TRPM3 mediates spontaneous pain and mechanical allodynia in a mouse model of chronic orofacial neuropathy

Summary

Trigeminal nerve injury can lead to chronic and difficult-to-treat orofacial neuropathic pain. Here, we uncover a key role for the cation channel TRPM3 in the chronic constriction injury of the infraorbital nerve (IoN-CCI) mouse model of trigeminal neuropathic pain. Wild-type (WT) mice develop spontaneous pain and mechanical allodynia for up to 6 weeks following IoN-CCI, whereas Trpm3^−/− mice do not develop such symptoms. Using longitudinal RNA sequencing (RNA-seq) analysis, we obtain a detailed time course of transcriptome alterations in trigeminal ganglia during progression of the IoN-CCI model; notably, gene expression regulation is not different between WT and Trpm3^−/− mice. Two structurally distinct TRPM3 antagonists, primidone and isosakuranetin, effectively reverse spontaneous pain and mechanical allodynia, whereas mavatrep, a potent TRPV1 antagonist, is without analgesic effect. These data indicate that TRPM3 is essential for ongoing pain and allodynia following trigeminal nerve injury, making it a potential target for treating trigeminally mediated neuropathic pain.

Graphical abstract

Highlights

• •TRPM3 is required for ongoing pain and mechanical allodynia after trigeminal injury
• •Injury-induced transcriptomic changes in trigeminal ganglia are TRPM3 independent
• •TRPM3 antagonists reduce trigeminal neuropathic pain; TRPV1 blockade is ineffective
• •Association between human trigeminal neuralgia and rare variants in the TRPM3 gene

Deseure et al. demonstrate that the cation channel TRPM3 is essential for the development of spontaneous pain or mechanical hypersensitivity in a mouse model of chronic trigeminal neuropathic pain. Their findings indicate that pharmacological TRPM3 inhibition may represent a therapeutic target to treat trigeminal neuralgia and related neuropathies in patients.

Introduction

Trigeminal neuralgia (TN) is a chronic neuropathic pain condition affecting the trigeminal nerve, which carries sensation from the face, eyes, mouth, and meninges.¹ TN typically manifests in episodes of severe, shock-like unilateral facial pain that may either start unprovoked or be triggered by light touch or mild cooling and is regarded to be one of the most painful disorders known.² It can arise spontaneously or follow facial trauma, dental procedures, vascular compression, or a tumor impacting on the trigeminal nerve. There is also evidence that in a subset of patients, TN is linked to genetic factors, including rare cases where the chronic pain has been associated to mutations in neuronal ion channels.^1,2,3,4,5 Treatments include anticonvulsants such as carbamazepine, the GABA_B agonist baclofen, opioids, or gabapentoids,^1,6,7 but these often provide insufficient relief and cause side effects leading to discontinuation in 1 of 4 patients.^1,7 Therefore, there is a high unmet medical need for improved therapies for TN and other trigeminal neuropathic pain conditions.

Potential molecular targets for treating trigeminal neuropathic pain include ion channels that regulate the sensitivity and excitability of trigeminal nociceptors.⁸ Among these, the capsaicin receptor TRPV1, a member of the TRP cation channel family, has been extensively investigated.^9,10,11,12 Chemical ablation (using potent ligands such as resiniferatoxin or capsaicin) or pharmacological silencing (using TRPV1-dependent entry of the sodium channel blocker QX-314)¹³ of TRPV1-positive nociceptors attenuates thermal and mechanical hypersensitivity in rodent models of trigeminal neuropathic pain.^{12,14,15,16,17} However, these strategies can markedly suppress normal nociception, including heat sensitivity,^13,18,19 which may represent a safety concern in the facial area. Surprisingly, genetic deletion of TRPV1 or treatment with TRPV1 antagonists produces only modest or no effects on mechanical hypersensitivity in trigeminal neuropathic pain model.^16,20,21,22 This suggests that TRPV1-expressing nociceptors rely on additional, TRPV1-independent signaling pathway that contributes critically to neuropathic pain.

One potential contributor is TRPM3, a heat-activated cation channel stimulated by ligands such as the endogenous neurosteroid pregnenolone sulfate (PS) and the synthetic agonist CIM0216.^23,24,25 Single-cell transcriptome studies across mammalian species show that TRPM3 is robustly expressed in various types of somatosensory neurons from dorsal root ganglia (DRG) and trigeminal ganglia (TG), including peptidergic and non-peptidergic nociceptors.^11,26,27 At the functional level, ∼70% of TRPV1-positive somatosensory neurons from mouse or human donors show TRPM3-mediated calcium responses.^24,28,29 Acute activation of TRPM3 directly drives pain: it is one of three major noxious heat sensors (along with TRPV1 and TRPA1), and local injection of PS or CIM0216 into the paw, cheek, bladder, or pancreas of rodents induces TRPM3-dependent nociceptive behavior.^{24,29,30,31,32,33} Moreover, TRPM3 expression and activity is upregulated in sensory neurons during inflammation and chemotherapy-induced neuropathy.^{32,33,34,35,36} In addition, genetic ablation or pharmacological inhibition of TRPM3 attenuates thermal hypersensitivity in mouse models of inflammatory or neuropathic pain.^{24,33,34,37,38} However, whether TRPM3 is involved in spontaneous pain and mechanical hyperalgesia—key symptoms of chronic nerve injury and TN—remains unresolved.

We investigated the role of TRPM3 in a mouse model of chronic neuropathic facial pain, the chronic constriction injury of the ipsilateral infraorbital nerve (IoN-CCI).^39,40 Genetic ablation or pharmacological inhibition of TRPM3 markedly suppressed spontaneous pain and mechanical hyperalgesia, with sustained effects over the full course of the 6-week-long IoN-CCI model. This provides direct evidence that TRPM3 antagonism has continued analgesic effects in a chronic neuropathic pain model. In contrast, the selective TRPV1 antagonist mavatrep was without effect in the IoN-CCI model. Longitudinal transcriptome analysis of the TG indicates that the lack of pain behaviors in the Trpm3^−/− mice is not reflected in global differences in gene expression. Together, these findings indicate that TRPM3 channel activity sustains ongoing pain and mechanical allodynia in a chronic trigeminal neuropathy model. In addition, we provide human genetic evidence that suggests a potential association between TRPM3 variants leading to enhanced TRPM3 activity and TN. Overall, these results support evaluation of TRPM3 antagonists in patients with trigeminal neuropathic pain.

Results

Trpm3^−/− mice do not develop mechanical allodynia or spontaneous pain in the IoN-CCI model of trigeminal neuropathic pain

To investigate a potential role of TRPM3 in trigeminal neuropathic pain, we subjected wild-type (WT) and Trpm3^−/− mice to a unilateral constriction of the infraorbital nerve (IoN-CCI) or a sham operation and assessed mechanical sensitivity and spontaneous pain behavior over a period of 6 weeks following the operation (Figure 1A).

Figure 1.

Trpm3^−/− mice do not develop mechanical allodynia or spontaneous pain behavior in the IoN-CCI model

(A) Experimental design: WT and Trpm3^−/− mice were subjected to the unilateral IoN-CCI procedure or to a sham procedure, and behavioral testing was performed before the procedure and at the indicated 7 time points after the procedure. Each group consisted of n = 10 animals.

(B) Time course (mean ± SEM) of the response score to von Frey stimulation to the ipsilateral side in IoN-CCI and sham-operated WT and Trpm3^−/− mice. The von Frey response score ranges from 0 to 4, where 0 indicates a lack of response, 1 indicates a non-aversive detection of the stimulus, and higher scores indicate increasingly aversive responses (see STAR Methods). Two phases can be discerned in the WT-IoN-CCI mice: an early phase (days 3–18 post-operation) of reduced mechanical responses and a late phase of increased mechanical responses.

(C and D) Comparison of the mean ipsilateral von Frey response score during the early (C) and late (D) phases of the IoN-CCI model. The mean basal (pre-operative) mechanical response score for animals in this experiment is indicated by the dashed line.

(E) Time course (mean ± SEM) of the number of IFG episodes, a measure of spontaneous pain behavior, in IoN-CCI and sham-operated WT and Trpm3^−/− mice.

(F and G) Comparison of the mean number of IFG episodes during the early (F) and late (G) phases. The mean basal (pre-operative) number of IFG episodes for animals in this experiment is indicated by the dashed line.

Scatterplots in (C),(D), (F) and (G) show data from individual animals, as well as the mean (horizontal line) ± SD (whiskers). Statistical comparisons were made using two-way repeated measures ANOVA (B and E) or two-way ANOVA (C, D, F, and G), using Holm-Bonferroni post hoc tests. ∗∗p < 0.01 and ∗∗∗p < 0.001 for the comparison of WT-IoN-CCI versus WT-Sham. ##p < 0.01 and ###p < 0.001 for the comparison between Trpm3^−/− IoN-CCI versus WT-IoN-CCI.

To test the effect of the IoN-CCI procedure on mechanically induced pain, we used a set of graded von Frey filaments to stimulate the facial area of both the ipsilateral and contralateral sides and scored behavioral responses on a scale between 0 and 4, where 0 indicates a lack of response, 1 indicates a non-aversive detection of the stimulus, and higher scores indicate increasingly aversive responses (see STAR Methods).^39,41 Note that, before the IoN-CCI procedure, all three used von Frey filaments were consistently detected by WT and Trpm3^−/− mice (i.e., a score of ≥1). Therefore, this assay is tailored to detect changes in severity of the mechanically evoked nociceptive response, rather than alterations in the mechanical detection threshold, which is extremely low in the facial area of rodent.⁴¹

WT-IoN-CCI animals showed a typical biphasic change in von Frey response score of the ipsilateral side (Figure 1B): after an initial period of reduced mechanical responsiveness (early phase: days 3–18 post-operation; Figures 1B and 1C), mechanical sensitivity was gradually restored and followed by a period (late phase; from day 25 post-operation on) of increased response scores that persisted until the end of the 6-week observation period, indicative of mechanical allodynia (Figures 1B and 1D). Trpm3^−/−-IoN-CCI animals also exhibited reduced mechanical responsiveness during the early phase (Figures 1B and 1C). However, during the late phase, the mechanical sensitivity returned to a similar level as the sham controls, without signs of mechanical allodynia (Figures 1B and 1D). Response scores to von Frey hairs applied to the contralateral side did not change in function of time and were not significantly different between genotypes (Figures S1A–S1C). When scrutinizing the response scores of the ipsilateral side to the individual filaments during the late phase, we found that response scores to the thinnest filament (0.02 g) were not increased in the IoN-CCI groups, whereas WT (but not Trpm3^−/−) mice showed significantly increased response scores to the two thicker filaments (0.16 and 0.4 g; Figures S2A–S2C). Taken together, these data indicate that Trpm3^−/− mice show normal mechanical responses of the facial area under control conditions but fail to develop mechanical allodynia in the IoN-CCI model.

To monitor spontaneous pain in the IoN-CCI model, we quantified the number of isolated face grooming (IFG) episodes and the total duration of IFG episodes during a 10-min observation period.³⁹ Compared to WT-sham controls, WT-IoN-CCI animals exhibited a significant increase in the number and total duration of IFG episodes (Figures 1E–1G and S3A–S3C). This increase was the highest during the early phase (days 3–18) and partly subsided thereafter, but remained at a more elevated level than sham controls until day 42 post-operation (Figures 1E–1G; Figure S3A–S3C). Strikingly, in Trpm3^−/−-IoN-CCI animals, the number and total duration of IFG episodes were not significantly different from those of WT- or Trpm3^−/−-sham animals over the entire 6-week post-operation period (Figures 1E–1G and S3A–S3C).

The difference in IFG between WT-IoN-CCI and Trpm3^−/−-IoN-CCI mice was not due to an overall change in grooming behavior, since there were no significant differences in the number and total duration of face grooming during body grooming episodes between genotypes or treatment groups (Figures S3D and S3E). Taken together, these data indicate that Trpm3^−/− mice are protected from developing spontaneous pain in the IoN-CCI model.

It is well-described that there are clear sex differences in pain processing and chronic pain pathophysiology.⁴² We, therefore, analyzed whether the changes in mechanical responses and spontaneous pain in the IoN-CCI model seen in WT and Trpm3^−/− mice differed between the male and female mice included in the study. This analysis did not reveal any clear sex-specific effects: male and female WT mice similarly exhibited spontaneous pain and mechanical allodynia, which was not observed in male and female Trpm3^−/− mice (Figure S4).

IoN-CCI induces robust and time-dependent transcriptional changes in trigeminal ganglia independent of TRPM3

To explain why Trpm3^−/− mice do not develop mechanical allodynia or spontaneous pain in the IoN-CCI model, we first considered the possibility that global TRPM3 deficiency may alter transcriptional programs and associated biological processes in TG that contribute to spontaneous pain and mechanical hypersensitivity following nerve injury. To evaluate this possibility, we isolated the ipsilateral and contralateral TGs from sham and IoN-CCI mice of both genotypes, isolated mRNA, and performed bulk RNA sequencing (RNA-seq) to quantify genome-wide gene expression. Animals were sacrificed before or at five different time points (days 3, 11, 18, 24, and 42) post-operation, to obtain a time course of transcriptional changes covering both the early and late phases of the IoN-CCI model (Figure 2A). In both genotypes, we did not detect any significant changes in gene expression between the ipsilateral ganglia of sham animals and the contralateral ganglia of sham or IoN-CCI animals, and hence the data from these groups were pooled to form one control group. The uniform manifold approximation and projection (UMAP) representation⁴³ of the principal components of individual samples (using the top 1,000 differentially expressed genes [DEGs]) showed a pronounced separation between control and IoN-CCI samples (Figure 2B), indicating that the nerve injury has a marked effect on the transcriptomic profile of the ganglia.

Figure 2.

IoN-CCI provokes robust time-dependent but TRPM3-independent changes in the transcriptome of trigeminal ganglia

(A) Experimental design: WT and Trpm3^−/− mice were subjected to the unilateral IoN-CCI procedure or to a sham procedure, and trigeminal ganglia (both ipsilateral and contralateral) were collected at the indicated time points.

(B) UMAP visualization of the RNA-seq profiles of 128 trigeminal ganglia from the indicated groups, time points, and genotypes. The inset zooms in on the ganglia from the IoN-CCI mice.

(C) Number of up- and downregulated DEGs for the statistical comparison of IoN-CCI versus sham, at different post-operative time points.

(D) Heatmap plots showing the time course of changes in gene expression for upregulated DEGs in IoN-CCI versus sham. Shown are the top 25 early DEGs (left), stable DEGs (middle), and late DEGs (right). The top graphs shows the mean ± SEM change in expression (expressed as log₂ (fold change)) relative to control in function of time for the DEGs in each category.

(E) Same as (D), but for downregulated DEGs. Note that our analysis did not yield any late downregulated DEGs.

(F) Volcano plot showing the overall difference in gene expression between Trpm3^−/− and WT mice. Each gene is represented by its p_adj (expressed as its negative logarithm) versus the difference in expression in Trpm3^−/− compared to WT (expressed as log₂ (fold change)).

(G) Plot showing the 648 genes whose expression is different between sham and the IoN-CCI mice, comparing the summed log₂(fold change) for Trpm3^−/− versus WT mice. Top up- and downregulated genes are indicated.

We used DESeq2⁴⁴ to statistically evaluate the effects of (1) treatment, (2) genotype, and (3) the interaction between the treatment and genotype on gene expression at the different time points post-IoN-CCI; as criteria to identify DEGs, we set a false discovery rate of 0.05 and a minimal change of 25%.

First, we used DESeq2 to evaluate the effect of treatment on gene expression. We found a total of 648 genes that were differentially expressed between IoN-CCI and control ganglia. These included 464 genes that were significantly upregulated and 184 genes that were significantly downregulated at minimally one time point post-operation (Figure 2C; Table S1). To identify transcriptional correlates of the biphasic behavioral changes in the IoN-CCI model (Figure 1), we assessed the temporal dynamics of the expression changes for the 648 DEGs. Based on their expression levels at different time points of the IoN-CCI model, they were classified as early DEGs (showing the strongest change in expression at the initial time points and normalizing toward the end of the IoN-CCI model), stable DEGs (showing a stable change in expression over the entire course of the IoN-CCI model), and late DEGs (showing the strongest change in expression toward the end of the IoN-CCI model). The top 25 DEGs in each class are shown in Figure 2D for upregulated DEGs and in Figure 2E for downregulated genes (note that there were no late downregulated DEGs), along with a time course of the mean changes in expression levels for all DEGs in each class (Figures 2D and 2E, top). A full list of all DEGs, including their annotation as early, stable, or late, is provided in Table S1, and the expression levels of all genes for every experimental group and time point are provided in Table S2.

Next, we used DESeq2 to evaluate the effect of genotype on gene expression. This analysis revealed only 18 genes that were differentially expressed between Trpm3^−/− and WT mice, including 10 genes that were significantly upregulated and 8 genes that were significantly downregulated in the Trpm3^−/− mice (Figures 2F, S5A, and S5B). TRPM3 was itself among the downregulated genes, showing ∼40% lower expression (Figure S5B and S5C), which can be attributed to the deletion of an essential exon in the Trpm3^−/− mice. However, expression of TRPM3 was stable, and we did not find any significant difference in TRPM3 expression between sham and IoN-CCI mice at any post-operative time point (Figure S5C). Moreover, TRPM3 deletion did not lead to compensatory changes in the expression of the related nociceptive TRP channels TRPV1 and TRPA1. The expression of these TRP channels and of PIEZO2, a key mechanosensitive ion channel in sensory neurons implicated in mechanical pain,⁴⁵ was not altered in the IoN-CCI-treated mice (Figure S5C). Overall, there was no overlap between the 18 genotype-specific genes and the 648 genes that were differentially expressed between IoN-CCI and control ganglia (Figures 2C–2E; Table S1).

Finally and most importantly, we used DESeq2 to test for a possible interaction between treatment and genotype. This analysis did not identify a single gene for which a statistically significant interaction effect was obtained. In line herewith, the 648 genes that showed differential expression between sham and IoN-CCI mice showed quantitatively similar changes in expression in WT and Trpm3^−/− mice (Figure 2G). We also performed histochemical analysis of transverse sections (Figure S6) of the ipsilateral infraorbital nerve of sham and IoN-CCI mice, to evaluate protein expression of genes with the most robust mRNA upregulation (Sprr1a, encoding the small proline-rich protein 1a [also known as Cornifilin-A], known to be involved in neurite outgrowth and regeneration)⁴⁶ or downregulation (Calb2, encoding the calcium-binding protein calretinin) in the nerve bundles. Increased protein expression of SPRR1 and decreased expression of calretinin was similarly observed in WT and Trpm3^−/− mice (Figure S6).

Overall, these expression data indicate robust and time-dependent transcriptional changes in TG, but do not reveal genes that are differentially regulated in the Trpm3^−/− mice compared to WT following IoN-CCI.

Gene Ontology enrichment analysis uncovers time-dependent biological processes in IoN-CCI independent of TRPM3

To gain insight into the biological processes modulated during the different phases of the IoN-CCI model and to assess whether any of these processes are differentially regulated between WT and Trpm3^−/− mice, we performed separate Gene Ontology (GO) enrichment analysis for early, stable and late DEGs (Figure 3A). Among the early DEGs (418 genes in total: 279 upregulated and 139 downregulated genes), the most significant enrichment was for the GO biological process “axonogenesis.” Mean expression changes over time for DEGs in this biological process showed an overall upregulation, peaking at the earliest time points and decaying toward the end of the IoN-CCI model (Figure 3B). This upregulation suggests neuronal remodeling and regeneration in trigeminal neurons during the early phase of IoN-CCI. Notably, the time course of these expression changes for WT and Trpm3^−/− mice overlapped (Figure 3B), indicating that transcriptomic programs promoting axonogenesis after nerve injury are not affected by TRPM3 deficiency. Among early DEGs, we also observed significant enrichment for the GO biological process “regulation of monoatomic ion transmembrane transport.” Genes in this category include numerous ligand- or voltage-gated cation channels, transporters, and associated regulatory proteins, suggesting potential alterations in neuronal excitability following IoN-CCI. Because CCI models involving sensory neurons from DRG often exhibit decreased expression and function of K⁺ channels,⁴⁷ we specifically analyzed DEGs annotated to the GO biological process “K⁺ channel activity” (a subset of the GO term “regulation of monoatomic ion transmembrane transport”). We observed substantial downregulation at early time points, with recovery toward the end of the IoN-CCI model (Figure 3B, right). WT and Trpm3^−/− mice again showed overlapping time courses, indicating that decreased K⁺ channel expression—a hallmark of many chronic pain models—is also present in IoN-CCI and is unaffected by TRPM3 deficiency.

Figure 3.

Gene enrichment analysis identifies time-dependent modulation of biological processes in IoN-CCI independent of TRPM3

(A) GO processes enriched in the early (left), stable (center), and late (right) DEGs among the 648 genes differentially expressed between IoN-CCI and sham. The top five enriched biological processes for each group are shown.

(B) Comparison between WT and Trpm3^−/− mice of the mean (± SEM) time-dependent changes in the expression (compared to control) of DEGs related to the biological pathways “axonogenesis” and “K⁺ channel activity,” top biological processes for early DEGs.

(C) Same as (B), for DEGs related to activation of immune response, the top biological process for stable DEGs.

(D) Same as (B), for DEGs that are related to extracellular matrix organization, the top biological process for late DEGs.

Among the stable DEGs (105 genes in total: 79 upregulated and 26 downregulated genes), there was a significant enrichment for GO biological processes related to immune responses (Figure 3A). In particular, we found a sustained upregulation, both in WT and in Trpm3^−/− mice, of DEGs attributed to the GO biological process “activation of immune response” (Figure 3C) as well as of specific aspects of the immune signature (Figure S7). These findings are consistent with a sustained inflammatory response of the affected TG during the 6 weeks after the IoN-CCI procedure.

Among the late DEGs (119 upregulated genes and no downregulated genes), significant enrichment was observed for the GO biological process “extracellular matrix organization” and other biological processes related to the extracellular matrix and collagen production (Figure 3A). Mean expression changes over time for DEGs in this GO biological process show minimal changes at the first time point, followed by a gradual increase and a robust and sustained upregulation at 18–42 days post-IoN-CCI, with overlapping data for WT and Trpm3^−/− mice (Figure 3D). These findings indicate that the late phase of the IoN-CCI model is associated with post-injury tissue remodeling and fibrosis.

Overall, these transcriptome analyses establish that the IoN-CCI model leads to robust and time-dependent transcriptional changes in the affected TG and uncovers signatures of key biological processes that are specifically altered during the different phases of model. Importantly, we did not find any differences between WT and Trpm3^−/− in the genes or biological processes affected by the IoN-CCI model that could explain the absence of spontaneous pain and mechanical allodynia in the absence of TRPM3. This finding also implies that spontaneous pain per se does not noticeably influence gene expression in the affected TG.

Effect of TRPM3 antagonists on spontaneous pain and mechanical allodynia in the IoN-CCI model

We found no transcriptomic changes that could explain the absence of mechanical allodynia and spontaneous pain in the Trpm3^−/− mice, leading us to hypothesize that TRPM3 activity is required to maintain both behaviors in the IoN-CCI model. To test this directly, we evaluated two structurally unrelated TRPM3 antagonists, isosakuranetin and primidone, in WT-IoN-CCI mice. An earlier study showed that both compounds attenuate heat-evoked responses in WT but not in Trpm3^−/− mice, supporting a TRPM3-specific mode of action. Isosakuranetin is a plant-derived flavanone that potently inhibits TRPM3 (IC₅₀ = 50 nM), with >200-fold selectivity over other sensory TRP channels and pain-relevant channels and receptors.^48,49 We used isosakuranetin at a dose of 2 mg/kg (intraperitoneally [i.p.]), expected to yield peak plasma concentrations in the low micromolar range peaking around 30 min post-dosing based on pharmacokinetic studies.⁴⁸ Primidone is a clinically approved drug used as an antiepileptic. As a barbiturate congener, it is classically considered a prodrug, being slowly metabolized in the liver to the barbiturate phenobarbital, a well-known sedative and antiepileptic drug that acts primarily via activation/potentiation of GABA_A receptors.⁵⁰ Primidone itself is not active at GABA_A receptors but directly inhibits TRPM3 at submicromolar concentrations (IC₅₀ = 600 nM).^51,52,53 This contrasts to the high micromolar concentrations of phenobarbital needed to potentiate or activate GABA_A receptors (typical EC₅₀ values in the range between 100 and 1,000 μM)^54,55 and its limited effects on TRPM3.⁵⁶ We used primidone at a low dose of 1 mg/kg (i.p.), far below the doses needed for GABA_A-dependent protection in rodent seizure models.⁵⁷ Based on published pharmacokinetic studies, this is expected to yield plasma concentrations of primidone in the 2–10 μM range during behavioral assessment, with negligible (submicromolar) levels of phenobarbital.^58,59 Therefore, in our experimental settings, isosakuranetin and primidone inhibit TRPM3 activity with substantial selectivity over their other known targets. Carbamazepine (50 mg/kg; i.p.), a non-selective inhibitor of voltage-gated Na⁺ channels and first-line treatment of TN patients served as positive control. On each experimental day, mice were dosed with the respective compounds or vehicle, and behavioral testing started 30 min following dosing (Figure 4A). A sham-operated group was included as a comparator.

Figure 4.

TRPM3 antagonists attenuate mechanical allodynia and spontaneous pain behavior in the IoN-CCI model

(A) Experimental design to evaluate the effect of the TRPM3 antagonists primidone (1 mg/kg) and isosakuranetin (2 mg/kg) on mechanical sensitivity and spontaneous pain in the IoN-CCI model (WT mice; n = 15 animals per group). Carbamazepine (50 mg/kg) was used as positive control.

(B) Time course (mean ± SEM) of the response score to von Frey stimulation to the ipsilateral side in IoN-CCI and sham-operated WT mice that received the indicated drugs or vehicle.

(C and D) Comparison of the mean ipsilateral von Frey response score during the early (C) and late (D) phases of the IoN-CCI model. The mean basal (pre-operative) mechanical response score for animals in this experiment is indicated by the dashed line.

(E) Time course (mean ± SEM) of the number of IFG episodes in IoN-CCI- and sham-operated WT mice that received the indicated drugs or vehicle.

(F and G) Comparison of the mean number of IFG episodes during the early (F) and late (G) phases of the IoN-CCI model. The mean basal (pre-operative) number of IFG episodes for animals in this experiment is indicated by the dashed line.

Scatterplots in (C),(D), (F) and (G) show data from individual animals, as well as the mean (horizontal line) ± SD (whiskers). Statistical comparisons were made using two-way ANOVA (C, D, F, and G), using Holm-Bonferroni post hoc tests. ∗∗p < 0.01 and ∗∗∗p < 0.001 for the comparison with vehicle-treated IoN-CCI mice, and ###p < 0.001 for the comparison with vehicle-treated sham mice.

In comparison to vehicle-treated animals, isosakuranetin, primidone, and carbamazepine significantly attenuated the mechanical allodynia that develops during the late phase of the IoN-CCI model, but had no effect on the reduction in mechanosensitivity of the ipsilateral side during the early phase (Figures 4B–4D). Isosakuranetin and primidone were without effect on the mechanical responses of the contralateral side (Figures S8A–S8C). isosakuranetin, primidone, and carbamazepine also robustly reduced the total duration and number of IFG episodes, both during the early and the late phases (Figures 4E–4G). The reduction in IFG in the isosakuranetin- and primidone-treated groups was not due to an overall change in grooming behavior, since there were no significant differences in the total duration and number of face grooming during body grooming episodes between vehicle-, isosakuranetin-, and primidone-treated mice (Figures S8D–S8F). In contrast, the carbamazepine-treated animals exhibited a sizable reduction in the total duration and number of face grooming during body grooming episodes, which was significant during the late phase of the experiment, in line with the known sedative effect of this drug (Figures S8D–S8F).

These results demonstrate that isosakuranetin and primidone reduce spontaneous pain and mechanical allodynia in the IoN-CCI model, supporting the hypothesis that TRPM3 channel activity is necessary to sustain pathological pain and hypersensitivity in this model of TN.

Differential effects of TRPM3 and TRPV1 antagonism in the IoN-CCI model and in carrageenan-induced inflammatory heat hyperalgesia

In trigeminal and dorsal root ganglion neurons, TRPM3 is often coexpressed with TRPV1, the receptor for capsaicin and marker of nociceptive neurons.^24,28 While TRPV1 antagonism has consistently been found to reduce heat hypersensitivity in rodent pain models, the channel’s contribution to mechanical hypersensitivity and ongoing pain in (trigeminal) neuropathic pain models is less well established.^14,15,16,22 To evaluate the role of TRPV1 in the IoN-CCI model, we juxtaposed the efficacy of primidone with that of mavatrep, a potent and selective TRPV1 antagonist that is currently in clinical development for chronic osteoarthritis pain.^9,60 Notably, in stark contrast to primidone (1 mg/kg; i.p.), mavatrep (10 mg/kg; i.p.) did not reduce mechanical hypersensitivity or spontaneous pain behavior at any time point during the IoN-CCI model (Figures 5A–5D).

Figure 5.

The TRPV1 antagonist mavatrep is inactive in the IoN-CCI model

(A) Experimental design to compare the effect of the TRPV1 antagonist mavatrep (10 mg/kg) and the TRPM3 antagonist primidone (1 mg/kg) on mechanical sensitivity and spontaneous pain in the IoN-CCI model (n = 15 animals per group). Mechanical responses for the contralateral side of IoN-CCI mice were used as control.

(B) Time course (mean ± SEM) of the response score to von Frey stimulation to the ipsilateral and contralateral sides in IoN-CCI-operated WT mice that received the indicated drugs or vehicle.

(C and D) Comparison of the mean ipsilateral von Frey response score during the early (C) and late (D) phases of the IoN-CCI model. The mean basal (pre-operative) mechanical response score for animals in this experiment is indicated by the dashed line.

(E) Time course (mean ± SEM) of the number of IFG episodes in IoN-CCI-operated WT mice that received the indicated drugs or vehicle.

(F and G) Comparison of the mean number of IFG episodes during the early (F) and late (G) phases of the IoN-CCI model. The mean basal (pre-operative) number of IFG episodes for animals in this experiment is indicated by the dashed line.

Scatterplots in (C),(D), (F) and (G) show data from individual animals, as well as the mean (horizontal line) ± SD (whiskers). Statistical comparisons were made using two-way repeated measures ANOVA (B and E) or one-way ANOVA (C, D, F, and G) and Holm-Bonferroni post hoc tests. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 versus vehicle-treated animals.

Importantly, at the same doses, both mavatrep and primidone significantly reduced heat hyperalgesia following carrageen-induced inflammation of the hind paw (Figure 6). Here, mavatrep showed a more pronounced effect than primidone (Figures 6B and 6C). These findings underscore that TRPV1 antagonism is highly effective at suppressing inflammatory heat hyperalgesia, whereas TRPM3 antagonism is clearly superior in alleviating the key features of trigeminal neuropathic pain.

Figure 6.

The TRPV1 antagonist mavatrep and the TRPM3 antagonist primidone reduce inflammatory heat hyperalgesia

(A) Experimental design to compare the effect of the TRPV1 antagonist mavatrep (10 mg/kg) and the TRPM3 antagonist primidone (1 mg/kg) on heat hyperalgesia in the carrageenan model of acute inflammation. Each group consisted of n = 8 mice.

(B) Latency to withdrawal of the hind paw in the Hargreaves test (mean ± SEM), measured before the procedure and 4 h after intraplantar injection of carrageenan. Then, animals were dosed with vehicle, primidone (1 mg/kg), or mavatrep (10 mg/kg) and retested 30 min later (i.e., 4.5 h post-carrageenan).

(C) Percentage reversal (mean ± SEM) of the carrageenan-evoked heat hypersensitivity by vehicle, primidone, and mavatrep. Reversal was determined for each individual animal using the following equation: Reversal (%) = (latency_4.5h−latency_4h)/(latency_pre-op−latency_4h) × 100%.

Statistical comparisons were made using two-way repeated measures ANOVA (B) or one-way ANOVA (C) and Holm-Bonferroni post hoc tests. ∗∗p < 0.01, ∗∗∗p < 0.001 versus vehicle-treated animals.

Association between rare variants in the TRPM3 gene and trigeminal neuropathy in patients

While classically the majority of TN cases in humans were considered to be sporadic and evoked by compression of the trigeminal nerve, there is growing evidence, including familial occurrences, that genetic factors may contribute to the pathogenesis of TN in a subset of patients. To date, we are not aware of any published genome-wide association studies that identify genetic associations with genome-wide significance for TN. However, based on the notion that hyperexcitability of TG neurons contributes to the pathophysiology of TN, a recent study performed whole-exome sequencing in 12 patients with a family history of TN to identify rare variants in a panel of 173 ion channels implicated in neuronal excitability.³ This analysis yielded 41 rare variants in ion channels, including one variant in the TRPM3 gene (Chr9:70536143G>A: leading to the amino acid substitution Ala1657V in TRPM3) in a 77-year-old woman with idiopathic TN and a TN-affected father.³ We performed a functional characterization of this variant following heterologous expression in HEK293, using patch-clamp electrophysiology and calcium imaging. Cells were transfected either with cDNA encoding WT human TRPM3 or a combination cDNA encoding WT and the A1657V mutant in a one-to-one ratio (with identical total cDNA), thereby mimicking the heterozygous condition in the affected patient. This analysis revealed a mild gain of function for the A1657V mutant, characterized by unaltered basal currents (Figures 7A and 7B), significantly larger current amplitudes following stimulation with the neurosteroid agonist PS (40 μM; Figures 7A and 7B), and a mild leftward shift of the concentration-response curve for PS activation in calcium imaging, characterized by a change in EC₅₀ value from 3.6 ± 0.3 to 2.2 ± 0.5 μM (Figure 7C). To evaluate whether rare variants in the TRPM3 gene are more generally associated with trigeminal neuropathic pain in a larger cohort, we analyzed data from the UK Biobank, in which 1,074 participants are diagnosed with TN (ICD-10 code: G50.0). As no single SNP in the TRPM3 gene reached genome-wide significant association (p < 5 × 10⁻⁸) with TN, we performed rare-variant collapsing analysis.⁶¹ Collapsing all missense variants, all variants affecting the 5′ or 3′ untranslated regions or splice region variants in the TRPM3 gene did not reveal any significant association with TN in the UK Biobank patients (Figure 7D). However, we found a significant association (P_adj = 0.0023; odds ratio, 9.1; 95% confidence interval [CI]: 2.3, 37.1]) between splicing (donor/acceptor) variants in TRPM3 and TN. The single splice donor variant Chr9:70537121C>T also showed a significant positive association (P_adj = 0.0005; odds ratio, 11.4; 95% CI: 2.8, 46.7) (Figure 7D). This splice donor variant is predicted to suppress the alternative splicing in exon 28 of the TRPM3 gene. Of note, earlier work has demonstrated that the alternatively spliced exon 28 encodes TRPM3 isoforms with a shorter C terminus, which show reduced channel activity when compared to isoforms containing the unspliced exon 28.⁶² Together, these findings suggest a potential association between TRPM3 variants leading to enhanced TRPM3 activity and TN, warranting further validation in larger cohorts.

Figure 7.

Human genetics of TRPM3 in TN

(A) Representative whole-cell current-voltage relations in HEK293 cells expressing WT human TRPM3 or a one-to-one ratio of WT and the A1657V variant, under basal conditions and following stimulation wit PS (40 μM).

(B) Statistical comparison of the current densities of the basal current (I_basal) and PS-activated current (I_PS) for recordings as in (A). Current density was quantified at +145 mV. ∗p < 0.05, Mann-Whitney test (n = 9 biological replicates per condition). Bars and whiskers indicate mean ± SEM.

(C) Concentration-response curves for the PS-evoked intracellular Ca²⁺ signal (measured using the genetically encoded indicator jRCaMP1b) in HEK293 cells expressing WT human TRPM3 or a one-to-one ratio of WT and the A1657V variant. Individual symbols show mean ± SEM. Solid lines represent a fit using the Hill equation, yielding EC₅₀ values of 3.6 ± 0.3 and 2.2 ± 0.5 μM, respectively (n = 3 biological replicates per condition).

(D) Odds ratios and 95% CIs for the association between the variants in the human TRPM3 gene and trigeminal neuralgia in participants of the UL Biobank. Variants were collapsed into the indicated groups based on their predicted effect.

Discussion

In this study, we demonstrated the key involvement of TRPM3 in pain-related behavior in the mouse IoN-CCI model of trigeminal neuropathy. In this model, which spans a duration of 6 weeks, chronic constriction of the infraorbital nerve results in a biphasic behavioral pain phenotype in WT mice: spontaneous pain with blunted mechanosensitivity in the first 2 weeks, followed by the emergence of mechanical allodynia and a persistent spontaneous pain behavior at later time points. Both spontaneous pain and mechanical allodynia were absent in Trpm3^−/− mice. In WT animals, systemic treatment with the TRPM3 antagonists isosakuranetin and primidone suppressed spontaneous pain and mechanical allodynia. These findings identify TRPM3 as a key driver of chronic trigeminal neuropathic pain in mice.

Earlier studies showed that Trpm3^−/− mice have reduced heat hyperalgesia in various mouse models of pathological pain, including local inflammation (CFA and carrageenan models) and nerve injury (chronic constriction injury of the sciatic nerve).^24,37,38 However, results regarding the effects of TRPM3 ablation on mechanical hypersensitivity have been mixed. Similarly to our present results in the IoN-CCI model, Trpm3^−/− mice were protected from the development of mechanical hypersensitivity in the oxaliplatin model of chemotherapy-induced neuropathy.³⁴ In contrast, in the context of inflammation (CFA and carrageenan) and CCI of the sciatic nerve, Trpm3^−/− mice were found to develop mechanical hypersensitivity to a similar level as WT mice.^24,38 The mechanisms underlying the variable involvement of TRPM3 in mechanical hypersensitivity in the IoN-CCI model compared to some of the reported models in the literature are currently unclear. One notable difference is that all these earlier studies examined DRG-mediated mechanical responses, while here we used a model that involves TG neurons. Recent single-cell RNA-seq studies indicate that the transcriptional identity of neuronal and non-neuronal cell types is largely conserved between TG and DRG^26,27 and TRPM3 is expressed at similarly high levels in both ganglion types.^24,26,27,63 Yet, there is evidence for differential gene expression changes following chronic constriction injury of the infraorbital (TG) versus sciatic (DRG) nerves, which may lead to differential regulation of mechanosensitivity.^64,65 In addition, approaches to analyze mechanical hypersensitivity differ significantly between studies. Vriens et al.²⁴ and Su et al.³⁸ used graded von Frey filaments, and based their assessment of mechanical sensitivity on whether specific filaments evoked a withdrawal response or not. Aloi et al.³⁴ used an electronic von Frey device, yielding a single value for the force at which a withdrawal reaction is induced. In the present study, we used a well-established scoring system to evaluate the intensity of the pain response to graded von Frey filaments, rather than the detection threshold.⁶⁶ Using this approach, we were able to evaluate whether mild mechanical stimuli evoke an exaggerated pain response, which is a typical feature of trigeminal neuropathic pain and TN in patients. Our findings indicate that in the late stage of the IoN-CCI model (day 25 and further), when WT mice show an increased pain score when stimulated on the ipsilateral site, Trpm3^−/− mice show baseline responses on both sides of the face. Thus, genetic ablation of TRPM3 preserves normal mechanosensitivity but precludes the development of painful mechanical hyperalgesia.

In addition to mechanical allodynia, patients with trigeminal neuropathic pain suffer from unprovoked pain.¹ The increased IFG behavior in the IoN-CCI model, which in WT mice persists during the entire 42 days of the model, represents a robust measure of unprovoked pain in mice. Strikingly, Trpm3^−/− mice did not show increased IFG, demonstrating that genetic elimination of TRPM3 suppresses chronic spontaneous pain-related behavior in a long-lasting neuropathic pain model.

To investigate whether the absence of pain-related behavior in the Trpm3^−/− mice correlated with differences at the transcriptome level, we performed bulk RNA-seq of TGs of WT and Trpm3^−/− at different time points following the IoN-CCI procedure. This analysis revealed significant changes in the mRNA expression of 648 genes in the ipsilateral TGs of the IoN-CCI animals. Within the top upregulated DEGs, we identified various genes that had been shown to be upregulated in other sensory nerve injury models, including the transcription factor Atf3; neuropeptides such as Npy, Nts, and Gal; and regeneration-associated Sprr1a.^{46,67,68,69,70} In line with previous work in neuropathic pain models,⁷¹ we also found robust upregulation of Cacna2d1. This gene encodes the α2δ1 subunit of voltage-gated calcium channels, the molecular target of gabapentinoid drugs pregabalin and gabapentin,⁷² widely used to treat neuropathic pain.^73,74 Concurrent with the biphasic behavioral changes, we could classify DEGs based on the time courses of the changes in gene expression. GO enrichment analysis revealed that many DEGs whose expression is increased in the early phase of the IoN-CCI model are implicated in axonogenesis, in line with the restoration of axons of the somatosensory neurons following the initial injury. We also found a significant downregulation during the early phase of genes implicated in the regulation of K⁺ channel activity, which is known to contribute to hyperexcitability of sensory neurons.⁴⁷ Among genes whose expression was stably altered during the entire duration of the model, we found a strong enrichment for genes related to immune responses. Late DEGs were enriched for genes involved in biological processes related to extracellular matrix organization and collagen production, indicative of post-injury tissue remodeling and fibrosis. Strikingly, despite the profound differences in pain-related behavior between WT and Trpm3^−/− mice, and the robust changes in gene expression invoked by the IoN-CCI procedure, we did not identify any difference in IoN-CCI-related DEGs or gene expression programs between the genotypes. Therefore, the absence of spontaneous pain and mechanical allodynia in the Trpm3^−/− mice is not the consequence of pronounced global differences in gene expression programs affected by the nerve injury. Moreover, it indicates that the spontaneous pain behavior by itself does not measurably influence TG gene expression. The most straightforward explanation for these findings is that TRPM3 channel activity is a key driver of spontaneous pain and mechanical allodynia in the IoN-CCI model and that these sensory abnormalities do not develop when TRPM3 activity is silenced, even if all other transcriptome programs occur independently of TRPM3 function.

In our experiments, genetic ablation of TRPM3 attenuated mechanical allodynia without affecting basal mechanosensitivity. These findings indicate that TRPM3 is not directly involved in the detection of mechanical stimuli, but rather acts as a determinant of excitability of mechanosensitive and nociceptive sensory neurons in the context of the IoN-CCI model, similar to its proposed role in the context of tissue inflammation.³⁵ TRPM3 is temperature sensitive, with significant activity at body core temperature, and its activity is further enhanced by circulating levels of neurosteroids such as PS.^24,75 Moreover, TRPM3 channel activity is under inhibitory control of a variety of G-protein-coupled receptors, via direct inhibitory binding of the G_βγ subunit of trimeric G proteins to the N terminus of TRPM3^{28,76,77,78,79} and depletion of membrane phosphoinositides.^80,81 Further research is needed to elucidate how these different signaling pathways impact on TRPM3 activity in the context of IoN-CCI and other neuropathies.

In addition to these signaling pathways, TRPM3 activity may also be affected by genetic variations. The causes of TN in humans are multifactorial, and there is growing evidence that genetic factors may contribute to pathogenesis in a subset of patients. We found that a TRPM3 variant from a patient with idiopathic TN and with a family history of TN³ exhibits a mild gain of function. Moreover, rare-variant collapsing analysis using data from the UK Biobank revealed a significant association between TN and splice donor and acceptor variants in the TRPM3 gene. In particular, a splice donor variant that suppresses the alternative splicing in exon 28, thereby eliminating the formation of a TRPM3 splice variant with reduced channel activity,⁶² showed a statistically significant association with TN. Overall, these data are suggestive of an association between TN and TRPM3 variants with increased channel activity. However, caution is needed due to the relatively low number of TN patients in the UK Biobank, warranting further validation in larger cohorts.

To test the hypothesis that pain-related behavior in the IoN-CCI model is driven by TRPM3 channel activity, we evaluated the effect of acute TRPM3 inhibition using two antagonists, isosakuranetin and primidone.^48,51 Both compounds were active, reducing spontaneous pain during the entire duration of the model and alleviating mechanical allodynia in the late phase of the model, without any obvious side effects. The positive control carbamazepine was equally active, but it also caused a reduction in combined face and body grooming, in line with its known sedative effect. The sustained analgesic activity over a period of 6 weeks suggests that TRPM3 inhibition may be a therapeutic option to treat chronic trigeminal pain in patients and thus encourages controlled studies to test the efficacy of TRPM3 antagonism in TN or other neuropathic pain conditions. Efforts are ongoing to develop more potent and specific TRPM3 antagonists,¹¹ which will facilitate assessment of the full potential of TRPM3 antagonism in treating trigeminal pain.

In comparison to TRPM3, the development of TRPV1-targeted drugs to treat pathological pain is much more advanced, with several antagonists tested in clinical pain studies. Contrary to the tested TRPM3 antagonists, mavatrep, a potent and selective TRPV1 antagonist showing efficacy in painful knee osteoarthritis patients,^9,60 did not attenuate spontaneous pain or mechanical allodynia in the IoN-CCI model. Yet, mavatrep was superior to primidone in attenuating heat hyperalgesia in the carrageenan inflammatory model. Summarizing these and earlier^20,21,34,38 findings underscores the differential involvement of TRPM3 and TRPV1 in neuropathic and inflammatory pain models, where TRPM3 is likely the more promising target for the treatment of pain associated with nerve injury.

In conclusion, we report that genetic ablation and systemic pharmacological inhibition of TRPM3 reduces spontaneous pain behavior and mechanical allodynia in the mouse IoN-CCI model of chronic trigeminal neuropathic pain. While the IoN-CCI procedure invoked substantial and time-dependent changes at the transcriptome level in the TG, these changes were not affected by the absence of TRPM3. These findings indicate that spontaneous pain and mechanical allodynia in the IoN-CCI model depend on ongoing TRPM3 channel activity. These results warrant further studies to evaluate whether TRPM3-targeted therapies can be developed to treat trigeminally mediated pain in patients.

Limitations of the study

Our study identifies an essential role for TRPM3 in the development of mechanical allodynia and spontaneous pain in the mouse IoN-CCI model of trigeminal neuropathic pain, but we have not identified how TRPM3 channel activity is regulated in trigeminal ganglion or whether its activity is modulated after the IoN-CCI procedure.

We found that TRPM3 deletion did not change the overall IoN-CCI-induced transcriptomic response in TG, but effects in rare cell populations cannot be excluded given the limits of bulk RNA-seq.

The TRPM3 antagonists isosakuranetin and primidone reproduced the Trpm3^−/− phenotype at low doses and without behavioral side effects. Nevertheless, we cannot exclude that their analgesic effects in the IoN-CCI model may involve other molecular targets.

Finally, it remains unknown whether the preclinical findings in mice will translate to clinically relevant pain relief in (trigeminal) neuropathic pain patients treated with TRPM3-targeted therapies. Selective TRPM3 antagonists have not yet been tested for analgesic efficacy in clinical trials; past efforts to develop selective antagonists for other TRP channels involved in pain signaling have not yet yielded effective new analgesics.^9,11 This underscores the challenges of achieving satisfactory pain relief by targeting a single molecular target.

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Thomas Voets (thomas.voets@kuleuven.be).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •This study generated transcriptomic data, which are available from the Gene Expression Omnibus (GEO) repository GSE316925.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We thank all members of the Laboratory of Ion Channel Research for support and discussions. This work was supported by grants from the Research Foundation Flanders (FWO; G0B9520N, G0B7620N, G055124N, and Scientific Network W000221N to T.V. and G.084515N and G.0B1819N to J.V.), the KU Leuven (C24M/21/028 to T.V. and C3/21/049 to J.V. and T.V.), the Queen Elisabeth Medical Foundation for Neurosciences (to T.V.), and the VIB (to T.V.). This research has been conducted using the UK Biobank Resource under application number 103552.

Author contributions

K.D., J.V., G.H., and T.V. conceived the study and designed experiments. K.D., I.D., and S.P. performed mouse behavioral data. K.L. and E.P. performed immunohistochemistry. K.D., S.P., and N.V.R. dissected trigeminal ganglia and isolated RNA for transcriptomic studies. A.S. and T.V. carried out data analysis and visualization of RNA-seq data. R.R. performed functional analysis of heterologously expressed mutant TRPM3. S.C. and S.M. analyzed data from the UK Biobank. K.D. and T.V. carried out analysis and visualization of behavioral data. T.V. and K.D. wrote the manuscript, with input from all authors.

Declaration of interests

T.V. and J.V. are inventors on patents and patent applications related to TRPM3 and received research funding from Biohaven.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-SPRR1A antibody (rabbit)	Novus Biologicals	Cat# NBP2-93464; RRID: AB_3463407
Anti-Calbindin 2 antibody (rabbit)	Abcam	Cat# ab108404; RRID: AB_10861236
Anti-PGP9.5 antibody (rabbit)	Dako	Cat# Z5116; RRID; AB_2622233
Biotinylated pig anti-rabbit	Dako	Cat# E0353; RRID: AB_2737292
Anti-Rabbit Immunoglobulins/HRP (goat)	Dako	Cat# P0448; RRID: AB_2617138
Chemicals, peptides, and recombinant proteins
Isosakuranetin	PhytoLab	N/A
Primidone	Sigma-Aldrich	P7423
Carbamazepine	Sigma-Aldrich	C4024
Mavatrep	Sanbio	HY-16935-100
Tween80	Sigma-Aldrich	P4780
λ-carrageenan	TCI	C2871
Hematoxylin & Eosin (H&E)	Sigma-Aldrich	HHS16
TransIT-LT1 transfection reagent	Mirus Bio	MIR 2306
Critical commercial assays
innuPREP RNA Mini Kit 2.0	Innuscreen GmbH	845-KS-2040250
Deposited data
Bulk RNAseq data of trigeminal ganglia	This paper	Gene Expression Omnibus (GEO) repository: GSE316925
Experimental models: Cell lines
Human: HEK293T	ATCC	CRL-3216
Experimental models: Organisms/strains
Wild type C57Bl/6J mice (10-week old)	Jackson Laboratory	JAX: 000664
Trpm3−/− C57Bl/6J mice (10-week old)	Vriens et al.24	N/A
Recombinant DNA
Plasmid: pCAGGSM2-hTRPM3-Ires-GFP	Burglen et al.82	N/A
Plasmid: pCAGGSM2-hTRPM3(A1657V)-Ires-GFP	This paper	N/A
Software and algorithms
Origin	OriginLab Corporation	OriginPro 2023b
STAR aligner	GitHub	v2.7.10b
RSEM	GitHub	v1.3.1
DESeq2	Bioconductor	v1.44.0
DOSE	Bioconductor	v4.0.0
enrichplot	Bioconductor	v1.26.6
QuPath	GitHub	V0.5.1
PatchMasterPro	HEKA Elektronik	V2x92
Other
Chromic catgut ligatures (6-0)	Ethicon	1836G
Vicryl Rapide sutures (6-0)	Ethicon	W9911

Experimental model and study participant details

Mice

Male and female wild type (N = 164) and Trpm3^−/− (N = 44) C57Bl/6J mice were used (10 weeks old at start of the experiment). They were housed in standard plastic mice cages (2–5 mice per cage) in a colony room with a room temperature of 21 ± 1°C, a relative humidity of 40 ± 10% and a 12h light/hour dark cycle. Water and food were available ad libitum. Mice were allowed to acclimate for 5 days to the housing conditions before they were habituated to the testing procedures. Animals were treated and cared for according to the guidelines for pain research in conscious animals of the International Association for the Study of PAIN and in line with the Flemish and European regulations for animal research and the ARRIVE guidelines. The protocols were approved by the local institutional Ethical Committees of the University of Antwerp and the KU Leuven.

Cell lines

HEK293T cells were grown in Dulbecco’s modified Eagle’s medium containing 10% (v/v) human serum, 2 mml-glutamine, 2 units/ml penicillin, and 2 mg/mL streptomycin at 37°C in a humidity-controlled incubator with 10% CO₂. HEK293T cells were obtained from ATCC (CRL-3216) and used only up to passage number 25 without further verifying their identity. The cells were tested monthly for the lack of mycoplasma. For patch-clamp experiments, HEK293T cells grown in 6-well plates were transfected with either 2 μg of pCAGGSM2 vector containing WT hTRPM3 (pCAGGSM2-hTRPM3-Ires-GFP) or 1 μg of pCAGGSM2-hTRPM3-Ires-GFP and 1 μg of pCAGGSM2-hTRPM3(A1647V)-Ires-GFP, using TransIT-LT1 transfection reagent (Mirus Bio, Madison, WI, USA). For calcium the fluorescence-based assay, cells were grown in T25 flasks, and co-transfected with 6 μg of the genetically encoded calcium indicator jRCAMP1b together with either 6 μg of pCAGGSM2-hTRPM3-Ires-GFP or 3 μg of both pCAGGSM2-hTRPM3-Ires-GFP and pCAGGSM2-hTRPM3(A1647V)-Ires-GFP.

Method details

IoN-CCI surgery

Mice were anesthetized with ketamine/xylazine (75/15 mg/kg, i.p., 10 mL/kg). The animal was placed in a custom pre-shaped foam pad and the left ear and whiskers were fixed to the foam pad with transparent surgical tape (Transpore). The area between the whisker pad and the left eye was shaved, taking care not to clip any of the whiskers or damage the eye, and a skin incision (4 mm) was made just below the infraorbital foramen. Under direct visual control using a surgical microscope, the nerve was exposed by blunt dissection and chromic catgut ligatures (6-0) were loosely tied around the approximately midway between the point where the nerve exits the skull and the point where the nerve branches out. In the study comparing WT versus Trpm3^−/− mice two ligatures were placed, whereas in the TRPM3 pharmacology a single ligature was placed. The incision was closed using synthetic absorbable 6-0 sutures and the animal was allowed to fully recover in a separate heated cage before being returned to the home cage. Mice were randomly assigned to the IoN-CCI group or the sham group. Sham surgery was performed exactly as described above except for the nerve ligation. A detailed description and visualization of the Ion-CCI method can be found in Deseure et al.³⁹

Carrageenan-induced inflammation model

λ-carrageenan powder was dissolved in sterile 0.9% sodium chloride solution to obtain a 1% solution.³⁸ Subcutaneous injections of 10 μL of the 1% λ-carrageenan solution into the dorsal surface of one of the hind paws were performed using BD Micro-Fine Demi 0.3mL 30G syringes (Becton, Dickinson and Company, Franklin Lakes, NJ, USA).

Behavioral testing

Behavioral testing in the IoN-CCI model was performed one day before IoN surgery and on post-operative days 3, 6, 11, 18, 25, 32 and 42. Mice were habituated to the test procedure one and two days before pre-operative testing (Pre-op). Habituation and testing were conducted in an acoustically isolated room with background noise from the ventilation units in the experimental room. All behavioral experiments were performed by researchers that were blinded for the experimental group.

For observation of face grooming behavior, mice were observed in a transparent plastic cage with a mirrored back (l x w x h, 12 × 12 × 17 cm). Behavior was video-recorded for 10 min, and the amount of time spent on face grooming (i.e., movement patterns in which paws contact facial areas⁶⁶) was determined from the recorded video by a researcher that was blinded for the experimental group or treatment. A distinction was made between isolated face grooming and face grooming during body grooming.⁸³ If a sequence was neither preceded nor followed by body grooming (i.e., movement patterns in which the paws, tongue or incisors are brought into contact with a body area other than the face or the forepaws⁶⁶), the episode was categorized as isolated face grooming. If body grooming was present before or after a sequence of face grooming actions, the episode was categorized as face grooming during body grooming.

For mechanical stimulation, animals were restrained as previously described.⁴¹ Briefly, an experimental animal’s tail was placed in a silicone holder that was magnetically attached to a metal plate on an elevated platform. Next, a three-walled plastic holder (65 × 25 × 23 mm) was placed over the animal so that only the head and front paws poke out, and a weight was placed on top of the holder. A graded series of three von Frey hairs (Pressure Aesthesiometer, Stoelting Co, Chicago, IL) were used. The force required to bend the filaments was 0.02 g, 0.16 g and 0.40 g, respectively. The stimuli were applied perpendicularly within the IoN territory, near the center of the vibrissal pad, on the hairy skin surrounding the mystacial vibrissae. Within each animal, stimuli were applied in an ascending order of intensity. After a stimulus intensity was applied to one side, it was applied to the other side before moving on to the next stimulus intensity. The order in which the ipsilateral and contralateral sides were stimulated, was randomized. The scoring system described by Vos et al.⁶⁶ was used to evaluate the response of the mice to the stimulation. The response was observed to belong to one of the following response categories: (score 0) no response; (score 1) detection = the mouse turns the head toward the stimulating object and the stimulus object is then explored; (score 2) withdrawal reaction = the mouse turns the head slowly away or pulls it briskly backward when the stimulation is applied, sometimes a single face wipe ipsilateral to the stimulated area occurs; (score 3) escape/attack = the mouse avoids further contact with the stimulus object by attacking the stimulus object, making biting and grabbing movements; (score 4) asymmetric face grooming = the mouse displays an uninterrupted series of at least three face-wash strokes directed toward the stimulated facial area. For each animal, and at every designated time, a mean score for the three von Frey hairs (0.02 g, 0.16 g and 0.40 g) was determined.

For thermal stimulation, we used the Hargreaves test to assess changes in heat sensitivity in the carrageenan model, using the Hargreaves apparatus Model 37570 (Ugo Basile S.R.L, Gemonio, Italy). Individually enclosed animals were placed on top of a glass surface, under which the infrared beam emitter was placed. Following a 30–60 min of habituation, the midplantar region of both hind paws was stimulated five times with the infrared beam. The withdrawal latency was automatically detected by the apparatus. The intensity of the beam was adjusted to obtain an average baseline latency of 10–12 s in naive wild-type animals. To avoid tissue damage, we set a 5-min interval between repeated stimulations of the same paw, and a 20-s beam disengagement cutoff.

Drugs administration

To test the effect of TRPM3 antagonists or Carbamazepine on behavior, mice were injected intraperitoneally (i.p.) with Isosakuranetin (PhytoLab; 2mg/kg), Primidone (Sigma; 1 mg/kg), Carbamazepine (Sigma; 50 mg/kg), Mavatrep (10mg/kg) or vehicle, which consisted of 0.5% Tween80 in isotonic NaCl (Sigma). Behavioral testing was performed 30 min after dosing.

Tissue isolation

Mice were euthanized at the indicated time points after IoN surgery with an overdose of pentobarbital (150 mg/kg, i.p.). The infraorbital nerve was exposed using blunt dissection and as much of the nerve adjacent to the ligation site was removed by transecting the nerve distally and proximally. The nerve tissue was placed in 4% PFA for 24 h and later transferred to PBS. The ipsilateral and contralateral trigeminal ganglia were dissected free, rinsed in PBS, transferred to an RNAse free container, snap frozen in liquid nitrogen and stored at −80°C.

Bulk RNA seq

RNA extraction was performed using column purification with innuPREP RNA Mini Kit 2.0 (Innuscreen GmbH, Berlin). Quality control of the RNA samples was performed using the Agilent Bioanalyzer, and only samples with a RIN score >7.5 were retained for further analysis. The RNA samples were sequenced on an Illumina NovaSeq 6000 (Flowcell 100 cycles kit 1.5 single reads). Reads were aligned against the mouse GENCODE gene set release M28 using the STAR aligner (v2.7.10b)⁸⁴ and gene and isoform expression levels were estimated using the RSEM (v1.3.1) software package.⁸⁵

Histology & analysis

Following isolation of the infraorbital nerve tissue 4 μm-thick slices were cut and stained using hematoxylin & eosin (H&E), or immunostained to address protein expression of the neuronal marker PGP9.5, and for SPPR1 and Calbindin2. For immunostaining, sections were subjected to a series of deparaffinization and rehydration, respectively toluene and 100% ethanol. Endogenous peroxidase was blocked with 3% Hydrogen Peroxidase in TBS buffer. Antigen retrieval was done with a pH 9 Tris solution at 95°C for 1h. The following primary antibodies were used: rabbit anti-SPRR1A primary Ab (1/500), rabbit anti- Calbindin 2 primary Ab (1/1000) and rabbit anti-PGP9.5 primary Ab (1/1000). Sections were incubated with primary antibody overnight at 4°C, followed by an incubation with peroxidase-coupled goat anti-rabbit secondary Ab (1/100;for SPRR1A and Calbindin 2) in normal mouse serum (1/25) for 30 min, or with biotinylated pig anti-Rabbit (1/400; for PGP9.5) in normal mouse serum (1/25) for 30 min followed by 30-min incubation with Streptavidin-peroxidase for 30 min. Positive signals were visualized using the chromogen 3,3-diaminobezidine (DAB Sigma). Reactions were stopped by immersion in water, and sections were counterstained in Mayer Haematoxylin.

Analysis of the immunohistological stainings was done with QuPath software. In brief, nerve bundles were identified and within these, four random region of interest (ROIs) were annotated (both in small and large diameter nerve bundles). The mean intensity of the DAB signal of the four ROIs was calculated using the Compute Intensity Feature of QuPath (resolution 2 μm, tile diameter 25 μm) and averaged per animal. At every analyzed time point, data from 3 different animals were obtained, and intensity values were then normalized to the mean intensity value for sham at day 3 for each genotype.

Functional assays of hTRPM3 activity

Whole-cell patch-clamp recordings were performed using an EPC-10 amplifier and PatchMaster software (HEKA Elektronik). Current measurements were done at a sampling rate of 20 kHz, and currents were digitally filtered at 2.9 kHz. In all measurements, 70% of the series resistance was compensated. The standard internal solution contained (in mM): 100 CsAsp, 45 CsCl, 10 EGTA, 10 HEPES, and 1 MgCl₂ (pH 7.2 with CsOH); the standard extracellular solution contained (in mm): 150 NaCl, 1 MgCl₂, and 10 HEPES (pH 7.4 with NaOH), and was supplemented with pregnenolone sulfate (PS; 40 μM) when indicated (Figures 7A and 7B). The standard patch pipette resistance was between 2 and 4 MΩ when filled with pipette solution. All experiments were performed at room temperature (22 ± 2°C). To evaluate TRPM3 channel activity at basal level and in the presence of PS (40 μM), a voltage ramp protocol was applied in which voltage changed from −150 mV toward +150 mV over 400 ms, from a holding potential at 0 mV. For quantification, currents were normalized to the membrane capacitance, to compensate for differences in cell size.

To obtain concentration–response curves for PS action, 96-well plates containing HEK293T-cells co-expressing the channel constructs and jRCAMP1b were washed and filled with extracellular solution containing 150 NaCl, 4 KCl, 2 CaCl₂, 2 MgCl₂ and 10 HEPES (pH 7.4 with NaOH), and placed in the Hamamatsu FDSS/μCell kinetic plate imager. jRCAMP1b fluorescence (excitation: 560 nm; emission: 590 nm) was monitored at 0.5 Hz, while PS at concentrations ranging from 0.1 to 100 μM was automatically pipetted into the wells from preloaded compound plates. At the end of the experiments, all wells were stimulated with a solution containing the calcium ionophore ionomycin (to a final concentration of 20 mM) and 20 mM CaCl₂, to produce saturation of the jRCAMP1b fluorescence for normalization. A concentration-response curve was obtained by normalizing the PS-evoked increase in normalized jRCAMP1b fluorescence to the maximal response. Experiments were performed at room temperature (22 ± 2°C).

Quantification and statistical analysis

The time courses of behavioral readouts are expressed as mean ± SEM. Scatterplots showing behavioral data from individual animals show the mean value (horizontal line) ± SD (whiskers). Normality was assessed using the Shapiro-Wilk test. Significance was evaluated using one-way ANOVA, two-way ANOVA and repeated measures ANOVA as indicated in the figure legends. The Holm-Bonferroni post-hoc test was used to assess significance between individual groups. Data analysis and statistics were performed using OriginPro 2023b.

Differential expression analysis was performed using the DESeq2 (v1.44.0) R package following the standard workflow outlined in the software vignette.⁴⁴ To identify differentially expressed genes (DEGs) we used a false discovery rate (FDR) of 0.05 and a minimal change of 25%. Genes that did not reach 50 reads in any of the samples were discarded from analysis. Enrichment analyses of gene sets were performed and visualized using the DOSE (v4.0.0) and enrichplot (v1.26.6) R packages respectively.⁸⁶ The RNAseq data included samples from male and female mice, which was confirmed based on expression of the Xist gene (X-inactive specific transcript, expressed exclusively in females). The study was not designed to have the statistical power to identify sex-specific genes when identifying DEGs for treatment or genotype. However, sex determination was included in the DESeq2 model design. To evaluate potential bacterial infection of the isolated ganglia, we also checked for the presence of bacterial signatures. The number of non-mouse reads that could be aligned to bacterial genomes was very low (<2 reads per million) and not different between sham and IoN-CCI mice, indicating that bacterial infection is unlikely to contribute to the behavioral and transcriptomic changes observed following IoN-CCI.

Rare-Variant Collapse Analysis was conducted using the UK Biobank Resource under application number 103552. The UK Biobank is a large, prospective cohort comprising over 500,000 participants aged 40–69 years recruited across the United Kingdom between 2006 and 2010.⁸⁷ Genetic and phenotypic data were obtained using UK Biobank’s standardized pipelines. Participants with ICD-10 coding G50.0, Trigeminal neuralgia, were identified using UK Biobank’s hospital episode statistics. TN was modeled as a binary trait, with cases defined as individuals having TN and controls defined as those without the code of interest.TRPM3 variants were identified from UK Biobank whole genome sequencing data. Variants were annotated using the Ensembl Variant Effect Predictor (VEP),⁸⁸ and each variant was assigned to a VEP-defined functional consequence bin reflecting its predicted molecular impact. Participants with ICD-10 G50.0 were segregated into cohorts representing the presence of one or more rare variants within a given VEP bin. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated for each VEP bin. To correct for multiple testing across genes and annotation categories, the Benjamini–Hochberg procedure was applied to control the false discovery rate (FDR). Adjusted p-values (P_adj) were computed and associations with P_adj < 0.05 were considered statistically significant.

Published: March 9, 2026

Supplemental information

Table S1. DEGs for the comparison between sham and IoN-CCI, related to Figure 2

Excel file containing a list of all significantly up- or downregulated genes, their cumulative log₂(fold change), time point of maximal (upregulated DEGs) or minimal (downregulated DEGs) expression, maximal absolute log₂(fold change), P_adj, and classification (EU, early up; SU, stable up; LU, late up; ED, early down; SD, sustained down)

Table S2. Genome-wide changes in expression in the IoN-CCI model, related to Figure 2

Excel file containing the mean and standard error of the mean of the normalized gene counts at every time point (0, 3, 11, 18, and 42 days), for each genotype (WT versus knockout [Trpm3^−/−]) and treatment (IoN-CCI versus sham)