EGFR activation sensitizes trigeminal NMDA receptors to promote pain and morphine analgesic tolerance in oral cancer

Abstract

Pain associated with oral cancer is debilitating. Opioids are the gold standard for cancer pain management, but tolerance and side effects limit their use. Epidermal growth factor receptor (EGFR) signaling is commonly amplified in oral tumors. Here, we found that EGFR activation contributed to both oral cancer pain and opioid tolerance by sensitizing trigeminal ganglion (TG) cells, the main sensory neurons innervating the face and mouth. EGFR ligands were secreted by oral squamous cell carcinoma (OSCC) cells and by peripheral glial cells cocultured with OSCC cells. In human OSCC and an orthotopic mouse model, EGFR was abundant in tumor-innervating TG nerves. Oral cancer pain and opioid tolerance in the mice were increased by EGFR ligands and reduced by EGFR inhibitors. In mice, the abundance of glutamate-type NMDA receptors (NMDARs) was also increased in both the TG and the brainstem. Upon activation by ligands or OSCC cell supernatant, EGFR phosphorylated the NMDAR subunit GluN2B, which increased electrical currents and sensitized pre- and postsynaptic NMDARs in the brainstem. This sensitization was also seen in the brainstems of mice receiving chronic morphine treatment and was mitigated by EGFR blockade. These findings suggest that EGFR-targeted cancer therapeutics may be repurposed to manage cancer pain and reduce opioid tolerance in patients with OSCC.

INTRODUCTION

Oral squamous cell carcinoma (OSCC) is a painful disease that severely impairs eating, drinking, and talking (1–5). Patients with OSCC are less opioid responsive and develop opioid tolerance quicker than patients with other chronic pain conditions (6, 7). Escalating doses of opioids are required as tolerance develops, causing not only severe adverse effects (6) such as addiction but also prolonged hospitalizations and increased readmission rate in patients (8). The mechanisms underlying oral cancer pain and opioid tolerance are not well understood.

Epidermal growth factor receptor (EGFR) is a member of the receptor tyrosine kinases (RTKs) subfamily named HER/ERBB that is aberrantly expressed in 80 to 100% of the OSCC cases (9–11). EGFR antagonism including antibodies and tyrosine kinase inhibitors (TKIs) are US Food and Drug Administration (FDA) approved to treat many cancers, including OSCC (12–14). Clinical studies report pain relief in patients with neuropathic pain, including those associated with cancer after treatment with EGFR inhibitors (15, 16). Human genetic studies find associations between painful disease conditions and EGFR and its ligands, such as epiregulin, heparin-binding EGF (HB-EGF), and transforming growth factor–α (TGFα) (17–19). In animal models, HB-EGF directly causes dorsal root ganglion (DRG) cell excitation and elicits pain-like behaviors, whereas epiregulin mainly works in conjunction with underlying inflammation or tissue injury to generate pain (17, 19, 20). EGFR has also been implicated in opioid tolerance in human (21) and animal studies (20, 22, 23). Although elegant mechanistic studies suggest that EGFR can activate various pathways—through the ion channel transient receptor potential vanilloid 1 (TRPV1), the kinase cascade phosphatidylinositol 3-kinase (PI3K)–AKT–mammalian target of rapamycin (mTOR), the protease matrix metalloproteinase–9 (MMP-9), or the oncogene KRAS [to tetrahydrobiopterin (BH4)]—to increase pain sensitivities in mice (17, 24), these studies were done at the spinal level of noncancer pain models. It has been shown that EGFR involvement in pain is ligand and disease dependent (17, 25, 26), which warrants the investigation of EGFR signaling both in the setting of oral cancer pain that involves the trigeminal system and in opioid analgesic tolerance.

In this study, we aimed to determine how EGFR signaling contributes to oral cancer pain and opioid tolerance. We found that EGFR activation sensitizes trigeminal ganglion (TG) neurons and enhances glutamate N-methyl-d-aspartate receptor (NMDAR) signaling, resulting in heightened cancer pain and diminished opioid analgesic efficacy. EGFR ligands abundant in the OSCC tumor microenvironment trigger calcium influx, NMDAR phosphorylation, and protein kinase C (PKC) up-regulation in TG neurons. Last, we showed that EGFR activation induces presynaptic and postsynaptic hypersensitivity of NMDARs in the trigeminal nucleus caudalis (TNc) of the brainstem. Together, these findings establish EGFR-mediated NMDAR sensitization as a central mechanism underlying oral cancer pain and opioid tolerance and highlight EGFR as a promising therapeutic target.

RESULTS

EGFR ligands are dysregulated in the OSCC tumor microenvironment

Of the seven EGFR ligands, we focused on HB-EGF and TGFα, because there is ample literature suggesting that HB-EGF, TGFα, or both are up-regulated in human oral cancerous (27–29) and painful tissues (19). We first quantified HB-EGF and TGFα concentrations in both human and mouse OSCC cells to determine whether the ligand release is cell line specific. Two human OSCC cell lines were used: KCCT-781 (KCCT) cells, which are sensitive to EGFR inhibitors, and HSC-3 cells, which are not sensitive to EGFR inhibitors (fig. S1). Precancerous human oral keratinocytes [dysplastic oral keratinocytes (DOKs)] were used as a control cell line. The KCCT cells released higher TGFα compared with precancerous DOK cells or HSC-3. HSC-3 cells released higher concentrations of HB-EGF compared with DOK or KCCT cells (Fig. 1, A and B). These data suggest that EGFR ligand release is cell line dependent. Given that Schwann cells (SCs) are the peripheral glia that are intimately associated with primary afferent neurons and modulate neuronal function and activity, we also examined whether human SCs, after stimulation with human OSCC cells, could increase the production of HB-EGF and/or TGFα. We found that SCs released undetectable or low levels of HB-EGF and TGFα at the baseline (Fig. 1, A and B). After a noncontact coculture with OSCC cells for 2 days, despite the difference in ligands produced by KCCT and HSC-3 cells, both OSCC cell lines stimulated SCs to produce more HB-EGF and TGFα (Fig. 1, A and B). Next, we quantified HB-EGF and TGFα in freshly isolated patient oral cancerous tissues using matched contralateral normal oral tissues as controls (Fig. 1C). HB-EGF, but not TGFα, was significantly up-regulated in patient tumors. Moreover, we quantified HB-EGF and TGFα in a syngeneic, orthotopic mouse model. The model was produced by inoculating a mouse OSCC cell line (MOC2 cells) into the mouse tongue (30, 31); tissues were harvested at day 14 after inoculation. The MOC2-inoculated tongue exhibited increased TGFα abundance compared with the sham, whereas that of HB-EGF was not different (Fig. 1D). Collectively, these data suggest that although EGFR signaling is dysregulated in the oral cancer, as previously shown (27, 28), the specific ligand released by the tumor microenvironment is highly cell line and context dependent.

Fig. 1. HB-EGF and TGFα released from cells in the tumor microenvironment.

(A and B) HB-EGF or TGFα concentration in the supernatant of DOKs, human OSCC, and OSCC-conditioned human Schwann cells (hSCs). One-way ANOVA, followed by Tukey’s multiple comparisons test. N = 3 independent experiments per group. (C) HB-EGF or TGFα concentration in patient OSCC tissues compared with the contralateral normal oral tissues. N = 6 patients; four males, two females, analyzed by paired t tests. (D) HB-EGF and TGFα concentration in mouse tongues 14 days after inoculation with MOC2 cells. N = 3 mice per group, analyzed by a t test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

EGFR is overexpressed in human nerves adjacent to OSCC and in TGs of cancer mice

It was previously shown that EGFR is expressed in mouse DRG neurons (17). Here, we examined whether EGFR is expressed in peripheral nerves in resected tumors from patients and mouse TGs. We conducted immunofluorescence staining in resected OSCC tumors from patients. Because we did not have matched normal human oral tissues as controls, we compared EGFR expression in nerves associated with perineural invasion (PNI) and those without. PNI, defined by cancer cells surrounding or invading the nerve, is a common pathological feature of OSCC and worsens pain and quality of life in patients (5, 32). We found significantly greater EGFR expression in glial fibrillary acidic protein (GFAP)–negative nerve areas from cases with PNI; in contrast, EGFR expression in GFAP-positive nerve areas (SCs) was not different from those from PNI-negative cases (Fig. 2, A to D), suggesting that EGFR was increasingly expressed mostly in neurons but not SCs. EGFR staining specificity was evaluated in normal human oral tissues (fig. S2), demonstrating robust EGFR positivity in the basal layer of keratinocytes and an absence of staining in muscle fibers. In naïve mouse TGs, we quantified EGFR expression in cells expressing the neuronal marker NeuN. EGFR was expressed in 69.73% of NeuN⁺ TG neurons (Fig. 2E). In the MOC2 oral cancer mouse model, we performed RNAscope in situ hybridization and quantified Egfr mRNA expression in NeuN⁺ TG neurons. In the sham mice, 55.5% of neurons expressed Eg fr, compared with 77.8% in the tumor-bearing mice (Fig. 2F; the validation of the mouse EGFR antibody is shown in fig. S3). These data suggest that EGFR is overexpressed in both human and mouse nerves associated with the tumor.

Fig. 2. EGFR expression in human nerves and mouse TGs.

(A and B) EGFR and GFAP staining in resected human OSCC tissues without PNI (A) or with PNI (B). CK, cytokeratin. T, tumor area. Red arrows, GFAP⁺ areas only. Yellow arrows, GFAP⁺/EGFR⁺ double-positive areas. Green arrows, EGFR⁺ areas only. (C and D) EGFR fluorescence intensity in GFAP⁻ and GFAP⁺ OSCC tissues with PNI (N = 12) and those without PNI (N = 8). Data were analyzed in two images per sample using a t test. (E) Representative staining for EGFR (red) and NeuN (green) in mouse TG sections. Images are representative of N = 3 mice; five images per mouse and three mice per group were quantified. High-powered mono-channel images are shown to the right. (F) RNAscope in situ hybridization of Egfr in NeuN⁺ TG neurons from sham and MOC2-inoculated oral cancer mice. Data were quantified from one or two images per mouse from four mice per group. Scale bars, 50 μm (A and G), 100 μm (B), and 20 μm (F). *P < 0.05 and ***P < 0.001.

TGFα induces mechanical allodynia and calcium influx in TG neurons

We then tested whether TGFα, like HB-EGF (19), can directly produce nociception and induce calcium influx in neurons. We observed a dose-dependent reduction in mechanical thresholds in both male and female mice 30 min after orofacial injection of TGFα (1.67 to 50 ng, 10 μl), with no sex-based differences in withdrawal thresholds (Fig. 3A). TGFα-induced reduction in mechanical thresholds lasted for at least 1 hour, which was completely reversed by the AG1478 EGFR-TKI (10 mg/kg) and partially reversed by gefitinib EGFR-TKI (50 mg/kg) in both male and female mice (Fig. 3, B and C). On the basis of its higher potency and solubility compared with gefitinib, AG1478 was used in all subsequent experiments. Next, we stimulated isolated TG neurons with TGFα (100 ng/ml); TG neurons were preincubated with either vehicle or AG1478 (10 μM) for 30 or 60 min before TGFα stimulation. AG1478 preincubation, in a time-dependent manner, reduced both peak response to TGFα and the percentage of responding neurons among KCl (50 mM)–responsive, viable neurons (Fig. 3, D to F) compared with vehicle preincubated neurons. These data showed that TGFα can directly act on TG neurons to modulate neuronal excitability and induce nociception.

Fig. 3. TGFα reduces paw withdrawal thresholds and induces calcium influx in TG neurons.

(A) Effect of increasing doses of TGFα on orofacial mechanical thresholds in male and female mice. N = 7 mice per treatment per sex. (B) Time course plot of von Frey paw withdrawal thresholds in male and female mice after tongue injection of TGFα (50 ng) with and without EGFR inhibitors (2 and 10 mg/kg subcutaneously). Veh, Vehicle; AG, AG1478; GF, gefitinib, N = 6 mice per treatment group per sex. (C) Area under the curve (AUC) of the data in (B). (D) Representative Ca²⁺ imaging traces of TG neurons in response to TGFα stimulation with and without EGFR inhibitor AG1478. Each line represents one cell; six neurons per group are shown. (E) Comparison of TGFα-induced calcium increase from baseline in KCl-responding neurons after preincubation with vehicle alone or with AG1478 (10 μM) for either 30 or 60 min (168 to 808 neurons from three independent experiments). (F) Percentage of TGFα-responding neurons after vehicle, 30-min AG1478 (10 μM), or 60-min AG1478 (10 μM) preincubation. One-way ANOVA, followed by Tukey’s multiple comparisons test: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. ns, not significant.

Suppressing EGFR signaling reduces oral cancer pain and calcium influx induced by oral cancer supernatant

We established an orthotopic syngeneic model of OSCC by inoculating a mouse OSCC line (MOC2 cells) into the mouse tongue. Facial von Frey was used to test referred pain, experienced by patients and mice with tongue SCC (4, 33). As expected, both male and female mice with MOC2 cell inoculation exhibited reduced facial von Frey thresholds over time, indicative of an increased mechanical allodynia (Fig. 4, A and B). Starting on postinoculation day (PID) 7, mice were treated with AG1478 [10 mg/kg, subcutaneously (sc)] or vehicle. AG1478 treatment significantly increased facial von Frey thresholds 1 hour after treatment tested at PIDs 7, 10, and 14 compared with vehicle-treated groups (Fig. 4, A and B). There was no difference in tumor volume between sex or treatment groups (Fig. 4, C and D). Collectively, these results show that oral cancer pain can be alleviated by EGFR blockade.

Fig. 4. EGFR inhibitor reduces mechanical allodynia in mice with OSCC and reduces TG neuron calcium influx induced by OSCC cell supernatant.

(A and B) Facial von Frey mechanical thresholds in MOC2-inoculated male (M) and female (F) mice treated with EGFR-TKI AG1478 (AG,10 mg/kg, subcutaneously injected in days marked with arrows) or with vehicle. Two-way ANOVA: ****P < 0.0001. (C and D) Tumor volume in vehicle- and AG1478-treated tongues in male and female mice. N = 6 mice per treatment group per sex. (E) Representative Ca²⁺ imaging traces of TG neurons in response to HSC-3 supernatant (sup) stimulation. Each line represents one cell. Twelve cells per group are shown. (F) Comparison of HSC-3 supernatant–induced calcium levels in KCl-responsive neurons at baseline or when preincubated with vehicle or AG1478 (10 μM). N = 263 (veh) and 198 (AG) neurons from three independent experiments. (G) Percentage of HSC-3 supernatant–responding neurons treated with vehicle or AG1478. N = 263 (veh) and 198 (AG) neurons from three independent experiments. (H) Representative Ca²⁺ imaging traces in TG neurons upon stimulation with MOC2 supernatant (sup). Each line represents one cell; 12 neurons per group are shown. (I) Comparison of MOC2 supernatant–induced calcium levels in KCl-responsive neurons at baseline and when preincubated with vehicle or AG1478 (10 μM). N = 261 (veh) and 184 (AG) neurons from three independent experiments. (J) Percentage of MOC2 supernatant–responding neurons treated with vehicle or AG1478. N = 261 (veh) and 184 (AG) neurons from three independent experiments. In (C) to (D and (F) to (J), analyzed by t tests: ***P < 0.001 and ****P < 0.0001.

Supernatant from oral cancer cells contains EGFR ligands and has been previously shown to evoke calcium influx in isolated mouse TG neurons (34, 35). We then tested whether EGFR inhibitors could directly act on neurons to prevent cancer supernatant–induced neuronal activation. Preincubation of TG neurons with EGFR inhibitor (AG1478, 10 μM) for 30 min significantly reduced calcium response and percentage of responders evoked by supernatant collected from either the human oral cancer cell line HSC-3 (Fig. 4, E to G) or MOC2 cells (Fig. 4, H to J). The calcium imaging experiments further confirm that EGFR inhibitors can act on neurons directly to suppress nociceptive signaling.

EGFR ligands HB-EGF and TGFα reduce, whereas EGFR inhibitors improve, morphine analgesia efficacy

To determine the role of EGFR in morphine analgesic tolerance, we used two models—a classical model, where morphine was injected repetitively in naïve mice, and a tumor-bearing model, where morphine was injected repetitively in mice with tongue SCC. In naïve mice, the administration of morphine (10 mg/kg sc, daily) resulted in analgesic tolerance at day 4 (Fig. 5A). EGFR ligands, HB-EGF or TGFα (0.1 mg/kg), when coinjected with morphine (10 mg/kg sc, daily), significantly accelerated the analgesic tolerance development to morphine, as shown by the reduced paw withdrawal and tail withdrawal latencies measured with the Hargreaves’s test (Fig. 5, A and B). The morphine + EGFR ligand groups became indistinguishable from the vehicle group earlier than the morphine-alone group (days 2 to 3 versus day 4). In contrast, EGFR TKI AG1478 (10 mg/kg), when coinjected with morphine, significantly slowed down the morphine tolerance development (day 8 versus day 4). Morphine also induced alteration in locomotion, such as increased track length and total walking, trotting, and running time, which were significantly reduced by AG1478 (Fig. 5, C and D). Total grooming duration (grooming time spent on the whole-body region) and time spent in the inner zone, a measure of anxiety-like behaviors, were reduced by morphine but were not significantly reversed by AG1478 (Fig. 5, E and F). Although not statistically significant, HB-EGF, but not TGFα, tends to attenuate the effect exerted by morphine on locomotion, grooming, and anxiety-like behaviors (Fig. 5, C to F). These results suggest that TGFα is specific to pain-like behaviors, whereas EGFR and HB-EGF may serve additional roles such as morphine-induced changes in locomotion.

Fig. 5. HB-EGF or TGFα reduces, whereas EGFR inhibition improves, morphine analgesia efficacy in mice.

(A and B) Paw and tail withdrawal latencies measured in mice treated subcutaneously with vehicle, morphine (10 mg/kg), or morphine plus either EGFR ligand (0.1 mg/kg) or EGFR inhibitor AG1478 (10 mg/kg) using the Hargreaves’s test. (C to F) Track length (C); total walking, trotting, and running times (D); total time spent grooming in any manner (E); time spent in the inner zone (F). N = 5 mice per treatment group were used in (A) to (F). (G and H) Facial von Frey thresholds in sham or MOC2-inoculated mice subcutaneously coinjected with morphine (10 mg/kg) plus either vehicle or the EGFR inhibitor AG1478 (10 mg/kg) daily for 9 days. (I) Volume of tumors on the tongue in mice used in (G) and (H). N = 6 per group per sex were used in (G) to (I). In (A) to (H), data were analyzed by one-way ANOVA, followed by Tukey’s multiple comparisons test. In (I), data were analyzed by t test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

In a separate test, mice were inoculated with MOC2 cells into the tongue and treated with daily morphine (10 mg/kg), in combination with AG1478 (10 mg/kg) or vehicle starting on PID 7. Mechanical thresholds were measured after morphine administration every day using the facial von Frey test. Compared with the sham mice treated with morphine, mechanical thresholds were significantly lower in MOC2-inoculated mice treated with morphine (Fig. 5, G to H). AG1478 significantly increased withdrawal thresholds of morphine-treated cancer and sham mice, indicative of reduced morphine tolerance during the observation period (Fig. 5, G to H). No statistically significant difference in tumor volume was observed in AG1478- or vehicle-treated tumor-bearing mice after chronic morphine administration (Fig. 5I). Thus, in both naïve and tumor-bearing mice, our data suggest that EGFR stimulation increases, whereas EGFR blockade reduces, opioid analgesic tolerance.

Glutamate release, NMDARs, and PKC proteins are increased in the TGs and brainstems of OSCC mice

Next, we questioned whether EGFR-mediated oral cancer pain and morphine tolerance could be signaled through NMDARs, which are critically involved in the amplification of pain signals and the development of opioid tolerance (36–38). First, we examined whether oral cancer increased NMDAR expression in the trigeminal nervous system. We measured the obligatory GluN1 subunit protein level in the TGs and brainstems collected from mice inoculated with MOC2 cells in the tongue using Western blotting analysis at PID 14. GluN1 expression was increased in both the TGs and brainstems of OSCC mice compared with the sham (Fig. 6A), suggesting that NMDARs are indeed up-regulated by oral cancer. Next, we used enzyme-linked immunosorbent assays (ELISAs) and demonstrated that both HB-EGF (50 ng/ml) and TGFα (50 ng/ml) stimulation increased glutamate release from cultured TG neurons (Fig. 6B). Because PKC is a postulated mechanism of NMDAR sensitization in chronic pain and morphine tolerance (39), we measured PKC protein expression in TG neurons stimulated with TGFα (50 ng/ml). TGFα significantly increased PKC expression in TG neurons (Fig. 6C). Mice implanted with MOC2 tongue cancer cells exhibited increased expression of PKC proteins in the TG and brainstem tissues at PID 14 (Fig. 6D). In the TG, PKC expression was reduced in tumor-bearing mice receiving EGFR blockade (AG1478, 10 mg/kg intraperitoneally) compared with the vehicle group (Fig. 6D). A pan-PKC blocker (AEB071) significantly reduced TGFα–induced mechanical allodynia (Fig. 6E), confirming a functional role of PKC in orofacial pain.

Fig. 6. NMDAR and PKC expression and glutamate release in the TG and TNc.

(A) Western blotting for GluN1 in TG and brainstem tissues from sham (N = 4) mice and mice with tongue OSCC (N = 5). Relative protein quantification of GluN1 in which β-tubulin was the loading control. (B) Glutamate release before and after application of HB-EGF (50 ng/ml) or TGFα (50 ng/ml) in cultured TG neurons from naïve mice (N = 3). (C) PKC expression in cultured mouse TG neurons in response to TGFα (50 ng/ml) or vehicle, assessed by ELISA. OD, optical density. Data are from neurons isolated separately from six mice per treatment group. (D) PKC protein quantification of TG and brainstem tissues collected from mice 14 days after sham or MOC2 inoculation in the tongue, with days 7, 10, and 14 therein treated with vehicle or AG1478. Data are from N = 5 mice per group. (E) Facial mechanical allodynia in naïve mice treated with TGFα (50 ng/10 μl injected in the tongue) with vehicle or the pan-PKC inhibitor AEB071 (10 mg/kg). Vehicle or AEB071 was intraperitoneally injected 1 hour before TGFα injection. Data are from six mice per group. Data were analyzed by t test in (A) to (C), one-way ANOVA with Tukey’s test in (D), and two-way ANOVA with Tukey’s test in (E): *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

EGFR physically interacts with NMDARs and phosphorylates NMDARs upon activation by TGFα

To determine how EGFR activation can influence NMDAR function, we first examined whether EGFR physically interacts with NMDAR, as previously shown in cortical neurons (40). We focused on the GluN2B subunit because it is known to be particularly important for neuropathic pain and opioid tolerance (36, 37). In cultured mouse TG neurons, we observed colocalization of EGFR and GluN2B by confocal microscopy (Fig. 7A) and the proximal ligation assay (PLA) (Fig. 7B). The application of TGFα to TG neuron cultures induced an increase in GluN2B phosphorylation, which was inhibited by the EGFR-TKI AG1478 (Fig. 7, C and D).

Fig. 7. EGFR interacts with and phosphorylates GluN2B in TG neurons.

(A) Representative images of primary mouse TG neurons stained for EGFR and GluN2B are shown in the top panels. In the bottom left, a random area (white line in the magnified image) was selected for an example of pixel intensity profile measurement. The representative fluorescence intensity profile from 15 neurons of two mice shows that EGFR and GluN2B intensity peaks at the same location. Scale bars are 10 μm (top) and 2 μm (bottom). (B) Representative immunofluorescence images of primary mouse TG neurons assessed for the proximity of EGFR to GluN2B by PLA signal (left image, yellow). Nuclei were stained with DAPI (blue). Right image is the negative control. Images are representative of 102 neurons from three mice. Scale bars, 10 μm. (C and D) Representative images (C) and analysis (D) of fluorescent signal intensity for phosphorylated GluN2B in cultured TG neurons treated with vehicle, TGFα (100 ng/ml) + vehicle, or TGFα (100 ng/ml) + AG1478 (10 μM) for 1 hour. Data are from 24 to 32 neurons from two mice per group, analyzed by one-way ANOVA with Tukey’s test: *P < 0.05 and ****P < 0.0001. Scale bars [(B) and (C)], 10 μm. a.u., arbitrary unit.

EGFR stimulation increases presynaptic NMDAR activity of SG neurons at the TNc

In mouse brainstem slices, we first recorded miniature excitatory postsynaptic currents (mEPSCs), which reflect quantal release of glutamate from presynaptic terminals (41, 42). Treatment with HB-EGF (100 ng/ml) or TGFα (100 ng/ml) in brainstem slices for 30 min significantly increased the frequency of mEPSCs in substantia gelatinosa (SG) neurons in the TNc. This increase was readily reversed by bath application of the NMDAR antagonist AP5 (50 μM) (Fig. 8, A to C). The amplitude of mEPSCs was not significantly affected by EGFR stimulation, with or without AP5 application (Fig. 8D).

Fig. 8. EGFR stimulation augments presynaptic NMDAR activity in brainstem SG neurons.

(A) Representative mEPSC traces of SG neurons from mouse TNc slices pretreated with vehicle, HB-EGF, or TGFα for 30 min, followed with AP5 application and washout. (B) Cumulative probability plots showing the effect of bath application of 50 μM AP5 (6 min) on mEPSCs of SG neurons from mouse TNc slices pretreated with vehicle, HB-EGF, or TGFα for 30 min. (C) Effects of treatment with vehicle, HB-EGF, or TGFα on the frequency of mEPSCs of SG neurons. (D) Effects of treatment with vehicle, HB-EGF, or TGFα on the amplitude of mEPSCs of SG neurons. Washout recordings were obtained 10 to 15 min after discontinuing AP5 application. Data are representative of (A) and (B) and means ± SEM (C) and (D) of n = 15 neurons from seven mice per group. *P < 0.05 and ***P < 0.001 by two-way ANOVA with Tukey’s post hoc test.

EGFR stimulation increases NMDAR activity at primary afferent central terminals that form synapses with SG neurons at the TNc

To determine specifically whether increased presynaptic NMDAR activity by EGFR activation occurs at primary afferents that form synapse with TNc neurons, we recorded EPSCs of SG neurons monosynaptically evoked from the dorsolateral edge of the TNc in brainstem slices. The baseline amplitude of evoked EPSCs of SG neurons was significantly larger in HB-EGF– or TGFα–treated slices than in vehicle-treated slices (Fig. 9, A and B). The bath application of 50 μM AP5 for 6 min rapidly reversed the increased amplitude of monosynaptic EPSCs of TNc neurons in HB-EGF– or TGFα-treated slices but had no such effect in vehicle-treated slices (Fig. 9, A and B). In addition, we recorded the paired-pulse ratio (PPR) of monosynaptically evoked EPSCs, a measure of the probability of glutamate release from presynaptic terminals (43). The baseline PPR of evoked EPSCs of SG neurons was significantly lower in HB-EGF– or TGFα–treated slices than in vehicle-treated slices (Fig. 9, C and D). In HB-EGF– or TGFα–treated slices, the bath application of 50 μM AP5 inhibited the first evoked EPSCs more than the second evoked EPSCs, resulting in an increase in the PPR of evoked EPSCs (Fig. 9D). AP5 application had no effect on the PPR of evoked EPSCs in neurons from vehicle-treated brainstem slices (Fig. 9D). These data suggest that EGFR activation augments presynaptic NMDAR activity at primary afferent terminals that form synapses with TNc neurons.

Fig. 9. EGFR stimulation increases activity of presynaptic NMDARs at primary afferent central terminals that form synapses with brainstem SG neurons.

(A) Representative EPSC traces show the effect of bath application of 50 μM AP5 for 6 min on evoked monosynaptic EPSCs of SG neurons from mouse TNc slices pretreated with vehicle, HB-EGF, or TGFα for 30 min. (B) Effects of vehicle, HB-EGF, or TGFα on the amplitude of evoked EPSCs in SG neurons. (C) Representative recording traces of PPR of evoked EPSCs of SG neurons from mouse TNc slices pretreated with vehicle, HB-EGF, or TGFα for 30 min. n = 16 neurons from six mice per group. (D) Effects of vehicle, HB-EGF, or TGFα on PPR of evoked EPSCs in SG neurons. n = 14 or 15 neurons from seven mice per group. *P < 0.05 and **P < 0.01 by two-way ANOVA with Tukey’s post hoc test.

EGFR stimulation augments postsynaptic NMDAR activity of brainstem SG neurons

To determine whether EGFR activation changes the postsynaptic NMDAR activity in the TNc, we recorded NMDAR currents elicited by puff application of 100 μM NMDA directly onto the recorded SG neurons (44). The amplitude of puff NMDAR currents was significantly greater in slices treated with HB-EGF or TGFα than in vehicle-treated slices (Fig. 10, A and B). These findings suggest that EGFR activation also increases the activity of postsynaptic NMDARs in the TNc.

Fig. 10. EGFR stimulation augments postsynaptic NMDAR activity of brainstem SG neurons.

(A and B) Representative traces (A) and amplitude of NMDA currents in (B) SG neurons pretreated with vehicle, HB-EGF, or TGFα in response to puff application of NMDA. n = 17 to 19 neurons from six mice per group, analyzed by one-way ANOVA with Tukey’s post hoc test, ***P < 0.001.

EGFR inhibition reverses presynaptic NMDAR hyperactivity in brainstem SG neurons caused by prolonged treatment with morphine

The effect of prolonged treatment with morphine on synaptic NMDAR activity has been studied predominantly in the spinal cord (37, 45). Here, we found that morphine treatment (10 mg/kg daily for 7 days) similarly increased the frequency of mEPSCs in the SG neurons in the TNc compared to vehicle-treated mice. This increase was normalized by bath application of AP5 (50 μM;, Fig. 11, A to C). Preincubation of brainstem slices with the EGFR inhibitor AG1478 (5 μM) substantially reduced the frequency of mEPSCs induced by chronic morphine (Fig. 11, A to C). The baseline amplitude of mEPSCs in SG neurons did not significantly differ among treatment groups (Fig. 11D), with or without AP5 application.

Fig. 11. EGFR inhibition reverses presynaptic NMDAR hyperactivity in brainstem SG neurons caused by prolonged treatment with morphine.

(A) Representative mEPSC traces of SG neurons in TNc slices isolated from morphine-treated or vehicle (control)–treated mice (10 mg/kg per day for 7 days). Slices were bath-incubated with vehicle or AG1478 (5 μM) for 30 min before recording at baseline, after application of NMDAR antagonist AP5 (50 μM), and after washout. (B) Cumulative probability plots of mEPSC interval and amplitude of SG neurons described in (A). (C and D) Frequency (C) and amplitude (D) of mEPSCs in SG neurons described in (A). n = 14 or 15 neurons from seven mice per group. *P < 0.05 and ***P < 0.001 by two-way ANOVA with Tukey’s post hoc test.

EGFR inhibition abrogates morphine treatment–induced increases in the activity of presynaptic NMDARs at primary afferent central terminals that form synapses with brainstem SG neurons

The amplitude of evoked monosynaptic EPSCs of SG neurons in TNc slices was much larger in chronic morphine–treated mice than in vehicle-treated mice (Fig. 12, A and B). Bath application of AP5 reversed this increase in evoked EPSCs by morphine treatment, consistent with findings from the spinal cord (37, 46). Treatment of brainstem slices with AG1478 reduced the amplitude of evoked EPSCs in morphine-treated mice (Fig. 12, A and B). Furthermore, the baseline PPR of evoked EPSCs of SG neurons was significantly lower in morphine-treated mice than in vehicle-treated mice, which was reversed by bath application of AP5 (Fig. 12, C and D), suggesting that morphine increased presynaptic NMDAR activity at the primary afferent central terminals. Treatment of brainstem slices with AG1478 inhibited morphine treatment–induced changes in the PPR of evoked EPSCs in SG neurons (Fig. 12, C and D). These data suggest that EGFR signaling is important for opioid-induced presynaptic NMDAR hyperactivity at primary afferent terminals that form synapses with TNc neurons.

Fig. 12. EGFR inhibition abrogates morphine-induced increases in the activity of presynaptic NMDARs at primary afferent central terminals that form synapses with brainstem SG neurons.

(A and B) Representative EPSC traces (A) show the effect of bath application of 50 μM AP5 on evoked monosynaptic EPSCs and (B) amplitude of SG neurons. (C and D) Representative EPSC traces (C) and PPR of evoked EPSCs in SG neurons after bath application of 50 μM AP5 In (A) to (D), TNc slices were isolated from morphine-treated or vehicle (control)–treated mice (10 mg/kg per day for 7 days). Slices were bath-incubated with vehicle or AG1478 (5 μM) for 30 min before recording at baseline and after application of NMDAR antagonist AP5 (50 μM). n = 15 or 16 neurons from eight mice per group. *P < 0.05 and **P < 0.01 by two-way ANOVA with Tukey’s post hoc test.

EGFR inhibition normalizes morphine treatment–induced postsynaptic NMDAR hyperactivity of brainstem SG neurons

In the spinal dorsal horn, prolonged treatment of morphine increased presynaptic, but not postsynaptic, NMDAR hyperactivity (37, 46, 47). Unexpectedly, we found that prolonged morphine treatment also resulted in increased postsynaptic NMDAR activity in the TNc, as shown by puff application of 100 μM NMDA directly onto the recorded SG neurons (Fig. 13A). Treatment with AG1478 significantly reduced the amplitude of puff NMDAR in SG neurons from morphine-treated mice (Fig. 13B). These findings suggest that opioid treatment causes postsynaptic NMDAR hyperactivity in the TNc by EGFR signaling.

Fig. 13. EGFR inhibition normalizes morphine treatment–induced postsynaptic NMDAR hyperactivity of brainstem SG neurons; graphical abstract.

(A and B) Representative NMDAR current traces (A) and quantitative analysis (B) show the effects of treatment with morphine and AG1478 on the amplitude of currents evoked by puff application of NMDA onto SG neurons in mouse TNc slices. TNc slices were obtained from mice treated with morphine (10 mg/kg per day) or vehicle (control) for 7 days. Slices were treated with 5 μM AG1478 or vehicle for 30 min before recordings. n = 14 or 15 neurons from seven mice per group. ***P < 0.001 by one-way ANOVA with Tukey’s post hoc test. (C) Graphic abstract summarizing the key mechanism of the study. EGFR ligands such as HB-EGF and TGFα are abundant in the OSCC tumor microenvironment, leading to heightened EGFR activation and expression in trigeminal neurons. This EGFR signaling increases calcium influx and up-regulates PKC in TG neurons. Moreover, EGFR directly interacts with NMDARs, enhancing presynaptic NMDAR function and promoting glutamate release, which subsequently amplifies postsynaptic NMDAR activity in the TNc. Together, these mechanisms contribute to the development of oral cancer pain and the emergence of morphine tolerance.

DISCUSSION

In this study, we demonstrated that EGFR signaling sensitizes trigeminal NMDARs, contributing to oral cancer pain and opioid tolerance. The oral cancer microenvironment releases EGFR ligands, enabling a crosstalk with primary afferent neurons that overexpress EGFR. HB-EGF and TGFα, which are differentially expressed in the tumor and oral cancer cells, are both acutely nociceptive. Each ligand reduced opioid analgesia efficacy and increased presynaptic and postsynaptic NMDAR activity in the TNc. EGFR inhibition reduces oral cancer pain and opioid tolerance as well as trigeminal NMDAR hypersensitivities induced by chronic morphine exposure. Mechanistically, EGFR colocalizes and interacts with NMDARs; EGFR activation promotes GluN2B phosphorylation, increases PKC recruitment, and elevates glutamate release from TG neurons. Key mechanisms of the study are summarized in Fig. 13C.

Oral cancer pain is driven largely by peripheral mechanisms in which cancer cells and other components in the tumor microenvironment sensitize and injure primary afferent neurons (48). Pain severity increases with tumor progression, suggesting that shared molecular drivers that could be targeted to manage both cancer and pain. EGFR is overexpressed in more than 80% of OSCC cases, and multiple EGFR ligands including HB-EGF and TGFα are implicated in oral cancer development and progression (14). We therefore hypothesized that dysregulated EGFR signaling originating from cancer cells amplifies EGFR activity in the tumor microenvironment, including in sensory neurons. Such amplification could heighten trigeminal neuron excitability and could prime the system for reduced opioid responsiveness (“pretolerance”) (20), helping explain the diminished analgesic efficacy and rapid tolerance seen in patients with oral cancer. Supporting this hypothesis, we found that both HB-EGF and TGFα promote oral cancer pain and morphine tolerance, whereas EGFR inhibition reduces both. These findings align with prior work on EGFR, pain, and opioid tolerance in spinal and DRG-based models (17, 20, 23, 26, 32).

EGFR sensitization of NMDARs may have broad implications in other pain conditions in the trigeminal system. Spinal NMDARs are well-established contributors to neuropathic pain caused by traumatic nerve injury or chemotherapy and chronic opioid-induced hyperalgesia and tolerance (37, 46, 47, 49, 50). Clinically, NMDAR antagonists and modulators—including ketamine, memantine and dextromethorphan—provide analgesic benefit in chronic pain (51). Increasing evidence also demonstrates that NMDARs in the trigeminal system drive both peripheral and central sensitization in neuropathic and inflammatory orofacial pain, as well as migraine (52, 53).

EGFR activation can lead to increased NMDAR activity by phosphorylation or overexpression of NMDAR subunits. In mice lacking HB-EGF, both EGFR and GluN1 expression are reduced in the prefrontal cortex (54). GluN2B has been identified as an EGFR substrate in glioma cells, where EGFR phosphorylates its COOH-terminal domain and enhances glutamate-NMDAR signaling (41). Similarly, in rat embryonic hippocampal neurons, EGF increases NMDAR-mediated calcium currents by up-regulating and phosphorylating GluN2B, resulting in long-term potentiation relevant to learning and memory (40). NMDARs can be phosphorylated by cyclic adenosine 3′,5′-monophosphate (cAMP)–dependent protein kinase, PKC, casein kinase 2, Ca²⁺/calmodulin-dependent protein kinase II, and tyrosine kinases Src and Fyn (38). PKC, in particular, increases the probability of NMDARs opening and decreases their affinity for extracellular Mg²⁺ (42), serving as a major contributor to opioid-induced presynaptic NMDAR hyperactivity in the spinal cord (37, 45, 55). PKC can directly phosphorylate the GluN1 and GluN2B subunits and indirectly potentiate NMDAR function via the PKC–cell adhesion kinase β/proline-rich tyrosine kinase 2–Src signaling (38). Our findings suggest that, in addition to direct EGFR-dependent phosphorylation of NMDARs, increased PKC activity may also contribute to EGFR-mediated NMDAR sensitization.

NMDARs expressed at primary afferent central terminals are normally in a quiescent state; they become tonically activated in opioid-induced hyperalgesia and neuropathic pain (38). Opioid-induced hyperalgesia and analgesic tolerance are mediated by presynaptic NMDAR hyperactivity in the spinal cord (37). Morphine treatment decreased the amplitude of evoked NMDAR-EPSCs and NMDAR currents of dorsal horn neurons elicited by puff NMDA application (37), suggesting a paradoxically decreased postsynaptic NMDAR activity. In our study, both presynaptic and postsynaptic NMDARs were potentiated by prolonged morphine exposure, similar to what has been reported for nerve injury–induced pre- and postsynaptic NMDAR hyperactivity in the spinal dorsal horn (49). Our findings reveal the difference in sensory processing between the DRG–spinal cord and TG-brainstem circuitry and suggest that EGFR-mediated NMDAR potentiation may rely on distinct opioid signaling mechanisms in spinal versus brainstem neurons.

Several questions remain. The current study focused on only two EGFR ligands, yet EGFR-driven pain signaling is highly ligand and context dependent. For example, epiregulin shows dichotomous effects in pain (17, 56). Ligands such as EGF needed to be administered repeatedly to induce mechanical hypersensitivity (20), in contrast with immediate effects seen for TGFα and HB-EGF (19). Supporting our study, mRNA encoding TGFα is increased in human pancreatic cancer, a severely painful cancer (19). Down-regulation of TGFα in DRGs correlates with pain reduction in a mouse model of osteoarthritis (57). Different ligands could coexist and exert synergistic, additive, or competitive actions on EGFR expressed in sensory neurons, modulating pain and NMDAR sensitivities. In addition, EGFR could engage other pathways not studied here. In the spinal level, EGFR signaling activated PI3K-AKT-mTOR, MMP-9, and TRPV1 (17) and recruited voltage-gated sodium channels Na_v1.8 and Na_v1.9 and L-type voltage-gated calcium channel Ca_v1.2 by a vesicular chaperone, low-density lipoprotein receptor-related protein 1 (LRP1) (58). In neuropathic pain models, EGFR and KRAS signaling trigger increased guanosine 5′-triphosphate (GTP) cyclohydrolase-1/6R-l-erythro-5, 6, 7, 8-tetrahydrobiopterin (GCH1) expression, leading to increased BH4 and persistent pain sensitivity (24). EGFR has also been shown to modulate cytokine release in glial cells in a DRG compression–induced pain model (26).

Our current findings provide new evidence linking trigeminal EGFR to synaptic NMDAR hyperactivity in oral cancer and opioid tolerance. This information not only extends mechanistic understanding of oral cancer pain and opioid tolerance but also suggests new targets for treating these conditions. FDA-approved EGFR inhibitors such as cetuximab are not very efficacious for head and neck cancer as a standalone treatment. Repurposing these inhibitors for oral cancer pain is appealing, because they are well tolerated and have established side-effect profiles. EGFR inhibitors may offer dual benefits by supporting tumor control while reducing pain and opioid requirements, without increasing the risk of tumor progression.

MATERIALS AND METHODS

Cell culture

Human OSCC cell lines HSC-3 and DOKs were purchased from American Type Culture Collection (ATCC). KCCT-871 (KCCT) cell line was a gift from H. Takahashi, Department of Biology and Function in Head and Neck, Yokohama City University, Japan. OSCC cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) with glucose (4.5 g/liter), l-glutamine, and sodium pyruvate (Gibco, Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS; HyClone), streptomycin sulfate (100 μg/ml), and penicillin G (100 U/ml). DOK cells were cultured in DMEM supplemented with hydrocortisone (5 μg/ml), 10% FBS, streptomycin sulfate (100 μg/ml), and penicillin G (100 U/ml; Lonza). MOC2 cells were purchased from Sigma-Aldrich and were grown in IMDM/F12 (2:1) (HyClone) with 5% FBS, 1% penicillin/streptomycin, EGF (5 ng/ml), hydrocortisone (40 ng/ml; Sigma-Aldrich), and insulin (5 μg/ml) (27). All cells were cultured at 37°C in 5% CO₂. Human SCs were purchased from ScienCell and grown in SC Media (SCM, ScienCell Research Laboratories). Cell passages between 4 and 10 were used for all the experiments. Cells were authenticated by ATCC and routinely tested for Mycoplasma (PlasmoTest Mycoplasma Detection Kit; InvivoGen).

Animals

All procedures involving animals were approved by the Institutional Animal Care and Use Committee (IACUC) at New York University and University of Texas Houston MD Anderson Cancer Center, in accordance with the Guide for the Care and Use of Laboratory Animals published by the US Institute for Laboratory Animal Research (8th edition). Six to 12-week-old C57BL/6 male and female mice (the Jackson Laboratories) were housed in a temperature-controlled, pathogen-free room on a 12:12-hour light/dark cycle (6:00 a.m. to 6:00 p.m.) with ad libitum access to food and water. For all experiments, animals were habituated to handling before behavioral testing.

Supernatant and cell lysate collection

OSCC cell supernatant was collected using a published method (34, 59, 60). Briefly, OSCC cells were cultured until 90% confluency. Media were replaced with fresh serum-free media 48 hours before collection; supernatant was centrifuged to remove dead cells. For cell lysate collection, once OSCC cells in culture reached 70% of confluency, they were homogenized in an ice-cold radioimmunoprecipitation assay (RIPA) buffer containing 10% protease inhibitor cocktail and kept at −80°C until further use. SC supernatant, cell lysate collection, or RNA extraction was performed after being cocultured with OSCC cells. The coculture was prepared using a noncontact coculture approach. A total of 100,000 SCs were seeded onto a six-well culture plate (Celltreat) in SCM with 1% FBS. Cell culture inserts containing SCM alone (control) or 10⁵ OSCC cells (treatment) were then placed on top of the SC culture. To prevent the contamination of released products from OSCC cells, 2 days following coculture, cell inserts containing either SCM or OSCC were discarded, and fresh media was added. Control SCs or SCs cocultured with OSCC cells were allowed to grow for another 48 hours before the supernatant was collected and centrifuged to remove cell debris. For cell lysate collection, cocultured SCs were homogenized in an ice-cold RIPA buffer containing 10% protease inhibitor cocktail and kept at −80°C until further use.

ELISA quantification

The following ELISA kits were used: human HB-EGF (R&D Systems), human TGFα (R&D Systems), glutamate (Abcam), and mouse PKC (Abcam). Total protein concentrations in cell supernatant or homogenized mouse and human tumor tissues were quantified using a QuantiPro bicinchoninic acid assay kit (Sigma-Aldrich). Six fresh patient (four males, two females) oral cancer tissues and matched contralateral normal oral tissues were included. All samples were run in duplicate. The optical density was read at 450 nm wavelength with the GloMax-Multi Microplate Multimode Reader (Promega).

PLA

The Duolink In Situ Orange Starter Kit (DUO92102, Millipore) was used for PLA. Primary mouse TG neurons were cultured for 2 days. Following 1-hour incubation with TGFα (100 ng/ml), neurons were fixed with 4% paraformaldehyde for 10 min. The fixed cells were incubated in blocking solution at 37°C for 60 min, followed by incubation with rabbit anti-EGFR (PA1-1110, Thermo Fisher Scientific) and mouse anti-GluN2B (MMS-5148, BioLegend), all diluted 1:500 in blocking solution, at 37°C for 120 min. After washing, cells were incubated at 37°C for 60 min with anti-rabbit PLA probe PLUS (1:5) and anti-mouse PLA probe MINUS (1:5). Cells were washed, and then the ligation reaction was performed at 37°C for 30 min, followed by the amplification reaction at 37°C for 100 min. After final washes, slides were mounted using mounting medium with 4′,6-diamidino-2-phenylindole (DAPI). The negative control represents the PLA without the primary antibody. Images were taken with a 63× objective on Leica SP8 confocal microscope.

Animal models and tumor volume measurement

For the syngeneic, tongue OSCC model, C57/BL6 mice were inoculated with 5 × 10³ MOC2 cells in 30 μl of 1:1 DMEM and Matrigel into the anterior lateral portion of the tongue. Control sham mice were injected with 30 μl of the mixture of DMEM and Matrigel only. Tumor volume was estimated by measuring maximal length, width, and thickness using a digital caliper at PID 14. EGFR inhibitor AG1478 (10 mg/kg) or its vehicle [10% dimethyl sulfoxide (DMSO) and 40% polyethylene glycol] was given 1 hour before von Frey test and was administered on PIDs 7, 10, and 13.

For morphine tolerance studies in naïve mice, HB-EGF (0.1 mg/kg; Gibco) TGFα (0.1 mg/kg; Gibco), EGFR inhibitor (AG1478; 10 mg/kg), or its vehicle were coinjected subcutaneously daily with morphine (10 mg/kg; Sigma-Aldrich) for nine consecutive days. Paw and tail latency were evaluated on day 0 (baseline) and then daily for eight consecutive days by using the Hargreaves’s test 30 min after subcutaneous coinjection. On day 9, locomotion and anxiety behaviors were performed in the box behavioral spectrometer. For morphine tolerance studies in mice with MOC2 tumors as mentioned above, chronic morphine (10 mg/kg sc, daily) together with EGFR blocker AG1478 (10 mg/kg sc, daily) or DMSO vehicle were started on day 7 after MOC2/vehicle inoculation and lasted for a period of 9 days.

Behavioral assays

For all behavior experiments, the animals were randomly assigned to treatment groups. The experimenter was blinded to treatment groups. A minimum of N = 5 is needed for an effect size d = 2.1, α = 0.05, and power = 0.8.

For von Frey testing, mice were allowed to acclimate to the behavior room, the experimenter, and the measuring device for 2 weeks. The orofacial region (between the eye and the whisker pad) was stimulated with a series of von Frey fibers with logarithmically incremental stiffness (TouchTest, North Coast Medical Inc.) using the up-and-down methods (61–63). The von Frey fibers were held perpendicular to the testing surface with sufficient force to cause buckling. A positive response was considered if the head was sharply withdrawn and if there was immediate flinching upon removal of fiber. The fibers were presented at least 5 s apart to allow the resolution of previous stimuli.

Thermal latency in the right hind paw and tails was recorded using a Hargreaves’s test; recording of three measurements were averaged for each mouse in each group. Twenty seconds were used as a cutoff to prevent any tissue damage. For anxiety-like behaviors, locomotion, and grooming behaviors, we used the box behavioral spectrometer (64). The box behavioral spectrometer is equipped with a charge-coupled device camera mounted in the center of the ceiling. Mouse movement is assessed by a floor mounted vibration sensor and 32 wall-mounted infrared transmitter and receiver pairs. Mice were individually placed in the center of the behavioral spectrometer, and their behavior including locomotion, grooming, ambulation, track length, and time spent in the inner zone was recorded and analyzed using a computerized video tracking system (Viewer3, BiObserve, DE) for 15 min.

IF staining

Deidentified surgical specimens with available PNI information (retrieved from pathology report) were processed for immunofluorescence (IF) staining under the patient consents and the Institutional Review Board (IRB) protocols approved by MD Anderson Cancer Center. Twelve PNI⁺ and 8 PNI⁻ patient samples (age range, 29 to 74, median age = 58; 58% males and females 42%) were used. Four-micrometer-thick formalin-fixed paraffin-embedded human OSCC sections were deparaffinized, rehydrated, antigen retrieved, and blocked. The slides were stained using the following antibody panels: EGFR (1:50 dilution; catalog no. 4267, Cell Signaling Technology), cytokeratin (dilution: 1:250; catalog no. ab27988, Abcam), and GFAP (dilution 1:600; catalog no. 3670, Cell Signaling Technology). All markers were sequentially applied and stained using their respective fluorophores in the Opal 7 kit (catalog no. NEL797001KT, Akoya Biosciences/PerkinElmer). The slides were scanned using a multispectral microscope (Vectra Polaris imaging system, Akoya Biosciences). Following scanning, image quantification was completed using Oncotopix Discovery (Visiopharm, version 2023.01). A deep learning classifier was trained using specific tissue features from annotated reference images to guide the identification of regions of interest (ROIs). Tissue and ROI boundaries were then delineated using image analysis smoothing. Aberrant signals—such as those caused by dust particles, tissue folds, or air bubbles—were excluded manually from the ROIs. EGFR fluorescence intensity was normalized to the GFAP⁺ and GFAP⁻ nerve areas.

MOC2-inoculated mice were transcardially perfused with 10% formalin. TGs were removed and postfixed overnight with 10% formalin at 4°C, followed with 30% sucrose solution. The TG was sectioned on a cryostat microtome to obtain sections (12 μm). Five sections from each mouse were used for imaging and quantification (N = 3 mice per group). Following washes with Dulbecco’s phosphate-buffered saline (DPBS), slides were incubated in 0.5% tween in DPBS for 30 min and blocked for 1 hour. The slides were incubated overnight with mouse anti-NeuN (1:500; Abcam) and rabbit anti-EGFR (1:500; Thermo Fisher Scientific) primary antibodies. The slides were washed again with PBS before incubation with anti-rabbit Alexa Fluor 647 (1:500; Thermo Fisher Scientific) and anti-mouse Alexa Fluor 488 (1:500; Invitrogen) secondary antibodies for 2 hours. Slides were mounted with antifade mountant with DAPI (Thermo Fisher Scientific). Imaging was performed using a Leica TCS SP8 laser-scanning confocal microscope (Leica microsystems) and analyzed using ImageJ software (National Institutes of Health).

For cultured neuron staining, neurons were fixed with 4% paraformaldehyde for 10 min, followed by incubation with the blocking solution at 37°C for 60 min. Primary antibodies including rabbit anti–phosphorylated-GluN2B (M2442, Millipore), chicken anti-Neun (A104330, Antibodies), rabbit anti-EGFR (Abcam, ab32562), and mouse anti-GluN2B (BioLegend, MMS-5148) were incubated at room temperature for 2 hours. The slides were washed with PBS before incubation with secondary antibodies anti-rabbit Alexa Fluor 647 (Thermo Fisher Scientific, A31573), anti-chicken Alexa Fluor 488 (Invitrogen, PIA32931), and anti-mouse Alexa Fluor 488 (Millipore, AP308PMI) diluted at 1:500 in blocking solution for another 2 hours. Then, slides were mounted with antifade mountant with DAPI. Imaging was performed using a Leica SP8 laser-scanning confocal microscope (Leica Microsystems). The cell counts and colocalization analysis were performed using ImageJ software (National Institutes of Health). Each treatment group included three replicates, with analysis performed on three fields per slide.

Western blot

The TG and brainstem tissues from MOC2 or vehicle-inoculated tongue OSCC mice were homogenized in RIPA lysis buffer (catalog no. 20-188, MilliporeSigma) with a protease inhibitor cocktail (catalog no. P8340, MilliporeSigma). The protein homogenate was then centrifuged at 10,000g for 15 min at 4°C to remove insoluble cell debris. The resulting supernatant was quantified. The same amount of protein for every sample was used for immunoblotting analysis of GluN1 expression as described previously (43). In brief, proteins were separated on 4 to 12% SDS NuPage bis-tris gels (catalog no. NP0336BOX, Thermo Fisher Scientific) and subsequently transferred to polyvinylidene fluoride membranes (catalog no. IPVH00010, MilliporeSigma). The membranes were then incubated with primary antibodies overnight at 4°C. This was followed by incubation with horseradish peroxidase–conjugated secondary antibodies (1:5000; anti-rabbit IgG, catalog no. 7074; anti-mouse IgG, catalog no. 7076, Cell Signaling Technology) for 1 hour at 22°C. The membranes were thoroughly washed after each antibody incubation. Immunoblots were developed using a chemiluminescence kit (catalog no. 34580, Thermo Fisher Scientific) or an enhanced chemiluminescence kit (catalog no. 34095, Thermo Fisher Scientific). The protein bands were visualized using an Odyssey Fc Imager (LI-COR Biosciences, Lincoln, NE) and quantified using ImageJ software. For GluN1 quantification, the intensity of the GluN1 bands was normalized to that of β-tubulin. The primary antibodies included rabbit anti-GluN1 (1:2000; catalog no. G8913, MilliporeSigma) and mouse anti–β-tubulin (1:10,000; catalog no. 32-2600, Thermo Fisher Scientific).

RNAscope in situ hybridization and quantification

mRNA transcripts were localized in the mouse TG using the RNAscope system (Advanced Cell Diagnostics) following the manufacturer’s instructions for fresh-frozen tissue, except for omitting the initial on-slide fixation step. Probe hybridization and detection were performed using the Multiplex Fluorescent Kit v2 according to the manufacturer’s protocol. The probe specific for mouse Egfr (Mm-Egfr, #1579451-C1) was used. Tissues were incubated with TSA Vivid Fluorophore 520 (1:1500; Advanced Cell Diagnostics, catalog no. 323271) for fluorescence detection. To identify total neurons, hybridized slides were blocked and incubated overnight at 4°C with guinea pig anti-NeuN antibody (1:500; EMD Millipore, catalog no. N90). Slides were then washed and incubated with goat anti-guinea pig Alexa Fluor 647 (1:1000) for 1 hour at room temperature. Following additional washes, slides were counterstained with DAPI (1 μg/ml, 5 min) and mounted using ProLong Gold Antifade Mountant (Thermo Fisher Scientific). Sections were imaged using a Leica SP8 confocal microscope equipped with an HCX PL APO 20× objective. The percentage of neurons positive for Egfr hybridization was quantified and normalized to the total number of NeuN-positive cells. RNAscope signals for Egfr mRNA were quantified using a custom macro in ImageJ (Fiji) according to the guidelines provided in the ACD Bio-Techne Technical Note. Neuronal nuclei were identified via NeuN immunofluorescence, and binary masks were generated to segment NeuN-positive cells. Egfr puncta were thresholded and separated using watershed algorithms and counted within NeuN masks using the “Analyze Particles” function. Neurons were classified as Egfr-positive based on a minimum puncta threshold, which was determined by the distribution of Egfr-negative cells. The proportion of Egfr-expressing neurons was calculated by dividing the number of Egfr-positive neurons by the total number of NeuN-positive cells, and results were expressed as the percentage of Egfr⁺ neurons.

Calcium imaging

TG neurons were collected and cultured from naive C57BL/6 mice. One hour before initiating the Ca²⁺ imaging experiments, the media were replaced with serum-free media. For experiments involving AG1478 (10 μM), the cells were preincubated with the drug for 30 or 60 min. The cultured TG neurons were then loaded with 1 μM Fura-2 AM (Thermo Fisher Scientific) for 30 min and washed with Hanks’ balanced salt solution. HSC-3 or MOC2 cells were cultured until 80% confluency in 75-cm tissue culture flasks. Media were replaced with 15-ml serum-free media, and cells were cultured for another 48 hours before supernatant was collected. Ca²⁺ imaging was conducted by applying serum-free media for 1 min, followed by TGFα or HSC-3/MOC2 cell supernatant for 2 min. This was followed by a wash with serum-free media for 1 min and then an application of 50 mM KCl for another 1 min. Fluorescence detection was performed with a Nikon Eclipse TI microscope equipped with a 20× fluor/numerical aperture 0.75 objective lens. Fluorescence images at 340- and 380-nm excitation wavelengths were captured and analyzed using Nikon TI Element Software. Analysis was applied only to the viable cells that were responsive to KCl.

Brainstem slice preparation

The mouse was deeply anesthetized with 3% isoflurane, and the whole brain was rapidly removed and immediately placed in ice-cold sucrose artificial cerebrospinal fluid presaturated with 95% O₂ and 5% CO₂. The artificial cerebrospinal fluid contained 234 mM sucrose, 3.6 mM KCl, 1.2 mM MgCl₂, 2.5 mM CaCl₂, 1.2 mM NaH₂PO₄, 25.0 mM NaHCO₃, and 12.0 mM glucose. The brainstem TNc slices were prepared as described previously with minor modifications. Briefly, brainstem was dissected and glued to a 30-degree angled agar block, and the agar block was put on the stage of a vibratome (Leica, Wetzlar, Germany). Transverse slices (300-μm thick) of the brainstem containing the TNc were cut in ice-cold sucrose artificial cerebrospinal fluid and preincubated in Krebs solution oxygenated with 95% O₂ and 5% CO₂ at 34 °C for at least 1 hour before being transferred to the recording chamber. The Krebs solution contained 117.0 mM NaCl, 3.6 mM KCl, 2.5 mM CaCl₂, 1.2 mM MgCl₂, 1.2 mM NaH₂PO₄, 11.0 mM glucose, and 25.0 mM NaHCO₃. Brainstem slices in a recording chamber were perfused constantly with Krebs solution at 3.0 ml/min at 34°C.

Whole-cell electrophysiological recordings

The SG neurons in the TNc were visualized using an upright fixed-stage microscope with differential interference contrast/infrared illumination. EPSCs were recorded using whole-cell voltage-clamp configuration at the holding potential of −60 mV. The impedance of the glass electrode was 4 to 6 Mohm when the pipette was filled with an internal solution containing 135.0 mM potassium gluconate, 2.0 mM MgCl₂, 0.5 mM CaCl₂, 5.0 mM KCl, 5.0 mM Hepes, 5.0 mM EGTA, 5.0 mM adenosine 5′-triphosphate (ATP)–Mg, 0.5 mM Na-GTP, and 10 mM QX314 (280 to 300 mOsm, adjusted to pH 7.25 with 1.0 M KOH). mEPSCs were recorded in the presence of 0.5 μM tetrodotoxin. To assess glutamate release from primary afferent nerves, EPSCs of SG neurons were electrically evoked via a bipolar stimulating electrode (0.2 ms, 0.5 mA, and 0.1 Hz) placed at the dorsolateral edge of the TNc in brainstem slices (65). Monosynaptic EPSCs of SG neurons were identified on the basis of the constant latency and absence of conduction failure of evoked EPSCs in response to 20-Hz electrical stimulation. To determine the PPR, a pair of electrical stimuli was generated at 50-ms intervals to evoke EPSCs. The PPR was expressed as the ratio of the amplitude of the second synaptic response to the amplitude of the first synaptic response (46, 47).

Postsynaptic NMDAR currents were elicited and recorded via puff application of 100 μM NMDA directly onto the neuron at a 30-degree angle using a positive pressure system (Toohey Company, Fairfield, NJ). The distance was 150 μm between the tip (5 μm in diameter) of the puff pipette and the recorded neuron (46). The internal solution contained 135.0 mM potassium gluconate, 2.0 mM MgCl₂, 0.5 mM CaCl₂, 5.0 mM KCl, 5.0 mM Hepes, 5.0 mM EGTA, 5.0 mM ATP-Mg, 0.5 mM Na-GTP, and 10 mM QX314 (280 to 300 mOsm, adjusted to pH 7.25 with 1.0 M KOH). To remove the magnesium block of NMDARs at negative holding potentials, MgCl₂ used in the extracellular solution was replaced with CaCl₂. All signals were filtered at 1 to 2 kHz and digitized at 10 kHz using an amplifier (MultiClamp 700B; Molecular Devices, San Jose, CA). During the recording, the input resistance was continuously monitored, and the recording was abandoned if the input resistance changed by more than 15%.

Tetrodotoxin and AP5 were purchased from Hello Bio. NMDA was obtained from MilliporeSigma. All drugs were prepared in artificial cerebrospinal fluid before the recordings and delivered via syringe pumps to reach their final concentrations in the recording chamber.

Statistical analysis

We used statistics software Prism 6.0 for all data analysis. Student’s t test or Mann-Whitney U test was used for two-group analysis. One-way analysis of variance (ANOVA) with Tukey’s post hoc analysis was used to compare multiple groups. Two-way ANOVA with one within-subject factor (time) and one between-subject factor (treatment) followed by Tukey’s post hoc tests were used to compare the effect of different treatments over time. P < 0.05 was considered statistically significant. Results are presented as means ± SEM.

Ethics approval and patient consent to participate

All procedures involving animals were approved by the IACUC at New York University (no. PROTO202000028) and University of Texas Houston MD Anderson Cancer Center (no. 00000660-RN04), in accordance with the Guide for the Care and Use of Laboratory Animals published by the US Institute for Laboratory Animal Research (8th edition).

For all human studies, patients were consented before data collection, and studies were conducted under IRB protocols approved by Loma Linda University (# 5200162) and MD Anderson Cancer Center (# LAB08-0848). Only deidentified information was processed and analyzed.