Introduction

Chronic pain is a disabling condition that extends beyond altered sensory processing and often includes anxiety, depression, and cognitive disturbances.1,2 Inflammatory and neuropathic pain arise from different initiating insults, yet both involve persistent nociceptive sensitization and may share downstream mechanisms.3,4 Recent rodent studies illustrate this convergence at the molecular and behavioral levels. Anandamide attenuated carrageenan-induced inflammation and chronic constriction injury-evoked mechanical and thermal hypersensitivity.5 Ethyl gallate likewise reduced neuropathic pain behaviors, pro-inflammatory cytokine production, and oxidative stress after peripheral nerve injury.6 These observations suggest some mechanistic overlap, but they do not establish whether the two pain states engage similar brain-wide activation patterns.

Pain processing is distributed across the somatosensory, thalamic, limbic, prefrontal, insular, and periaqueductal regions.7,8 Much of this framework, however, was developed through studies of a limited number of predefined regions. Such approaches cannot fully capture the coordinated neural activity underlying the sensory and affective dimensions of chronic pain. Recent work has begun to address this limitation at the circuit and network levels. In awake mice, inflammatory and neuropathic pain both produced sustained amplification of parabrachial neuronal responses. The parabrachial nucleus may therefore serve as a shared ascending node for nociceptive and aversive signaling.9 Resting-state functional magnetic resonance imaging has also identified pain-associated connectivity changes across retrosplenial, somatosensory, prefrontal, insular, thalamic, parabrachial, and periaqueductal networks.10 Together, these findings support a distributed network model of chronic pain.

Targeted circuit studies further indicate that inflammatory and neuropathic pain can recruit different pathways. CFA-induced inflammatory pain reduced the activity of hypothalamic paraventricular oxytocin neurons and their projections to the central amygdala, contributing to pain-associated anxiety-like behavior.11 Following SNI, enhanced excitatory transmission from the ventral posterolateral thalamus to the hindlimb primary somatosensory cortex promoted mechanical hypersensitivity.12 Distinct prelimbic projections to the basolateral amygdala and periaqueductal gray separately regulated anxiety-like behavior and hyperalgesia in SNI mice.13 Peripheral nerve injury also increased activity in the paraventricular thalamus–basolateral amygdala pathway, thereby promoting nociceptive hypersensitivity and anxiety-like behavior.14 These studies provide mechanistic resolution at the pathway level, but most examine one circuit within a single pain model. A direct brain-wide comparison of inflammatory and neuropathic pain therefore remains lacking.

The immediate-early gene c-Fos is widely used as a proxy for recent neuronal activation and enables relatively unbiased surveys of regional brain engagement.15,16 Automated imaging, atlas registration, and quantitative analysis now permit c-Fos-positive neurons to be mapped across many anatomically defined regions.15,17 Although these methods are well suited to brain-wide comparisons, few studies have applied them to inflammatory and neuropathic pain under matched experimental conditions. It also remains unclear whether the two models generate overlapping regional profiles or differ in their principal co-activated nodes. Addressing these questions requires an analysis that extends beyond predefined pain-related regions and considers interregional co-activation.

We therefore compared two established mouse models of chronic pain: complete Freund’s adjuvant (CFA)-induced inflammatory pain and spared nerve injury (SNI)-induced neuropathic pain. Behavioral testing assessed nociceptive hypersensitivity and anxiety-like behavior, followed by brain-wide c-Fos mapping across 50 anatomically defined subregions registered to the Allen Mouse Brain Atlas. We then compared regional activation profiles and constructed interregional co-activation networks for each model. This analysis was designed to identify shared and model-associated patterns of brain activation and network organization without inferring directed or causal circuit interactions.

Materials and MethodsAnimals

C57BL/6J mice (8–10 weeks old, male, n=8 per group) were obtained from Chengdu Dossy Experimental Animals Co., Ltd. All mice were housed in a controlled environment with unrestricted access to food and water under a 12 h/12 h light/dark cycle at 24 ± 2 °C. All animal experiments in this study were performed in accordance with the Guide for the Care and Use of Laboratory Animals published by the National Institutes of Health (NIH). The study protocol was approved by the Animal Ethics Committee at Chengdu University of Traditional Chinese Medicine (Approval Number: 2025011).

Study Design

To investigate the pattern of c-Fos expression throughout the brain under chronic pain, we established mouse models of inflammatory pain induced by Complete Freund’s Adjuvant (CFA) and neuropathic pain induced by Spared Nerve Injury (SNI). The Hargreaves test and Von Frey test were employed to evaluate hyperalgesia in CFA and SNI mice, respectively. Additionally, the elevated plus maze (EPM) and open field test (OFT) were utilized to assess anxiety-like behavior. To evaluate depression-like behavior, the sucrose preference test (SPT), forced swimming test (FST), and tail suspension test (TST) were performed 90 min before sample collection (Figure 1A). Coronal brain sections were collected at 7 predefined levels (Figure 1B). Raw c-Fos labeling images were aligned to the Allen Mouse Brain Atlas, followed by brain region segmentation, signal detection, and analysis (Figure 1C–E). All behavioral tests were completed before tissue collection, and the final behavioral assessments were performed approximately 90 min before perfusion to match the expected peak window of c-Fos induction. Therefore, c-Fos signals were interpreted as reflecting chronic pain-related brain activation under the influence of the immediately preceding behavioral assessment.

Figure 1 Experimental design and workflow for brain-wide c-Fos mapping and data analysis. (A) Experimental timeline. After 7 days of adaptation, mice received complete Freund’s adjuvant (CFA) injection or underwent spared nerve injury (SNI) surgery on day 0. Nociceptive behaviors were assessed using the Hargreaves and von Frey tests during the 14-day experimental period. Emotional and affective behaviors were evaluated using the elevated plus maze (EPM), open-field test (OFT), sucrose preference test (SPT), forced swim test (FST), and tail suspension test (TST) on day 14. Brain samples were collected 90 min after behavioral testing and subsequently processed for immunostaining and data analysis. (B) Schematic of brain sectioning and representative coronal levels selected for analysis at +1.69, +0.94, −1.07, −1.79, −3.39, −4.35, and −5.33 mm relative to bregma. Representative regions of interest are highlighted in cyan. (C) Representative example of image registration. The original coronal fluorescence image (upper panel) was aligned with the corresponding section of the Allen Mouse Brain Atlas (lower panel) for anatomical localization and regional quantification. Scale bars, 2000 μm. (D) Representative images of c-Fos signal detection before (upper panel) and after (lower panel) signal identification. c-Fos-immunoreactive signals are shown in red, cell nuclei are shown in blue, and detected signals are numbered. Scale bars, 20 μm. (E) Regional c-Fos density data were standardized as Z-scores and subjected to hierarchical clustering and network analysis to characterize brain-wide co-activation patterns.

Chronic Pain-Like Mouse ModelComplete Freund’s Adjuvant Injection

The inflammatory pain mouse model was generated as previously reported.18 Before intra-plantar injection of 20 µL Complete Freund’s Adjuvant (CFA), the plantar surface of the right hind paw was cleaned with 75% ethanol. An equal volume of saline was injected into the right hind paw of control mice.

Spared Nerve Injury Surgery

Neuropathic pain mice were established as described previously.19 Mice were anesthetized with sodium pentobarbital (40 mg/kg) via intraperitoneal injection. After hair removal from the left lateral thigh region and a small skin incision, the femoral muscle was carefully dissected to expose the sciatic nerve and its three major peripheral branches: the common peroneal nerve (CPN), the tibial nerve (TN), and the sural nerve. Subsequently, a surgical knot tied with nylon suture was used to secure the CPN and TN. To prevent regrowth, 2 mm segments were excised from both the CPN and TN. Sham-operated mice underwent the same surgical procedure without nerve ligation or sectioning. After the procedure, the incision was sutured with nylon thread, and mice were administered cephalosporin as prophylaxis against infection.

Pain-Like Behavior AssessmentHargreaves Test

Each mouse was habituated in a transparent square box (5 cm × 5 cm × 5 cm) for 10 min per day for 3 days before the first testing. The thermal pain threshold was measured using a Hargreaves apparatus (PL-200, TaiMeng, China). A beam of radiant heat was directed at the right hind paw (the side affected by CFA injection), and the paw withdrawal latency (PWL) was recorded over three separate trials. The average of the three trials was calculated for each mouse, with an interval of at least 5 min between trials. A cut-off latency of 20 seconds was used to avoid tissue damage.

Von Frey Test

A series of Von Frey filaments (range: 0.02–2 g) were applied according to the “up-down” method,20 and the 50% paw withdrawal threshold was used to determine neuropathic pain hyperalgesia. Filaments were gently applied perpendicularly to the plantar surface of the right hind paw ipsilateral to SNI surgery until the filament bent (≤5 sec). Withdrawal, flinching, licking, or shaking the paw was considered a positive response. The test started with a 0.02 g filament; a negative response prompted use of the next higher filament, while a positive response prompted use of the next lower filament. The testing environment and timeline were consistent with those of the Hargreaves test.

Anxiety-Like Behavior AssessmentElevated Plus Maze (EPM)

The EPM apparatus consisted of four arms (two enclosed, two open) each measuring 30 cm × 6 cm, converging on a central square platform (6 × 6 cm).21 The platform was elevated 50 cm above the ground, with the enclosed arms featuring 20 cm high walls extending into the open arms. Mice were carefully placed on the central platform facing one of the open arms and allowed to explore freely for 10 minutes. Movements were recorded by a digital camera positioned above the maze and analyzed using Ethovision XT v11.5 video tracking software (Noldus BV) to quantify the time spent in open and closed arms. The experimental environment was maintained at 24 ± 1 °C with a background noise level < 50 dB.

Open Field Test (OFT)

An open box (40 cm × 40 cm × 30 cm) equipped with a data collection and analysis system was used for the OFT.21 The observation area was divided into peripheral zones (10 × 10 cm) at each corner and a central zone (20 × 20 cm). Mice were gently placed in the center of the box and allowed to explore freely for 10 minutes. Movements were recorded by a high-definition digital camera positioned above the arena and analyzed using Ethovision XT v11.5 software to evaluate total distance traveled and time spent in each zone.

Depression-Like Behavior AssessmentSucrose Preference Test (SPT)

The SPT was performed as previously described.22 Mice were singly housed during the test. During the adaptation period, mice were exposed to two identical bottles of 1% sucrose solution for 2 days, followed by two bottles of water for 2 days. Mice underwent 24 h of water deprivation after acclimation, then were allowed to consume either 1% sucrose solution or water for 12 h. Bottle positions were rotated midway through the experiment to mitigate placement bias. Sucrose preference was calculated as the ratio of sucrose solution intake to total fluid intake, expressed as a percentage.

Forced Swimming Test (FST)

As described previously,22 mice were carefully placed in a plastic cylinder (20 cm height, 15 cm diameter) filled with water to a depth of 12 cm (temperature: 23–25°C). Observation lasted for 6 min, with immobility defined as floating with minimal or no effort to maintain equilibrium. Immobility time was measured during the last 4 minutes of the test period.

Tail Suspension Test (TST)

The TST box was constructed of polyethylene (30 × 30×30 cm).22 Mice were suspended in the center of the box with tape positioned 0.1 cm from the tip of the tail, maintaining approximately 5 cm between the mouse’s nose and the bottom of the box at rest. Mouse activity was recorded with a digital camera for 6 min, and immobility time was recorded during the last 4 min of the session.

Sampling

Mice received an intraperitoneal injection of sterile pharmaceutical-grade 1% sodium pentobarbital. Complete loss of corneal and hindpaw withdrawal reflexes was verified to confirm lethal anesthetic depth; humane euthanasia was achieved via pentobarbital overdose combined with transcardial perfusion fixation prior to thoracotomy. Animals were sequentially transcardially perfused with ice-cold 0.9% sterile saline, followed by cold 4% paraformaldehyde diluted in phosphate-buffered saline (PFA-PBS, pH 7.3–7.4). Whole brains were rapidly dissected and post-fixed in fresh 4% PFA-PBS at 4 °C for 12 h. Brains were subjected to serial cryoprotection in graded 10%, 20%, and 30% sucrose-PBS solutions at 4 °C. Afterwards, 20 μm-thick coronal frozen sections were generated spanning seven standardized anteroposterior coordinates relative to Bregma: 1.69 mm, 0.94 mm, −1.07 mm, −1.79 mm, −3.39 mm, −4.35 mm, and −5.33 mm.

Immunofluorescence

Brain slices were washed three times with PBS (10 min each) and blocked with 0.3% Triton X-100 in PBS containing 2% goat serum albumin (GSA) for 2 h at room temperature. After incubation with primary antibodies at 4°C overnight, slices were washed three times in PBS (10 min each) and incubated with secondary antibodies at 37°C for 2 h. Slices were then washed three times in PBS (10 min each) and mounted with DAPI/Antifade solution (Cat #S2110, Solarbio, China). Fluorescence images were captured using a laser scanning confocal microscope (FV3000, Olympus Japan) and analyzed with Image J software (National Institutes of Health, USA). The following reagents were used: rabbit anti-c-Fos (1:800, Cat #2250, Cell Signaling Technology, USA), coraLite594-conjugated goat anti-rabbit IgG(H+L) (1:400, Cat #SA00013-4, Proteintech, USA).

Quantification of c-Fos Signals

c-Fos analysis consisted of two steps: image registration and signal detection, both performed using Image J software. The “BigDataViewer/Big Warp” toolkit in Image J was used to align brain slice images with the Allen Mouse Brain Atlas for precise brain region segmentation. First, fluorescence staining images of coronal slices were aligned with the reference atlas. After manual alignment of the outer boundaries of brain sections and characteristic inner boundaries, the toolkit performed geometric transformation of the slices, enabling accurate image matching and segmentation of 50 brain subregions across 7 major divisions (cerebral cortex [CTX], thalamus [TH], hypothalamus [HY], amygdaloid complex [AMC], hippocampus [Hip], midbrain [MB], hindbrain [HB]). The number of c-Fos-positive cells in each brain region was manually counted using the Image J counting tool on each digitized brain section. To ensure accuracy and precision, three researchers independently counted the same brain regions in a blinded manner (samples were randomly coded to conceal group allocation), and the average value was taken as the final result. Inter-rater reliability was assessed using the intraclass correlation coefficient (ICC), with ICC > 0.85 indicating excellent consistency. c-Fos density was calculated as the number of c-Fos-positive nuclei per atlas-defined regional area. For bilateral regions, values from the left and right hemispheres were averaged for each animal before group-level analysis. For each predefined anteroposterior level, anatomically matched sections were selected across animals, and the same atlas-defined region of interest was applied consistently within each level. The actual inter-rater reliability across counted regions exceeded the predefined criterion, with ICC values ranging from 0.86 to 0.94.

Hierarchical Clustering

Relative c-Fos densities were calculated by dividing the c-Fos density of each brain region in the model group by that of the respective control group, followed by log-transformation to obtain log relative c-Fos densities. Values were further normalized to reduce variance, with a value of zero indicating no difference compared to the control group. After data normalization, hierarchical clustering was performed by calculating pairwise Euclidean distances between brain regions (smaller distances indicating greater similarity). Iterative grouping was performed, with cluster boundaries defined using the complete linkage method to ensure homogeneity within clusters. To clarify the statistical significance of differential c-Fos expression, Z-values were calculated for each condition (CFA and SNI) by dividing the log relative c-Fos density by the standard error. Positive Z-values indicate higher expression than the control group, while negative Z-values indicate lower expression.

Network Generation

To assess changes in functional brain connectivity, 37 brain regions from 6 major brain divisions (CTX, CNU, TH, HY, MB, HB) were selected. Correlation matrices were generated by calculating interregional Pearson correlation coefficients for c-Fos expression in these 37 regions. Average correlations were calculated to assess changes in functional connectivity between major brain divisions. Weighted undirected networks were constructed using a Pearson’s correlation coefficient threshold of r ≥ 0.85 (corresponding to a one-tailed significance level of p < 0.05), determined based on permutation tests to optimize network topology. Nodes in the network represent brain regions, and correlations exceeding the threshold were considered connections. Theoretical graph analysis was performed using R software (version 4.5.0), and network visualization was performed using Cytoscape (version 3.10.3). The permutation procedure randomly reassigned animal labels 1000 times to estimate the null distribution of interregional correlations and to support selection of a conservative network threshold.

Hub Region Identification

Network centrality was assessed using degree centrality, betweenness centrality, and eigenvector centrality to identify potential hub regions. Degree centrality counts the number of edges connected to a node, indicating direct influence. Betweenness centrality measures the number of shortest paths through a node, highlighting its role as a network connector. Eigenvector centrality assesses node influence based on the centrality of its connections, with higher weights assigned to nodes connected to well-connected neighbors. Hub regions were defined as nodes with centrality values in the top 10% of the network, validated by comparison with 1000 randomly generated networks to confirm statistical significance.

Statistical Analysis

For normally distributed data, one-sample t-tests and one-way analysis of variance (ANOVA) with Tukey’s post-hoc tests were used for single-sample and multi-group comparisons, respectively. For non-normally distributed data, Wilcoxon signed-rank tests and Kruskal–Wallis tests followed by Dunn’s tests were used. c-Fos densities in both experimental and control groups were log-transformed. Z-values for each brain region were calculated by normalizing these log-transformed values to the mean and standard deviation of the respective control group. Z-scores were compared to the null distribution (mean = 0) and adjusted for multiple comparisons using the Benjamini-Hochberg method (false discovery rate = 5%). Pearson correlation coefficients (R) were converted to Z-scores using Fisher’s Z-transformation, and between-group means were statistically compared. All statistical analyses were performed using GraphPad Prism 9.5.1 (GraphPad Software, USA) and R software (version 4.5.0). A p-value < 0.05 was considered statistically significant.

ResultsCFA and SNI Models Recapitulate Nociceptive Hypersensitivity Accompanied by Anxiety-Like Phenotypes

Following CFA injection, mice showed persistently reduced paw withdrawal latency in the Hargreaves test throughout the 14-day observation period (Figure 2A). CFA-treated and control mice did not differ in sucrose preference or immobility in the TST and FST (Figures 2B–D). Representative locomotor traces from the OFT are shown in Figure 2E. CFA-treated mice spent less time in the central zone (Figure 2F), whereas the total distance traveled remained unchanged (Figure 2G). Representative locomotor traces from the EPM are shown in Figure 2H. CFA-treated mice exhibited reduced open-arm exploration (Figure 2I), without a significant change in total distance traveled (Figure 2J). Similarly, SNI mice showed persistently reduced mechanical withdrawal thresholds compared with sham-operated mice throughout the 14-day observation period (Figure 3A). SNI and sham-operated mice did not differ in sucrose preference or immobility in the TST and FST (Figure 3B–D). Representative OFT locomotor traces are shown in Figure 3E. SNI mice spent less time in the central zone (Figure 3F), whereas the total distance traveled remained unchanged (Figure 3G). Representative EPM locomotor traces are shown in Figure 3H. SNI mice exhibited reduced open-arm exploration (Figure 3I), without a significant change in total distance traveled (Figure 3J).

Figure 2 The pain-like behaviors and anxiety-like behaviors in CFA model. (A) Changes in PWLs in each group (n = 8 per group; ****p < 0.0001 vs Control). (B) Preference for sucrose water in the SPT (n = 8 per group; ns, not significant vs Control). (C) Immobility time in the TST (n = 8 per group; ns vs Control). (D) Immobility time in the FST (n = 8 per group; ns vs Control). (E) Representative traveling tracks in the OFT. (F) Time spent in the central area of the OFT (n = 6 for Control, n = 5 for CFA; ****p < 0.0001 vs Control). (G) The total traveling distance in the OFT (n = 6 for Control, n = 5 for CFA; ns vs Control). (H) Representative traveling tracks in the EPM. (I) Time spent in the open arms of the EPM (n = 6 for Control, n = 5 for CFA; ****p < 0.0001 vs Control). (J) The total traveling distance in the EPM (n = 6 for Control, n = 5 for CFA; ns vs Control).

Figure 3 The pain-like behaviors and anxiety-like behaviors in SNI model. (A) Changes in 50% Von Frey threshold in each group (n = 8 per group; ****p < 0.0001 vs sham SNI). (B) Preference for sucrose water in the SPT (n = 8 per group; ns vs sham SNI). (C) Immobility time in the TST (n = 8 per group; ns vs sham SNI). (D) Immobility time in the FST (n = 8 per group; ns vs sham SNI). (E) Representative traveling tracks in the OFT. (F) Time spent in the central area of the OFT (n = 8 per group; ****p < 0.0001 vs sham SNI). (G) The total traveling distance in the OFT (n = 8 per group; ns vs sham SNI). (H) Representative traveling tracks in the EPM. (I) Time spent in the open arms of the EPM (n = 8 per group; ****p < 0.0001 vs sham SNI). (J) The total traveling distance in the EPM (n = 8 per group; ns vs sham SNI).

Collectively, these findings demonstrate that CFA- and SNI-induced chronic pain was accompanied by persistent hypersensitivity and anxiety-like behavior, without detectable depression-like behavior under the present testing conditions.

Identification of Brain Regions Activated by CFA Induced Chronic Pain

The mouse brain was categorized into 7 major regions (CTX, TH, HY, AMC, Hip, MB, HB) and 50 distinct subregions. Z-scores were calculated to compare c-Fos expression between model and control groups. Compared to controls, 28 of the 50 brain regions were significantly activated by CFA-induced inflammatory pain (Figure 4A and B). Activated regions included primary sensory and motor areas such as the primary somatosensory cortex for the hindlimb (S1HL), secondary somatosensory cortex (S2), barrel field of the primary somatosensory cortex (S1BF), somatosensory cortex for the trunk (S1Tr), primary visual cortex (V1), mediolateral area of the secondary visual cortex (V2ML), primary motor cortex (M1), and secondary motor cortex (M2). Higher cortical regions, including the prelimbic cortex (PrL), infralimbic cortex (IL), cingulate cortex 1 (Cg1), and cingulate cortex 2 (Cg2), were also significantly activated. Interestingly, the dorsal endopiriform nucleus (Den), associated with olfactory perception, was activated. In the thalamus, the paraventricular nucleus (PVT) was activated. Hypothalamic subregions activated included the paraventricular nucleus (PVN), ventral part of the paraventricular hypothalamic nucleus (Pav), dorsomedial hypothalamic nucleus (DM), medial magnocellular part of the paraventricular nucleus (PaLM), lateral hypothalamic nucleus (LH), and compact part of the dorsomedial hypothalamic nucleus (DMC). Midbrain regions, including the ventrolateral periaqueductal gray (vLPAG), dorsal raphe nucleus (DR), and ventral tegmental area (VTA), showed marked activation. The locus coeruleus (LC) in the hindbrain was also activated. Activation of diverse regions such as the dentate gyrus (DG), central amygdaloid nucleus (CeA), basomedial amygdala posterior part (BMP), and basolateral amygdala (BLA) underscores the wide-ranging effects of CFA-induced chronic inflammatory pain on various neural networks.

Figure 4 Brain regions exhibiting significant activation by CFA. (A) The 28 brain regions exhibited significant CFA activation (adjusted p < 0.05). Following the identification of brain regions significantly activated by CFA, a further selection criterion was applied to retain only the top 40% of regions as ranked by effect size (Cohen’s d). P-values are adjusted for multiple comparisons (*P < 0.05, **P < 0.01, ***P < 0.001). (B) Representative immunohistochemical staining of brain regions in CFA mice (scale bar = 100 μm).

Abbreviations: DMPAG, dorsomedial periaqueductal gray; DLPAG, dorsolateral periaqueductal gray; LPAG, lateral periaqueductal gray; vLPAG, ventrolateral periaqueductal gray; PVN, paraventricular hypothalamic nucleus; PaLM, medial magnocellular part of paraventricular hypothalamic nucleus; Pav, ventral part of paraventricular hypothalamic nucleus; DM, dorsomedial hypothalamic nucleus; DMC, compact part of dorsomedial hypothalamic nucleus; LH, lateral hypothalamic nucleus; PVT, paraventricular thalamic nucleus; S1HL, primary somatosensory cortex for the hindlimb; S1BF, barrel field of primary somatosensory cortex; S1Tr, somatosensory cortex for the trunk; S2, secondary somatosensory cortex; V1, primary visual cortex; V2ML, mediolateral area of secondary visual cortex; M1, primary motor cortex; M2, secondary motor cortex; PrL, prelimbic cortex; IL, infralimbic cortex; Cg1, cingulate cortex 1; Cg2, cingulate cortex 2; Den, dorsal endopiriform nucleus; CeA, central amygdaloid nucleus; BMA, basomedial amygdala anterior part; BMP, basomedial amygdala posterior part; BLA, basolateral amygdala; DG, dentate gyrus; DR, dorsal raphe nucleus; VTA, ventral tegmental area; LC, locus coeruleus; Aq, aqueduct; 4V, fourth ventricle; D3V, third ventricle; MHb, medial habenular nucleus; LHb, lateral habenular nucleus; MD, mediodorsal thalamic nucleus; LAD, lateral dorsal thalamic nucleus; RSA, retrosplenial agranular cortex; RSG, retrosplenial granular cortex; CA1, cornu ammonis 1; CA2, cornu ammonis 2; CA3, cornu ammonis 3; MeA, medial amygdaloid nucleus; ACo, anterior cortical amygdaloid nucleus.

Identification of Brain Regions Activated by SNI Induced Chronic Pain

Using the same criteria as for CFA, 34 brain regions were activated by SNI-induced neuropathic pain (Figure 5A and B). These included S1HL, S2, S1BF, S1Tr, V1, retrosplenial agranular cortex (RSA), retrosplenial granular cortex (RSG), dorsal part of the secondary visual cortex (V2L), Den, M1, and M2 in the CTX. Activated limbic system regions included the ventral part of the lateral septal nucleus (LSV), intermediate part of the lateral septal nucleus (LSI), basomedial amygdala anterior (BMA), lateral part of the substantia nigra (SNr), cornu ammonis 1 (CA1), cornu ammonis 2 (CA2), PrL, IL, CeA, and VTA. Midbrain regions, including the dorsomedial periaqueductal gray (DMPAG), dorsolateral periaqueductal gray (DLPAG), lateral periaqueductal gray (LPAG), vLPAG, and DR, exhibited significant activation. In the TH, HY, and HB, activation of the lateral habenular nucleus (LHb), PVT, DMC, DM, PVN, and LC was markedly elevated. These results suggest that SNI-induced chronic pain predominantly influences the limbic system, while also activating cortical regions associated with sensory processing.

Figure 5 Brain regions exhibiting significant activation by SNI. (A) The 34 brain regions exhibited significant SNI activation (adjusted p < 0.05). Following the identification of brain regions significantly activated by SNI, a further selection criterion was applied to retain only the top 40% of regions as ranked by effect size (Cohen’s d). P-values are adjusted for multiple comparisons (*P < 0.05, **P < 0.01, ***P < 0.001). (B) Representative immunohistochemical staining of brain regions in SNI mice (scale bar = 100 μm). (C) A Venn diagram displays 20 brain regions co-activated by SNI and CFA.

Abbreviations: RSA, retrosplenial agranular cortex; RSG, retrosplenial granular cortex; V2L, dorsal part of the secondary visual cortex; LSV, ventral part of the lateral septal nucleus; LSI, intermediate part of the lateral septal nucleus; LHb, lateral habenular nucleus; DMPAG, dorsomedial periaqueductal gray; DLPAG, dorsolateral periaqueductal gray; LPAG, lateral periaqueductal gray; SNr, lateral part of the substantia nigra; CA1, cornu ammonis 1; CA3, cornu ammonis 3; other abbreviations as in Figure 4.

The 20 brain regions co-activated by both CFA and SNI included S1HL, S1Tr, S1BF, V1, M1, M2, Den, PrL, IL, Cg2, PVT, PVN, DMC, DM, vLPAG, VTA, DR, CeA, BMA, and LC. Cg1, DG, LH, Pav, BLA, BMP, V2ML, and PaLM were specifically activated by CFA, while S2, RSA, RSG, V2L, CA1, CA3, LSV, LSI, LHb, LPAG, DLPAG, DMPAG, and SNr were specifically activated by SNI (Figure 5C). Our findings not only confirmed classical sensory modulation areas (eg, S1HL, PVT, PVN, PAG, LC) and emotional regulation regions (eg, PrL, IL, CeA, VTA, CA1, CA3) but also expanded understanding of chronic pain effects by revealing activation of previously unexplored regions, including Den, LSV, BMA, BMP, S1BF, S1Tr, V1, V2L, DMC, Pav, and PaLM.

Comparison of c-Fos Activation Patterns in Response to CFA and SNI

Hierarchical clustering of the 50 brain regions was performed to uncover common activation patterns and potential associations. c-Fos density in each region was normalized to the average value of the respective control group, followed by logarithmic transformation. Hierarchical clustering revealed distinct activation patterns for CFA and SNI (Figure 6A and B), with optimal clustering at a threshold of 0.7, confirmed by high silhouette coefficients (Figure 6C). Based on this threshold, two distinct clusters were identified for both CFA and SNI (Figure 6D and E). In CFA mice, Cluster 1 (mean Z-value = 0.02) showed activation similar to the control group, while Cluster 2 showed significantly increased activation. Brain regions within the CTX, TH, HY, and AMC, including S1HL, PVT, CeA, and PVN, displayed similar activation patterns (Figure 6F). SNI hierarchical clustering identified two distinct clusters, with Cluster 2 exhibiting the highest mean Z-value (5.27). This cluster centered on cortical regions and limbic systems (Figure 6G), including IL, RSA, RSG, S1HL, S2, and PrL, indicating that SNI-induced chronic pain primarily targets higher-order brain functions while also influencing fundamental nociceptive processes through effects on the thalamus, hypothalamus, and midbrain.

Figure 6 Comparison of c-Fos activation patterns in response to CFA and SNI. (A and B) Hierarchical clustering of log-relative c-Fos density for CFA and SNI, based on Euclidean distance and complete linkage. Clusters are defined by a dendrogram cut-off ratio of 0.7 and labeled numerically. (C) For both CFA and SNI, a dendrogram tree height ratio of 0.7 (dashed line) was identified as optimal, as indicated by peak Silhouette values. (D and E) Cluster Z-scores for CFA and SNI conditions are shown as mean ± SEM. Statistical analysis was performed using one-way ANOVA with Tukey’s post hoc test (ns, not significant; *p < 0.05). (F and G) Analysis of c-Fos expression is visualized by heatmaps, which display the Z-scores of its log-transformed relative density across the clustered brain regions. The red box in each heatmap highlights the cluster exhibiting the highest mean Z-score under its respective condition.

Abbreviations: CTX, cerebral cortex; TH, thalamus; HY, hypothalamus; MB, midbrain; Hip, hippocampus; AMC, amygdaloid complex; HB, hindbrain.

Network Generation and Hub Identification

Functional networks were constructed by examining c-Fos expression covariance across groups, incorporating the 50 brain subregions activated by CFA and SNI. A network model was created using a Pearson’s correlation coefficient threshold of > 0.85, and comprehensive inter-regional correlations were calculated for both CFA and SNI. Nine distinct activation modules were identified in both models (Figure 7A and B). In the 9 CFA modules, 6 hub regions were identified: DMC in Module 1, PVT in Module 2, insular cortex (CI) in Module 3, DM in Module 6, posterior part of the paraventricular hypothalamic nucleus (PaPo) in Module 7, and globus pallidus externa (GPe) in Module 8. In the 9 SNI modules, 7 hub regions were identified: S1HL in Module 1, M2 in Module 2, GPe in Module 3, CeA in Module 5, CA3 in Module 6, mediolateral area of the secondary visual cortex (V2ML) in Module 7, and substantia nigra (SN) in Module 8.

Figure 7 Generation of pain-induced networks and identification of hub regions. Network diagrams depict significant positive inter-regional correlations (Pearson’s r > 0.85, p < 0.05). Legend: Edge thickness scales with correlation strength; node size represents the degree of connectivity; node color (from dark blue to dark red) encodes betweenness centrality (darker red indicates higher betweenness centrality). (A) CFA-induced functional network with 9 modules and 6 hub regions. (B) SNI-induced functional network with 9 modules and 7 hub regions.

Abbreviations: DMC, compact part of dorsomedial hypothalamic nucleus; PVT, paraventricular thalamic nucleus; CI, insular cortex; DM, dorsomedial hypothalamic nucleus; PaPo, posterior part of paraventricular hypothalamic nucleus; GPe, globus pallidus externa; S1HL, primary somatosensory cortex for the hindlimb; M2, secondary motor cortex; CeA, central amygdaloid nucleus; CA3, cornu ammonis 3; V2ML, mediolateral area of secondary visual cortex; SN, substantia nigra; other abbreviations as in Figures 4 and 5.

Discussion

Using a common c-Fos mapping and covariance-analysis pipeline, we found that CFA and SNI produced convergent regional recruitment but different network organizations. Both models activated 20 common regions and induced persistent nociceptive hypersensitivity accompanied by anxiety-like behavior. However, their hub distributions differed markedly. The CFA network showed greater centrality in thalamic and hypothalamic regions, whereas the SNI network involved a broader combination of sensory, motor, limbic, and hippocampal regions. This distinction is clinically relevant because chronic pain and mental disorders frequently coexist and contribute substantially to disability.23,24 Nevertheless, these patterns represent model-associated co-activation architectures and should not be interpreted as directed or causal neural circuits.

The relative subcortical emphasis of the CFA network is consistent with the established role of ascending thalamic and hypothalamic pathways in inflammatory pain. Parabrachial projections transmit nociceptive information to forebrain regions involved in sensory, autonomic, and affective responses.25 The PVT contributes to persistent pain and its emotional consequences, whereas the hypothalamus integrates nociception with endocrine, autonomic, and defensive responses.26–28 Ascending pathways involving the dorsal raphe, thalamocortical system, and parabrachial complex further regulate the transmission and appraisal of nociceptive signals.29–31 In our analysis, PVT, DMC, DM, and PaPo occupied hub positions. The presence of CI and GPe among the CFA hubs also indicates that this organization was not exclusively thalamic or hypothalamic. Collectively, these findings suggest that CFA-induced inflammatory pain is associated with coordinated recruitment of systems regulating nociception, homeostasis, and defensive behavior. Because the roles of DMC, DM, and PaPo in inflammatory pain remain incompletely characterized, their hub status should be considered hypothesis-generating.

The SNI network displayed a more distributed hub profile involving S1HL, M2, GPe, CeA, CA3, V2ML, and SN. This pattern is compatible with evidence that neuropathic pain engages cortical and limbic circuits responsible for sensory discrimination, aversion, cognition, and pain-related learning. The CeA encodes the unpleasant dimension of pain and contributes to persistent affective responses.32 Hippocampal dysfunction has also been implicated in neuropathic pain-associated cognitive impairment, while motor cortical circuits can influence pain perception and analgesia.33,34 Thus, the centrality of CeA, CA3, and M2 in the SNI network is consistent with previous circuit-level observations. The additional involvement of S1HL and V2ML suggests that neuropathic pain may be accompanied by broader reorganization of sensory cortical processing.

Our findings also show partial agreement with previous brain-wide investigations of neuropathic pain. A longitudinal imaging study conducted two weeks after SNI reported altered activity in thalamic and olfactory regions, together with changes in thalamic–motor cortical connectivity. Some of these alterations were supported by c-Fos measurements.35 Both studies therefore indicate that neuropathic pain involves distributed cortical–subcortical reorganization rather than isolated activation of canonical nociceptive nuclei. However, the specific regions assigned the greatest network importance were not identical. This difference may reflect variation in species, brain parcellation, imaging modality, behavioral context, or the temporal properties of the measured signals.

The SNI-associated pattern is further supported by studies of individual pain circuits. Enhanced S1HL–BLA signaling has been associated with neuropathic pain-related affective dysfunction.36 Reorganization of thalamocortical and cingulate pathways contributes to persistent sensory hypersensitivity, whereas somatosensory cortical projections to the CeA link nociceptive processing with autonomic and immune responses.37,38 These observations are consistent with the simultaneous involvement of S1HL and CeA in our SNI network. The centrality of CA3 may additionally indicate hippocampal participation in the contextual or mnemonic dimensions of neuropathic pain. However, c-Fos covariance alone cannot determine the specific behavioral process represented by this region.

Despite their different hub distributions, CFA and SNI produced increased c-Fos expression in 20 common regions. These included areas associated with sensory discrimination, such as S1HL, PVT, and vLPAG, and regions involved in affective and motivational processing, including PrL, IL, CeA, and VTA.13,26,32,36–51 Their common recruitment may contribute to the persistent hypersensitivity and anxiety-like behavior observed in both models. Importantly, common regional activation does not imply that the same neuronal populations, synaptic mechanisms, or circuit directions were engaged. Moreover, the absence of significant changes in the SPT, TST, and FST indicates only that depression-like behavior was not detected under the present experimental conditions. It does not exclude the possibility that depressive phenotypes could emerge at later stages or be detected using other behavioral paradigms.

The brain-wide analysis also identified pain-associated activation in comparatively understudied regions, including Den, LSV, and BMP. Den has been linked primarily to olfactory and limbic processing, although interactions between olfaction and nociception have been reported.52,53 The lateral septum participates in reward, stress, and social behavior,54,55 while BMP receives hippocampal input and may integrate contextual information with amygdala-dependent responses.56 Their recruitment may therefore reflect broader effects of persistent pain on salience, arousal, contextual processing, or behavioral state. These interpretations remain provisional because the available evidence does not establish pain-specific functions for these regions. Cell-type-resolved recording and manipulation will be required to determine whether they directly regulate nociceptive or affective behavior.

Overall, the comparison supports a relative, rather than absolute, distinction between the two models. CFA was associated with greater network prominence of thalamic and hypothalamic nodes, whereas SNI involved a more widely distributed sensory and corticolimbic hub configuration. These patterns may reflect persistent peripheral inflammatory input in CFA and injury-induced plasticity within sensory–affective circuits after SNI. However, the terms “bottom-up” and “top-down” should be regarded only as anatomical shorthand for the observed hub distributions. The present data do not establish the direction of information flow or hierarchical control within either network.

Several limitations should be considered. First, c-Fos is an indirect marker of recent neuronal activation and does not identify the contributing cell types, synaptic mechanisms, or inhibitory and excitatory components. The animals underwent their final behavioral assessments before tissue collection. Consequently, the measured c-Fos patterns may reflect both the ongoing pain state and neuronal activity associated with recent exploration, arousal, anxiety, or motor behavior. Second, the networks were derived from inter-animal correlations in regional c-Fos counts. Such covariance does not constitute direct functional connectivity. Network properties may also depend on the selected correlation threshold and the subset of regions included in the analysis. Finally, causal involvement of the identified regions was not examined through electrophysiological recording or region- and cell-specific manipulation.

Conclusions

In summary, comparative brain-wide c-Fos mapping revealed distinct but partially overlapping neuronal activation patterns in CFA- and SNI-induced pain. CFA was associated predominantly with subcortical activation, whereas SNI showed greater cortico-limbic involvement, with both models engaging a shared set of brain regions. Because c-Fos mapping and correlation-based network analysis are observational, these findings do not establish the functional or causal involvement of the identified regions. Instead, we provide a descriptive brain-wide activation map and identify candidate regions and networks for future cell-type- and circuit-specific functional validation.

Animal Ethics

The Ethics Committee at Chengdu University of Traditional Chinese Medicine provided full approval for this research (No.2025011).

Data Sharing Statement

Data is made available from the corresponding author upon reasonable request.

Author Contributions

Xiaolu Fan conceived the study, designed the methodology, and drafted the original manuscript. Liuxuan He and Ruizhu Zhou curated the data and performed the formal analyses. Shuai Hou conducted the investigations and validation. Mengling Cheng provided resources and supervised the study. All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising, or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Funding

Natural Science Foundation of Hubei Province (No. 2026AFB801).

Disclosure

The authors declare no competing interests.

References

1. Bäckryd E, Alföldi P. Långvarig smärta – relationen till ångest och depression är komplex [Chronic pain – the relationship with anxiety and depression is complex]. Lakartidningen. 2023;120:23010.

2. Hearn JH, Munday I, Bullo S, Rogers K, Newton-John T, Kneebone I. Metaphorical markers of pain catastrophizing, depression, anxiety, and pain interference in people with chronic pain. J Pain. 2025;26:104733. doi:10.1016/j.jpain.2024.104733

3. Ji RR, Nackley A, Huh Y, Terrando N, Maixner W. Neuroinflammation and central sensitization in chronic and widespread pain. Anesthesiology. 2018;129(2):343–17. doi:10.1097/ALN.0000000000002130

4. Kosek E, Clauw D, Nijs J, et al. Chronic nociplastic pain affecting the musculoskeletal system: clinical criteria and grading system. Pain. 2021;162(11):2629–2634. doi:10.1097/j.pain.0000000000002324

5. Sharma H, Khan S, Lohani A, Chandra P, Sachan N, Baldi A. Anandamide as a therapeutic target for alleviating neuropathic pain and inflammation in rat models. Curr Neurovasc Res. 2025;22(2):167–181. doi:10.2174/0115672026391315250822063941

6. Pawar HD, Chaudhari S, Barote PG, Nakhate KT, Sherikar A, Goyal SN. Ameliorative potential of ethyl gallate in a rat model of chronic constriction injury-induced neuropathic pain. Curr Neurovasc Res. 2025;22(5):410–423. doi:10.2174/0115672026409563251204065841

7. Woo CW, Chang LJ, Lindquist MA, Wager TD. Building better biomarkers: brain models in translational neuroimaging. Nat Neurosci. 2017;20(3):365–377. doi:10.1038/nn.4478

8. Kragel PA, Koban L, Barrett LF, Wager TD. Representation, pattern information, and brain signatures: from neurons to neuroimaging. Neuron. 2018;99(2):257–273. doi:10.1016/j.neuron.2018.06.009

9. Smith JA, Ji Y, Lorsung R, et al. Parabrachial nucleus activity in nociception and pain in awake mice. J Neurosci. 2023;43(31):5656–5667. doi:10.1523/JNEUROSCI.0587-23.2023

10. Nasseef MT, Ma W, Singh JP, et al. Chronic generalized pain disrupts whole brain functional connectivity in mice. Brain Imaging Behav. 2021;15(5):2406–2416. doi:10.1007/s11682-020-00438-9

11. Li YJ, Du WJ, Liu R, et al. Paraventricular nucleus–central amygdala oxytocinergic projection modulates pain-related anxiety-like behaviors in mice. CNS Neurosci Ther. 2023;29(11):3493–3506. doi:10.1111/cns.14282

12. Yan Y, Zhu M, Cao X, et al. Thalamocortical circuit controls neur