A glutamatergic circuit from the piriform cortex to the mediodorsal thalamus regulates trigeminal neuropathic pain and comorbid cognitive impairment in mice

Abstract

Background

Trigeminal neuropathic pain (TNP) is a chronic, debilitating pain disorder affecting the orofacial region. Patients often experience concomitant cognitive impairment, severely impacting their quality of life. The principal neural circuit mechanisms that link TNP to cognitive dysfunction remain unclear. The piriform cortex (PIR), primarily recognised for its role in olfaction, has not been systematically investigated in trigeminal neuropathic pain processing or pain-related cognitive dysfunction.

Methods

This study utilised pseudorabies virus-mediated multisynaptic retrograde tracing to map whole-brain anatomical connectivity associated with orofacial sensory pathways. c-Fos immunofluorescence staining and in vivo fibre-optic calcium signalling recordings were used to detect alterations in glutamatergic neuronal activity within the PIR of TNP model mice. Chemogenetic and optogenetic techniques were employed to manipulate PIR glutamatergic (PIR^Glu) neurons in a bidirectional manner, validating their functional roles in pain and cognitive regulation. Anterograde viral tracing was integrated with circuit-specific chemogenetic intervention to elucidate downstream regulatory pathways.

Results

This study identified the PIR as a previously unrecognized brain region anatomically associated with trigeminal neuropathic pain-related pathways, providing evidence for its potential involvement in TNP. PIR^Glu neurons in TNP model mice exhibited marked burst-like activation, with in vivo calcium activity intensity exhibiting considerable synchronisation with facial pain-sensitive behaviour. Chemogenetic inhibition of PIR^Glu neurons significantly increased pain thresholds and ameliorated cognitive deficits in TNP mice, whereas selective activation of PIR^Glu neurons in sham mice recapitulated the pain–cognition comorbidity phenotype, demonstrating the essential role of these neurons in mediating pain-related cognitive dysfunction. Viral tracing identified the mediodorsal thalamus (MD) as a critical downstream target of PIR^Glu neurons. Electrophysiological recordings further confirmed that PIR^Glu neurons establish direct excitatory monosynaptic connections with MD neurons, with fast glutamatergic transmission predominantly mediated by AMPA receptor-dependent signalling. Specific disruption of the PIR^Glu–MD circuit simultaneously alleviated pain and cognitive impairment in model mice, while inhibition of PIR^Glu neurons markedly reduced calcium activity in downstream MD neurons.

Conclusion

This study highlights the PIR^Glu–MD circuit as an important pathway contributing to TNP–cognitive impairment comorbidity. Furthermore, it augments our understanding of the non-olfactory physiological functions of the PIR, offering a novel potential target for investigating the mechanisms underlying the association of chronic pain with cognitive impairment.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s10194-026-02486-w.

Introduction

Trigeminal neuropathic pain (TNP) is a chronic and debilitating condition that affects the orofacial region, characterised by spontaneous burning pain, allodynia, and hyperalgesia [1–3] In addition to persistent somatic pain, TNP is often complicated by severe comorbidities, including anxiety, depression, and cognitive dysfunction, which collectively impair patients’ quality of life and significantly complicate clinical management [4–7]. Despite advancements in TNP research, the underlying neural circuit mechanisms—particularly the complex networks driving its diverse comorbidities—remain unclear.

TNP pathophysiology involves sensitisation within the peripheral and central nervous systems [8, 9]. Nociceptive signals are relayed from the trigeminal ganglion to the spinal trigeminal nucleus caudalis in the brainstem and subsequently projected to several higher brain regions, forming an intricate ‘pain matrix.’ Increasing evidence indicates that neuronal hyperactivity and neuroplasticity modifications in regions such as the anterior cingulate cortex (ACC) [10, 11], insular cortex (IC) [12–14], and primary somatosensory cortex (S1) [15, 16] are instrumental in pain perception and affective processing. However, functional alterations in these canonical pain-related brain areas insufficiently explain the complex cognitive impairments observed in chronic pain states. This knowledge gap suggests that other, underexplored cortical regions potentially play pivotal roles in developing TNP and its associated cognitive comorbidities.

The piriform cortex (PIR) is among the most ancient cortical structures in the mammalian brain. Traditionally viewed as the primary olfactory cortex responsible for processing and encoding odour information [17–19], its three-layered architecture and highly developed intrinsic associative fibre system provide a unique, non-topographic distributed network organisation distinct from the six-layered columnar neocortex [20]. This structural characteristic underpins the robust capacity of the PIR for information integration and pattern separation. Recent evidence suggests that the functional role of the PIR extends well beyond olfactory processing. It is involved in learning and memory [21], consciousness state modulation [22], emotional processing [23], and sleep regulation [24, 25], while also exhibiting high intrinsic excitability attributed to epileptogenesis [26]. Importantly, human neuroimaging studies have demonstrated PIR activation during intranasal trigeminal stimulation, with PIR activity correlated with stimulus intensity, suggesting that this olfactory-associated cortex may also participate in trigeminal sensory processing [27, 28]. Direct experimental evidence further indicates that the PIR can participate in pain modulation, as activation of a specific anterior PIR–IC circuit mediates analgesic effects [29]. Furthermore, the PIR is associated with higher cognitive functions; for instance, PIR–medial prefrontal cortex synaptic plasticity has been implicated in social communication [30, 31]. In addition to its cortical interactions, anatomical and electrophysiological studies have demonstrated direct glutamatergic connectivity between the PIR and the MD [32, 33], a higher-order thalamic nucleus closely linked to cognitive integration and cortical communication [34, 35]. Moreover, a recent study identified an anterior PIR^GABA–MD corticothalamic circuit that bidirectionally regulates pain sensitivity and mediates innate fear-induced analgesia, further highlighting the relevance of piriform–MD pathways in pain modulation [36]. However, whether PIR neurons and their downstream circuits contribute to TNP-associated pain–cognition comorbidity remains unknown.

In the present study, we employed a mouse model of TNP induced by chronic constriction injury of the infraorbital nerve (CION) to investigate the role of PIR glutamatergic neurons in pain–cognition comorbidity. Our findings identify the PIR as a previously underappreciated component of TNP-related networks and uncover a PIR^Glu–MD excitatory circuit that contributes to TNP-induced pain hypersensitivity and cognitive impairment.

Materials and methods

Animals

All experimental procedures were performed in accordance with protocols approved by the Animal Ethics Committee of the Affiliated Hospital of Jiaxing University (Approval ID: JXYY2026-047). Male C57BL/6J mice aged 8–10 weeks, procured from Shanghai VTL Co., Ltd. (Shanghai, China), were used for the experiments. Male mice were used throughout the study to reduce variability potentially associated with sex-related hormonal fluctuations and to maintain consistency across behavioural and circuit experiments. Mice were maintained under controlled conditions with a 12-h light/dark cycle, temperature regulated at 22–26 °C, and relative humidity between 40% and 70%.

Establishment of a mouse model of chronic constriction injury of the infraorbital nerve (CION)

As described previously, trigeminal neuropathic pain was induced in mice via chronic constriction injury of the infraorbital nerve (CION). Briefly, mice were anaesthetised by intraperitoneal (ip) injection. Under aseptic conditions, a 2 mm incision was made in the palatal buccal mucosa. The right infraorbital nerve was exposed via blunt dissection and loosely ligated using 5 − 0 chromic catgut suture. For the sham-operated control group, mice underwent the identical surgical procedure for nerve exposure without the ligation step. Following surgery, mice were allowed to recover in a clean enclosure on a heating pad to maintain normal thermoregulation.

Von Frey test

One day prior to formal behavioural testing, all mice underwent a 15-min acclimatisation session resting on the gloved palm of the experimenter. During the test session, graded Von Frey filaments with calibrated bending forces (0.02, 0.07, 0.16, 0.4, 1, and 2.0 g; purchased from Stoelting) were applied perpendicularly to the V2 and V3 regions of the mouse’s facial pad. A positive behavioural response was defined as rapid head retraction triggered by filament bending, with the retraction posture maintained for a minimum of 1 s (≥ 1 s). The 50% mechanical withdrawal threshold was calculated using the Dixon up–down method based on the log force values corresponding to each von Frey filament. Briefly, a negative response was followed by stimulation with the next higher-force filament, whereas a positive response was followed by stimulation with the next lower-force filament. After the first reversal in response pattern, four additional stimulations were performed, and the final response sequence was used to determine the 50% withdrawal threshold. Mice with a reduced response threshold were deemed to exhibit mechanical hyperalgesia.

Acetone test

A total of 0.05 mL of acetone was applied to the ipsilateral V3 or V2 facial region using a Hamilton microsyringe. The cumulative duration of acetone-evoked facial wiping was recorded during a 2-min observation period immediately after acetone application. An increase in facial wiping duration was considered indicative of orofacial cold hypersensitivity.

Ugo basile orofacial pain test

The apparatus consists of an animal activity cage, a test unit, and software. The test unit includes a stimulation module, water reward system, and infrared sensor. The stimulation module has two types: mechanical and thermal, delivering noxious stimuli to the oral-facial region. The mechanical stimulation module comprises three components with distinct intensity levels. Prior to testing, animals were deprived of water for at least 12 h before being placed in the activity cage. Each mouse was afforded 10 min to acclimatise, followed by a 10-min period of unrestricted licking. When the mouse’s mouth contacted the nipple of the drinking bottle, the built-in infrared sensor automatically recorded the time and duration. Pain-related behaviour was evaluated using the total contact time and total number of contacts recorded by the ORO software.

New object recognition test

During the familiarisation phase, two identical objects were placed in the testing arena, and each mouse was allowed to explore freely for 10 min. After a 2-h retention interval, one of the familiar objects was replaced with a novel object, and the mouse was allowed to explore for an additional 10 min. Cognitive performance was evaluated using the time spent exploring the novel object and the recognition index. The recognition index was calculated as follows: recognition index = [time spent exploring the novel object/(time spent exploring the novel object + time spent exploring the familiar object)]. For longitudinal NOR assessments, different sets of objects were used at each postoperative time point to minimise carry-over effects associated with object familiarity.

Y-maze test

The Y-maze apparatus comprises three opaque plastic arms (labelled A, B, and C) positioned at 120° angles to one another. Mice were placed in the central chamber and allowed to explore freely for 10 min. An entry into each arm was counted when all four limbs had fully entered. Sequential entry into three distinct arms (for example, ABC, ACB, BCA, BAC, CAB, or CBA) was considered an alternation. The alternation fraction was calculated as follows: spontaneous alternation rate (%) = [(number of spontaneous alternations)/(total arm entries − 2)] × 100. Between consecutive animals, the maze was cleaned with 75% ethanol and allowed to dry completely to minimise olfactory interference. ANY-maze software was used for recording and analysis.

Open-field test (OFT)

An open-field apparatus (41 × 41 × 30 cm) was employed to evaluate anxiety-related behaviours and locomotor activity. During an 8-min observation period, exploratory behaviour and locomotion were quantitatively assessed. Data were collected and processed using ANY-maze software, analysing the duration and distance traversed in the central area, as well as the total locomotor distance covered throughout the session.

Elevated plus maze (EPM) test

Anxiety-like behaviour in mice was assessed using the EPM test, conducted at a height of 50 cm above the floor. The apparatus comprised four arms: two enclosed arms (35 cm long × 5 cm wide × 15 cm high) and two open arms (35 cm long × 5 cm wide), connected by a central platform (5 cm long × 5 cm wide). Each mouse was positioned in the center facing the open arms and given 8 min to explore freely. The ANY-maze behavioral analysis system was used to measure entry into the arms and the duration spent in the open arms. The percentage of entries into the open arms, duration spent there, and distance covered were statistically analysed. After each test, the apparatus was cleaned thoroughly with 75% alcohol and dried to eliminate any lingering odors.

Viral injection and stereotaxic surgery

All viral injections were performed under isoflurane inhalation anaesthesia. For stereotaxic brain injections, AAV vectors, including AAV-VGLUT1-Cre-WPRE, AAV-EF1α-DIO-GCaMP6s-WPRE-hGH polyA, rAAV-CaMKIIα-GCaMP6s, AAV-CaMKIIα-hM4Di-mCherry, AAV-CaMKIIα-mCherry, rAAV-CaMKIIα-hChR2(H134R)-mCherry, rAAV-hSyn-DIO-hM4Di-EGFP, rAAV-CaMKIIα-Cre-WPRE-hGH, rAAV2/Retro, rAAV-VGLUT1-mCherry and rAAV-VGLUT1-Cre, rAAV2/Retro, were delivered into the corresponding target brain regions. Unless otherwise stated, AAV vectors were used at titres of ≥ 5.00 × 10¹² viral genomes (vg)/mL. For retrograde trans-synaptic tracing from the infraorbital nerve, PRV-CMV-EGFP (BrainVTA, P01001, 2.00 × 10⁹ IFU/mL) was used. Mice were anaesthetised with isoflurane, and the left infraorbital nerve was carefully exposed. A blunt Hamilton syringe was inserted into the infraorbital nerve at an appropriate angle and advanced 2–3 mm proximally along the nerve. Approximately 2 µL of PRV-EGFP was slowly injected over 2 min. The needle was partially withdrawn and left in place for an additional 1–2 min before complete withdrawal to minimise viral reflux. The wound was then sutured. Three to five days after PRV injection, mice were transcardially perfused, and brain tissues were collected for subsequent histological analysis. Using a stereotaxic apparatus (RWD Life Science, Shenzhen, China), the animals were positioned, and a heating pad with temperature control ensured their body temperature remained constant. The stereotaxic injection coordinates relative to the bregma were as follows: PIR: AP + 0.3 mm, ML −3.2 mm, and DV − 5.0 mm from the dural surface; MD: AP − 1.23 mm, ML 0.4 mm, and DV − 3.00 mm from the dural surface. A craniotomy was performed at each target coordinate, and 150 nL of the viral vector was injected at a constant rate of 0.3 nL/s via a glass micropipette connected to a microinjection pump (RWD Life Science, Shenzhen, China). After injection, the needle was held stationary for 5 min before being retracted by 0.5 mm. Following an additional 10 min, it was gradually withdrawn from the brain tissue.

Thereafter, optical fibres (diameter 200 μm, NA 0.37; Qiaoxingke, Nanjing, China) were implanted for fibre photometry recordings and in vivo optogenetic manipulation. For PIR recordings or manipulation, an optical fibre with a length of 5.0 mm was implanted at the following coordinates relative to bregma: AP + 0.30 mm, ML -3.2 mm, and DV − 4.85 mm. For recording calcium activity in the MD, an optical fibre with a length of 3.25 mm was implanted above the MD at the following coordinates relative to bregma: AP − 1.23 mm, ML -0.4 mm, and DV − 3.00 mm. The fibre was secured in position using dental acrylic. Mice were housed individually for at least 7 days after surgery to ensure complete recovery.

Immunofluorescence staining

To assess c-Fos protein expression, viral infection loci, and the precise anatomical location of implanted fibres or catheters, we prepared frozen brain tissue sections for subsequent histological analysis. For c-Fos immunofluorescence analysis, mice were sacrificed at the fourth postoperative week after behavioural phenotyping, corresponding to the chronic phase of TNP when stable pain-like behaviours and cognitive impairment had been established. Mice were deeply anaesthetized with sodium pentobarbital, then underwent sequential transcardial perfusion with 0.9% sterile saline, followed by 4% paraformaldehyde (PFA) dissolved in 0.1 mol/L phosphate-buffered saline (PBS, pH 7.4). Following perfusion, whole brains were rapidly harvested, post-fixed in 4% PFA solution at 4 °C overnight, and then processed for gradient dehydration. Dehydration was performed using a 20%–30% (w/v) sucrose gradient prepared in PBS, with brains maintained at 4 °C in each gradient solution until they fully sank to the bottom of the container. Serial coronal brain sections with a thickness of 30 μm were cut using a CM1860 cryostat (Leica, Wetzlar, Germany), and then mounted onto adhesive glass slides.

Sections underwent three 10-minute washes with PBS and were blocked at room temperature for 2 h with a solution of 10% donkey serum and 0.2% Triton X-100. Thereafter, they were incubated overnight at 4 °C with primary antibodies against c-Fos (mouse, 1:200, ab208942, Abcam), Glu (rabbit, 1:200, 50049s, Cell Signalling Technology, Danvers, MA, USA), GABA (rabbit, 1:200, A2052, Sigma), and VGLUT1 (rabbit, 1:400, 47181, Cell Signalling Technology, Danvers, MA, USA) diluted in blocking buffer. After incubation with the primary antibodies, the sections were washed three times with PBS and subsequently incubated at room temperature for 2 h with Alexa Fluor 488- and 594-conjugated secondary antibodies (A10067, A11037; Carlsbad, CA, USA). Cells were counterstained with a nuclear stain (4’,6-diamino-2-phenylindole [DAPI]) for 10 min at room temperature. Following three additional washes with PBS, slides were mounted using an anti-fluorescence quencher, and images were captured with a fluorescence microscope (Olympus). For each mouse, c-Fos-positive cells were quantified in three to four fields, and the mean value across fields was used as one biological replicate. Image acquisition and data analysis were performed by personnel unaware of the experimental conditions.

Fibre photometry

We captured fluorescence signals using a 488-nm laser beam, which was sequentially reflected via a dichroic mirror, focused through an objective lens, and transmitted through an optical chopper. Emitted GCaMP6s fluorescence was collected in turn, passed through a bandpass filter, and detected via a photomultiplier tube. The resulting voltage signals underwent low-pass filtering and were digitized at a sampling frequency of 500 Hz. For acclimatisation, all experimental mice were habituated to the fibre optic jumper for 10 min daily across three consecutive days. During facial pain threshold testing, we performed synchronous recording of behavioural responses and GCaMP6s fluorescence signal acquisition. We calculated relative fluorescence changes (ΔF/F) with the formula (F − F₀)/F₀, where F₀ denotes the mean baseline fluorescence signal within a 4-s control window immediately preceding facial pain testing. ΔF/F values were plotted as heatmaps or trial-averaged traces, with peak and mean ΔF/F values quantified across a 10-s time window. Specifically, peak and mean ΔF/F values were extracted from the maximum and average fluorescence change values within the target window, respectively. To quantify the global neuronal calcium response, we computed the area under the curve (AUC) of the ΔF/F trace within the predefined target time window. For data normalisation, the raw AUC value of each individual mouse was normalised to the mean raw AUC value of the corresponding control group to generate normalised relative AUC values. All statistical analyses and data visualisation were performed using GraphPad Prism 10 (GraphPad Software, San Diego, CA, USA).

During in vivo electrophysiological recording with microelectrode implantation

All mice were anesthetized via isoflurane inhalation anaesthesia, with the toe pinch reflex monitored continuously to confirm stable anesthetic depth throughout the procedure. Core body temperature was maintained at a constant 37 °C using a thermostatically controlled heating blanket. Mice were then secured in a stereotaxic frame, and sterile ophthalmic ointment was applied to the corneal surface to prevent desiccation during surgery. A midline scalp incision was made from the glabellar forehead region to the cranial midpoint between the bilateral ears, to fully expose the skull surface.

The custom 2 × 4-channel microneedle array microelectrodes (Kedou (Suzhou) Brain-computer Technology Co., Ltd., Suzhou, China) were implanted through a small craniotomy bone flap created above the target PIR region (stereotaxic coordinates: AP: +0.3 mm, ML: -3.2 mm). Stainless steel screws were anchored to the cranial surface, and the reference electrode was tightly wound around one of the skull screws for stable electrical grounding. Each microelectrode was slowly lowered to a final depth of 4.8–5.25 mm below the dural surface, with real-time electrophysiological recording performed during implantation to verify optimal positioning within the PIR.

Neuronal signals were acquired using a NeuroLego electrophysiological acquisition system (Jiangsu Brain Medical Technology Co., Nantong, China) through a headstage connected to an 8-channel amplifier. Signals were amplified with a fixed gain of 2000×, and a high-pass filter of 300 Hz was used to capture action potentials. After stable electrode placement was confirmed, the microelectrode array was fixed with light-curing adhesive and reinforced with dental cement to prevent displacement. Recordings were initiated after at least 10 days of postoperative recovery. Spike sorting was performed offline using Offline Sorter software (Plexon) based on principal component analysis and waveform templates. Neuronal firing characteristics were analysed using NeuroExplorer software (version 5.0; Plexon). Raster plots were generated by aligning spike events to the onset of behavioural events or stimulation epochs. Firing activity was quantified as spike counts or firing rates within predefined time windows and compared between baseline and event-related periods.

In vivo chemogenetic manipulation

For chemogenetic experiments, mice received an intraperitoneal dose of clozapine-N-oxide (CNO, 3.3 mg/kg; C0832, Sigma) 45 min before behavioral tests, three weeks after the virus was injected. For within-subject behavioural comparisons, the same mice were tested under PBS and CNO conditions on separate days in a fixed PBS-to-CNO sequence, with the testing intervals shown in the corresponding experimental timelines. The same testing schedule was applied to both hM4Di-expressing and mCherry control mice.

In vivo optogenetic manipulation

Channelrhodopsin-2 (ChR2) optogenetic stimulation was achieved using a 473 nm blue-light laser generator (Newdoon, Hangzhou, China). Light was transmitted via an interconnect cable to the optical fibre implanted in the PIR. For somatic stimulation, the laser intensity was set to 5 mW, with photostimulation applied during facial pain testing, anxiety-like behavioural assessment, and cognitive behavioural testing. To minimise tissue heating during prolonged stimulation, blue-light stimulation was delivered as 10-Hz trains of 5-ms pulses, using alternating 20-s stimulation and 20-s non-stimulation intervals [37].

Whole-cell patch-clamp recordings

Mice were anaesthetised with 2% isoflurane and rapidly decapitated. The brains were quickly removed and placed in ice-cold oxygenated artificial cerebrospinal fluid (ACSF). Coronal brain slices containing the target brain region were prepared at a thickness of 300 μm using a vibratome (Leica VT1200S, Leica, Germany) at 4 °C. Slices were then transferred to oxygenated ACSF and incubated at room temperature for at least 1 h before recording. ACSF was continuously bubbled with 95% O₂ and 5% CO₂ and contained the following components in mM: 124 NaCl, 25 NaHCO₃, 2.5 KCl, 1 NaH₂PO₄, 2 CaCl₂, 1 MgSO₄, and 10 glucose. Whole-cell recordings were performed in voltage-clamp mode using a MultiClamp 700B amplifier and an Axon patch-clamp recording system (Molecular Devices, USA). Data were acquired and analysed using Clampex and Clampfit 10.2 software. Patch pipettes were pulled using a P-97 micropipette puller (Sutter Instrument, USA) and had a resistance of 4–6 MΩ when filled with internal solution. For optogenetically evoked EPSC recordings, a cesium-based internal solution was used, containing the following components in mM: 112 cesium gluconate, 10 HEPES, 5 TEA-Cl, 5 QX-314-Br, 3.7 NaCl, 2 MgATP, 0.3 Na₃GTP, and 0.2 EGTA. The pH was adjusted to 7.2 with CsOH, and the osmolarity was adjusted to approximately 290 mOsm.

For optogenetically evoked EPSC recordings, MD neurons were voltage-clamped at − 70 mV, and ChR2-expressing PIR projection terminals were stimulated with 5-ms blue-light pulses delivered through the microscope objective using a pE-300 ultra illumination system (CoolLED Ltd., UK). To examine whether the light-evoked EPSCs were monosynaptic, baseline EPSCs were first recorded under ACSF perfusion. Tetrodotoxin (TTX, 1 µM; Sigma-Aldrich, USA) was then applied to block action potential-dependent synaptic transmission, followed by co-application of 4-aminopyridine (4-AP, 1 mM; Sigma-Aldrich, USA) to restore terminal depolarisation-dependent neurotransmitter release. To determine the receptor mechanism underlying the fast glutamatergic currents, light-evoked EPSCs were recorded before and after bath application of CNQX (20 µM; Sigma-Aldrich, USA). All recordings were performed only after achieving a gigaohm seal resistance greater than 1 GΩ. Recordings were included for analysis only when the series resistance was less than 35 MΩ and remained stable throughout the experiment. Cells were excluded if the series resistance changed by more than 15% during recording. Major salts and chemicals used in ACSF and internal solutions, including NaCl, KCl, NaHCO₃, NaH₂PO₄, CaCl₂, MgSO₄, glucose, cesium gluconate, HEPES, EGTA, MgATP, Na₃GTP, TEA-Cl, QX-314-Br, CsOH, TTX, 4-AP, and CNQX, were purchased from Sigma-Aldrich unless otherwise indicated. Recording pipettes were obtained from Wuhan Braintree Biotechnology Co., Ltd.

Statistical analysis

Statistical comparisons between two groups were performed using paired or unpaired two-tailed Student’s t-tests, as appropriate. For comparisons involving multiple groups, experimental conditions, or time points, one-way or two-way analysis of variance (ANOVA), including repeated-measures ANOVA where appropriate, was applied, followed by Šídák’s multiple-comparisons test. Correlations between behavioural variables were assessed using Pearson’s correlation analysis. Statistical significance was defined as follows: *p < 0.05, **p < 0.01, ***p < 0.001. Data are presented as the mean ± standard error of the mean (SEM). The specific statistical tests and sample sizes used for individual experiments are provided in the corresponding figure legends. Statistical analyses and graphing were performed using GraphPad Prism 10.0 (GraphPad Software, San Diego, CA, USA). Offline analysis of in vitro electrophysiological data was performed using Clampfit software (version 10.2; Molecular Devices, USA), whereas in vivo electrophysiological data were analysed using NeuroExplorer software (Plexon, USA). All figures were assembled by the authors using Adobe Illustrator.

Results

The CION-induced TNP model exhibits persistent pain-like behaviours, delayed-onset anxiety-like behaviours, and cognitive impairment

We first established a mouse model of TNP induced by chronic constriction injury of the infraorbital nerve (CION) and conducted a 4-week longitudinal behavioural phenotyping assessment (Fig. 1A). On the sensory dimension, compared with sham mice, CION mice exhibited significant facial mechanosensitivity from the first postoperative day, as reflected by a markedly reduced Orofacial response threshold in the von Frey test that persisted over the following 4 weeks (Fig. 1B). Acetone testing also revealed pronounced cold hypersensitivity in CION mice, with a significantly increased Duration wiping compared with sham mice (Fig. 1C). Furthermore, stimulus-evoked operant pain-related behaviour was assessed using the orofacial stimulation test. CION mice showed a significantly reduced Total contact time, whereas Total contact numbers did not differ significantly from sham mice, suggesting that model mice actively minimised contact duration with the stimulus owing to facial pain rather than showing reduced drinking motivation (Fig. 1D–F). These multidimensional pain assessments collectively support the successful establishment of persistent pain-like behaviours in the CION-induced TNP model [38].

Fig. 1.

The CION model induces persistent pain and cognitive impairment. (A) Experimental timeline. (B) Orofacial response threshold of mice in the sham and CION groups at different time points, as assessed by the von Frey test. (C) Duration wiping in response to acetone stimulation in sham and CION mice at different time points. (D) Total contact time during the 10-min orofacial stimulation test session. (E) Total contact numbers during the 10-min orofacial stimulation test session. (F) Representative schematic of drinking/contact behaviour in the orofacial stimulation test. (G) Time spent on novel object during the NOR test phase. (H) Recognition index calculated based on the exploration time of novel and familiar objects. (I) Total time spent exploring both novel and familiar objects during the NOR test phase. (J) Representative locomotor activity heatmaps of sham and CION mice during the NOR test phase. Square: familiar object; circle: novel object. Colour intensity represents the total residence duration of mice at each location, with darker colour indicating longer residence time. (K) Spontaneous alternation in the Y-maze test. An arm entry was considered valid only when all four paws of the mouse had completely entered the arm. Spontaneous alternation was calculated as follows: number of correct alternations / (total number of arm entries − 2) × 100. (L) Total arm entries during the Y-maze test session. (M) Representative locomotor activity heatmaps of mice from each group during the 10-min Y-maze test session. Colour intensity represents the total residence duration of mice at each location, with darker colour indicating longer residence time. Data are presented as the mean ± SEM; n = 8 mice per group. Data were analysed by two-way repeated-measures ANOVA followed by Sidak’s multiple-comparisons test. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001

To evaluate anxiety-like behaviours after CION, mice were subjected to the OFT and EPM at different postoperative time points (Supplementary Fig. 1A). In the OFT, CION mice showed a decreasing trend in Center zone time and Center zone distance compared with sham mice, whereas total distance was not significantly altered, suggesting that the behavioural changes were not attributable to impaired locomotor activity. Representative movement heatmaps further showed reduced exploration of the center zone in CION mice (Supplementary Fig. 1B–E). In the EPM, CION mice exhibited significantly decreased open arms time percentage, open arms entries percentage, and open arms distance percentage from the second postoperative week onwards, accompanied by representative movement traces showing reduced open-arm exploration (Supplementary Fig. 1F–I). These results indicate that CION induces anxiety-like behaviours beginning in the second postoperative week. This finding aligns with clinical observations wherein chronic pain frequently co-occurs with emotional disorders such as anxiety and depression [39–40].

To evaluate cognitive function after CION, an independent cohort of mice was assessed over four consecutive postoperative weeks using the NOR and Y-maze tests. In the NOR test, CION mice showed a significant reduction in Time spent on novel object and Recognition index at the fourth postoperative week compared with sham mice, whereas Total time was not significantly altered, indicating impaired recognition memory without an obvious change in overall object exploration. Representative heatmaps further showed reduced exploration of the novel object in CION mice (Fig. 1G–J). In the Y-maze test, CION mice displayed significantly decreased Spontaneous alternation at the fourth postoperative week, while the Total arm entries were not significantly changed, suggesting impaired spatial working memory rather than reduced exploratory activity. Representative heatmaps also illustrated altered exploratory patterns in CION mice (Fig. 1K–M). These results indicate that CION-induced chronic TNP leads to cognitive dysfunction by the fourth postoperative week, particularly affecting recognition memory and spatial working memory, which may be associated with persistent pain-related disturbances in higher-order brain regions involved in learning and memory [41–42].

Overall, our study demonstrates that the CION model effectively replicates the persistent pain symptoms of chronic TNP, along with associated anxiety and cognitive impairments. Importantly, anxiety-like behaviours emerged from the second postoperative week, whereas cognitive impairment became evident only at the fourth postoperative week, indicating distinct temporal profiles of the affective and cognitive comorbidities following CION.

The PIR is anatomically linked to ION-related trigeminal circuits and activated under TNP conditions

To identify brain regions anatomically associated with the ION-related trigeminal pathway, PRV-mediated retrograde trans-synaptic tracing following injection into the infraorbital nerve (ION). Following PRV injection into the ION terminal field, viral labelling was observed at the injection site and in the trigeminal ganglion (TG), confirming successful retrograde tracing from the peripheral trigeminal pathway (Fig. 2A, B). PRV-labelled neurons were further detected in both the posterior and anterior piriform cortex (pPIR and aPIR), indicating that the PIR is anatomically linked, through multisynaptic pathways, to ION-associated trigeminal circuits (Fig. 2B). In addition, PRV-labelled neurons were observed in multiple pain-, emotion-, and cognition-related regions, including the IC, lateral septal region, M1, NAc, paraventricular hypothalamic nucleus, lateral hypothalamus, amygdala, paraventricular thalamic nucleus, LHb, VTA, magnocellular part of the red nucleus, ventral hippocampus, periaqueductal grey, dorsal raphe nucleus, parabrachial nucleus, principal sensory trigeminal nucleus, and rostral ventromedial medulla (Supplementary Fig. 2B). These tracing results provide an anatomical basis for further examining the potential involvement of the PIR in TNP-related behavioural alterations.

Fig. 2.

These findings indicate that PIR^Glu neuronal activity is dynamically associated with orofacial mechanical stimulation under TNP conditions. (A) Schematic showing unilateral PRV injection into the infraorbital nerve (ION). (B) Representative fluorescence images showing PRV labelling at the injection site, trigeminal ganglion (TG), posterior piriform cortex (pPIR), and anterior piriform cortex (aPIR), from left to right. Scale bar = 100 μm. Images are representative of n = 3 mice. (C) Representative immunofluorescence images of c-Fos staining in the PIR of sham and CION mice at the fourth postoperative week. Scale bar = 100 μm. (D) Quantification of c-Fos-positive cells in the PIR. (E) Representative immunofluorescence images showing co-labelling of c-Fos with GABA, VGLUT1, or Glu in the PIR of CION mice. Nuclei were counterstained with DAPI. Scale bar = 100 μm. (F) Quantification of c-Fos-positive cells co-labelled with GABA, VGLUT1, or Glu in the PIR. (G) Schematic of stereotaxic microinjection of a glutamatergic neuron-specific calcium indicator virus (rAAV-VGLUT1-Cre; rAAV-EF1α-DIO-GCaMP6s) into the PIR of mice, along with a representative fluorescence image confirming virus expression at the injection site. Nuclei were counterstained with DAPI (blue). Scale bar = 100 μm. (H) Heatmap showing GCaMP6s calcium responses in PIR^Glu neurons aligned to facial mechanical stimulation. (I) Representative fibre photometry traces of GCaMP6s calcium signals (normalised to ΔF/F) in PIR^Glu neurons of sham and CION mice. (J) Quantitative statistics of the AUC of the calcium signal ΔF/F traces. Data are presented as the mean ± SEM. For c-Fos and co-labelling analyses, n = 5 mice per group; three to four brain sections were analysed per mouse. For PRV tracing, n = 3 mice. For fibre photometry recordings, n = 8 mice per group. *p < 0.05, **p < 0.01, ***p < 0.001

To determine whether the PIR is activated during the chronic phase of TNP, we examined c-Fos expression in the PIR at the fourth postoperative week, when stable pain-like behaviours and cognitive impairment had been established. Immunofluorescence staining revealed that the PIR of CION mice contained significantly more c-Fos-positive cells than that of sham mice (Fig. 2C, D). A broader analysis of pain-related brain regions further showed increased c-Fos-positive signals in the mPFC, ACC, S1, LS, IC, MD, VPM, PAG and VTA of CION mice, whereas the PVH displayed detectable c-Fos expression in both groups without a significant between-group difference (Supplementary Fig. 3A, B). Double-labelling analysis in the PIR showed that most c-Fos-positive cells were co-labelled with the glutamatergic neuronal marker VGLUT1, suggesting that PIR glutamatergic neurons are preferentially recruited under TNP conditions (Fig. 2E, F). To further monitor the real-time response of PIR^Glu neurons to facial mechanical stimulation, we injected a virus encoding the calcium indicator GCaMP6s into the PIR and recorded neuronal population calcium activity via implanted optical fibres (Fig. 2G). Facial mechanical stimulation evoked transient calcium responses in PIR^Glu neurons, and the response amplitude and AUC were significantly greater in CION mice than in sham mice (Fig. 2H–J). These findings indicate that PIR^Glu neuronal activity is dynamically associated with orofacial mechanical stimulation under TNP conditions.

Chemogenetic inhibition of PIR^Glu neurons alleviates pain-like behaviours and cognitive impairment in CION mice

To examine the functional contribution of PIR^Glu neurons to TNP-related pain and cognitive impairment, we employed a cell-type-preferential chemogenetic loss-of-function strategy in CION mice. AAV-CaMKIIα-hM4Di-mCherry or control AAV-CaMKIIα-mCherry was stereotaxically injected into the unilateral PIR (Fig. 3A, B). In vivo electrophysiological validation demonstrated that intraperitoneal administration of CNO significantly reduced the mean firing frequency of PIR^Glu neurons in hM4Di-expressing mice, confirming the reliable inhibitory efficacy of our chemogenetic manipulation (Figs. 3C, F).

Fig. 3.

Chemogenetic inhibition of PIR^Glu neurons alleviates pain-like behaviours and cognitive impairment in CION mice. (A) Schematic portraying the experimental workflow for the chemogenetic inhibition of PIR^Glu neurons. (B) Schematic of the stereotaxic microinjection of the glutamatergic neuron-specific inhibitory ‘Designer Receptors Exclusively Activated by Designer Drugs’ virus into the PIR, along with a representative fluorescence image verifying virus expression at the injection site. (C) Schematic of in vivo extracellular electrophysiological recording in the PIR of CION mice following chemogenetic manipulation. (D) Representative action potential waveforms of PIR^Glu neurons in the PBS vehicle control and CNO treatment groups after chemogenetic inhibition. (E) Raster plots depicting the action potential firing of PIR^Glu neurons in the PBS (upper panel) and CNO (lower panel) groups after chemogenetic inhibition. (F) Quantitative analysis of the firing frequency (spikes/s) of PIR^Glu neurons in the PBS and CNO groups. (G) Orofacial response threshold in the von Frey test after PBS or CNO administration in hM4Di-expressing and mCherry control CION mice. (H) Duration wiping in response to acetone stimulation after PBS or CNO administration. (I) Total contact time during the orofacial stimulation test after PBS or CNO administration. (J) Representative schematic of drinking/contact behaviour in the orofacial stimulation test. (K) Total contact numbers during the orofacial stimulation test after PBS or CNO administration. (L) Time spent on novel object during the NOR test after PBS or CNO administration. (M) Recognition index in the NOR test after PBS or CNO administration. (N) Total time spent exploring both novel and familiar objects during the NOR test. (O) Representative locomotor activity heatmaps during the NOR test. Colour intensity represents the total residence duration of mice at each location, with darker colour indicating longer residence time. (P) Spontaneous alternation in the Y-maze test after PBS or CNO administration. (Q) Total arm entries during the Y-maze test. (R) Representative locomotor activity heatmaps, with colour intensity signifying total residence duration. Data are presented as the mean ± SEM. Electrophysiological data were analysed using an unpaired two-tailed Student’s t-test (n = 8 neurons from 4 mice). Behavioural data were analysed using two-way repeated-measures ANOVA followed by Šídák’s multiple-comparisons test (n = 8 mice per group). *p < 0.05, **p < 0.01, ***p < 0.001

For orofacial pain-related behavioural assessments, in hM4Di-expressing CION mice, CNO administration significantly increased the Orofacial response threshold in the von Frey test and reduced acetone-evoked Duration wiping compared with PBS treatment (Fig. 3G, H). In the orofacial stimulation test, CNO significantly increased Total contact time, whereas Total contact numbers remained unchanged (Fig. 3I–K). These effects were not observed in mCherry control mice following CNO administration, supporting that the behavioural improvement was dependent on hM4Di-mediated inhibition rather than nonspecific CNO effects.

Cognitive performance was then assessed using the NOR and Y-maze tests. In hM4Di-expressing CION mice, CNO treatment increased Time spent on novel object and Recognition index in the NOR test, whereas Total time was not significantly altered (Fig. 3L–O). In the Y-maze test, CNO significantly increased Spontaneous alternation, while Total arm entries remained unchanged (Fig. 3P–R). In contrast, CNO did not significantly affect these cognitive indices in mCherry control mice. These results indicate that chemogenetic inhibition of PIR^Glu neurons alleviates pain-like behaviours and improves recognition memory and spatial working memory in CION mice.

We also examined whether chemogenetic inhibition of PIR^Glu neurons could influence anxiety-like behaviours related to TNP and general locomotor activity, using the EPM and OFT. No significant changes were noted in the time spent in the open arms of the EPM or in the central area of the OFT in mice after CNO administration compared with that under PBS control conditions. Moreover, total distance travelled in the OFT remained unchanged between the two conditions, indicating that our chemogenetic manipulation did not affect general locomotor activity (S4A–I). These results suggest that PIR^Glu neurons do not play a direct role in regulating the delayed anxiety-like behaviour associated with TNP.

Collectively, these findings indicate that the hyperactivity of PIR^Glu neurons is essential for maintaining a persistent pain state and associated cognitive dysfunction in the TNP model while not directly modulating comorbid anxiety-like behaviour.

Optogenetic activation of PIR^Glu neurons evokes pain-like behaviours and impairs cognitive task performance

To determine whether acute activation of PIR^Glu neurons can elicit pain-like behaviours and cognitive impairment, we conducted cell-type-specific optogenetic gain-of-function assays. Technically, we stereotaxically injected AAV-CaMKIIα-hChR2-mCherry or control AAV-CaMKIIα-mCherry into the unilateral PIR of sham wild-type mice, with an optical fibre implanted above the PIR after virus injection (Fig. 4A). For optogenetic manipulation, we delivered 473-nm blue light stimulation (5 mW, 10 Hz, and 5 ms pulse width) during behavioural tests. All mice were tested sequentially under two conditions—Light OFF and Light ON—to assess the effect of optogenetic activation within each group (Fig. 4A, B).

Fig. 4.

Optogenetic activation of PIR^Glu neurons induces pain-like behaviours and cognitive impairment in sham mice. (A) Experimental timeline for optogenetic activation of PIR^Glu neurons in sham mice. (B) Schematic showing unilateral stereotaxic injection of AAV-CaMKIIα-hChR2-mCherry or control AAV-CaMKIIα-mCherry into the PIR, together with representative fluorescence images showing viral expression at the injection site. Scale bar = 100 μm. (C) Schematic of optogenetics-coupled in vivo electrophysiological recording and representative action potential waveforms. (D) Representative firing raster plots of PIR^Glu neurons in ChR2-expressing mice under Light OFF and Light ON conditions. (E) Quantification of firing frequency under Light OFF and Light ON conditions. (F) Orofacial response threshold in the von Frey test in mCherry and ChR2 groups under Light OFF and Light ON conditions. (G) Duration wiping in response to acetone stimulation under Light OFF and Light ON conditions. (H) Total contact time during the orofacial stimulation test under Light OFF and Light ON conditions. (I) Representative schematic of drinking/contact behaviour in the orofacial stimulation test. (J) Total contact numbers during the orofacial stimulation test under Light OFF and Light ON conditions. (K) Time spent on novel object during the NOR test under Light OFF and Light ON conditions. (L) Recognition index in the NOR test. (M) Total time spent exploring both novel and familiar objects during the NOR test. (N) Representative locomotor activity heatmaps during the NOR test. Colour intensity represents the total residence duration of mice at each location, with darker colour indicating longer residence time. (O) Spontaneous alternation in the Y-maze test. (P) Total arm entries during the Y-maze test. (Q). Representative locomotor activity heatmaps during the Y-maze test. Data are presented as the mean ± SEM. Electrophysiological data were analysed using an unpaired two-tailed Student’s t-test (n = 8 neurons from 4 mice). Behavioural data were analysed using two-way repeated-measures ANOVA followed by Šídák’s multiple-comparisons test (n = 8 mice per group). *p < 0.05, **p < 0.01, ***p < 0.001

In vivo electrophysiological validation showed that blue-light stimulation significantly increased the firing frequency of PIR^Glu neurons in ChR2-mCherry-expressing mice (Fig. 4C–E), confirming the effectiveness of the optogenetic activation protocol.

For pain-related behavioural assessments, a series of orofacial pain tests was sequentially performed under Light OFF and Light ON conditions. In the von Frey test, no significant difference in Orofacial response threshold was observed between Light OFF and Light ON conditions in the mCherry control group. In contrast, blue light stimulation significantly reduced Orofacial response threshold in the ChR2 group compared with the Light OFF baseline (Fig. 4F). In the acetone test, blue light stimulation significantly increased Duration wiping in the ChR2 group, whereas no significant change was observed in the mCherry control group (Fig. 4G). In the orofacial stimulation test, blue light stimulation significantly decreased Total contact time in the ChR2 group, while Total contact numbers remained unchanged; these effects were not observed in the mCherry control group (Fig. 4H–J). These findings indicate that acute activation of PIR^Glu neurons is sufficient to elicit orofacial pain-like hypersensitivity.

Cognitive functions were subsequently evaluated using the NOR and Y-maze tests under Light OFF and Light ON conditions. In the NOR test, blue light stimulation significantly reduced Time spent on novel object and Recognition index in the ChR2 group, whereas Total time was not significantly altered; no significant changes were observed in the mCherry control group (Fig. 4K–N). In the Y-maze test, the ChR2 group exhibited a significant decrease in Spontaneous alternation under the Light ON condition compared with the Light OFF baseline, while Total number of arm entries remained unchanged. No significant changes in these indices were observed in the mCherry control group (Fig. 4O–Q). These data indicate that acute activation of PIR^Glu neurons impairs recognition memory and spatial working memory without affecting overall exploratory activity.

We also investigated whether optogenetic activation of PIR^Glu neurons could trigger anxiety-like behaviours or affect general locomotor activity using the EPM and OFT under Light OFF and Light ON conditions. In the EPM, no significant differences were observed in open arms time percentage, open arms entries percentage, or open arms distance percentage between Light OFF and Light ON conditions in either the mCherry control group or the ChR2 group. Similarly, the OFT showed no significant differences in Center zone time, Center zone distance, or total distance between Light OFF and Light ON conditions in either group (Supplementary Fig. 5A–I). These findings suggest that acute activation of PIR^Glu neurons does not directly induce anxiety-like behaviour or alter locomotor activity.

Collectively, these results indicate that acute activation of PIR^Glu neurons is sufficient to evoke pain-like behaviours and impair cognitive task performance in sham mice, but does not acutely induce anxiety-like behaviour.

Viral tracing dissection of the downstream projection network of PIR^Glu neurons

Having established the involvement of PIR^Glu neurons in TNP-related pain and cognitive impairment, we further examined the downstream circuitry that may contribute to this dual regulatory function. To systematically map the efferent projection network of PIR^Glu neurons, we utilised two complementary anterograde viral tracing strategies: unilateral injection of AAV-VGLUT1-mCherry into the PIR for anterograde axonal labelling and HSV-EGFP injection into the same region for early anterograde trans-synaptic tracing, with tissue collection performed 24 h after HSV injection (Fig. 5A, B). Because viral injections were performed unilaterally, projection laterality was described relative to the PIR injection side. The tracing results revealed broad AAV-labelled efferent fibres mainly in ipsilateral downstream regions implicated in pain modulation, emotional regulation and cognitive functions, including the prelimbic cortex (PrL), infralimbic cortex (IL), primary somatosensory cortex (S1), nucleus accumbens (NAc), submedius thalamic nucleus (Sub), medial amygdala (MeA), VTA region and mediodorsal thalamus (MD), with overlapping HSV-EGFP signals detected in these regions (Fig. 5C). Within the MD, labelled fibres were mainly localized to the central subdivision of the ipsilateral MD, with additional labelling extending into the medial subdivision, corresponding to MDC and MDM. This distribution is broadly consistent with previous anatomical studies describing piriform cortical projections to the central segment of the MD [32, 43]. Among these downstream regions, the ipsilateral MD was selected for further validation because of its established involvement in higher-order cognitive processing, its anatomical connection with the PIR and the prominent fibre labelling observed in MD subdivisions.

Fig. 5.

Efferent projection mapping of PIR^Glu neurons and electrophysiological validation of the PIR^Glu–MD connection. (A) Schematic of two complementary anterograde tracing strategies: anterograde axonal labelling with AAV-VGLUT1-mCherry and early anterograde trans-synaptic tracing with HSV-EGFP (Brain tissues were collected 24 h after HSV-EGFP injection). (B) Representative fluorescence image showing viral expression at the PIR injection site. Scale bar = 100 μm. (C) Representative fluorescence images showing predominantly ipsilateral efferent projections from PIR^Glu neurons to downstream brain regions implicated in pain modulation, emotional regulation and cognitive functions, including the mediodorsal thalamus (MD), prelimbic cortex (PrL), infralimbic cortex (IL), primary somatosensory cortex (S1), nucleus accumbens (NAc), submedius thalamic nucleus (Sub), medial amygdala (MeA) and ventral tegmental area (VTA). Images are representative of n = 3 mice. (D) Schematic of whole-cell patch-clamp recording from MD neurons in acute brain slices combined with optogenetic stimulation of ChR2-expressing PIR projection terminals.(E) Representative light-evoked excitatory postsynaptic current (oEPSC) traces recorded from MD neurons under ACSF, TTX, and TTX + 4-AP conditions. (F) Quantification of oEPSC amplitudes under ACSF, TTX, and TTX + 4-AP conditions. (G) Representative oEPSC traces recorded from MD neurons before and after bath application of CNQX. (H) Quantification of oEPSC amplitudes before and after CNQX application. Data are presented as the mean ± SEM. For whole-cell patch-clamp recordings, oEPSC amplitudes under ACSF, TTX, and TTX + 4-AP conditions were analysed using one-way repeated-measures ANOVA followed by Šídák’s multiple-comparisons test, whereas oEPSC amplitudes before and after CNQX application were analysed using a paired two-tailed Student’s t-test (n = 8 cells from 5 mice). *p < 0.05, **p < 0.01, ***p < 0.001

To determine whether PIR^Glu neurons form functional excitatory synaptic connections with MD neurons, we performed whole-cell patch-clamp recordings in acute brain slices containing the MD while optogenetically stimulating ChR2-expressing PIR projection terminals (Fig. 5D). A 5-ms blue-light pulse reliably evoked excitatory postsynaptic currents (oEPSCs) in MD neurons. Bath application of TTX abolished the light-evoked oEPSCs, whereas subsequent addition of 4-AP in the presence of TTX restored the oEPSCs, supporting the presence of functional monosynaptic excitatory connections from PIR^Glu neurons to MD neurons (Fig. 5E, F). Moreover, the light-evoked oEPSCs were blocked by CNQX, indicating that the fast excitatory glutamatergic transmission at the PIR^Glu–MD synapse is predominantly mediated by AMPA receptor-dependent currents (Fig. 5G, H). Together, these tracing and electrophysiological results identify the MD as a functionally connected downstream target of PIR^Glu neurons and provide a circuit basis for subsequent pathway-specific manipulation.

Specific blockade of the PIR^Glu–MD circuit ameliorates orofacial mechanical allodynia and cognitive impairment in TNP model mice

Based on anterograde tracing and electrophysiological validation identifying the MD as a prominent, functionally connected downstream target of PIR^Glu neurons, we employed a projection-defined chemogenetic strategy to examine the functional contribution of MD-projecting PIR^Glu neurons to TNP-related pain and cognitive impairment. To selectively manipulate MD-projecting PIR^Glu neurons, rAAV-hSyn-DIO-hM4Di-EGFP or the control vector rAAV-hSyn-DIO-EGFP was injected into the PIR of CION mice, while retrograde rAAV-CaMKIIα-Cre-WPRE-hGH (AAV2/Retro) was concurrently injected into the MD (Figs. 6A-B). Following sufficient viral expression, hM4Di receptors were activated via intraperitoneal CNO injection, after which we conducted behavioural assessments to evaluate alterations in pain perception and cognitive function.

Fig. 6.

Chemogenetic inhibition of MD-projecting PIR^Glu neurons alleviates orofacial mechanical hypersensitivity and cognitive impairment in CION model mice. (A) Experimental timeline for circuit-specific chemogenetic inhibition of the PIR^Glu–MD pathway in CION model mice. (B) Schematic of dual-site stereotaxic microinjection (PIR + MD) and representative fluorescence images verifying virus expression in the PIR and accurate injection targeting in the MD. Scale bar = 100 μm. (C) Orofacial response threshold in the von Frey test after CNO administration in control and PIR^Glu–MD pathway inhibition mice. (D) Duration wiping in response to acetone stimulation after CNO administration. (E) Total contact time during the orofacial stimulation test after CNO administration. (F) Total contact numbers during the orofacial stimulation test after CNO administration. (G) Representative schematic of drinking/contact behaviour in the orofacial stimulation test. (H) Recognition index. (I) Time spent on novel object during the NOR test. (J) Representative locomotor activity heatmaps during the test phase. (K) Total time spent exploring both novel and familiar objects during the NOR test. (L) Spontaneous alternation percentage. (M) Total arm entries. (N) Representative locomotor activity heatmaps during the test session. (O) Schematic of the viral strategy for specific labeling and in vivo fiber photometry recording of MD-projecting PIR^Glu neurons. (P) Schematic of dual-site stereotaxic virus injection for calcium signal recording and representative fluorescence image showing specific labeling of MD-projecting PIR^Glu neurons in the PIR. Scale bar = 100 μm. (Q) Heatmap showing GCaMP6s calcium activity in MD-projecting PIR^Glu neurons in the Cre and EGFP control groups. (R) Representative ΔF/F traces recorded from MD-projecting PIR^Glu neurons in the Cre and EGFP control groups. (S) Quantification of the AUC of the ΔF/F traces. Data are presented as the mean ± SEM. For behavioural tests, n = 8 mice per group. For fibre photometry recordings, n = 5 mice per group. Two-group comparisons were performed using unpaired two-tailed Student’s t-tests. *p < 0.05, **p < 0.01, ***p < 0.001

The pain-related behavioural assessments showed that chemogenetic inhibition of MD-projecting PIR^Glu neurons significantly increased the Orofacial response threshold in the von Frey test, indicating attenuation of mechanical hypersensitivity in CION mice (Fig. 6C). In the orofacial stimulation test, inhibition of these neurons significantly increased both Total contact time and Total contact numbers compared with the control group (Fig. 6E–G), indicating reduced pain-related avoidance of facial contact with the mechanical stimulus. In contrast, Duration wiping in the acetone test was not significantly altered following inhibition of MD-projecting PIR^Glu neurons (Fig. 6D), suggesting that this neuronal population preferentially contributes to mechanical rather than cold hypersensitivity in the CION model.

In terms of cognitive function testing, the NOR test demonstrated that CION mice receiving chemogenetic inhibition of MD-projecting PIR^Glu neurons showed significantly higher Recognition index and Time spent on novel object than controls, whereas Total time was not significantly altered, indicating improved recognition memory without a general increase in object exploration (Fig. 6H–K). In the Y-maze test, the inhibition group displayed significantly higher Spontaneous alternation than the control group, while Total number of arm entries was not significantly altered (Fig. 6L–N). This suggests that chemogenetic inhibition of MD-projecting PIR^Glu neurons improves spatial working memory without obvious changes in overall arm-entry activity.

To characterise the activity of MD-projecting PIR^Glu neurons in CION mice, rAAV-hSyn-DIO-GCaMP6s was injected into the PIR, whereas rAAV2/Retro-VGLUT1-Cre or the control vector rAAV2/Retro-VGLUT1-EGFP was injected into the MD, followed by fibre photometry recording from the PIR (Fig. 6O, P). Robust calcium transients were detected in the Cre group, whereas only minimal signals were observed in the EGFP control group. Consistently, the AUC of the ΔF/F traces was significantly greater in the Cre group than in the EGFP control group (Fig. 6Q–S). These results indicate that MD-projecting PIR^Glu neurons exhibit detectable calcium activity in CION mice.

To further address the possibility that improved cognitive performance was secondary to pain relief or altered exploratory motivation, we performed correlation analyses between pain-related and cognitive behavioural indices. No significant correlation was observed between Recognition index and Orofacial response threshold, or between Total time and Recognition index. Similarly, Spontaneous alternation was not significantly correlated with either Orofacial response threshold or Total arm entries (Supplementary Fig. 6A-D). These results suggest that the improvement in cognitive performance after PIR^Glu–MD circuit inhibition is unlikely to be explained solely by increased pain threshold, enhanced object exploration, or altered arm-entry activity.

Chemogenetic inhibition of upstream PIR^Glu neurons attenuates mechanically evoked calcium signals in the downstream MD of TNP model mice

To clarify the functional regulation of downstream MD neuron activity by upstream PIR^Glu neurons and to establish causal links between the circuit’s anatomical connectivity and functional effects, we employed in vivo fibre photometry to record calcium signal dynamics in MD neurons following chemogenetic inhibition of upstream PIR^Glu neurons.

We stereotaxically injected AAV-CaMKIIα-hM4Di-mCherry into the PIR region of CION-induced TNP model mice to achieve targeted chemogenetic inhibition of PIR^Glu neurons. Concurrently, rAAV-CaMKIIα-GCaMP6s was injected into the MD region, along with the implantation of an optical fibre recording probe to detect in vivo calcium signals in MD excitatory neurons. Following viral expression, baseline mechanically evoked calcium responses in the MD were recorded. Thereafter, intraperitoneal administration of CNO activated hM4Di receptors to inhibit PIR^Glu neurons. Additionally, calcium responses in the MD were monitored during mechanical stimulation of the orofacial region post-CNO intervention (Figs. 7A-B).

Fig. 7.

Chemogenetic inhibition of upstream PIR^Glu neurons attenuates mechanically evoked calcium responses in downstream MD neurons in CION mice. (A) Schematic of the experimental design for in vivo fiber photometry recording of MD glutamatergic neuronal calcium signals during chemogenetic inhibition of PIR^Glu neurons in CION model mice. (B) Schematic of bilateral stereotaxic virus injection and optical fiber implantation in the PIR and MD, and representative fluorescence images verifying virus expression and fiber placement. Scale bar = 100 μm. (C) Representative GCaMP6s calcium signal heatmaps of MD neurons evoked by orofacial mechanical stimulation before and after chemogenetic inhibition of PIR^Glu neurons. (D) Representative GCaMP6s calcium signal traces (ΔF / F₀) of MD neurons evoked by mechanical stimulation after PBS or CNO treatment. (E) Quantitative analysis of the area under the curve of mechanically evoked calcium responses in MD neurons after PBS or CNO treatment. Data are presented as the mean ± SEM; n = 8 mice per viral group. AUC data were analysed using two-way repeated-measures ANOVA followed by Sidak’s multiple-comparisons test. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001

Calcium signal recordings revealed that in CION model mice, orofacial mechanical stimulation induced a robust calcium signal response in MD neurons. Following CNO-mediated chemogenetic inhibition of upstream PIR^Glu neurons, mechanically stimulated calcium signals (ΔF/F) in the MD region displayed a marked and transient attenuation, with peak calcium events and the area under the response curve reduced compared with pre-CNO responses (Figs. 7C–E). This finding indicates that PIR^Glu neurons functionally facilitate mechanically evoked activity in downstream MD neurons, offering further in vivo functional evidence for the involvement of the PIR^Glu–MD circuit in TNP–cognitive impairment comorbidity.

Discussion

Chronic pain is increasingly recognised as a central nervous system disorder involving widespread remodelling of brain networks [44–45]. Clinical observations reveal that patients with trigeminal neuralgia, in addition to experiencing severe pain hypersensitivity in the orofacial region, frequently suffer from significant cognitive comorbidities, such as impaired working memory and reduced cognitive flexibility, forming a complex pathological relationship [46–47]. For years, understanding the interconnections and common mechanisms within these brain networks has posed a major challenge in neuroscience. In the present study, we combined trans-synaptic anatomical tracing, in vivo fibre photometry, electrophysiological recording, and precise circuit manipulation to demonstrate the contribution of the PIR^Glu–MD excitatory pathway to TNP-related mechanical hypersensitivity and cognitive impairment. Notably, manipulation of PIR^Glu neurons did not significantly affect anxiety-like behaviours, suggesting that this pathway may preferentially contribute to the cognitive rather than affective dimension of TNP. The findings not only highlight a previously underappreciated role for traditional olfactory brain regions in the regulation of orofacial pain and cognition but also suggest that the diverse complications of chronic pain may be governed by partially dissociable neural circuits that differ in their temporal emergence and functional roles.

The orofacial region is innervated by the trigeminal system, in which noxious information from peripheral trigeminal afferents is transmitted through the trigeminal ganglion and brainstem trigeminal nuclei, relayed by thalamic sensory nuclei such as the ventral posteromedial nucleus, and ultimately represented in somatosensory cortical areas, including S1 [48–50]. This canonical ascending pathway provides the major anatomical framework for understanding the sensory-discriminative component of orofacial pain. However, chronic orofacial pain is increasingly recognised as a network-level disorder in which persistent nociceptive input engages not only classical sensory relays but also brain regions involved in sensory integration, salience detection, affective regulation and cognition. In this context, our PRV-mediated tracing extends the anatomical landscape of ION-associated trigeminal circuits by showing that the PIR is included within a polysynaptically connected orofacial pain-related network. Because PRV tracing reflects propagation across connected multisynaptic circuits, these findings are best interpreted as evidence that the PIR is anatomically linked to trigeminal afferent pathways, while the precise synaptic order and hierarchical position of the PIR within this network remain to be determined.

This interpretation is consistent with the emerging view that the PIR is not restricted to odor coding but may participate in multisensory and state-dependent processing. Previous human imaging studies have shown that intranasal trigeminal chemical stimulation can recruit cortical regions overlapping with olfactory networks, including the piriform cortex, suggesting anatomical and functional convergence between olfactory and trigeminal chemosensory processing [28]. In pain models, whole-brain network analyses have also identified PIR-related activity changes, supporting the possibility that the PIR may be engaged as part of a broader pain-responsive network [51]. Consistently, a recent study of lavender oil odour-induced analgesia identified an anterior piriform cortex–insular cortex circuit, in which aPIR^Glu neurons engage IC^GABA and IC^Glu neuronal populations to modulate pain sensitivity, further supporting a functional link between the PIR and pain-related cortical networks [29]. In the present study, PIR^Glu neurons showed increased c-Fos expression and enhanced calcium responses during facial mechanical stimulation under TNP conditions. These findings indicate that PIR^Glu neurons are recruited during TNP-related nociceptive processing. Importantly, our chemogenetic and optogenetic results further suggest that PIRGlu activity is functionally involved in orofacial mechanical hypersensitivity and cognitive impairment. Therefore, we propose that the PIR may act as a modulatory interface linking persistent trigeminal nociceptive activity with higher-order cognitive circuits, while its precise hierarchical position relative to classical trigeminal sensory relays remains to be clarified in future studies.

To clarify the downstream circuit mechanisms by which abnormal PIR excitation contributes to cognitive impairment, we screened for key downstream projection targets. The results suggest that the MD is a central effector region in this regulatory pathway [52]. As a critical higher-order relay nucleus within the thalamocortical network, the MD actively engages in various executive functions, including working memory, cognitive flexibility, and decision-making [34, 53]. It maintains close and interdependent anatomical and functional connections with the mPFC, a partnership essential for maintaining normal cognitive function [54, 55]. We observed that specific blockade of the PIR^Glu- MD loop in CION model mice not only significantly alleviated mechanical allodynia but also reversed the associated deficits in cognitive flexibility and working memory. This behavioural improvement coincided with the effective suppression of abnormally elevated calcium activity in downstream MD neurons. These data establish an in vivo functional link between PIR^Glu activity and MD neuronal recruitment under TNP conditions. However, because the present study did not directly measure MD–mPFC communication, cortical oscillatory dynamics, synaptic strength, or long-term plasticity within the MD–mPFC axis, the downstream thalamocortical mechanism remains hypothetical and requires further validation. In this context, persistent PIR^Glu hyperactivity may impose excessive excitatory drive on MD neurons, thereby perturbing MD-dependent thalamocortical processing required for cognitive performance. Given previous evidence that MD inputs can amplify prefrontal cortical connectivity and sustain task-related representations, abnormal signal flow from the MD may further disrupt neural oscillatory rhythms within the mPFC and alter the firing patterns of pyramidal neurons, ultimately impairing the neural network infrastructure essential for executive function [55, 56]. Future studies combining pathway-specific electrophysiology, MD–mPFC activity recordings, and synaptic plasticity assays will be required to determine whether TNP induces maladaptive thalamocortical coupling or LTP-like plasticity within this circuit.

Notably, in its dynamic observation of the pathological process, this study has revealed a significant dissociation between the emotional and cognitive comorbidities induced by chronic pain, particularly concerning the temporal and spatial localisation of neural circuits. In our model, anxiety-like behaviour was observed as early as the second week post-surgery, while cognitive impairment did not reach statistical significance until the fourth week. This temporal dissociation implies that comorbid symptoms may follow distinct developmental trajectories and pathophysiological mechanisms. Functional intervention experiments further validated the functional separation of the two circuits: inhibiting PIR^Glu neurons during the chronic phase of a stable anxiety phenotype or acutely activating this neuronal population in normal mice did not effectively modulate anxiety-related behaviour. This negative result is important because it suggests that, although PIR^Glu neurons are recruited by orofacial nociceptive-related stimulation, they appear preferentially contribute to the sensory–cognitive dimension of TNP rather than to its affective–motivational component. This interpretation is consistent with findings from a chronic social stress model, in which PIR-related alterations were also not simply coupled to generalized anxiety-like behavioural expression [37]. Our results support the view that the PIR^Glu–MD circuit preferentially mediates cognitive dysfunction, whereas pain-related affective abnormalities may depend more strongly on classical limbic and salience-related circuits [57–59]. We hypothesise that early negative emotions triggered by sustained pain input depend more on rapid adaptive changes within the limbic system. For instance, circuits involving the central nucleus of the amygdala and the LHb reportedly mediate depression-like behaviour in chronic pain [6, 60]. In contrast, the emergence of cognitive deficits potentially requires a longer accumulation of noxious stimuli to induce more profound structural synaptic remodelling within the thalamocortical circuit. Research indicates that input from the MD is crucial for sustaining activity in the prefrontal cortex during cognitive tasks such as working memory [61]. Overall, when exposed to persistent noxious stimuli, the brain may exhibit a modular, hierarchical response, with distinct but partially dissociable circuits regulating emotional and cognitive responses. This functional division underscores the complexity of central nervous system regulation and may help explain why current clinical analgesics often struggle to alleviate patients’ emotional and cognitive complications simultaneously.

Furthermore, this study indicates that the PIR is not a functionally uniform structure; it comprises diverse neuron types, each with distinct regulatory functions. Recent research demonstrates that, during acute stress from innate fears (such as the scent of a predator), GABAergic neurons in the aPIR mediate an endogenous analgesic response by projecting to the MD, thereby prioritising the animal’s survival and defence needs [36]. The analgesic effects of plant essential oils also depend on the regulatory role of the PIR inhibitory circuit [29]. This mechanism underlines the protective role of the PIR–MD pathway in maintaining physiological homeostasis. Nevertheless, in contrast to this adaptive mechanism, chronic dysregulation in TNP results in enhanced recruitment of PIR^Glu neurons and increased functional engagement of the PIR^Glu–MD pathway. This finding aligns with recent evidence suggesting that abnormal activation of the MD and its associated corticothalamic circuits significantly promotes hyperalgesia and comorbid negative emotions such as aversion and anxiety [52]. Excessive excitatory input may disrupt the excitatory/inhibitory balance within the local microcircuitry of the MD. Therefore, the functional consequence of PIR–MD signalling may depend strongly on disease state, neuronal subtype, and the balance between excitatory and inhibitory PIR outputs.

Overall, this study identifies the PIR^Glu–MD projection as an important circuit contributing to TNP-related mechanical hypersensitivity and cognitive impairment. These findings expand the current understanding of the non-olfactory functions of the PIR and support the view that pain chronicity involves multidimensional reorganisation of distributed brain networks. However, several limitations should be acknowledged. First, only male mice were used in this study, and whether this circuit operates similarly in females remains to be determined. Second, although our results demonstrate PIR^Glu–MD functional connectivity, the downstream MD–mPFC mechanism was not directly examined and warrants further investigation.

Conclusions

This study identifies the PIR^Glu–MD circuit as an important pathway contributing to TNP-related orofacial mechanical hypersensitivity and cognitive impairment. PRV-mediated multisynaptic retrograde tracing showed that the PIR is anatomically associated with ION-related trigeminal circuits, while in vivo fibre photometry revealed that PIR^Glu neurons are recruited during orofacial mechanical stimulation under TNP conditions. Chemogenetic inhibition of PIR^Glu neurons alleviated pain-like behaviours and cognitive deficits in CION-induced TNP model mice, whereas optogenetic activation of PIR^Glu neurons in sham mice recapitulated pain-like behaviours and cognitive impairment. These findings indicate that PIR^Glu neuronal activity plays an important role in TNP-related pain and cognitive impairment.

Viral tracing and ex vivo electrophysiological recordings further identified the MD as a functionally connected downstream target of PIR^Glu neurons, with fast excitatory synaptic transmission predominantly mediated by AMPA receptor-dependent currents. Circuit-specific inhibition of MD-projecting PIR^Glu neurons attenuated mechanical hypersensitivity and improved recognition and working memory performance, while inhibition of upstream PIR^Glu neurons reduced mechanically evoked calcium responses in downstream MD neurons. Notably, manipulation of PIR^Glu neurons did not significantly affect anxiety-like behaviours, suggesting that the PIR^Glu–MD circuit may preferentially contribute to the cognitive rather than affective dimension of TNP.

Overall, our study enhances the current understanding of the non-olfactory roles of the canonical olfactory cortex, specifically the PIR, in pain–cognition integration during chronic pain states. Furthermore, it offers innovative mechanistic insights into cognitive impairments associated with chronic pain and identifies the PIR^Glu–MD circuit as a promising therapeutic target for developing precision treatments for TNP and its associated cognitive dysfunction (Fig. 8).

Fig. 8.

Schematic summary of the PIR^Glu–MD pathway in TNP-related pain and cognitive impairment. PRV-mediated tracing from the ION-associated orofacial region identifies the PIR as part of an extended trigeminal-related network. Under TNP conditions, PIR^Glu neurons show enhanced activity during orofacial mechanical stimulation. PIR^Glu neurons form a monosynaptic excitatory projection to the MD, with fast glutamatergic transmission mediated mainly by AMPA receptor-dependent currents. Chemogenetic inhibition of PIR^Glu neurons alleviates pain-like behaviours and cognitive impairment, whereas optogenetic activation of PIR^Glu neurons induces pain-like behaviours and cognitive deficits in sham mice. Circuit-specific inhibition of MD-projecting PIR^Glu neurons reduces mechanically evoked MD calcium responses, attenuates mechanical hypersensitivity, and improves cognitive performance

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

A designed the study, performed the core experiments, analyzed the data, drafted the original manuscript, and prepared all figures. B, C, D, E, F, G, and H participated in experimental operation and data acquisition. I and J contributed to data analysis and figure preparation. K* and L* conceived the study, provided full supervision throughout the project, acquired funding, revised the manuscript, and finalized the version for submission. All authors have read and approved the final version of the manuscript for publication.

Funding

This study was supported by the Medical and Health Science Program of Zhejiang Province (No. 2025HY1177), the Natural Science Foundation of Zhejiang Province (No. LKLY25H090005), and the Key Clinical Specialty of Zhejiang Province (Anesthesiology, No. 2023-ZJZK-001).

Data availability

Data are available from the corresponding authors Qing-he Zhou, Ph.D or Long-sheng Xu, Ph.D upon request.

Declarations

Ethics approval and consent to participate

All animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee of Jiaxing University Affiliated Hospital, and were performed in accordance with the Guide for the Care and Use of Laboratory Animals. The approved protocol number is JXYY2026-047.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.