AMPA receptor, neuroimaging, and chronic pain

Abstract

^{AMPA receptors (AMPARs) are key molecular mediators of fast excitatory neurotransmission and synaptic plasticity in the central nervous system. Increasing evidence indicates that maladaptive regulation of AMPAR trafficking, subunit composition, and phosphorylation contributes to central sensitization underlying chronic pain. Experimental models demonstrate that inflammation or nerve injury induces GluA2 internalization and insertion of calcium-permeable AMPARs in the spinal dorsal horn and supraspinal regions, such as the anterior cingulate cortex, insula, amygdala, and nucleus accumbens. These alterations enhance excitatory transmission and sustain pain perception. Translational advances in molecular neuroimaging have enabled visualization of such mechanisms in vivo. The positron emission tomography (PET) radioligand [11C]K-2 allows quantitative mapping of AMPAR density in the human brain, while FDG-PET captures pain-related metabolic abnormalities within the pain matrix, including the anterior cingulate, insula, and thalamus. Integrating AMPAR-PET with MRI provides a multiscale framework linking molecular receptor dynamics with network-level reorganization. Furthermore, longitudinal PET and MRI studies demonstrate partial reversibility of structural and functional brain alterations following effective treatments, suggesting that pain-related neuroplasticity is modifiable. These insights highlight the potential of AMPAR-targeted imaging biomarkers to objectively characterize chronic pain mechanisms, guide therapeutic stratification, and evaluate treatment response. Establishing standardized multimodal imaging protocols and large collaborative databases will be essential to translate these discoveries into mechanism-based diagnostics and interventions for chronic pain .}

Introduction

Excitatory glutamatergic AMPARs constitute a critical molecular substrate for information transfer within the central nervous system [1, 2]. Synaptic plasticity associated with memory and learning involves increased presentation of AMPARs at synaptic sites (long‑term potentiation, LTP), whereas the converse occurs during memory extinction (long‑term depression, LTD) [2, 3]. Sustained alterations in AMPAR expression have been consistently observed in the brains and spinal cords of diverse animal models of chronic pain [4–6]. Numerous studies demonstrate that increased AMPAR protein abundance and AMPAR-dependent synaptic currents exacerbate pain‑related behaviors, and that interfering with these increases can relieve pain in animal models [5]. However, until recently, it was unknown whether AMPAR density changes in the brains of patients with chronic pain, because no method could quantify AMPARs in vivo in humans.

We recently developed the PET radioligand [¹¹C]K‑2, which enables in vivo visualization of AMPARs in the human brain [7]. Extending this work to patients with chronic pain and conducting parallel imaging in animal models provides a translational framework for understanding pain pathophysiology. Moreover, alterations in AMPAR densities may contribute to changes in brain morphology and large‑scale neural networks detectable with MRI. Integrative PET–MRI studies may, therefore, yield a comprehensive account of chronic pain mechanisms.

Here, we summarize core biology of AMPARs relevant to chronic pain in animal models and synthesize what neuroimaging—particularly MRI and PET—has revealed about chronic pain in the human brain.

Structure and function of AMPA receptors

Information transfer between neurons occurs at synapses through the release and reception of neurotransmitters. In excitatory neurotransmission, glutamate is released from presynaptic terminals following action potential propagation, causing exocytosis of synaptic vesicles into the synaptic cleft. Released glutamate binds to glutamate receptors on the postsynaptic membrane, allowing cation influx, depolarizing the postsynaptic membrane, and eliciting excitation [1]. Glutamate receptors include N-methyl-D-aspartate (NMDA), AMPA, and kainate receptors. In the mammalian central nervous system, the majority of fast excitatory synaptic transmission is mediated by AMPARs [2]. AMPARs comprise four subunits (GluA1–GluA4) and generally assemble as heterotetramers. In adult principal neurons, including those in the hippocampus and cerebral cortex, GluA1 and GluA2 predominate, with GluA1/2 or GluA2/3 heteromers being most abundant [8].

Developmental regulation and calcium permeability

In many neurons, GluA1 expression is developmentally regulated, with low expression in early postnatal periods. In contrast, the transiently high expression of GluA4 promotes formation of GluA4 homomers in neonates, which are calcium‑permeable, because incorporation of GluA2 reduces calcium permeability [9]. This heightened calcium permeability is implicated in synaptic maturation and circuit development. During neonatal periods, rapid network organization occurs in response to external inputs, largely via synaptic changes: AMPARs accumulate at synapses that regulate specific neural activities, increasing synaptic transmission efficiency. The abundance of GluA4 homomers contributes to elevated intracellular calcium during this phase. Similar dynamics are observed in adulthood. For instance, during LTP, increased glutamate release at specific synapses can transiently recruit calcium‑permeable GluA1 homomers [10]. The ensuing calcium influx triggers phosphorylation of proteins that modify AMPARs, and GluA1/2 and GluA2/3 receptors are subsequently inserted into synapses, increasing synaptic AMPAR content.

Subunit expression is tightly regulated developmentally and varies by brain region and cell type. For example, GluA1 mRNA is evident in postnatal day‑5 cortex but is scarcely translated into protein; likewise, GluA4 mRNA is detectable in postnatal day‑14 cortex with minimal protein translation [9]. Viral gene transfer experiments have shown limited synaptic trafficking after neonatal GluA1 introduction, whereas GluA4 expression yields robust trafficking. In the cerebellum, GluA4 expression remains dominant into adulthood, in marked contrast to the cerebral cortex, where protein‑level GluA4 is not observed at the same stage [11].

Synaptic plasticity and underlying mechanisms

AMPAR dynamics are reversible and bidirectional, underpinning synaptic plasticity that remodels neural networks. Although LTP is often studied in vitro [12], similar AMPAR trafficking occurs in vivo with learning and experience. In rodents, spatial learning drives AMPAR trafficking in the hippocampus and requires phosphorylation of mediators such as α‑calcium/calmodulin-dependent kinase II (CaMKII) [13]. Sensory experience—for example, whisker input—induces GluA1 trafficking in the barrel cortex, enhancing sensory function [14]. Conversely, LTD removes AMPARs from synapses and contributes to memory erasure [15, 16]. During memory updating, mGluR1 activation can remove AMPARs, followed by trafficking of newly inserted receptors [17].

Mechanistically, strong postsynaptic input favors transient formation of calcium‑permeable GluA1 homomers, permitting calcium entry; depolarization also relieves Mg2 + block of NMDA receptors, allowing substantial calcium influx. Calcium activates protein kinase C (PKC) and protein kinase A (PKA) [3]: PKC phosphorylates GluA1 at Ser831 and PKA at Ser845, altering electrostatic charge and subsequent structural change to promote interactions with motor proteins that deliver GluA1 to synapses. CaMKIIα can also phosphorylate Ser831. LTD often begins with low‑frequency stimulation that elicits smaller, NMDA‑dependent calcium signals to activate PP1 and calcineurin [18]. Furthermore, synaptic removal of AMPA receptor can also be induced by antagonizing voltage-dependent calcium channels or AMPA receptors itself, which indicates multiple sources of calcium influx contributing to LTD. Endocytosis from clathrin‑coated pits removes AMPARs—particularly GluA2/3‑containing receptors [19]—with endocytic proteins binding via PSD-95/ Dlg/ZO-1 (PDZ) domains; Glutamate receptor interacting protein (GRIP)-1 participates in cellular uptake [20].

AMPA receptors and chronic pain

Trafficking, subunit switching, and phosphorylation‑dependent modulation of AMPARs are indispensable for physiological plasticity in learning and memory. In chronic pain, these mechanisms become dysregulated and drive maladaptive central sensitization. Below, we outline spinal and supraspinal AMPAR alterations that contribute to chronic pain onset and maintenance.

AMPA receptor dysfunction underlying central sensitization

Chronic pain encompasses persistent inflammatory, tissue/nerve injury-related, and nociplastic conditions characterized by central nervous system dysfunction [4]. Central sensitization—a specific form of plasticity marked by increased membrane excitability, enhanced synaptic efficacy, and disinhibition—amplifies nociceptive input and can generate spontaneous pain [21, 22]. AMPARs mediate rapid excitatory neurotransmission along nociceptive pathways [5]. Initial AMPAR‑mediated depolarization relieves NMDA receptor Mg2 + block, enabling NMDA activation. Resultant LTP, expressed as enhanced AMPAR function, sustains pain states [6].

Pain‑related AMPA receptor changes in the spinal dorsal horn

In the dorsal horn, GluA1 is enriched in laminae I–II, whereas GluA2 is distributed throughout but abundant in medial layer II [23, 24]. GluA2‑containing AMPARs are relatively calcium‑impermeable, while GluA2‑lacking receptors (e.g., GluA1 homomers or GluA1/3 heteromers) are calcium‑permeable [25]. In chronic pain, a critical pathological change occurs: internalization of GluA2 subunits from the membrane and insertion of GluA1 subunits into the membrane [26]. This compositional switch increases the proportion of calcium-permeable AMPA receptors in the postsynaptic membrane, promoting calcium influx into dorsal horn neurons. Consequently, further synaptic plasticity and activation of signaling cascades are triggered, enhancing nociceptive neurotransmission and forming and maintaining chronic pain pathologies, such as hyperalgesia and allodynia. PKC‑dependent phosphorylation of GluA2‑Ser880 reduces its interaction with GRIP and promotes this switch [5, 26]. PKCα is particularly implicated in chronic pain maintenance; intrathecal inhibition by Go6976 suppresses mechanical allodynia and thermal hyperalgesia dose-dependently in inflammatory models [27]. N-ethylmaleimide-sensitive factor (NSF) also plays an important role in central sensitization [28]. NSF is involved in the transport of GluA2 subunits to cell membrane and contributes to spinal central sensitization through changes in GluA2 subunit composition, triggered by inflammation.

AMPARs undergo rapid constitutive cycling and lateral diffusion between extra‑ and perisynaptic domains [29–31]. Painful stimulation regulates these processes in spinal nociceptive neurons [32, 33]. Visceral pain stimulation has been reported to induce marked trafficking of GluA1 subunits from cytoplasm to cell membrane [32].

Inflammatory pain models using Complete Freund’s Adjuvant (CFA) demonstrate consistent patterns of AMPA receptor changes in the spinal dorsal horn [34]. CFA injection causes marked changes in AMPA receptor composition and trafficking. The time course of these changes closely correlates with behavioral pain indices: PKC activation begins within hours of inflammation onset, with GluA2 internalization reaching maximum at 24 h. Electrophysiological analysis has confirmed that dorsal horn neurons after CFA treatment show increased amplitude of AMPA receptor-mediated excitatory postsynaptic currents and exhibit inward rectification characteristics typical of calcium-permeable AMPA receptors [34–36].

Neuropathic pain models reveal unique patterns of AMPA receptor changes following peripheral nerve injury. In the sciatic nerve ligation model, upregulation of AMPA receptor expression is observed, with AMPA receptor immunoreactivity labeling in the superficial laminae peaking at day 14 after nerve ligation and decreasing by day 35 [37]. This temporal pattern matches the changes in mechanical and thermal hyperalgesia. Nerve injury increases GluA1 subunit expression in laminae I–II neurons immunohistochemically while promoting GluA2 subunit internalization [38], which indicates a shift from GluA2-containing to GluA2-lacking calcium-permeable AMPA receptors. In the spinal nerve ligation (SNL) model, studies with LY3130481, a selective inhibitor of transmembrane AMPA receptor regulatory protein γ8-containing AMPA receptors, showed that single administration had no effect; however, 5 days of repeated administration produced tactile allodynia suppression effects comparable to gabapentin [39]. These results demonstrate the importance of AMPA receptors in neuropathic pain maintenance while suggesting that sustained receptor function inhibition is necessary for therapeutic effect development.

Pain‑related supraspinal AMPA receptor changes

Chronic pain engages brain regions that encode sensory, affective, and cognitive dimensions of pain. The anterior cingulate cortex (ACC), insula, nucleus accumbens, and amygdala exhibit distinct AMPAR plasticity patterns.

Anterior cingulate cortex (ACC)

The anterior cingulate cortex (ACC) plays a key role in the emotional and cognitive dimensions of pain. Peripheral nerve injury enhances AMPA receptor-mediated excitatory postsynaptic currents (EPSCs) in pyramidal neurons in layers II–III and V of the ACC [40, 41], accompanied by altered rectification properties and increased sensitivity to calcium-permeable AMPA receptor inhibitors.

Biochemical studies show increased membrane-bound GluA1-containing AMPA receptors, with no significant change in GluA2/3. Peripheral inflammation similarly elevates GluA1 [42]. This enhanced AMPAR-mediated transmission facilitates NMDA receptor activation, and nerve injury further upregulates GluN2B subunit-containing NMDA receptors in the ACC (39, 40). Genetic studies reveal that GluA1, but not GluA2, is essential for LTP in the ACC: peptides that interfere with GluA1 trafficking inhibit early LTP, and GluA1-deficient mice lack LTP, whereas GluA2-deficient mice do not [43, 44].

In chronic visceral pain in the ACC, GluA1 and GluA2/3 expression increases [45], likely via enhanced glutamate release and NMDA-dependent AMPA receptor trafficking. Imaging studies in IBS patients show visceral stimulation activates the ACC, insula, and amygdala (56–59), supporting the ACC’s dual role in sensory-discriminative and affective pain processing. Elevated GluA1 and GluA2/3 in membrane fractions suggest receptor redistribution [45].

Chronic pain involves impaired LTD and enhanced LTP [46–49]. Both NMDA- and mGluR-dependent LTD exist, but repetitive stimulation-induced LTD is suppressed after injury, suggesting LTP saturation and LTD loss. Presynaptic and postsynaptic LTP can coexist [50]: postsynaptic LTP may encode pain unpleasantness, whereas presynaptic LTP may encode pain-related anxiety, reflecting distinct but additive mechanisms [45].

Insula

The insula is critically involved in pain perception and sensory processing. Functional neuroimaging studies in humans demonstrate that the insula is activated by noxious stimuli, and direct electrical stimulation elicits pain sensations, supporting its essential role in acute pain processing [51].

In neuropathic pain models, AMPA receptor-mediated excitatory synaptic transmission in the insula is enhanced. Following peripheral nerve injury, excitatory transmission mediated by both AMPA and NMDA receptors increases. Notably, GluA1 subunits—but not GluA2/3—are selectively upregulated at postsynaptic membranes. This depends on PKA-mediated phosphorylation of GluA1 at serine 845. A-kinase anchoring protein 79/150 (AKAP79/150) and PKA translocate to synapses after nerve injury, and genetic deletion of adenylyl cyclase 1 (AC1) inhibits this translocation and the subsequent upregulation of synaptic GluA1-containing AMPA receptors [52]. Pharmacological inhibition of calcium-permeable AMPA receptors in the insula reduces nerve injury-induced behavioral sensitization, underscoring their functional relevance.

Nucleus accumbens

Functions in reward processing and emotional regulation and is implicated in the affective dimensions of chronic pain [53]. In both inflammatory and neuropathic pain models, increased GluA1 expression and formation of GluA2-lacking AMPA receptors have been reported.

In chronic neuropathic pain, increased GluA1 expression in the nucleus accumbens contributes to GluA2-lacking receptor formation [53, 54]. Transmission through these receptors appears to have a protective function, mitigating pain-induced depressive symptoms [53, 54]. Time-course analyses show no changes 1–2 days after CFA-induced inflammatory pain, a selective increase in GluA1 at day 7, and recovery to baseline levels by day 14 [53]. Electrophysiological analyses confirm increased rectification indices of AMPA receptor-mediated EPSCs in nucleus accumbens neurons at day 7, indicating the formation of calcium-permeable AMPA receptor currents [53, 55, 56]. These findings suggest that persistent inflammatory pain promotes GluA2-lacking AMPA receptor formation in the nucleus accumbens, providing an endogenous mechanism for modulating pain-related depression [53].

Amygdala

The amygdala, crucial for emotional memory and fear/anxiety regulation, receives enhanced nociceptive input from the parabrachial nucleus in chronic pain [57, 58]. In inflammatory pain models, increased GluA1 expression has been reported in the central nucleus of the amygdala (CeA), contributing to enhanced pain sensitivity [59, 60]. Stress can suppress hippocampal LTP while facilitating amygdala LTP [61], highlighting region-specific differences in synaptic plasticity. Increased calcium‑permeable AMPARs in the basolateral amygdala contribute to depression‑like behaviors under chronic stress [62].

In chronic pain models, ascending nociceptive input from the spinal cord via the brainstem is enhanced in the amygdala. Inflammatory and neuropathic pain both increase excitatory synaptic input from the parabrachial nucleus to the central amygdala [57, 58]. However, Fu et al. demonstrated that pain-related synaptic plasticity in the CeA is mediated primarily by PKA and ERK signaling pathways acting on NMDA receptors, rather than direct alterations in AMPAR subunit composition [58]. In neuropathic pain, the lateral/basolateral amygdala-CeA pathway exhibits enhanced long-term depression (LTD) mediated by GluA2-containing AMPAR endocytosis, which contributes to comorbid aversive and depressive symptoms [63]. Insights from stress research show that stress loads that impair hippocampal LTP enhance amygdala LTP [61], highlighting region-specific differences in AMPA receptor regulation and synaptic plasticity.

Taken together, these findings demonstrate that chronic pain induces region-specific alterations in AMPAR subunit composition across multiple components of the pain matrix. While a common pattern of increased GluA1 expression and/or decreased GluA2 expression leading to enhanced calcium permeability is observed, the specific pain models, time courses, and signaling mechanisms show notable regional diversity (Table 1).

Table 1.

Regional AMPA receptor alterations in chronic pain models. This table summarizes region-specific alterations in AMPA receptor subunit composition, calcium permeability, and associated signaling mechanisms observed in various animal models of chronic pain. Data are derived from biochemical, electrophysiological, and immunohistochemical studies in rodent models

Region	Pain model	AMPAR subunit changes	Ca2⁺ permeability	Signaling molecules	References
Spinal dorsal horn	Inflammatory (CFA)	↑GluA1 membrane insertion, ↓GluA2 (internalization)	Increased	PKC, NSF, TNF-α, IL-1β	[25, 33–35]
	Neuropathic (SNL/nerve injury)	↑GluA1 (laminae I-II), ↓GluA2 (internalization)	Increased	PKCα, TARP γ8	[26, 36–38]
	Acute noxious stimulation	GluA1 translocation to plasma membrane	Variable	−	[22, 31, 32]
Anterior cingulate Cortex (ACC)	Neuropathic (nerve injury)	↑GluA1 membrane-bound, GluA2/3 unchanged	Increased	GluN2B, PKA, ERK	[39, 40, 43]
	Inflammatory	↑GluA1 synaptic expression	Increased	−	[41]
	Chronic visceral pain	↑GluA1 and GluA2/3 expression	Variable	NMDA-dependent	[44]
Insula	Neuropathic (nerve injury)	↑GluA1 (postsynaptic), GluA2/3 unchanged	Increased	PKA, AC1, AKAP79/150	[51]
Nucleus accumbens	Neuropathic	↑GluA1, formation of GluA2-lacking receptors	Increased	−	[52, 53]
	Inflammatory (CFA, day 7)	↑GluA1, GluA2-lacking receptor formation	Increased	−	[52]
Amygdala	[CeA] Inflammatory (CFA)	↑GluA1 expression	Increased	PKA, ERK, MeCP2-mediated transcriptional regulation	[57–59]
	[BLA–CeA pathway] Neuropathic (SNL)	Enhanced LTD via GluA2 endocytosis, ↑pGluA2-Ser880	Reduced	GluA2 endocytosis, LTD mechanism	[61]

↑ increased, ↓ decreased, — not reported or not applicable, AC1 adenylyl cyclase 1, AKAP A-kinase anchoring protein, AMPAR α-amino-3-hydroxy- 5-methyl-4-isoxazolepropionic acid (AMPA) receptor, CFA Complete Freund’s Adjuvant, CP-AMPAR calcium-permeable AMPA receptor, ERK extracellular signal-regulated kinase, IL interleukin, NSF N-ethylmaleimide-sensitive fusion protein, PKA protein kinase A, PKC protein kinase C, SNL spinal nerve ligation, TARP transmembrane AMPA receptor regulatory protein, TNF tumor necrosis factor, CeA central nucleus of amygdala, BLA basolateral amygdala, LTD long-term depression

Pharmacological interventions targeting AMPA receptors

Although AMPA receptor antagonists are promising for chronic pain treatment, their clinical development remains challenging. Preclinical studies have identified AMPA receptors as key mediators of nociceptive transmission and central sensitization, making them attractive therapeutic targets [4, 5]. While selective calcium-permeable AMPA receptor inhibitors and broader antagonists have been developed, clinical translation has proven difficult.

Clinical trials have shown mixed results. NS1209 improved certain neuropathic pain symptoms, including brush-evoked and cold allodynia, but failed to reduce spontaneous current pain, the primary endpoint [64]. The main limitation of nonselective antagonists is their widespread CNS distribution, causing dose-limiting side effects such as visual disturbances (95%) and sedation (40%) [39].

More selective approaches, including targeting TARPγ8-containing receptor subtypes, have shown analgesic efficacy without motor impairment in preclinical models, suggesting the potential to inhibit pain pathways while preserving normal function [65]. However, many clinical studies directly targeting AMPA receptors have yielded limited benefits across neurological diseases.

Chronic pain and neuroimaging

The molecular mechanisms of AMPA receptor dysregulation, including receptor trafficking, subunit switching, and phosphorylation-dependent modulation, operate at synaptic and cellular levels to drive central sensitization. While animal models have provided detailed molecular characterization in specific regions, neuroimaging techniques have the potential to reveal AMPAR-related changes across broader pain matrix regions in living human brains [66]. Advances in neuroimaging have transformed our understanding of chronic pain pathophysiology, particularly regarding nociplastic pain, and offer emerging possibilities for objective diagnosis and treatment monitoring.

A paradigm shift in chronic pain

The transmission of nociceptive signals from peripheral nerves to the cortex has long been established. Recent evidence, however, emphasizes altered brain function as a key factor amplifying pain sensitivity and promoting widespread pain. Pain once considered idiopathic or psychosomatic may actually reflect central neuroplastic changes.

This paradigm shift has been driven by advances in neuroimaging, which enable objective visualization of central alterations in vivo. The concept of nociplastic pain—a distinct pain category—likely would not have emerged without these data. Neuroimaging, thus, provides a powerful means to elucidate chronic pain pathophysiology.

Psychological factors, such as fear, anxiety, anger, and stress, contribute to pain sensitization and persistence. Previously, pain unexplained by conventional tests was attributed to psychosocial causes, but these symptoms are increasingly recognized as manifestations of central plasticity. Objective assessment of such changes is becoming clinically important.

Although nociplastic pain is now acknowledged, the lack of diagnostic tools remains a challenge. Clinicians often suspect it when patients exhibit signs of central sensitization (e.g., allodynia, widespread pain), psychosocial distress, or pain resistant to conventional therapy. Conditions like fibromyalgia and somatic symptom disorder are characteristic examples. However, diagnoses are still based on exclusion, lacking objective measures of neuroplasticity. Incorporating neuroimaging into clinical assessment could clarify underlying mechanisms and guide targeted treatment strategies.

MRI studies: brain alterations in primary chronic pain

Despite the increasing recognition of nociplastic mechanisms, a few neuroimaging studies have directly investigated nociplastic pain, reflecting the novelty of the term and the absence of standardized diagnostic criteria. Therefore, the following sections will interpret nociplastic pain in the context of neuroplastic changes observed in primary chronic pain and highlight relevant neuroimaging findings in this patient population.

Nociceptive input from peripheral sensory nerves ascends via the spinothalamic tract to the thalamus, and then projects to the primary somatosensory cortex (S1). Activity in S1 is further transmitted to the secondary somatosensory cortex (S2), posterior parietal regions, and the posterior insula. The posterior insula projects to the anterior insula, anterior cingulate cortex (ACC), amygdala, prefrontal cortex (PFC), and parahippocampal gyrus [67]. These interconnected regions collectively comprise the pain matrix.

Previously, the possibility of structural brain changes was suggested in chronic pain using MRI, reporting significant reductions in gray-matter density within the dorsolateral prefrontal cortex (DLPFC) and thalamus in patients with chronic low back pain compared to healthy controls [66]. Subsequent voxel-based morphometry (VBM) studies have demonstrated structural changes in multiple brain regions across various chronic pain conditions, many of which overlap with areas activated by acute nociceptive stimuli, consistent with the pain matrix. Other neuroimaging studies have demonstrated that chronic pain is associated with both structural and functional alterations within this network. For example, patients with chronic low back pain exhibit reduced gray-matter density in the bilateral DLPFC and somatosensory cortices (S1/S2) [68, 69]. In chronic tension-type headache, reductions in gray-matter density have been observed in the ACC, posterior cingulate cortex (PCC), and anterior insula, with these changes correlating with disease duration [69]. Longitudinal analyses have further revealed reductions in hippocampal gray-matter density in patients whose subacute back pain transitioned to chronic pain, suggesting limbic structural alteration associated with pain chronification [70]. Similarly, patients with fibromyalgia exhibit reduced gray-matter density in the left parahippocampal gyrus, PCC, left insula, and medial PFC (mPFC) [71].

Furthermore, functional MRI studies have revealed altered pain-related activity in patients with chronic pain. Several studies have demonstrated heightened activity in the somatosensory cortices (S1/S2), ACC, insula, and PCC compared with healthy controls [72]. In contrast, task-based fMRI have identified reduced responsiveness within reward-related regions such as the nucleus accumbens. Baliki et al. reported that chronic back pain patients exhibited abnormal nucleus accumbens responses to thermal stimuli, showing negative offset activation where healthy controls displayed positive responses predicting pain relief [73]. These finding highlight that chronic pain involves hypoactivity in reward-related circuits alongside hyperactivity within the pain matrix. Moreover, patients with chronic prostatitis/chronic pelvic pain syndrome have shown decreased gray-matter density in the ACC [74], as well as altered resting-state activity in the ACC, insula, and mPFC [75].

Collectively, these findings indicate that chronic pain is accompanied by alterations in both brain structure and function within various regions of the pain matrix. Figure 1 and Table 2 summarize alterations in AMPAR, gray-matter density (GMD), and blood oxygen level-dependent (BOLD) activity across representative brain regions in patients with chronic pain.

Fig. 1.

Summary of brain regions with functional and/or structural alterations in people with chronic pain brain schema of the representative regions. Hot-colored regions represent alterations of AMPA receptor (AMPAR), and the others with reduced gray-matter density (GMD) were colored cold ACC anterior cingulate cortex, NAc nucleus accumbens, PHG parahippocampal gyrus, DLPFC dorsolateral prefrontal cortex, mPFC medial prefrontal cortex, PCC posterior cingulate cortex, S1/S2 primary and secondary somatosensory cortices.

Table 2.

List of the representative brain regions. Summary of neuroimaging and molecular studies demonstrating alterations in AMPA receptor expression, structural gray-matter changes, and functional BOLD signal changes in various brain regions of patients with chronic pain conditions. Data compiled from human neuroimaging studies and preclinical pain models where applicable

Brain region	AMPAR	GMD	BOLD	References
Anterior cingulate cortex (ACC)	↑	↓	↑/↓	[41, 44, 68, 71, 73, 83, 84]
Insula	↑	↓	↑	[50, 51, 68, 70, 71, 83]
Nucleus accumbens (NAc)	↑	−	↓	[52–55, 72]
Amygdala	↑	↓	−	[56–60, 86]
Hippocampus	−	↓	−	[69, 86]
Parahippocampal gyrus (PHG)	−	↓	−	[70]
Dorsolateral prefrontal cortex (DLPFC)	−	↓	−	[65, 67, 68, 84]
Medial prefrontal cortex (mPFC)	−	↓	↓	[70]
Posterior cingulate cortex (PCC)	−	↓	↑	[68, 70, 71]
Thalamus	−	↓	−	[65]
Somatosensory cortices (S1/S2)	−	↓	↑	[67, 68, 71, 83]

↑ increased or upregulated, ↓ decrease, —not reported or not applicable, AMPAR α-amino-3-hydroxy- 5-methyl-4-isoxazolepropionic acid (AMPA) receptor, GMID gray-matter density, blood oxygen level-dependent signal

PET imaging studies: measurement of neuronal activity related to chronic pain

Positron emission tomography (PET) provides quantitative assays of molecular and physiological processes relevant to chronic pain. Glucose metabolism imaging with 18F‑fluorodeoxyglucose (FDG‑PET) is particularly informative. FDG‑PET measures regional glucose metabolism as a proxy for aggregate synaptic activity. Multiple clinical studies have revealed metabolic abnormalities within the pain matrix across diverse chronic pain phenotypes. In fibromyalgia, FDG-PET showed no significant voxel-level differences in whole-brain metabolism between groups. However, subgroup analysis revealed that patients with poor prognosis exhibited significantly increased metabolism in the right thalamus and bilateral lentiform nucleus compared to those with good prognosis, and a positive correlation was observed between tender point count and metabolism in the right parahippocampal gyrus [76]. Another FDG-PET study in macrophagic myofasciitis identified a distinct spatial pattern of glucose hypometabolism, particularly marked in patients with frontal-subcortical circuit dysfunction, with hypometabolic regions including bilateral anterior cingulate and medial frontal cortices, suggesting that FDG-PET can stratify heterogeneous cohorts and identify subgroups with distinct metabolic patterns [77]. Beyond demonstrating regional abnormalities, FDG-PET offers practical clinical utility in guiding pain management. A landmark PET/MRI study of patients with chronic neuropathic pain demonstrated focal increased FDG uptake (two-to-four times background) in affected nerves and muscles at sites of pain in 58 of 65 patients. Remarkably, imaging results led to changes in the management plan—additional diagnostic testing or new interventions—in 40 of 65 patients, including 16 mild modifications and 36 significant modifications such as new invasive procedures, which had not been anticipated by referring physicians [78]. Collectively, these studies demonstrate that FDG-PET can detect metabolic signatures associated with pain symptomatology and prognosis, and can identify previously unrecognized pain generators to enable targeted therapeutic interventions.

Another informative approach is receptor-specific PET, including AMPA-PET with [¹¹C]K-2. Receptor-specific PET extends beyond metabolic surrogates by quantifying discrete signaling systems. Because AMPARs mediate fast excitatory transmission potentiated during central sensitization, AMPA‑PET offers a mechanistically anchored biomarker. Using [¹¹C]K‑2, in vivo AMPAR binding can be visualized and quantified across cortical and subcortical regions, with reversible kinetics and acceptable dosimetry demonstrated in humans [7, 79]. We completed a clinical trial using [¹¹C]K-2 PET imaging to elucidate mechanisms underlying chronic postoperative pain. Thirty-six chronic pain patients were enrolled and compared with age- and sex-matched healthy controls selected from a cohort of healthy volunteers through propensity score matching. Voxel-based whole-brain analysis revealed markedly increased AMPA receptor densities in widespread brain regions in chronic pain patients compared to healthy participants, with particularly prominent increases observed in the occipital and parietal lobes. These findings represent the first in vivo quantification of AMPA receptor alterations in human chronic pain patients (Sakai et al., unpublished). First‑in‑human evaluation of ¹⁸F‑labeled derivatives further supports clinical translation [80]. In translational designs that pair human imaging with rodent models, alterations in [¹¹C]K‑2 binding can be mapped onto histological/electrophysiological readouts of AMPAR abundance and function, enabling cross‑species validation. Coupling AMPA‑PET with task‑based or resting‑state fMRI integrates receptor availability with effective connectivity, providing a systems‑level view of how local receptor changes propagate along large‑scale networks. Importantly, PET can stratify heterogeneous cohorts: FDG‑PET can identify subgroups with cortical hypermetabolism versus thalamic hypometabolism that may respond differentially to neuromodulation or pharmacotherapy, while AMPA‑PET can quantify target engagement for glutamatergic modulators. Longitudinal PET permits within‑patient assessment of plasticity by relating changes in binding or metabolism to clinical improvement. Together, FDG‑PET and AMPA‑PET thus provide complementary windows on systems‑level activity and receptor‑level signaling central to chronic pain [7, 76, 78–82].

Treatment‑induced neuroplasticity: evidence from neuroimaging

A critical clinical question is whether the structural and functional brain alterations observed in patients with chronic pain are reversible. Accumulating evidence suggests that successful pain relief can normalize or even restore structural and functional changes within pain matrix regions. For example, previously reported that reduced gray-matter density in the ACC, insula, and S2 before knee replacement surgery in patients with osteoarthritis, with postoperative increases observed in these regions following pain relief [83]. Furthermore, decreased gray-matter density was observed in the ACC and DLPFC three months after injury compared with the acute phase in patients with post-traumatic headache; however, these changes largely recovered to baseline after one year, concurrent with pain alleviation [84].

Surprisingly, during a cognitive attention task, patients with chronic low back pain exhibited greater activation in the left DLPFC compared with healthy controls, and this hyperactivation decreased after pain improvement [85]. Furthermore, cortical thickness in the left DLPFC, which was decreased during the pain condition, recovered following treatment. Our previous research also reported increased gray-matter density of the amygdala and hippocampus after participation in an interdisciplinary pain management program [86].

FDG-PET is well positioned to demonstrate how treatments exert biological effects in chronic pain. Because regional glucose metabolism reflects the integrated activity of local neuronal populations, longitudinal FDG-PET can track therapy-related normalization of aberrant circuit activity. In interventional paradigms, reductions in hypermetabolism within nociceptive integration areas (e.g., ACC, anterior insula, and amygdala) after surgery, pharmacotherapy, or cognitive-behavioral interventions would suggest decreased excitatory drive consistent with reduced central sensitization. Conversely, increased metabolism in frontoparietal control networks may reflect restored top-down modulation. FDG-PET enables within-subject quantification, facilitating responder analyses where metabolic changes correlate with improvements in pain intensity, interference, and affective burden. When combined with MRI, FDG-PET can relate metabolic normalization to changes in functional connectivity and gray-matter metrics, providing a multimodal account of systems-level plasticity [83–85]. The capacity of FDG-PET/MRI to pinpoint metabolic abnormalities at pain-generating sites has direct implications for precision management: by localizing hypermetabolic foci associated with inflammation or aberrant neuronal activity, clinicians can target specific anatomic structures with minimally invasive interventions (e.g., nerve blocks and focal ablation) rather than relying on systemic pharmacotherapy. This approach is particularly valuable in patients who have failed standard treatments, as demonstrated by studies showing that FDG-PET/MRI findings alter management in over half of such cases [78]. Thus, FDG-PET offers a sensitive, quantitative biomarker to track treatment mechanisms, guide precision management by identifying individuals most likely to benefit from specific modalities, and enable objective evaluation of therapeutic efficacy in chronic pain.

AMPA-PET also holds promise for elucidating pathological mechanisms in chronic pain and guiding therapeutic development. Our group is currently conducting a study comparing AMPAR distribution between chronic pain patients and healthy controls using propensity score matching for age and sex, followed by voxel-based analysis. This approach enables the identification of brain regions with altered AMPAR density across the entire brain, rather than being limited to pre-specified regions of interest. By quantifying region-specific alterations in AMPAR density and examining their relationships with pain symptomatology and functional disability, this work may provide critical validation of translational relevance. If AMPA-PET findings in patients align with region-specific receptor alterations documented in animal models—such as increased AMPAR expression in ACC, insula, nucleus accumbens, and amygdala under chronic pain conditions [40–42, 45, 53, 87]—this concordance would establish the clinical applicability of preclinical discoveries and validate AMPAR density as a quantitative imaging biomarker of central sensitization. Conversely, discordance between human and animal findings would reveal novel mechanisms specific to human chronic pain pathophysiology, potentially identifying new therapeutic targets. Beyond mechanistic insights, AMPA-PET could inform precision treatment strategies. Region-specific increases in AMPAR density might guide targeted neuromodulation approaches such as repetitive transcranial magnetic stimulation (rTMS), which has demonstrated good-level evidence of analgesic efficacy in chronic pain treatment [88]. Additionally, longitudinal AMPA-PET could objectively assess pharmacodynamic responses to glutamatergic modulators by quantifying changes in receptor occupancy or density following treatment, thereby accelerating drug development through early proof-of-mechanism studies and dose optimization based on central target engagement. Ultimately, integrating AMPA-PET with multimodal neuroimaging will establish a comprehensive framework linking molecular receptor dynamics to circuit-level reorganization and clinical outcomes in chronic pain. Ultimately, integrating AMPA-PET with multimodal neuroimaging will establish a comprehensive framework linking molecular receptor dynamics to circuit-level reorganization and clinical outcomes in chronic pain. However, to fully realize the diagnostic and therapeutic potential of AMPA-PET, large-scale studies employing standardized imaging protocols across multiple centers will be necessary to build harmonized datasets. Data-driven approaches, including machine learning and deep learning algorithms, may then help identify robust, generalizable imaging biomarkers that can stratify patients, predict treatment responses, and guide mechanism-based therapeutic selection.

Conclusion and future directions

We outlined chronic pain mechanisms from a molecular perspective, emphasizing AMPARs as central mediators. Changes in AMPAR expression and composition propagate beyond local synapses to reshape large‑scale networks. Accordingly, integrated analyses that combine quantitative PET measures of AMPARs with MRI‑based structural and functional metrics in chronic pain are warranted. Although imaging findings consistent with nociplastic pain are accumulating, definitive imaging criteria are lacking. Data‑driven approaches—including machine and deep learning—may help prioritize robust, generalizable biomarkers, but will require large, harmonized datasets acquired with standardized protocols across centers. Building such resources is essential for establishing diagnostic criteria and for enabling mechanism‑based, personalized treatment of chronic pain.

Author contributions

All authors have contributed to the preparation of the manuscript.

Funding

The work of Tomoyuki Miyazaki was supported by AMED, under Grant No. JP23ek0610025h0003; JSPS, under Grant No. JP23H03003; Yokohama City University, the 2024–2026 Strategic Research Promotion Grant (SK202404); and Japan Science and Technology Agency, under Grant No. JPMJMS239F.

Data availability

Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.

Declarations

Conflict of interest

The authors declare no competing interests.