain slices were washed in PBS twice for 10 min, followed by incubation in blocking buffer (5% goat serum, 5% BSA and 0.1% Triton X-100 in PBS) for 40 min at room temperature. The samples were then incubated at 4°C for 24 h with primary antibody diluted in PBS containing 1% BSA. After the incubation, the samples were washed three times with washing buffer (0.1% Tween-20 in PBS) and incubated with the Alexa Fluor conjugated secondary antibodies for 2 h at room temperature. The sections were washed for three times, mounted and imaged using a Zeiss LSM800 confocal microscope or Keyence BZ-X810 microscope. Antibodies used were as followed: rabbit anti-cFos (1:1000, Synaptic Systems, #226003), rabbit anti-ChAT (1:1000, Milliopore, #ZRB1012–25UL), Donkey anti-rabbit Alexa Fluor 488 (Thermo Scientific, #A21206), Donkey anti-rabbit Alexa Fluor 568 (Thermo Scientific, # A10042).

Stereotaxic surgeries

All stereotaxic surgeries were performed using a small-animal stereotaxic instrument (David Kopf Instruments, model 940) under general anesthesia by isoflurane (0.8 l min⁻¹; isoflurane concentration 1.5%) in oxygen. A feedback heater was used to maintain the body temperature of the mice. The eyes of the mice were kept moist using ophthalmic ointment throughout the surgery. A small craniotomy above the target brain region was performed using a dental drill. The viruses and CTB were injected into the target regions using a glass micropipette (10–20 μm in diameter at the tip) that was connected to a nanoliter injector (Nanoject III, Drummond Scientific, #3–200-207). For each site, 120 nl of viruses were injected at a flow rate of 1 nl s⁻¹ unless otherwise specified, allowing an additional 5 min for viral particles to diffuse before the pipette was slowly withdrawn. After the withdrawal, the wound was sutured and the mice were allowed to recover in a warm blanket before being transferred to the housing cages. The mice were treated with meloxicam (Covetrus, Cat. No. 049756) subcutaneously daily for 3 days and monitored for another 2 days.

The coordinates of viral injection and implantation sites are based on previous literature and The Mouse Brain in Stereotaxic Coordinates (third edition) relative to the bregma (anterior-posterior, mediolateral, dorsoventral axis in mm): PL (1.94, 0.45, −2.50), PTN (−0.46, 0.30, −3.70), MD (−1.22, 0.50, −3.20), VM (−1.34, 0.75, −4.20), PAG (−4.75, 0.55, −2.70), NAc (+1.20, 0.60, 4.50), BLA (−1.34, 2.90, −4.50), HDB (−0.10, 1.45, −5.50).

Chemogenetic and optogenetic manipulations

For chemogenetic manipulations of the PL mPFC Foxp2⁺ neurons, AAV5-hSyn-DIO-mCherry, AAV5-hSyn-DIO-hM4D(Gi)-mCherry or AAV5-hSyn-DIO-hM3D(Gq)-mCherry was bilaterally injected into the PL mPFC of Foxp2-Cre mice. The behavioral assays were performed 3 weeks later, and the mice were intraperitoneally injected with CNO (Cayman, Cat# 16882) at 1 mg/kg (for activation) or 5 mg/kg (for inactivation) in saline 20 min before each behavioral test as previously described^{63, 64}. For persistent chemogenetic inactivation, the mice were intraperitoneally injected with CNO twice per day for 5 consecutive days. For persistent inactivation of the mPFC Foxp2⁺ neurons, rAAV2/9-hSyn-DIO-EYFP or rAAV2/9-hSyn-DIO-TeNT-EYFP was bilaterally injected into the mPFC of Foxp2-Cre mice and behavioral assays were performed 5 weeks later.

For optogenetic activation of the terminals of PL mPFC Foxp2⁺ neurons, rAAV2/9-EF1a-DIO-EYFP or rAAV2/9-EF1a-DIO-ChR2-EYFP was bilaterally injected into the PL mPFC of Foxp2-Cre mice, followed by the bilateral implantation of optical cannulas (diameter: 200 μm; N.A., 0.37; length, 4.0 mm; Inper Inc., China) 100 μm above the PTN/MD/VM region and secured with dental cement (Parkell, no. S380). For optogenetic activation of the cholinergic inputs to PL mPFC, the same viruses were injected into the HDB region of ChAT-Cre mice, and the optical cannulas were implanted above the PL mPFC. For all optogenetic manipulation, the mice were connected to the optical fiber for acclimation for 3 days before the behavioral tests and the light pulses (473 nm, 1.5 mW at the tip of the fiber, 20 Hz, 5-ms pulse) at 10-s ON/OFF cycle were generated by a 473-nm laser (OEM laser/OptoEngine) and used throughout the behavioral tests controlled by a waveform generator (Keysight).

Neuronal tracing

For CTB retrograde tracing, 120 nl of CTB-555 was unilaterally injected into the MD, mPFC, PAG, NAc or amygdala in individual mice, respectively. The brain tissues were collected 1 week later for smFISH. For specific axon labeling of the mPFC Foxp2⁺ neurons, pAAV1-hSyn-DIO-mGFP-Synaptophysin-mRuby was unilaterally injected into the mPFC of Foxp2-Cre mice. For specific axon labeling of the HDB cholinergic neurons, the same virus together with rAAV2/9-ChAT-Cre-EGFP was injected into the HDB region of WT mice. The brain tissues were collected 3 weeks after the surgery.

For monosynaptic retrograde rabies tracing, 150 nl mixture of AAV helpers (rAAV2/9-EF1a-DIO-mCherry-TVA and rAAV2/9-EF1a-DIO-RVG) was unilaterally injected into the PL mPFC of the Foxp2-Cre mice. Four weeks later, the same mice were injected with 250 nl of rabies virus (RV-ENVA-ΔG-EGFP) to the same location. The brain tissues were collected 7 days later for imaging and examination.

Intracranial drug delivery

For intracranial drug delivery into the PL mPFC, the double cannula (Guide cannula: M3.5, C.C 1.0mm, C=2.5mm; Injector: G1=0.5mm; Dummy cannula: G2=0, mates with M3.5; RWD Life Science) with dust caps targeting both sides were directly implanted 0.5 mm above the PL mPFC and were secured with dental cement. Two weeks later, an injector (G2=0, mates with M3.5; RWD Life Science) filled with mineral oil and ABT-594 (Cayman Chemical, Cat. No. 22822) was connected to a pump (RWD Life Science) to slowly infuse the drug into both sides of mPFC (200 nl/site, 5 min) and 5 min were allowed for drug diffusion before the injector was slowly removed. The behavioral tests were performed another 5 min later. For intracranial drug delivery combined with chemogenetic inhibition, AAV5-hSyn-DIO-mCherry or AAV5-hSyn-DIO-hM4D(Gi)-mCherry was bilaterally injected into the mPFC, followed by the implantation of the double cannula. Three weeks later, the behavioral tests were performed 20 min after CNO injection (i.p.) and 5 min after ABT594 infusion (intracranially).

In vivo miniscopic calcium imaging and data processing

To probe the activity of the PL mPFC Foxp2⁺ neurons in response to pain, in vivo miniscopic calcium imaging was performed as previously described^{64, 65}. Firstly, AAV1-hSyn-DIO-jGCaMP7s was unilaterally injected into the PL mPFC of the Foxp2-Cre mice. Two weeks later, the mice were anesthetized and fixed on a stereotaxic frame. The skull above the target region was carefully removed using a dental drilling and the brain tissue above the target region was aspirated using a 27-gauge needle with a blunt tip. Then an integrated GRIN lens (diameter 1.0 mm, length 4.0 mm; Inscopix) connected to a data acquisition system for monitoring the calcium signal was slowly lowered (~100 μm/min) down to 300 μm above the target region. The GRIN lens was further inserted to obtain an optimal field of view to image calcium signals and then the GRIN lens was secured to the mice’s skull with dental cement attached to a baseplate cover for protection. Saline was continuously added to avoid bleeding or drying of the tissue during the surgery. After the surgery, the mice were individually housed for 2 weeks before being used in imaging experiments. Prior to calcium imaging recording, the mice were connected to the microscope through the head-fixed baseplate and acclimated to the handling of the experimenter and the testing chamber for 30 min each day for 3 days. On the day of data acquisition, the mice were connected to the microscope and allowed to move freely in the testing chamber. After 10 min of acclimation, the calcium images were obtained for 20–30 min using nVoke system (Inscopix) at 20 Hz with 0.6–1.2 mW of LED power and 1.5 of gain. For individual mice, the imaging parameters were kept consistent across days.

For tail pinch stimulation, after acclimatation the mice were placed in the testing chamber without stimulation for 2 min and then an alligator clip was applied to the tail for 1 min, after which the clip was removed and the mice were allowed to recover for 9 min, and the calcium signals were recorded during the entire process. For formalin stimulation, after acclimatation the mice were recorded in the testing chamber without stimulation for 5 min, and then 20 μl of formalin was injected to the plantar of the mice, after which the mice were recorded for another 30 min in the chamber.

To detect the spontaneous activity of the neurons, the mice were acclimated to the microscope and testing environment with fixed brightness, temperature and noise level one hour before the recordings. To avoid the potential effects of light-dark cycle on the neuronal activity, all the recordings were performed in the afternoon and at the fixed time points for each individual animal. After the acclimation, the calcium activities were recorded for 15 minutes for the baseline. Then the mice were subjected to saline or CFA injection under isoflurane anesthesia. One or seven days after the injection, the calcium activities were recorded again at the same time point after acclimation without any additional stimulation.

For data processing, the calcium traces were extracted and analyzed using the Inscopix Data Processing Software (version 1.9, Inscopix). The raw images were cropped, temporally downsampled (2x), spatially downsampled (2x), and further spatially filtered as well as motion-corrected using the default settings. The signals of putative individual cells were identified using an extended constrained non-negative matrix factorization (CNMF-E) algorithm and were normalized to ΔF over noise (settings: cell diameter: 8 pixels; minimum pixel correlation: 0.85; minimum peak-to-noise ratio: 15). The identified individual cells were then manually checked to be included based on their anatomy (size, shape, location), signal-to-noise ratio as well as overlap in spatial location and signal with other cells. The calcium signal traces were further deconvoluted using OASIS (online active set method to infer spikes) methods to detect the potential calcium transients with the default noise ranges and spike signal-to-noise ratio of 8.0. Across the different days of recording for chronic pain, we first analyzed the calcium images to identify individual neurons in each recording and compared the spontaneous calcium activities of all the identified neurons among different days. Then the identified individual neurons were longitudinally registered with minimum correlation of 0.9 and the matched neurons were further manually checked to be the same neurons across different days based on the locations and other visible landmarks (such as blood vessels).

To identify the neurons with significant change in the firing rate in response to tail pinch stimulation, the calcium traces were extracted and deconvoluted using OASIS to identify the calcium transients. We pooled together the identified transients of 2 min before and 1 min during the tail pinch stimulation, and used a permutation method to create 10,000 shuffled distributions of the transients. Then the shuffled transient change was calculated by subtracting shuffled 0–2 min from that of the 3^rd min. The neurons with actual transient change ranked higher than 99.5^th percentile of the shuffled transient change were classified as activity-increased neurons, and those ranked lower than 0.5^th percentile were classified as activity-decreased neurons. For formalin test, the calcium signals of −5 min to 0 min and that of 0 min to 5 min were compared. For CFA-induced chronic inflammatory pain, the calcium signals of the baseline, CFA D1 and CFA D7 were compared using the same method.

Behavioral assays

von Frey test.

The mechanical sensory threshold of the mice was measured with a series of von Frey filaments (ranging from 0.04–2.0 g, Stoelting) using the Up-Down method as previously described⁶⁶. Briefly, the mice were placed in a plastic chamber with a metal mesh floor and allowed to move freely. After 20 min of acclimation of the mice, the von Frey filaments were applied to the mid-plantar surface of the hindpaw through the mesh floor and kept for 2 s, or until there was a positive response (paw withdrawal, flinching, licking or shaking). To avoid the potential false positive responses caused by movement, the measurement was conducted again if a positive response was followed by other voluntary behaviors such as locomotion, exploratory or grooming behaviors.

Hot plate test.

The hot plate test was used to measure thermal nociception of the mice. After being acclimated in the testing room for 20 min, the mice were placed on a hot plate (Ugo Basile) set at 52 °C and the latencies to flinching or licking of hindpaws were measured. All animals were tested for 3 times with a minimum interval of 30 min. To avoid tissue injury, a cutoff latency was set at 30 s for the test.

Formalin test.

Formalin test was performed as previously described⁴². Briefly, 20 μl of 2% formalin (Sigma-Aldrich, HT5011–1CS) was injected subcutaneously into the plantar surface of one hind paw of the mice with a 30-gauge syringe as quickly as possible. The licking behaviors were measured for 45 min starting the injection of formalin and the licking duration were calculated in 5-min bins. The licking duration of the mice during phase 1 (defined as 0 – 5 min after formalin injection) and phase 2 (defined as 15 – 45 min after formalin injection) were analyzed.

Complete Freund’s adjuvant (CFA) injection.

For CFA-induced inflammatory pain, 15 μl of CFA (Sigma, F5881–10ML) was subcutaneously injected into the plantar surface of the right hind paw of the mice under brief isoflurane anesthesia.

Conditioned place preference.

The conditioned place preference test was used to measure the negative affection of pain and the effects of pain relief as previously described⁶⁴. The testing apparatus consists of an infrared photobeam detectors for tracking the mice trajectories and two compartments, one of which is with rod style floor and the other one is with grid style floor (Med Associates). For chemogenetic manipulations, individual mice were placed in the chamber and allowed to freely explore the entire apparatus for 30 min (pretest) on day 1. On day 2–4, mice injected with saline were placed in the compartment with grid floor (unconditioned compartment) for 30 min in the morning, while mice injected with CNO were placed in the opposite compartment with rod floor (conditioned compartment) of the same apparatus for 30 min in the afternoon (with a minimum interval of 6 h between the two conditioning). On day 5, like day 1, the mice were placed in the chamber and allowed to freely explore the entire apparatus for 30 min (posttest). For optogenetic activation, the pretest and posttest on the first and last day were similarly performed. On day 2–4, the mice were connected with optical fiber without light stimulation and placed in the compartment with grid floor for 30 min in the morning, and in the afternoon, the mice were connected with the optical fiber with light stimulation as described above, and placed in the compartment with rod floor for 30 min. For ABT-594 treatment, the pretest and posttest on the first and last day were similarly performed. On day 2–4, the mice were intracranially injected with ACSF and placed in the compartment with grid floor for 30 min in the morning, and in the afternoon, the mice were injected with ABT-594 and placed in another compartment for 30 min. To calculate the CPP score, the time that mice spent in the conditioned compartment during pretest was subtracted from the one during posttest.

Statistics

The mice were randomly assigned to the groups and the investigators were blind to the allocation of the groups or the outcome assessment. No statistical method was used to predetermine the sample size, but our sample sizes are similar to those reported in previous publications. Mice that, after histological inspection, had the location of the viral injection (reporter protein), cannula implantation, or the optic fiber(s) outside the area of interest were excluded. The data are presented as means ± SEM. Statistical analyses were performed using GraphPad Prism (version 8.0). Student’s t-test and one-way or two-way RM ANOVA test with Tukey’s or Sidak’s post hoc multiple comparisons were applied to determine statistical differences. Statistical significance was set at *P < 0.05, **P < 0.01, ***P < 0.001.

Extended Data

Extended Data Fig. 1. Foxp2 specifically marks the L6 corticothalamic neurons in the mPFC.

a, smFISH for Ntsr1 (red) in the mPFC. Scale bar: 200 μm. b, UMAP showing the results of an integrated analysis of single-cell RNA sequencing and spatial transcriptomics with neuron types and layer information (left). Predicted expression patterns of Foxp2 (middle left) and Oprk1 (middle right). L, layer; CT, corticothalamic neurons; ET, extratelencephalic neurons; IT, intratelencephalic neurons; NP, near-projecting neurons. Predicted expression levels of Foxp2 and Oprk1 in the mPFC neuronal subtypes with layer information (right). c, Diagram of the injections of CTB into the contralateral mPFC (cPFC) of wildtype mice (left). Representative images of the mPFC in mice with CTB injected into cPFC (middle, left); the enlarged images of the indicated regions (middle, top); injection site of the CTB (middle, bottom). Percentage of the Foxp2⁺ or Foxp2⁻ cells of all CTB⁺ neurons (top right); Percentage of the CTB⁺ or CTB⁻ cells of all Foxp2⁺ neurons (bottom right). Scale bars: middle left, 200 μm; middle top, 20 μm; middle bottom, 500 μm. d, e, Same as c but for CTB injected into the amygdala (d) or NAc (e). f, Representative image showing the virus expression in the mPFC of Foxp2-Cre mice as indicated in the Figure 1H. g-m, The mGFP and synaptophysin-fused mRuby signals labeling the axon puncta of the mPFC Foxp2⁺ neurons in the known targets of the mPFC, including striatum (g), NAc (h), amygdala (i), PAG (j), HPC (k), HY (l) and VTA (m).

3V, third ventricle; AQ, cerebral aqueduct; BMA, basomedial amygdala; CeA, central amygdala; DG, dentate gyrus; Hip, hippocampus; HY, hypothalamus; IPN, interpeduncular nucleus; LV, lateral ventricle; MRN, midbrain reticular nucleus; NAc, nucleus accumbens; SNr, substantia nigra, reticular part; VTA, ventral tegmental area.

Extended Data Fig. 2. Expression pattern of Foxp2 in the mPFC along the anterior-posterior axis.

a, smFISH for Foxp2 (red), Etv1 (white) and Ctgf (green) in the mPFC at 2.3 mm anterior to the bregma (left) and the percentages of the Foxp2⁺ neurons co-expressing Etv1 or Ctgf (right). Scale bar, 200 μm. b, c, Same as a but for the mPFC region 2.1 mm (b) or 1.7 mm (c) anterior to the bregma. d, Quantification of the normalized fluorescence signals of Etv1 (blue), Foxp2 (red) and Ctgf (green) in the mPFC from the pia to the white matter at the indicated sections of mPFC along the anterior-posterior axis.

Data are represented as mean ± SEM. Each dot indicated one mouse.

Extended Data Fig. 3. In vivo miniscopic calcium imaging of the mPFC Foxp2⁺ neurons.

a, Frequency of the calcium events of the identified neurons during baseline (BL) and during tail pinch stimulation. The numbers of the neurons with increased or decreased frequency are indicated. b, Amplitude of the calcium events of the identified neurons during BL and during tail pinch stimulation. Only the neurons that exhibited calcium events during both sessions were included (n = 34). c, Licking duration of the mice before and after formalin injection. d, Same as a but for the formalin stimulation comparing BL with phase 1 (left) or BL with phase 2 (right). e, Same as b but for the formalin stimulation comparing BL with phase 1 and phase 2. Only the neurons that exhibited calcium events during all sessions were included (n = 196). f, Representative field of view images of the mice during recording in baseline, 1 day and 7 days after CFA injection. Some neuronal contours in the same locations are identified in different days and longitudinally registered as the same individual neurons, which are labeled in the images

Data are represented as mean ± SEM. Each dot indicates an individual neuron. ns, no significant difference.

Extended Data Fig. 4. Effects of inactivating the mPFC Foxp2⁺ neurons on pain-related behaviors.

a, Diagram of the virus injection (left) and a representative image of hM4Di-mCherry expression in the PL mPFC of Foxp2-Cre mice (right). Scale bar: 100 μm. b, Paw withdrawal threshold and latency of the mice assessed by VFT (left) and HPT (right). c, Diagram of formalin test (left). Licking duration of the mice in response to formalin injection after CNO treatment and summarized licking duration in the phase 1 and phase 2 (right 3 panels). d, Diagram of the conditioned place avoidance (CPA) test (left). The time that mice (CFA treated) spent in the chamber paired with CNO treatment (middle) and CPA scores (right) of the mice in CPA test. e, Representative images of the cFos expression in the mPFC after CNO treatment in the mCherry-expressing, hM4Di-mCherry-expressing or hM3Dq-mCherry-expressing mice. Right: quantification of the cFos⁺ cells in the mPFC L6. Scale bar: 100 μm.

Data are represented as mean ± SEM. Each dot indicated one mouse. *p < 0.05, **p < 0.01, ns, no significant difference.

Extended Data Fig. 5. Inputs of the mPFC Foxp2⁺ neurons.

a, Monosynaptic retrograde rabies tracing of the PL mPFC Foxp2⁺ neurons (top left) and whole brain mapping. Representative images of brain regions with RV-EGFP labeled cells, including IC (top, 2^nd), TH (top, 3^rd), Hip (top, 4^th), BLA (bottom left), HY (bottom, 2^nd), PAG (bottom, 3^rd) and PBN (bottom, 4^th). Scale bars: 200 μm. b, Quantification of the retrogradely labeled cells in different brain regions.

Data are represented as mean ± SEM. Each dot indicated one mouse. 3V, third ventricle. Aq, cerebral aqueduct; BLA, basolateral amygdala; BMA, basomedial amygdalar nucleus; CeA, central amygdalar nucleus; CP, caudoputamen; CLA, claustrum; DR, dorsal raphe nucleus; HDB, horizonal diagonal band of Broca; HY, hypothalamus; IC, insula cortex; LA, lateral amygdalar nucleus; LC, locus ceruleus; LH, lateral hypothalamic area; LPO, lateral preoptic area; MA, magnocellular nucleus; MD, mediodorsal thalamus; OT, olfactory tube; PAG, periaqueductal gray; PBN, parabrachial nucleus; Pir, piriform area; Re, nucleus of reuniens; scp, superioi cerebellar pedunles; SI, substantial insominata; TH, thalamus; vHip, ventral hippocampus; VM, ventromedial thalamus; ZI, zona incerta.

Extended Data Fig. 6. Cell type-specific expression patterns of acetylcholine receptor in the mPFC.

a, Left, representative images of smFISH experiments in the mPFC for Foxp2 (red), Chrna4 (green) and Chrnb2 (white). An enlarged image of the indicated region is inserted in the bottom left box. Right: Quantification of Chrna4⁺Chrnb2⁺Foxp2⁺ cells among all the Foxp2⁺ cells. Scale bars: 200 μm; inserted box, 20 μm. b, Left: same as a but for Oprk1 (green), Chrna4 (red) and Chrnb2 (white). Right: Quantification of Chrna4⁺Chrnb2⁺Oprk1⁺ cells among all the Oprk1⁺ cells. c, Left: smFISH of Oprk1 (red) and Chrm1 (green) in the mPFC. Right: quantification of the Chrm1⁺Oprk1⁺ cells among all the Oprk1⁺ cells. Scale bars: 200 μm; inserted box, 20 μm. d, Same as c but for Oprk1 (red) and Chrm2 (green) in the mPFC.

Extended Data Fig. 7. Effects of activating the α4β2 nAChR in the mPFC.

a, Dosage test of ABT-594. Paw withdrawal threshold of the mice assessed in VFT before and after the intracranial injection of ABT-594 (2.5 pmol/site) (left) or ABT-594 (25 pmol/site) (right) into the mPFC compared to ACSF treatment. b, Schematic diagram for CPP test paired with ACSF or ABT-594 treatment (left) and the time in the chamber paired with ABT-594 treatment before and after the conditioning (middle) as well as the CPP scores of the mice (right).

Data are represented as mean ± SEM. Each dot indicated one mouse. *p < 0.05, **p < 0.01, ***p < 0.001; ns, no significant difference.

Data availability

All data generated or analyzed during this study are included in this article and its Supplementary Information files.