HPV+ Head and Neck Cancer-derived small Extracellular Vesicles communicate with TRPV1+ neurons to mediated Cancer Pain

Introduction

Nearly 57% of patients with head and neck squamous cell carcinoma (HNSCC) reported tumor-related pain at the time of presentation [52; 60]. Pain is often the main symptom motivating HNSCC patients to seek medical intervention [55; 75]. Patients report spontaneous and evoked pain hypersensitivity [15; 72]. Cancer pain is associated with poor survival outcomes [27; 29; 59; 88]. Despite the high prevalence and severity of HNSCC-associated pain, current treatments rarely provide adequate pain relief [13; 15; 50] because the underlying mechanisms promoting cancer pain are not fully defined.

HNSCC results mostly from human papilloma virus infection (HPV+) or smoking/drinking (HPV-). HPV+ and HPV− cancers are molecularly distinct and differ in clinical outcomes [67]. While the incidence of HPV− has remarkably decreased in recent years, [23] the incidence of HPV+ head and neck cancers is increasing [54]. However, previous research on oral cancer pain have been done mainly on models of HPV− cancer and pointed to perineural invasion and cytokine related mechanisms in HNSCC-associated pain [10; 72; 78; 99]. As HPV+ and HPV− cancers represent different diseases, the mechanisms underlying cancer pain are likely to be different as well. Indeed, we showed that cytokines are not involved in HPV+ cancer pain [30].

HNSCC tumor cells produce and secrete small extracellular vesicles (sEVs; size 30–150 nm) [93; 96] which play a role in intercellular communication under normal and pathological settings [3; 58]. Once released, sEVs act on local or distant target cells and modulate signaling pathways [14; 25]. sEVs contribute to various processes in cancer progression such as tumor metastasis and immune modulation [7; 41; 102; 104]. In addition, HNSCC-derived sEVs induce neurite outgrowth in cultured sensory neurons from humans and mice [2; 53] and promote tumor innervation by Transient Receptor Potential V1 expressing (TRPV1+) neurons in preclinical models [53]. These data indicate that sensory neurons can uptake HNSCC-derived sEVs and change their physiology in response. Although the role of HNSCC-derived sEVs in cancer cell-neuron communication is established, the contribution of sEVs to cancer pain remains unknown.

To model HNSCC in wildtype (WT) mice, we injected a previously characterized murine model of HPV+ HNSCC called mEERL cells. These cells were derived from oropharyngeal cells isolated from C57Bl/6 mice which were modified to stably express oncogenes. As such, when injected into immunocompetent mice, tumors grow with characteristics that are faithful to the human disease [53; 94]. We previously showed that implantation of mEERL cells in WT mice induces spontaneous and evoked pain [31]. Interestingly in contrast to previous works on HPV− cancers [98], in this model there is no neuroinflammation in the spinal cord and blocking IL-1 signaling does not alleviate cancer pain. Similar to human HNSCC, mEERL tumors are innervated by TRPV1+ neurons [53]. TRPV1+ neurons are critical for thermal sensitivity and nociception [11]. Therefore, TRPV1+ neurons are an attractive target for treating pain associated with HPV+ HNSCC.

This study explores the communication between HPV+ HNSCC-derived sEVs and TRPV1+ neurons and its contribution to pain.

Experimental procedures

Animal procedures were approved by Michigan State University Institutional Animal Care and Use Committee (IACUC) and were in accordance with National Institutes of Health Guidelines. Experiments were performed on male C57Bl/6J, Trpv1^−/− (JAX# 000664 and #003770, Jackson Laboratory, Bar Harbor, ME) and for calcium imaging; B6.129-Trpv1^tm1(cre)Bbm/J (JAX# 017769) [12] were bred with 6(129S4)-Gt(ROSA)26Sor^{tm1.1(CAG-tdTomato/GCaMP6f)Mdcah}/J (JAX# 031968) [21] mice to obtain genetically encoded calcium indicator GCaMP6f-tdTomato fusion protein expression specifically in TRPV1+ neurons (TRPV1^Cre:Salsa6f mice). These mice were subsequently called Trpv1^Cre:GCaMP6. Mice were bred and housed in the Michigan State University Animal Care Facility prior to the start of behavioral testing. Animals had ad libitum access to food and water and were on a 12 hr regular light/dark cycle. Mice were randomized to treatment groups. Solely male mice were injected with cancer cells.

HNSCC model:

To produce HNSCC tumors, we used a well-established murine model of human papillomavirus (HPV +) induced oropharyngeal squamous cell carcinoma [19; 31; 33; 53; 80]. This model consists of oropharyngeal epithelial cells from C57Bl/6 male mice that stably express HPV16 viral oncogenes, E6 and E7, activated H-Ras and luciferase (mEERL cells). These cells were cultured and injected subcutaneously into the right hindleg of 10–16-week-old male mice to produce tumors. Injection into the leg instead of the oral cavity is justified by the 3Rs guideline in an effort to reduce distress, inability to feed, and easier assessment of tumor growth. Only male mice were injected with tumor cells because the cell line was generated from male mice and cannot be implanted in immunocompetent female mice. Tumor size was monitored using Vernier digital calipers as previously described [31].

mEERL culture:

mEERL cells were cultured in Dulbecco’s modified Eagle’s medium F-12 GlutaMax media (#10565018; Fisher Scientific, Hampton, NH) that contained 10% fetal bovine serum (FBS; #26-140-079; Fisher) and 1X penicillin streptomycin (Pen-Strep; #15070063; Fisher) as previously described [31].

Generation of mEERL Rab27a^−/+ and Rab27b^−/− cells:

to compromise the release of sEVs in cancer cells, Rab27a and Rab27b, two genes encoding proteins essential for exosome release, were knocked out using CRISPR-Cas9 technology in mEERL cells. The gene editing approach and attenuation of sEV release were validated in a previous publication [53].

Behavioral testing was performed by experimenters blinded to experimental conditions. Mechanical hypersensitivity was assessed by stimulation of the hind paw with calibrated von Frey filaments from Stoelting using the Dixon up-down method [20] following 45 min of habituation to the testing boxes as previously described [31; 36]. Mouse grimace scale (MGS) assesses non-evoked pain using facial features [48] such as orbital tightening, cheek bulge, nose bulge, and ear position. Each feature is evaluated on a 3-point scale, and then the sum is analyzed as previously described [31]. MGS has been validated by several laboratories and is sensitive to opioid analgesics [6; 17; 32; 44].

Hotplate:

Mice were individually habituated in the apparatus with the plate set to 30°C for 2 min. The temperature of the hotplate was then increased to 52°C. Mouse was placed on the plate and the latency to jump within 60 sec was recorded. If mice did not jump, the score was 60 sec.

TRPV1+ neuron chemo-ablation:

Subcutaneous injection of resiniferatoxin (RTX, R8756; Sigma- Aldrich, St. Louis, MO) in the flank was performed on 4-week-old mice with escalating doses of 30 μg/kg at day 1, 70 μg/kg at day 2 and 100 μg/kg at day 3. Control mice received a vehicle solution consisting of DMSO with Tween80 in phosphate-buffered saline (PBS). Denervation was evaluated using the hot plate method [42; 66].

Small extracellular vesicle purification:

sEVs were purified following a method previously validated [62]. Briefly, sEVs were harvested from mEERL cells cultured 24 h in media containing exosome-depleted serum (A2720803, Invitrogen, Carlsbad, CA). Culture media was collected, centrifuged and steriflip filtered sterilized. This solution was then placed in VIVASPIN 100kDa ultrafiltration tubes (VS2001/VS2041; Sartorius, Göttingen, Germany) and centrifuged for 50 min at 3000G to concentrate volume down to ~50–150ul. The concentrated solution was run through IZON columns (SP2; IZON Science, Christchurch, New Zealand) and 200 μl fractions were collected. Fractions 6, 7, 8 are enriched for vesicles and have minimal protein contamination. The sEVs were concentrated by adding them to a Amicon 10kDa centrifugal filter tube (UFC801024; Millipore Sigma, St. Louis, MO) and spinning 50 min at 3000G at room temperature. Purified sEVs were quantified using NanoSight NS300 (Malvern Panalytical, Westborough, MA) as described previously [63]. Additionally, the expression of sEV markers is assessed by western blot. Isolated sEVs yield ~10⁶ particles/μl or 0.1–0.2 μg of protein/μl).

Western blotting:

All supplies from Invitrogen, Carlsbad, CA unless otherwise mentioned. Two ug of total protein from sEV samples were mixed with NuPAGE LDS Sample Buffer (NP0007; ThermoFisher, Waltham, MA) under reducing conditions and loaded onto a NuPAGE Novex Bis-Tris 4–12% pre-cast Gel (NP0321BOX; ThermoFisher). An equivalent sample of protein from a mEERL cell tumor, lysed in RIPA buffer, was also run as a control. Proteins were separated under reducing conditions in NuPAGE MES running buffer for 35 min at 200 V constant voltage. After separation the proteins were then transferred to PVDF membrane using NuPAGE transfer buffer with 10% methanol at 30 V constant voltage for 1 hour. After transfer the membrane was blocked for 1 hour in Pierce Clear milk blocking solution and then incubated at 4°C overnight with primary antibody Anti-CD9 antibody [EPR23105–125] (1:2000; ab263019; Abcam, Cambridge, United Kingdom) and Anti-TSG101 antibody [EPR7130(B)] (1:2000; ab125011; Abcam), which was prepared in the Pierce clear milk blocking solution. CD9 and TSG101 are ranked #1 and #11 as the best sEV marker respectively (http://exocarta.org/exosome_markers_new).

The next day the membrane was washed 3 times for 10 min with Tris Buffered Saline with 0.05% Tween20 (TBST) and then incubated for 1 hour at room temperature with a Goat anti-Rabbit IgG antibody with an HRP tag (1:4000; ab205718; Abcam) in TBST. The membrane was then washed 4 times with TBST for 5 min and then once with Tris Buffered Saline (TBS) for 5 min. The membranes were then incubated for 5 min at room temperature with Supersignal West Dura HRP substrate, excess substrate was then drained, and the membrane placed in a clear plastic folder. The membrane was then imaged using a Aplegen Omega Lum G (#8418-10-0005; GMI Inc, Ramsey, MN).

Primary neuron culture:

Trigeminal ganglia (TG) were extracted aseptically from 8-week-old male mice in Hanks’ buffered salt solution (HBSS; #24020117; Fisher) on ice to be cultured using a previously used procedure [38]. The TGs were dissociated enzymatically at 37°C, first with collagenase A (1 mg/ml; #10103578001; Sigma-Aldrich) for 25 minutes and then collagenase D (1 mg/ml; #11088858001; Sigma-Aldrich) that included papain (30 μg/ml; #10108014001; Sigma-Aldrich) for 20 minutes. Afterward, a trypsin inhibitor (1 mg/ml; #10109886001; Sigma-Aldrich) that contained bovine serum albumin (bovine serum albumin, 1 mg/ml; Fisher) was applied, and the ganglia were mixed to allow further dissociation with a polished Pasteur pipette. The tissue was then filtered through 70-μm nylon cell strainer (CLS431751; Sigma-Aldrich) and resuspended in Dulbecco’s modified Eagle’s medium F-12 GlutaMax media (#10-565-018; Fisher Scientific) that contained 10% fetal bovine serum (#26-140-079; Fisher Scientific) and 1X penicillin streptomycin (#15070063; Thermo Scientific, Waltham, MA). The media also contained nerve growth factor (1:1000; #93928-24-6; Sigma-Aldrich). Neurons were cultured for 4 days on Poly-L-Lysine German Glass Coverslips #1, 12 mm (#72292–02; Fisher Scientific) in a 24-well tissue culture plate (#09-761-146; Fisher Scientific) at 37°C with 95% air and 5% CO₂. On the day of the experiment, purified sEVs were diluted into Dulbecco’s modified Eagle’s medium F-12 plus GlutaMax media and added directly onto the neurons and incubated for 24 hours.

Immunocytochemistry:

After sEVs treatment, the cells were washed with PBS and fixed with 10% formalin in PBS for 30 minutes. Cells were blocked with 10% normal goat serum and labeled with anti-TRPV1 (1:1000; PA1–29770; Thermo Scientific), anti-peripherin, mouse monoclonal (1:500; P5117; Sigma) and activating transcription factor 3 (ATF3) (1:1000; ab207434; Abcam) overnight at 4°C. Next, cells were washed and incubated with flurochrome-conjugated secondary antibodies (Alexa Fluor, anti-rabbit 488 (1:1000; A-11008) and anti-mouse 568 (1:1000; A-11004; Thermofisher, Waltham, MA) and counterstained with a DNA stain, 4′,6-diamidino-2-phenylindole (DAPI) (1:50,000; D1306; Fisher) and mounted with Prolong Gold (P36930; Invitrogen, Carlsberg, CA). Staining was visualized using a fluorescent microscope (Nikon Eclipse Ni-U, Minato City, Tokyo, Japan) and a confocal microscope (Nikon A1 CLSM, Minato City, Tokyo, Japan). Images were created using ImageJ (National Institute of Health, Bethesda, MD).

Immunohistochemistry:

Sensory ganglia were removed, placed in 4% formalin overnight, transferred to 30% sucrose for cryoprotection for 24 hrs, then mounted in Optimal Cutting Temperature (OCT; #23-730-571; Fisher Scientific) compound. TG and DRG sections were cut into 20 μm slices using a cryostat and mounted onto positively charged (Superfrost plus; #12-550-15; Fisher) slides for immunohistochemistry. Following three 5-minute washes in 1X PBS, the slides were then put into a permeabilization solution containing 10% normal goat serum (NGS; #S13150H; R&D Systems, Mineapolis, MN) and 0.2% Triton X 100 (#9002-93-1; Sigma) in PBS for 30 minutes. This was followed by another series of 5-minute washes in PBS and 1 hr in a blocking solution containing 10% NGS and 0.01% Na azide (#18-613-272; Fisher) in PBS. Following another PBS wash, the slides were incubated overnight in a primary antibody solution made from the blocking solution. The next day, the slides were washed again in PBS then incubated in a secondary antibody solution also made from the blocking solution for 1 hr. Following a PBS wash, counterstaining with DAPI, a PBS wash, and a wash in deionized H₂O, Prolong Gold mounting media was used to mount coverslips. The primary antibodies used were anti-TRPV1 (1:500; ab203103; Abcam), ATF3 (1:1000; ab207434, Abcam), anti-peripherin, and mouse monoclonal (1:500; P5117, Sigma). Secondary antibodies used were Alexa Fluor anti-rabbit 488, anti-rabbit 568 and anti-mouse 568 (1:1000; Thermofisher). Staining was visualized using a fluorescent microscope (Nikon Eclipse Ni-U) and images were generated using ImageJ.

Nascent protein synthesis “Click Chemistry”:

Protein synthesis in vitro was measured using Click iT chemistry, as previously established [56]. TG neurons were isolated, cultured, and treated with sEVs as described above. Following treatment, cells were incubated in methionine-free DMEM (21013024; Gibco) for 30 minutes followed by a 2-hr incubation in 50μM Click iT AHA (L Azidohomoalanine) (C10102; Invitrogen) in methionine-free DMEM. 50 ng/mL nerve growth factor (NGF) treatment was used as a positive control as it induces an increase in protein synthesis [56]. The cells were then washed with PBS and fixed with cold methanol. This was followed by incubation in 3% bovine serum albumin (BSA; BP9700100; Fisher Scientific) in PBS for 15 min. Cells that incorporated AHA were labeled with Alexa 488–alkyne conjugate (1:200; A10267; Invitrogen) by incubating fixed cells with the conjugate for 30 min. The cells were then washed once with 3% BSA in PBS, followed by another wash with PBS. Cells were then blocked with 10% normal goat serum and labelled with anti-peripherin mouse monoclonal as previously described.

sEVs labeling:

Purified sEVs were labeled using the ExoGlow-Protein EV Labeling Kit (EXOGP400A-1, System Biosciences, Palo Alto, CA). sEVs were suspended in 500 μL PBS and 1 μL of the 500X labeling dye was added. This solution was incubated at 37°C with shaking for 20 minutes. Following this incubation, 167 μL ExoQuick-TC (EXOTC10A-1, System Biosciences, Palo Alto, CA) was added to the solution followed by an overnight incubation at 4°C. The following day, the sEVs solution was centrifuged for 10 min (10,000 rpm), the supernatant was aspirated, and the remaining pellet was resuspended in PBS prior to plantar injection.

Calcium Imaging:

Mice Trpv1^Cre:GCaMP6 were used for calcium imaging. TG neurons from these mice were cultured using the previously detailed methods [38]. Neurons were treated for 24 hrs with purified sEVs. Coverslips were placed in a custom-designed imaging chamber and 37°C HEPES-buffered physiological salt solution was superfused across. This solution contains NaCl (134 mM), HEPES (10mM), KCl (6 mM), glucose (7 mM), MgCl₂ (1.2 mM), and CaCl₂ (2 mM). Images were recorded using an Andor Zyla 4.2 PLUS scMOS digital camera (Andor, Oxford Instruments, UK) mounted on a variable-zoom Nikon SMZ18 stereomicroscope with 1X SHR Plan Apo objective (NA=0.15). Calcium events were recorded at 10 Hz over the course of 2.5 minutes using μManager software (http://www.micro-manager.org).

Digital image analysis:

Cell Profiler 4.0 [82]. https://doi.org/10.1186/s12859-021-04344-9) and GraphPad Prism 9 (GraphPad) were used to analyze calcium events and expression of ATF3 and AHA in cultured TG. Briefly, for calcium oscillations, cells were identified using an image of the tdTomato expression to identify cells that were TRPV1 positive. Then, the integrated intensity of GCaMP6 fluorescence was determined for each TRPV1+ cell in every frame of the recording. This intensity data was plotted for each cell and the data for cells with a stable baseline and at least 1 calcium oscillation were transferred to GraphPad for analysis. The data were corrected by dividing all values by the baseline and then the area under the curve was calculated for each cell.

ATF3+ cells were analyzed by first separating the 3 channels in the image and then identifying cells using either the Peripherin or TRPV1 channels. Then the % of peripherin+ or TRPV1+ cells that were also ATF3+ using the relateobjects command.

AHA intensity was examined using Cell Profiler 4.0. Briefly, the image was first separated into the 3 channels and then all peripherin positive cells were identified using the peripherin channel. Then only cells that were peripherin positive were then measured in the AHA channel for the integrated intensity of AHA fluorescence.

Drug administration:

Narciclasine (NCLS, sc-361271; Santa Cruz, Dallas, TX), was dissolved in Dimethyl sulfoxide (D128–1; Fisher Scientific) then diluted in 45% w/v 2-Hydroxypropyl-β-cyclodextrin (sc-203461A; Santa Cruz) for i.p. injection at a dose that produces an analgesic effect in preclinical models of persistent pain [35] [37]. Rapamycin (1 mg/kg; #37094; Sigma-Aldrich) was dissolved in DMSO then diluted to 10% in saline. GW4869 (1.25 mg/kg i.p.; D1692; Sigma-Aldrich) is a neutral sphingomyelinase inhibitor used for blocking the release of sEVs from the multivesicular bodies [45; 49; 91]. GW4869 was dissolved to 5 mg/mL in DMSO then diluted to 1.25 mg/kg in PBS. 50 mg QX-314 bromide (#1014; Tocris Bioscience, Bristol, United Kingdom) was dissolved in 1 ml solution (50% saline + 50% pure water) to produce a 5% solution. This solution was diluted in saline to appropriate concentration and injected intraplantar. Ketoprofen (15 mg/kg; K1751; Sigma-Aldrich) was dissolved in 100% ethanol (5 mg in 0.5 mL) then diluted in saline for i.p. injections.

Use of publicly available RNA-seq data and IPA:

Human DRG RNA-sequencing data from naïve DRGs were directly downloaded from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7305999/ [92]. Human DRG RNA sequencing data from DRG exposed to human HNSCC-derived sEVs were downloaded from https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE134220 and correspond to GSM3939063, GSM3939065, and GSM3939067 [2]. To normalize the datasets, we matched several hundred genes from the two RNA-seq datasets, and we used the matched housekeeping genes for normalization. We scaled the expression levels of all RNA species in each individual sample. Therefore, the median of ratios between sample and consensus was 1. Genes with missing expression data in one dataset were omitted. To identify gene-enrichment pathways, the data were analyzed through QIAGEN Ingenuity Pathway Analysis [46]. QIAGEN Ingenuity Pathway Analysis library identified canonical pathways that were most significant to the data set. The significance of the association between the data set and the canonical pathway was measured in two ways: 1) A ratio of the number of molecules from the data set that map to the pathway divided by the total number of molecules that map to the canonical pathway is displayed; and 2) A right-tailed Fisher’s Exact Test was used to calculate a p-value determining the probability that the association between the genes in the dataset and the canonical pathway is explained by chance alone.

Statistical Analysis:

Data are shown as mean ± standard error of the mean and the number of animals or samples used in each analysis are given in figure legends. GraphPad Prism 9 was used to analyze data for statistical tests, which are displayed in figure legends. Repeated measures two-way ANOVAs, One-way ANOVA, unpaired t-test and non-parametric Mann-Whitney tests were used based on experimental design. Statistical significances are indicated as follow * = p < 0.05, ** = p < 0.01, and *** = p < 0.001.

Results

Blocking cancer-derived sEVs attenuates pain

Implantation of mEERL cells into WT mice induced evoked and spontaneous pain after 4 days [31], while tumors are not detectable before 18 days (Supplemental figure 1). We assessed the effects of blocking cancer-derived sEVs on cancer pain. Administration of the sEV release inhibitor GW4869 (1.25 mg/kg) attenuates pain hypersensitivity and facial grimacing in tumor-bearing mice (Figure 1A–C). Blockage of sEV release is confirmed by the reduction of circulating sEVs (Figure 1D). Additionally, implantation of mEERL Rab27a^+/− and Rab27b^−/− cells, cancer cells that release a limited amount of sEVs significantly delayed the development of pain hypersensitivity and spontaneous pain (Figure 1E–G) compared to parental mEERL cells. However, tumor sizes were similar (Figure 1H).

Figure 1. Inhibition of sEV release alleviates cancer pain.

(A) Daily GW4869 treatment (beginning on day 3 post tumor implantation, yellow bar) alleviated pain hypersensitivity in tumor-bearing mice (n=5/group; 2-way ANOVA drug effect F (1,8) = 20.2, p<0.002). von Frey was assessed 24 h after GW4869 injection. (B) Representative images of MGS test. (C) GW4869 reduced MGS on day 15 (n=5/group; unpaired t-test 2-tail t=5.4, df=8, p=0.0006). (D) GW4869 reduced the number of circulating particles (n=4/group; unpaired t-test 2-tail t=3.3, df=6, p=0.0162) measured by Nanosight. (E) Genetic deletion of Rab27a and Rab27b in mEERL cells attenuated pain hypersensitivity in tumor-bearing mice (n=7/group; 2-way ANOVA cell effect F (2,18) = 47.2, p<0.0001). (F) Representative images of MGS test. (G) Genetic deletion of Rab27a and Rab27b in mEERL cells attenuated MGS (n=7/group; 1-way ANOVA F (2,17) = 32.9, p<0.0001). (H) Tumor volume, t-test df = 10, p=0.75.

Cancer-derived sEVs are sufficient to induce pain hypersensitivity

Since compromising the release of mEERL-derived sEVs attenuates cancer pain, next we determined whether injection of purified sEVs is sufficient to produce pain hypersensitivity in WT mice. The size and quality of purified sEVs were assessed by nanosight and western blot; CD9 and TSG101 are both well-established markers for sEVs (Figure 2A,B) [18; 22; 95]. First, to confirm that isolated sEV injected into the hind paw are taken up by DRG neurons, sEVs were stained blue prior to injection. DRGs harvested 1, 3 and 5 hrs post-injection showed an uptake of the blue dye confirming that sEVs and/or their contents trafficked from the paw to the DRG (Figure 2C–E). The fact that the dye reached the cell bodies of the DRG in 3 h does not mean that all content of the sEVs travel similarly. Intraplantar injections of sEVs also induced de novo expression of activating transcription factor 3 (ATF3) a stress-induced protein and neuronal injury marker associated with neuropathic pain [64; 65; 85] in the lumbar DRGs. ATF3 is similarly upregulated in sensory ganglion of tumor-bearing mice (Figure 2F). Additionally, injection of isolated sEVs into the paw caused pain hypersensitivity up to 24 hours in a dose-dependent manner (Figure 3A and Supplemental figure 2). Pain hypersensitivity in response to injection of isolated sEVs is not reduced by injection of ketoprofen (clinically used non-steroid anti-inflammatory drug) at a dose classically used in preclinical models [4] (Figure 3B). To determine whether TRPV1 is activated by purified sEVs, we injected a positively charged lidocaine derived QX-314. QX-314 inhibits voltage-gated sodium channel and neuronal activity only in cells with TRPV1 channel opened [8; 83]. Treatment with QX-314 alleviates pain hypersensitivity induced by injection of sEVs into the paw, indicating that TRPV1 channels are opened in presence of cancer-derived sEVs. (Figure 3C). Consistently, Trpv1^−/− mice did not develop pain hypersensitivity in response to isolated sEVs (Figure 3D).

Figure 2. Isolated mEERL-derived sEVs reached the sensory neurons.

(A) Representative histogram plot of purified sEVs by the Nanosight. (B) Representative image of the protein level of CD9 and TSG101 in protein extract from purified sEVs and tumor. Two μg of protein, except (*) denotes 30 μg of protein from tumor samples. (C) Diagram of sEV isolation from mEERL cell culture and intraplantar injection. (D) Lumbar dorsal root ganglion sections of PBS- (Ctrl) and Exoglow stained sEV-injected mice. The blue indicates the presence of Exoglow dye in neuronal cell body. Neuron cell bodies are marked with NeuN in red. (E) Quantification of percentage of neurons (red) with blue dye (Exoglow), One-way ANOVA with Tukey’s correction for multiple test F(1.136, 2.273) = 109.0, p=0.0057. (F) Dorsal root ganglion sections were stained with anti-Peripherin (peripheral neurons) and anti-ATF3 in PBS-, isolated sEVs-, and mEERL cell-treated mice. Quantification of number of neuron (peripherin, red) positive for ATF3 (green) One-way ANOVA with Tukey’s correction for multiple test F(2,7) = 9.35, p=0.011.

Figure 3. Isolated mEERL-derived sEVs induced pain hypersensitivity in naïve mice.

(A) Mechanical sensitivity monitored after intraplantar injection of sEVs (0; 5×10⁵; 10⁶, and 10⁷ particles). Data are compared to saline. One-way ANOVA 30 min F (3,35) = 16.6, p<0.0001; 60 min F (3,35) = 27. 98, p<0.0001; 24h F (3,36) = 6.4, p=0.0013. (B) No analgesic effect of ketoprofen (15 mg/kg i.p.) in sEV-injected mice (1M particles, 1 h, n=7/group). (C) QX-314 relieved pain hypersensitivity induced by injection of mEERL-derived sEVs (10⁶ particles, 1 h) (n=8/group, One-way ANOVA F (4,35) = 32.9, p<0.0001. (D) Isolated sEVs do not induce pain hypersensitivity in mice genetically lacking TRPV1 (n=7–8/group), Two-way ANOVA with Bonferonni’s correction, sEVs × genotype interaction F(1, 26) = 39.4, p<0.0001.

Cancer-derived sEVs increase calcium influx in TRPV1+ neurons

Because sEVs induce pain hypersensitivity that is alleviated by blocking TRPV1 signaling, we assessed whether purified cancer-derived sEVs directly impact TRPV1+ neurons. We performed calcium imaging on cultured TG neurons from Trpv1^Cre:GCaMP6 mice exposed to isolated sEVs. Cultured TG neurons showed spontaneous calcium influx as indicated by “fluorescence flashing”. Peaks indicating calcium influx are more frequent and larger in mEERL-derived-sEV-treated cultures indicating that cultured TG neuron incubated with cancer-derived sEVs showed enhanced intracellular calcium events following 24-hr sEVs treatment compared to control TG neurons treated with PBS (Figure 4 and supplemental videos). This suggests that sEVs increase the excitability of TRPV1+ neurons.

Figure 4. Cancer-derived sEVs induces calcium influx in cultured TRPV1+ neurons.

(A) Calcium events were recorded in cultured TG neurons from Trpv1^Cre:GCaMP6 mice following 24 h incubation with vehicle or purified sEVs. Cancer-derived sEVs induced an increase in cellular calcium events in TRPV1+ neurons. Each color line represents a different TRPV1+ cell. (B) Quantification of calcium events: area under the curve of graphs (n=14–16 cells/groups, Mann-Whitney p=0.013).

Ablation of TRPV1+ neurons prevent cancer pain

In order to assess the role of TRPV1+ neurons on cancer pain, naïve mice were treated with RTX to ablate TRPV1-expressing neurons. The ablation was confirmed by the absence of reaction in the hot plate test and a drastic reduction of Trpv1 expression in sensory ganglia (Figure 5A,B). Four weeks after RTX treatment, mice were injected with mEERL cells into the right hindleg. Von Frey (Figure 5C) and mouse grimace scale testing (Figure 5D,E) indicated that RTX treatment prevents evoked and spontaneous cancer pain but did not affect baseline mechanical sensitivity. TRPV1+ neuron ablation did not alter tumor volume (Figure 5F).

Figure 5. Chemo-ablation of TRPV+ neurons prevents cancer pain.

(A) RTX-treated mice are unresponsive to the hotplate (veh n=6, RTX n= 13; unpaired t-test t=18.3, df=17, P<0.0001). (B) RTX treatment reduced the expression of Trpv1 in sensory ganglia (n=4 Veh and n=8 RTX, unpaired t-test t=11.45 and df =10, p<0.0001). (C) Chemo-ablation of TRPV1+ neurons blocks pain hypersensitivity in tumor-bearing mice (n=6–8/group, 2-way ANOVA treatment effect F (2,17) = 112.5, p<0.0001). (D) Chemo-ablation of TRPV1+ neurons alleviates facial grimacing in tumor-bearing mice (n=6/group, one-way ANOVA F (2,15) = 63.4, p<0.0001). (E) Representative image of mouse facial grimacing. (F) Tumor volume, One-way ANOVA p=0.02, multiple comparison: Veh + mEERL vs. RTX + mEERL p=0.84.

Cancer-derived sEVs induce ATF3 expression in TG neurons

ATF3 is a transcription factor associated neuronal reprogramming and TRPV1 activation [9; 69].To determine the impact of cancer-derived sEVs on ATF3 expression, we exposed mouse TG neurons to purified sEVs for 24 hours and stained for ATF3. Purified sEVs increased ATF3 expression in TRPV1+ cells in a dose-dependent manner (Figure 6). Furthermore, staining with anti-Peripherin antibody indicated that all Peripherin+ cells expressing ATF3 are TRPV1+ as well (Supplemental Figure 3). Interestingly, purified cancer-derived sEVs triggered de novo expression of ATF3 exclusively in TRPV1+ neurons, which may lead to transcriptional changes in nociceptors.

Figure 6. mEERL-derived sEVs induce ATF3 expression in sensory neurons.

(A) Representative images of cultured TG following 24 h exposure to purified sEVs (1 μL equivalent to ~0.2 μg of protein) and stained for ATF3 and co-stained for TRPV1. (B) sEVs induced an increased numbers of ATF3+/TRPV1+ cells (n=6/group, One-way ANOVA F (2,15) = 4.13, p=0.037).

Cancer-derived sEVs activate the translation initiation pathways to induce pain hypersensitivity

To further investigate the potential mechanism of nociception induced by HNSCC-derived sEVs and identify clinically relevant targets, we took advantage of publicly available human RNA-sequencing data. RNA sequencing performed from unstimulated cultured human DRGs (5 different cultures) [92] and exposed to human HNSCC-derived sEVs (3 cultures) [2]. After removing genes that were not present in both datasets or had missing expression data, we generated a list of 9369 genes. We selected the genes with 0.2>log2FC>3 and a p-value<0.001 to obtain 1716 genes (Supplementary table). To infer the functional role of these differentially expressed genes, we performed gene-enrichment pathway analysis. Ingenuity Pathway analysis (IPA) data indicate that sEVs induce several canonical pathways linked with the initiation of protein translation such eukaryotic initiation factor (eIF) 2 and 4 signaling, mammalian target of rapamycin (mTOR) signaling, and p706K signaling. (Table 1). All of these pathways, involved in the regulation of the nascent translation, contribute to nociception [68; 79]. To test whether cancer-derived sEVs also trigger the translation of nascent protein in our mouse model, we used the methionine analog AHA to label newly translated proteins in cultured TG neurons and nerve growth factor (NGF) as positive control [56]. Incubation with sEVs increased AHA fluorescence in peripherin+ cells (TG neurons) indicating an enhancement of translation of nascent proteins (Figure 7A,B), aligning our preclinical model with human sequencing data: cancer-derived sEVs induced translation of nascent protein in sensory neurons. These pathways are blocked by AMP-activated protein kinase (AMPK) activation and mTOR inhibition [26; 37; 57] (Supplementary Figure 4). Consistently, in tumor bearing mice, inhibition of mTOR directly through rapamycin attenuates pain hypersensitivity (Figure 7C). Moreover, administration of AMPK activator narciclasine (NCLS) [40; 101] prevents evoked and spontaneous cancer pain (Figure 7D–F). NCLS reduced tumor volume as previously shown (Figure 7G). Given that the preventive effect of NCLS is observed before the tumor is detectable, it is unlikely that reduced tumor volume mediates the prevention of cancer pain. Moreover, NCLS also prevented pain hypersensitivity induced by injection of purified mEERL-derived sEVs (Figure 7H).

Table 1.

Top 5 Ingenuity Canonical Pathways

Pathways	−log(p-value)	Ratio	Overlap
EIF2 Signaling	2.09E+01	3.51E-01	35.1 %	71/202
mTOR Signaling	1.06E+01	2.72E-01	27.2 %	53/195
Regulation of eIF4 and p70S6K Signaling	9.29E+00	2.74E-01	27.4 %	45/164
Coronavirus Pathogenesis Pathway	6.81E+00	2.39E-01	23.9 %	42/176
CXCR4 Signaling	5.80E+00	2.33E-01	23.3 %	37/159

Figure 7. Blocking the translation initiation pathways alleviates cancer pain.

(A) Representative confocal images of intensity correlation analysis of fluorescently labeled AHA click iT chemistry with neuronal marker (peripherin+ cells). (B) sEVs induce translation of nascent proteins in cultured TG neurons at 2 hrs. (n=20–23 cells/group, One-way ANOVA F (3,85) = 6.873, 2 hr sEV p = 0.0074). NGF was used as positive control. (C) Rapamycin treatment attenuated cancer-induced mechanical hypersensitivity (n=7/group, 2-way ANOVA rapamycin effect F(1,12) = 33.1, p<0.0001). (D) Narciclasine treatment prevents cancer-induced mechanical hypersensitivity (n=4–5/group, 2-way ANOVA NCLS effect F (1,63) = 385.4, p<0.0001). (E) Representative images of mouse facial grimacing. (F) Narciclasine treatment alleviates facial grimacing in tumor-bearing mice (n=4/group, Mann-Whitney p=0.03). (G) Tumor volume at termination (day 28), unpaired t-test df=6, p=0.041). (H) Narciclasine treatment prevents sEVs-induced mechanical hypersensitivity (n=6/group, 2-way ANOVA NCLS × sEVs interaction effect F (1,10) = 8.75, p<0.014).

Discussion

One of the novel key findings of this study is that TRPV1+ neurons are directly stimulated by cancer-derived sEVs and these sEVs are necessary for HPV+ cancer pain. Cancer-derived sEVs induced translation of nascent proteins in nociceptors to mediate cancer pain (Figure 8). Moreover, our findings indicate a major role of TRPV1+ neurons and the TRPV1 channel, consistently with previous reports [28; 74]. Our study is one of the first to decipher the mechanisms of cancer pain in HPV+ models.

Figure 8: Graphical abstract.

HPV+ cancer cells release small extracellular vesicles (sEVs) that trigger an increase of calcium, ATF3 upregulation, a modification of the transcriptome, activation of eIF pathways/nascent translation of protein in TRPV1+ neurons to mediate cancer pain.

Communication between tumors and TRPV1+ neurons impacts tumor progression [71] and tumor immunology [5; 76; 87]. We show that sEVs released by cancer cells are a key mediator of communication to TRPV1+ neurons in line with previous reports [2; 53; 90]. Blocking cancer-derived sEV release alleviates cancer pain in tumor-bearing mice. This result coupled with the fact that injection of isolated sEVs trigger pain hypersensitivity indicates that sEVs play a critical role in HPV+ cancer pain. Additionally, administration of ketoprofen indicates that potential inflammation in response to sEVs is not driving the pain. However, it remains unclear mechanistically how sEVs mediate nociception. sEVs may either release their cargo into neurons or activate receptors on nociceptors [47; 70; 86].

ATF3 is a transcriptional factor de novo upregulated after neuronal injury and associated with transcriptional reprogramming, regeneration, and TRPV1 activation [9; 69; 103]. This observation is in line with the ability for HNSCC-derived sEVs to induce neurite outgrowths [2; 51; 53] and suggests that cancer pain has a neuropathic component [98]. The upregulation of ATF3 in response to sEVs supports the idea that sEVs drastically impact neuronal gene expression to switch to a transcriptome facilitating axon elongation and neuronal sensitization. Upregulation of ATF3 in TG neurons has also been reported in HPV− oral cancers [34].

Observation from human RNA-seq data and our data point out that sEVs induce eIF pathway/translation. ATF3 also plays a key role in axonal translation [39]. Nociception resulting from enhanced nascent translation is well-established [43; 56] indicating that translation inhibition is a promising target for cancer pain. Activation of mTOR pathways may also facilitate tumor innervation [53; 90; 97].

Our data do not demonstrate a direct activation of TRPV1 channels by sEVs. One might suggest that sEVs surface proteases [73] activate Protease activated receptor 2 (PAR₂) which mediates nociception [81; 89], eIF pathways [84], and increases ATF3 expression [24]. Then PAR₂ activates TRPV1 [1] as shown in HPV− oral cancer pain [47; 77; 86].

Unlike previous reports ablation of Rab27 or TRPV1+ neurons did not alter tumor volumes (ref 53; 71). Our experiments were performed at an early stage when tumors are barely detectable, we study early stages to understand patients’ pain experiences prior to cancer diagnosis and treatment.

Our study provides several promising potential targets for treating HNSCC pain. In addition to RTX which is already in clinical trials [61], inhibiting TRPV1+ neuron activity (QX-314 and TRPV1 antagonists) or interfering with sEV release from cancer cells are valuable therapeutic strategies. Blocking the eIF pathways/translation might be the most promising because AMPK activators like NCLS inhibits metastasis and tumor growth as well [16; 100] and AMPK activators are well known for their analgesic effects and some like metformin are already FDA-approved [43].

Taken together we define the role of cancer-sEVs in cancer pain and identify several druggable targets.

Limitations of the study

To limit harm to the animal’s wellbeing, cancer cells were injected into the leg. Although the leg and oropharyngeal cavity have anatomical differences that may affect outcomes, the similarity between in vitro and in vivo data indicates that the findings are likely relevant to the oral environment. Moreover, HNCSS is 3–4-time more prevalent in men than women regardless of alcohol and smoking consumption but addressing sex as a biological variable is extremely challenging in this study because: the present cancer cells (mEERL) have been generated by clonal selection from C57 male mice, therefore these cells cannot be injected in C57 immunocompetent female mice. Creating a female-compatible tumor cell line would require clonal selection as well, which may result in different mutations compared to the mEERL cells. Comparing data between males and females would involve analyzing two different cell lines that differ not only in sex but also in other genetic characteristics. Therefore, it would be difficult to separate the effects of sex from the effects of specific cell clones.

Supplementary Material

Supplementary Materials: movies

Supplementary video. Examples of fluorescence oscillations in cultured TG from Trpv1^Cre:GCaMP6 mice.

Supplementary Materials: figures, tables

Supplementary figure 1. Evolution of tumor growth over time. Male mice are injected with 200,000 mEERL cells into the hind leg. Tumor sizes are measured with digital calipers.

Supplementary figure 2. Analysis of pain hypersensitivity induced by isolated mEERL-derived sEVs. (A) Mechanical sensitivity monitored after intraplantar injection of sEVs (0; 5×10⁵; 10⁶, and 10⁷ particles). Same animals as Fig. 3. Stars indicate statistical difference from time-matched saline treated mice. Two-way ANOVA mixed-effect model (because the inequal number of animals in each group) followed by Bonferroni’s correction for multiple comparison: effect of the number of particles F(3, 141) = 38.1, p<0.0001; effect of time F(2341, 110) = 30.1, p<0.0001; interaction time × sEVs F(9, 141) = 5.82, p<0.0001. (B) Table of Bonferroni’s multiple comparisons test.

Supplementary figure 3. Cancer-derived sEVs induced ATF3 in Peripherin+/TRPV1+ cells. Cultured TG neurons from mice stained for ATF3/peripherin/TRPV1 following 24 h exposure to purified sEVs from mEERL cells.

Supplementary figure 4. AMPK/mTOR/eIF signaling pathways promoting protein translation. Rapamycin is an mTOR inhibitor and narciclasine is an AMPK activator.

Inclusion and Diversity

One or more of the authors of this paper self-identifies as an underrepresented ethnic minority in science. One or more of the authors of this paper self-identifies as a member of the LGBTQ+ community. While citing references scientifically relevant for this work, we also actively worked to promote gender balance in our reference list.

Data availability

Data will be available upon reasonable request.