MMP1 and PRSS23 induce PAR₂ biased agonism in painful oral cancers

Abstract

Protease-activated receptor 2 (PAR₂) mediates oral cancer pain. Patients with metastatic (N + ) cancers report greater pain. PAR₂ is activated by N-terminal proteolytic cleavage. Here we show that proteases encoded by genes overexpressed in N+ cancers from patients with pain (matrix metallopeptidase 1, MMP1 and serine protease 23, PRSS23) elicit protease-specific receptor redistribution (trafficking) and signaling that differs from that promoted by proteases encoded by genes not differentially expressed (transmembrane serine protease matriptase, ST14 and cathepsin S, CTSS). Mixtures of the proteases prepared to model the oral cancer microenvironment revealed that ST14-mediated PAR₂ activation predominated at low protease concentrations. At high concentrations, MMP1 and PRSS23 prevailed over the greater potency of ST14. We propose that PAR₂ activation in oral N+ cancers from patients with pain is driven by high levels of MMP1 and PRSS23. Our study informs design of signaling and location-specific antagonists to provide more efficacious analgesia.

Studies in vitro of protease mixtures modeling the oral cancer microenvironment elicit trafficking and signaling of PAR2, an oral cancer pain mediator, that is dictated by the relative potencies and efficacies of the proteases to cleave the receptor

Introduction

Oral cancer patients experience pain at the site of the cancer^1–5. Pain is often the symptom that motivates oral cancer patients to seek treatment^6,7. Seventy percent of patients self-report pain⁷, but the severity and nature of oral cancer pain varies among patients. Some patients report sensitivity to touch and interference with function. Others experience spontaneous pain and functional pain (i.e., pain when eating, talking and drinking), as well as sensitivity to touch and interference with function⁷. Patients with metastatic (N + ) cancers report experiencing greater pain than patients with non-metastatic (N0) cancers⁷. Clinical experience supports the hypothesis that oral cancer pain originates in the tumor microenvironment (TME). Surgical resection of the cancer relieves pain^5,7.

PAR₂, encoded by F2RL1, a member of the family of G-protein coupled receptors (GPCRs), mediates oral cancer pain^8–15. PAR₂ is activated by protease cleavage that unmasks an N-terminal tethered ligand¹⁶. Trypsin was identified as the first PAR₂-activating protease¹⁷. Trypsin and other proteases that cleave PAR₂ at the same site (³⁶R ↓ S³⁷) lead to “canonical” PAR₂ trafficking and signaling. Proteases can cleave PAR₂ at different (non-canonical) sites and expose distinct tethered ligands that promote protease-specific PAR₂ trafficking and signaling (biased agonism)¹⁶ and/or they can disarm PAR₂ by removing the canonical cleavage site¹⁸.

The oral cancer TME is enriched with dysregulated proteases that can both promote cancer¹⁹ and activate PAR₂⁷ at canonical and non-canonical sites. To date, one canonical protease, transmembrane serine protease 2 (TMPRSS2) and two non-canonical proteases, CTSS and legumain (LGMN), have been reported to produce oral cancer pain via PAR₂^9,20,21. None of these proteases, however, are overexpressed in N+ cancers from patients reporting high levels of pain. In addition, there are conflicting reports regarding whether TMPRSS2 is overexpressed or downregulated in oral cancer^20,22. Thus, it is unclear to what extent PAR₂-mediated oral cancer pain is driven by biased agonism.

To better understand how PAR₂ trafficking and signaling might be altered in N+ oral cancers from patients reporting high pain compared to patients with N0 oral cancers^7,23, we studied four proteases: two encoded by genes overexpressed in N+ oral cancers from patients reporting high levels of pain, MMP1 and PRSS23⁷, together with ST14, a canonical protease with oncogenic roles in oral cancer^24–27, and CTSS, a non-canonical protease that mediates oral cancer pain²¹. PRSS23 is a little studied serine protease and a member of the S1A family of serine proteases²⁸. Since PAR₂ is cleaved and activated by a number of serine proteases²⁹, PRSS23 is predicted to cleave PAR₂. While trafficking and signaling of PAR₂ promoted by CTSS have been reported previously^21,30,31, detailed studies of MMP1 and PRSS23 are incomplete or lacking. We show that proteases encoded by genes overexpressed in the TME of N+ cancers from patients reporting pain promote PAR₂ trafficking and signaling in vitro that differs from that of the previously studied proteases in relation to oral cancer pain. Additionally, we demonstrate by studying protease mixtures as might exist in the TME, that PAR₂ trafficking and signaling is dictated by the concentrations of proteases and their relative potencies (inverse of EC₅₀) and efficacies to cleave the receptor^18,32,33. Greater knowledge of PAR₂ trafficking and signaling promoted by a mixture of proteases, as expected in a patient’s oral cancer, could lead to improved PAR₂-targeted therapeutic approaches to alleviate oral cancer pain.

Results

The oral cancer microenvironment is enriched in proteases

Analysis of our published bulk RNA sequencing data (GSE156178) of human oral cancers, including N0, N+ and normal tissues⁷, revealed altered transcription of proteases in cancers compared to normal tissue (Fig. 1A). Transcription of genes encoding proteases cleaving PAR₂ at the canonical trypsin site included kallikreins (KLK5, KLK6, KLK14), ST14 and FURIN. Transcription of genes encoding proteases that cleave PAR₂ at non-canonical sites included MMP1, MMP13, LGMN, and CTSS. FURIN, MMP1, and MMP13 gene transcripts were overexpressed in cancers compared to clinically normal oral tissue, while MMP1 was also overexpressed in oral cancers from patients reporting high pain compared to N0 cancers and normal tissue⁷. Overexpression of CTSS relative to normal tissue has been reported in cohorts of oropharyngeal and oral cancers, as well as in a cohort of mostly male oral cavity cancer patients, the majority of whom were users of betel nut (areca)³⁴. The etiology of oropharyngeal³⁵ and areca associated oral cancers³⁶, however, differs from oral cancers in the western hemisphere. Transcript levels of TMPRSS2 were reduced in N+ and N0 cancers compared to normal tissue (Fig. 1A, Supplementary Table S1), consistent with previous reports of downregulated expression of TMPRSS2 in head and neck, oral and lung cancers^22,37, although elevated TMPRSS2 immunoreactivity in human oral cancers compared to normal tissue has been reported²⁰.

Fig. 1. Transcription levels of selected proteases that cleave PAR₂.

A Shown is the N-terminus of PAR₂ from amino acids 28 to 70 together with genes encoding proteases that cleave at the canonical (red arrow) and non-canonical sites (gray arrows). Transcript levels of genes encoding proteases were obtained from bulk RNA sequencing of human N0 (n = 12) and N+ (n = 7) oral cancers and clinically normal oral tissues (n = 5). Genes considered expressed (base mean >30 counts) are shown in bold black font. Genes significantly overexpressed or with reduced expression (adjusted P < 0.05) in cancers vs. normal oral tissue are shown in red and blue bold font, respectively. Non-expressed genes are shown in gray. B Hierarchical clustering of human cancers (n = 19) and clinically normal oral tissues (n = 5) according to transcript levels of proteases. Normalized counts were ln(N + 1) transformed. Rows and columns were clustered by Euclidean distance and Ward linkage. No scaling was applied to rows. The patients are shown in columns and proteases in rows.

We performed hierarchical clustering of patients, according to expression of the abovementioned PAR₂-cleaving protease gene transcripts and included PRSS23, also elevated in painful N+ cancers. The analysis revealed clusters enriched in normal tissue samples and N0 and N+ cancer samples from patients reporting greater pain (Fig. 1B). These observations from a small dataset suggest that non-canonical signaling of PAR₂ might be favored in oral cancers and more prevalent in N+ oral cancers from patients with pain. We note that the abundance of individual canonical and non-canonical proteases differs among patients (Supplementary Fig. S1), suggesting that PAR₂ activation and signaling in oral cancer is patient-specific.

Oral cancer proteases elicit different trafficking of PAR₂

We studied PAR₂ trafficking in HEK-293 cells (HEK-PAR₁KO) in which PAR₁ (F2R) had been deleted (Supplementary Fig. S2, Supplementary Data Table S1). The cells were treated with active recombinant proteases ST14, CTSS, PRSS23 and MMP1. Trypsin was included as a positive control, as it has been previously studied in detail^38,39. We confirmed that the recombinant proteases were active by the cleavage of Förster resonance energy transfer (FRET) peptide substrates (Materials and Methods section Table 1, Supplementary Fig. S3, Supplementary Data Table S2). We used Bioluminescence Resonance Energy Transfer (BRET) sensors (Materials and Methods section Table 2) with PAR₂ fused to a bioluminescent energy donor, Renilla luciferase (RLuc8), or a fluorescent energy acceptor, GFP. Intracellular sites were labeled with Venus/GFP (acceptors) on the marker proteins; Kras (plasma membrane), Rab5 (early endosomes), and RLuc8 (donor) on giantin (cis-Golgi) (Fig. 2A). The BRET ratio between the donor and acceptor (BRETΔ) is a measure of the distance between PAR₂ and the different subcellular markers. Positive BRETΔ indicates proximity of PAR₂ to the marker and negative BRETΔ distancing. HEK-PAR₁KO cells co-transfected with the BRET pairs were treated with graded concentrations of trypsin (1 µM–10 pM), ST14 (316–1 nM), CTSS, PRSS23 and MMP1 (1 µM–1 nM) (Supplementary Fig. S4).

Table 1.

Proteolytic activity kits and reagents

Name	Catalogue #	Company	Description
SensoLyte® 520 MMP-1	AS-71150	Anaspec	MMP1 activity
SensoLyte® 520 Cathepsin S	AS-72099	Anaspec	Cathepsin S activity
SensoLyte® Rh110 Matriptase	AS-72241	Anaspec	Matriptase activity
SensoLyte® 520 green protease	AS-71124	Anaspec	Serine protease activities, used for PRSS23
PEI STAR™	7854	R&D Systems	Transfection reagent
TFLLR-NH2	1464	Bio-Techne	PAR1 agonist
2-Furoyl-LIGRLO-amide	3015	Bio-Techne	PAR2 agonist
Fura-2 AM	F1221	Invitrogen	Intracellular Ca2+ indicator
AZ3451	HY-112558	MedChemExpress	PAR2 antagonist
GB88	HY-120261	MedChemExpress	PAR2 antagonist
Pitstop 2	HY-115604	MedChemExpress	Clathrin inhibitor
Dyngo® 4a	ab120689	Abcam	Dynamin inhibitor
Coelenterazine h	301	Nanolight technology	Renilla luciferase substrate
Prolume purple	369	Nanolight technology	Renilla luciferase substrate
Phorbol 12,13-dibutyrate	4153	Tocris	Protein kinase C activator
Forskolin	1099	Biotechne	Adenylyl cyclase activator
3-Isobutyl-1-methylxanthine	17018	Sigma	Phosphodiesterase inhibitor
Ionomycin calcium salt	1704	Tocris	Ca2+ ionophore
HBSS (10X)	14065056	ThermoFisher	Buffer for BRET/FRET assays

Table 2.

Plasmids

Plasmid	Catalogue #	Company	Description
PAR2-RLuc8100	Obtained from the laboratory of Dr. Nigel Bunnett, NYU Dentistry Dept. of Molecular Pathobiology	N/A	PAR2 with RLuc8 tag
Flag-PAR2-HA54			PAR2 with HA and Flag tags
PAR2-GFP38			PAR2 with GFP tag
Kras-Venu101			Plasma membrane marker
Rab5-Venus101			Early endosomes marker
Giantin-RLuc839			Cis-Golgi marker
CAMYEL49			cAMP BRET sensor
Cytosolic EKAR51			Cytosolic ERK FRET sensor
EKAR51	Gift from Karel Svoboda (plasmid #18681)	Addgene	Nuclear ERK FRET sensor
pcDNA3.1(+)-ExRai-AKAR250	Gift from Jin Zhang (plasmid #161753)	Addgene	PKA FRET sensor

Fig. 2. Oral cancer proteases elicit different trafficking of PAR₂.

A Shown are the BRET sensors specific to the plasma membrane (Kras-Venus), early endosomes (Rab5-Venus) and cis-Golgi (Giantin-RLuc8). The BRET donor (RLuc8) is shown in blue, while acceptors (Venus/GFP) are shown in yellow and green, respectively. Representative traces of PAR₂ trafficking after stimulation with proteases show PAR₂ distancing from B1 the plasma membrane and trafficking to C1 early endosomes and D1 cis-Golgi. The respective dose-response curves are shown for the individual proteases trypsin (B2; n = 12, C2; n = 8 D2; n = 13), ST14 (B2; n = 7, C2; n = 7 D2; n = 6), CTSS (B3; n = 6, C3; n = 6, D3; n = 6), PRSS23 (B4; n = 10, C4; n = 8, D4; n = 8) and MMP1 (B5; n = 15, C5; n = 9, D5; n = 8). Arrows indicate protease addition. The proteases trypsin (coral), ST14 (charcoal), CTSS (purple), PRSS23 (green) and MMP1 (light blue) are displayed in solid circles. The yellow rectangles highlight the proteases that promoted PAR₂ proximity to the examined locations. Data are shown as mean ± SEM. The complete set of kinetic plots at all doses are shown in Supplementary Fig. S4. Supporting data are provided in Supplementary Data Table S3.

Three distinctPAR₂ trafficking profiles were observed in response to the proteases (Fig. 2): (a) canonical trafficking of the receptor into early endosomes by trypsin and ST14, (b) PAR₂ retention at the plasma membrane after treatment with CTSS and (c) PAR₂ distribution into the cis-Golgi elicited by PRSS23 and MMP1. Trypsin and ST14 promoted rapid and sustained internalization of PAR₂ at low concentrations, with half-maximal responses (EC₅₀) of 15 nM and 10 nM, respectively (Fig. 2B2). By contrast, CTSS promoted a modest but transient internalization at the maximum concentration tested (1 µM, Figure 2B3) and did not promote intracellular redistribution of the receptor (Figures 2C3 and 2D3). The overexpressed proteases were much less potent than ST14 and trypsin and drove PAR₂ internalization only at high concentrations (Figs. 2B4 and 2B5). PRSS23 (EC₅₀: 4.5 µM) promoted sustained internalization at concentrations 450 times higher than ST14, while MMP1 (EC₅₀: 292 nM) elicited transient internalization at concentrations 30 times higher than ST14. Following PAR₂ internalization, trypsin and ST14 distributed PAR₂ into early endosomes (Figure 2C1), with trypsin displaying a higher efficacy (Figure 2C2). MMP1 promoted PAR₂ distancing from early endosomes while PRSS23 promoted distancing only at the highest concentration tested (Figures 2C3 and 2C4) and distribution to the cis-Golgi (Figures 2D3 and 2D4), but with different kinetics and efficacy. MMP1 promoted robust and transient distribution to the cis-Golgi, whereas PRSS23 promoted minimal cis-Golgi distribution only at high concentrations.

Oral cancer proteases promote biased signaling driven by intracellular pools of PAR₂

PAR₂ activation by proteolytic cleavage leads to downstream signaling that depends on the cleaved region of the N-terminus and the exposed tethered ligand. PAR₂ signaling involves the recruitment of the α and βγ subunits of G proteins, as well as recruitment of β-arrestins 1 and 2, ultimately culminating in PAR₂ internalization, redistribution to intracellular compartments or degradation⁴⁰.

Oral cancer pain is attributed to the release of pain mediators from the tumor⁴¹. It has been described as a form of nociceptive pain⁴² that differs from inflammatory and neuropathic pain with respect to responses to anti-inflammatory drugs and pain associated neurochemical changes^43,44. To investigate the activation of PAR₂ by the selected proteases, we focused on the α subunits of G proteins (Gα_q, Gα_s, and Gα_i/o and Gα_12/13), which are recruited by PAR₂ following activation by canonical and non-canonical proteases¹⁶. We opted to study Gα_q and Gα_s signaling, as PAR₂ signaling with these G-proteins has received the most attention with respect to oral cancer pain, while Gα_12/13, also recruited by PAR₂, is associated with inflammation⁴⁵. PAR₂ recruitment of Gα_q results in release of intracellular Ca²⁺, while Gα_s stimulates cAMP production. Downstream effectors include protein kinase A (PKA), which is directly activated by cAMP and indirectly activated by Ca^2+46,47. Both pathways converge in ERK phosphorylation (Fig. 3A). To determine whether signaling by proteases involved PAR₂, we first pre-treated cells with the PAR₂ antagonists AZ3451 (10 µM) or GB88 (10 µM). We also investigated whether signaling involved internalization of PAR₂ by pre-treatment with inhibitors of the proteins dynamin (Dyngo-4a, 30 µM) and clathrin (Pitstop 2, 30 µM), reported to mediate PAR₂ endocytosis⁴⁸.

Fig. 3. Oral cancer proteases elicit different PAR₂ signaling pathways.

A Scheme of PAR₂ Gα protein-coupled signaling from the plasma membrane and intracellular locations with the site of action of the PAR₂ inhibitors AZ3451 and GB88 and endocytosis inhibitors Dyngo-4a and Pitstop 2 indicated. Direct and indirect activation are indicated by black and gray arrows, respectively. Representative traces of PAR₂ signaling after stimulation with proteases are shown for B1 intracellular Ca²⁺ mobilization (trypsin n = 16, ST14 n = 14, CTSS n = 7, PRSS23 n = 6 and MMP1 n = 8), C1 cAMP levels (trypsin n = 7, ST14 n = 8, CTSS n = 8, PRSS23 n = 9 and MMP1 n = 16), D1 PKA activation (trypsin n = 16, ST14 n = 10, CTSS n = 6, PRSS23 n = 7 and MMP1 n = 14), E1 nuclear ERK phosphorylation (trypsin n = 11, ST14 n = 7, CTSS n = 6, PRSS23 n = 7 and MMP1 n = 6) and F1 cytosolic ERK phosphorylation (trypsin n = 14, ST14 n = 8, CTSS n = 7, PRSS23 n = 6 and MMP1 n = 8). The proteases trypsin (coral), ST14 (charcoal), CTSS (purple), PRSS23 (green) and MMP1 (light blue) are displayed in solid circles. Arrows indicate protease addition. Effect of PAR₂ antagonists and endocytic inhibitors on intracellular Ca²⁺ mobilization (B2, AZ3451 n = 9, Dyngo/Pitstop n = 6), cAMP levels (C2, AZ3451/GB88/Dyngo n = 6 and Pitstop n = 8 and C3, AZ3451/GB88 n = 6, Dyngo n = 7 and Pitstop n = 11), PKA activation (D2, AZ3451 n = 10, Dyngo n = 8 and Pitstop n = 11, D3 AZ3451/Dyngo/Pitstop n = 7 and D4 AZ3451/Dyngo/Pitstop n = 7), nuclear ERK phosphorylation (E2, AZ3451 n = 7, Dyngo n = 13 and Pitstop n = 10 and E3, AZ3451/Dyngo/Pitstop n = 5) and cytosolic ERK phosphorylation (F2, AZ3451 n = 6, Dyngo n = 10 and Pitstop n = 9 and F3, AZ3451 n = 6, Dyngo n = 5 and Pitstop n = 5). Pretreatments are shown as follows: AZ3451 (open squares), GB88 (solid squares), Dyngo 4-a (open diamonds) and Pitstop 2 (open rectangles). Max values for the positive controls: Ca²⁺ (Ionomycin: 0.87), cAMP (Forskolin: 0.35), PKA (Forskolin: 0.08), nuclear ERK (phorbol 12,13-dibutyrate, PDBu: 0.08) and cytosolic ERK (PDBu: 0.07). Data are shown as mean ± SEM. The complete set of dose response curves are shown in Supplementary Figs. S5 and S6. Supporting data are provided in Supplementary Data Tables S4–S5.

To study PAR₂ signaling, we transfected HEK-PAR₁KO cells with FLAG-PAR₂-HA and the respective FRET and BRET sensors described in the Materials and Methods section Table 3. For Gα_q signaling, we measured changes in intracellular Ca²⁺ levels using the calcium sensor Fura-2 AM. Gα_s activity was evaluated using the cAMP sensor CAMYEL⁴⁹. The convergent downstream effectors were measured using the ExRai-AKAR2 sensor for PKA activation⁵⁰ and two sensors to measure phosphorylation of cytosolic (cytosolic EKAR) and nuclear (EKAR) ERK⁵¹.

Table 3.

BRET and FRET sensors

PAR2 variant	Sensor	Assay	Measurement	Excitation/Emission
PAR2-RLuc8	Kras-Venus	BRET	Plasma membrane trafficking	Em1: 480 ± 40 Em2: 530 ± 15
	Rab5-Venus	BRET	Early endosomes trafficking
Flag-PAR2-HA	CAMYEL	BRET	cAMP
PAR2-GFP	Giantin-RLuc8	BRET	Cis-Golgi	Em1: 395 ± 40 Em2: 510 ± 15
	Cytosolic EKAR	FRET	Cytosolic ERK	Exc: 405 ± 20 Em1: 480 ± 40 Em2: 535 ± 50
	EKAR	FRET	Nuclear ERK	Exc: 405 ± 20 Em1: 480 ± 40 Em2: 535 ± 50
	ExRai-AKAR2	FRET	PKA	Exc1: 420 ± 20 Exc2: 475 ± 40 Em1: 535 ± 25

Cells were treated with graded concentrations of trypsin (1 µM–10 pM), ST14 (316 nM–1 nM), CTSS, PRSS23 and MMP1 (1 µM–1 nM) (Supplementary Figs. S5 and S6). Trypsin was included as a control, as it has been previously studied in detail^39,48,52,53.

We observed four distinct signaling outcomes (Fig. 3). (a) Trypsin and ST14 elicited increases in intracellular Ca²⁺ levels (Fig. 3B1) but not cAMP (Fig. 3C1), consistent with Gα_q recruitment, as previously reported for trypsin^30,54. Downstream, we observed increased PKA activation (Fig. 3D1) and nuclear translocation of phosphorylated ERK (Fig. 3E1). We confirmed for trypsin that these responses were driven by intracellular PAR₂, as they were attenuated by pre-treatment of cells with the PAR₂ antagonists and endocytosis inhibitors (Figs. 3B2, D2 and E2 and Supplementary Fig. S6A1, C1 and D1). Similar studies were not undertaken with ST14, due to the high cost of the recombinant ST14 – an omission, which is justified by the identical trafficking and signaling of the two canonical proteases. In addition, trypsin and ST14 increased cytosolic ERK phosphorylation (Fig. 3F1, Supplementary Figs. S5E, and S6E1 and E2), as previously reported for trypsin⁴⁸; however, the response to trypsin was not attenuated by treatment with PAR₂ antagonists and endocytosis inhibitors (Fig. 3F2). By contrast, the PAR₂ antagonist I-343 has been reported to block trypsin-induced cytosolic ERK phosphorylation⁴⁸. While trypsin and ST14 displayed the same signaling profile, modest differences in efficacy were noted for PKA activation (trypsin > ST14, Supplementary Figs. S5C1 and C2 and S6C1 and C2).

(b) CTSS stimulation only increased cytosolic ERK phosphorylation (Fig. 3F1), Supplementary Figs. S5E3 and S6E3). We observed only a modest, non-significant decrease in CTSS-induced cytosolic ERK phosphorylation after treatment with AZ3451, Dyngo-4a and Pitstop 2 (Fig. 3F3, Supplementary Fig. S6E).

(c) Activation of PAR₂ by MMP1 elicited increases in cAMP levels (Fig. 3C1 and Supplementary Fig. S5B5), but not intracellular Ca²⁺ levels (Fig. 3B1 and Supplementary Fig. S5A5), consistent with Gα_s recruitment. MMP1 activated PKA, promoted robust nuclear translocation of phosphorylated ERK (Fig. 3E1 and Supplementary Fig. S5D5) as reported previously¹⁸ but no cytosolic ERK phosphorylation (Fig. 3F1, Supplementary Fig. S5E5)¹⁸. (d) Following PAR₂ activation by PRSS23, we observed increased cAMP levels and PKA activation (Fig. 3C1, D1 and Supplementary Figs. S5B4, C4 and S6B4 and C4), but not ERK phosphorylation (Fig. 3E1, F1 and Supplementary Figs. S5D4–E4 and S6D4–E4). We note that MMP1 and PRSS23 promoted PKA activation with much lower efficacy than ST14 (Supplementary Table S2). The increases in cAMP levels (PRSS23 and MMP1), activation of PKA (PRSS23 and MMP1) and nuclear translocation of phosphorylated ERK (MMP1) were driven by intracellular PAR₂ (Fig. 3C2, C3, D3, D4, E3 and Supplementary Fig. S6B4, B5, C4, C5 and D5). The observed differences in downstream signaling promoted by MMP1 and PRSS23 suggest that PRSS23- and MMP1-activated PAR₂ could signal through different pathways.

PAR₂ trafficking and signaling promoted by a mixture of oral cancer proteases reflects the relative potencies and concentrations of the individual proteases in the mixture

The oral cancer TME is replete with canonical and non-canonical proteases that promote PAR₂ trafficking and signaling with different potencies, efficacies and kinetics. To better understand PAR₂ trafficking and signaling in the oral cancer TME, we studied PAR₂ trafficking and signaling in response to mixtures of equal concentrations of active recombinant ST14, CTSS, PRSS23 and MMP1 over a 316 nM–1 nM concentration range (Supplementary Figs. S4–S5). We acknowledge that equal concentrations do not accurately reflect the proportions of proteases in the TME. Moreover, we note that the relative concentrations of proteases vary among patients, so that a single protease cocktail cannot represent all patients (Supplementary Fig. S1). To assess the effect of the mixture, we compared the kinetic activities of the individual proteases, the kinetics of the protease mixtures at high (316 nM) and low (31.6 nM) concentrations and the respective dose-response curves (Fig. 4).

Fig. 4. The intracellular distribution of PAR₂ elicited by a mixture of oral cancer proteases reflects trafficking promoted by the individual proteases at a given concentration.

Shown are representative kinetic traces for the individual proteases that show the greatest contribution to PAR₂ trafficking from the (A1) the plasma membrane to (B1) early endosomes and C1 cis-Golgi, as measured using BRET sensors specific to subcellular compartments after stimulation with individual proteases. The kinetic traces for PAR₂ trafficking stimulated by the high concentration mixture (316 nM) of ST14, CTSS, PRSS23 and MMP1 are shown together with the sum of PAR₂ trafficking of individual proteases in (A2; n = 6) the plasma membrane, (B2; n = 6) early endosomes and (C2; n = 6) cis-Golgi and to the low concentration of the mixture (31.6 nM) in (A3; n = 6) the plasma membrane, (B3; n = 6) early endosomes and (C3; n = 6) cis-Golgi. In the respective dose-response curves (A4, B4, C4), high and low concentrations are indicated with a black arrow. The proteases ST14 (charcoal and light gray), PRSS23 (dark and light green) and MMP1 (light blue and cyan) are displayed in solid circles. The mixtures are shown in dark yellow hexagons and the sum of individual proteases are shown in yellow triangles. mean ± SEM. Arrows indicate addition of individual proteases or the mixture of proteases. The kinetic traces shown in A1, B1 and C1 are repeated from Fig. 2, with a slight adjustment to the scale to facilitate the comparison of individual proteases and the mixture. The complete set of dose response curves for the mixtures are shown in Supplementary Fig. S4. Supporting data are provided in Supplementary Data Table S3.

Trafficking elicited by the mixture of proteases reflected the potencies and abundance of different proteases. Distancing from the plasma membrane appeared to be mediated by the combined activities of the proteases (Fig. 4A1–A4). At low concentrations of the mixture, PAR₂ trafficking to endosomes (Fig. 4B1–B4) was favored by the greater potency of ST14, while at high concentrations, trafficking to the cis-Golgi predominated (Fig. 4C1–C4) and was mediated by the increased concentrations of the less potent PRSS23 and MMP1. Summing the activities of the proteases in the mixture (assuming equal contributions) accurately predicted the observed trafficking. An exception is the distancing from the plasma membrane at the low concentration.

The activity of the mixture with respect to signaling (Ca²⁺ influx, cAMP generation, PKA activation, phosphorylation of cytosolic ERK and nuclear ERK translocation) can also be understood by comparing the relative potencies of individual proteases at high and low concentrations and their ability (efficacy) to trigger PAR₂ signaling (Fig. 5). At high concentration of the mixture, when activation of PAR₂ by the proteases with lower potency and efficacy (MMP1 and PRSS23) is favored, the signaling activity predicted from summing the proteases is greater than or equal to that observed by the mixture (Fig. 5B2, B4, C2, C4, D2, D4). Summing the activity of MMP1 and PRSS23 at 316 nM better predicts the observed cAMP increases (Fig. 5B2 and B4), while the activity of MMP1 and ST14 at 316 nM predicts nuclear ERK phosphorylation (Fig. 5D2 and D4). By contrast, signaling reflects the activity of ST14 at low concentration when activation of PAR₂ by ST14 predominates and there is no contribution from MMP1 and PRSS23 (Figs. 5A3, A4, C3, C4, D3, D4, E3 and E4). We note, however, that the observed production of cAMP by the low concentration of the mixture (Fig. 5B3) is likely to originate from sources other than PAR₂ activation by the proteases, as none of the studied proteases elicits cAMP production at the low concentration. Similarly, with respect to phosphorylation of cytosolic ERK, summing the activity of high concentrations of CTSS and ST14 predicts the observed increase in phosphorylation of cytosolic ERK (Fig. 5E2 and E4), whereas at low concentrations the potency and efficacy of ST14 appear to drive signaling (Fig. 5E3 and E4).

Fig. 5. Signaling elicited by a mixture of oral cancer proteases reflects the potency and efficacy of the individual proteases.

PAR₂ signaling was measured using calcium, BRET or FRET assays. Shown are the kinetic traces for the individual proteases that show the greatest contribution to eliciting changes in (A1) intracellular Ca²⁺, (B1) cAMP levels, (C1) activation of PKA, (D1) phosphorylation of nuclear and (E1) cytosolic ERK. The kinetic traces for PAR₂ signaling stimulated by the high concentration mixture (316 nM) of ST14, CTSS, PRSS23 and MMP1 are shown together with the sum of PAR₂ signaling of individual proteases in (A2; n = 8) intracellular Ca²⁺, (B2; n = 7) cAMP levels, (C2; n = 8) activation of PKA, (D2; n = 7) phosphorylation of nuclear and (E2; n = 6) cytosolic ERK and to the low concentration of the mixture (31.6 nM) in (A3; n = 8) intracellular Ca²⁺, (B3; n = 7) cAMP levels, (C3; n = 8) activation of PKA, (D3; n = 7) phosphorylation of nuclear and (E23; n = 6) cytosolic ERK. In the respective dose-response curves (A4, B4, C4, D4, E4), high and low mixture concentrations are highlighted in red. The proteases ST14 (charcoal and light gray), CTSS (dark and light purple) PRSS23 (dark and light green) and MMP1 (light blue and cyan) are displayed in solid circles. The mixtures are shown in dark yellow hexagons and the sum of individual proteases are shown in yellow triangles. n = 6–15, mean ± SEM. Arrows indicate addition of individual proteases or the mixture of proteases. The kinetic traces shown in the first column are repeated from Fig. 3 to facilitate the comparison of individual proteases and the mixture. The complete set of dose response curves for the mixtures are shown in Supplementary Fig. S5. Supporting data are provided in Supplementary Data Table S4. Max values for the positive controls: Ca²⁺ (Ionomycin: 0.87), cAMP (Forskolin: 0.35), PKA (Forskolin: 0.08), nuclear ERK (PDBu: 0.08) and cytosolic ERK (PDBu: 0.07).

The observations with mixtures of proteases suggest that PAR₂ trafficking and signaling differ among N + , N0 and normal oral tissues depending on expression levels and potencies of the proteases. In N+ cancers expressing high levels of MMP1 and PRSS23, cAMP production and PAR₂ distribution in the cis-Golgi might predominate. In N0 cancers, with low protease expression levels, calcium mobilization will be driven by canonical PAR₂ signaling and endosomal distribution following activation by the more potent ST14. Downstream signaling may depend on the greater potency of ST14 compared to the other tested proteases at lower concentrations.

Expression of PAR₂ activating proteases in the oral cancer TME

Our experiments modeled the impact of pericellular proteases acting on PAR₂. This receptor is widely expressed by cells in the TME, and multiple cell type-specific PAR₂ activation pathways may contribute to oral cancer pain. Sensory neurons extending into the TME will be in close proximity to fibroblasts and epithelial (cancer), endothelial and immune cells (Supplementary Fig. S7A). To understand which cell types express the PAR₂-cleaving proteases, we queried a single-cell RNA sequencing dataset (GSE164690) of human oral cancers depleted for immune cells⁵⁵ (Supplementary Fig. S7B–D). Expression of F2RL1 was detected in epithelial cells, fibroblasts and endothelial cells (Supplementary Fig. S7C). Expression of ST14 was restricted to epithelial cells, MMP1 was expressed by epithelial cells and fibroblasts, PRSS23 by epithelial cells, fibroblasts and endothelial cells and CTSS by fibroblasts, endothelial cells and immune cells (macrophages²¹) (Supplementary Fig. S7B–D). Mixtures of proteases released from these cells have the potential to activate PAR₂ on the neurons innervating the TME, as well as on other cell types in the TME. Moreover, the local composition and concentrations of proteases are expected to determine PAR₂ trafficking and signaling according to the relative potencies and efficacies of the proteases, as demonstrated by our mixture experiments (Figs. 4 and 5). Accordingly, on the one hand, there is potential for direct activation of PAR₂ on neurons by canonical and non-canonical proteases. On the other hand, PAR₂-mediated sensitization of neurons innervating the cancer may be indirect. In either scenario, these observations support the development of PAR₂ therapeutics targeting multiple intracellular compartments and signaling pathways (Fig. 6).

Fig. 6. PAR₂ signaling varies with protease expression levels in the TME.

A Depicts the TME comprising immune and cancer epithelial cells, lymph and blood endothelial cells and fibroblasts. Protease expression is dysregulated in cancer. Colored circles represent proteases, and the number of circles reflects their concentration in the TME. ST14 (black) and CTSS (purple) transcript levels do not vary according to nodal or pain status. Painful N+ cancers express higher levels of PRSS23 and MMP1. B Canonical PAR₂ signaling and biased agonism elicited by the different concentrations of the studied proteases. The dark-colored circles represent the most abundant proteases, while proteases with decreased opacity are expressed at low concentrations. In N0 and N+ cancers, PAR₂ cleavage will be favored by the potency of ST14 and low levels of expression of PRSS23 and MMP1. By contrast, the high concentrations of PRSS23 and MMP1 in painful N+ cancers will shift PAR₂ activation and signaling to include biased agonism.

Discussion

Oral cancer patients report a variety of pain experiences and levels of discomfort at the primary site of the cancer. The differences are attributed to inter-tumor heterogeneity. The cancers differ in dysregulated pathways and genes that underlie oncogenesis, hence the expression and activation of different pain mediators. Here, we asked how the trafficking and signaling of PAR₂, a pain mediator, might be altered by the dysregulated expression of PAR₂-activating proteases in the TME. Our study was motivated by the observation that MMP1 and PRSS23 were overexpressed in N+ cancers from patients who reported high levels of pain. We demonstrate that MMP1 and PRSS23 promote non-canonical trafficking of PAR₂ to the cis-Golgi and increase cAMP production. By contrast, ST14 is not differentially expressed in oral cancers at the transcript level. The ST14 encoded protease cleaves PAR₂ at the canonical site, promotes trafficking to endosomes and Gα_q mediated Ca²⁺ signaling, and elicits PAR₂ trafficking and signaling with higher potency and efficacy than MMP1 and PRSS23. A mixture of proteases, prepared to model the oral cancer microenvironment, revealed that ST14-mediated trafficking and signaling predominated at low protease concentrations. At high concentrations, MMP1 and PRSS23 prevailed over the greater potency of ST14.

PAR₂ is present in all cell types in the TME, including nociceptive trigeminal neurons innervating the cancer. We do not know which PAR₂-expressing cell types contribute to oral cancer pain. Targeted deletion of PAR₂ on cancer cells¹⁰ or nociceptive neurons¹¹ reduced nociceptive behavior in mouse oral cancer xenograft models, demonstrating that oral cancer pain can be mediated by proteases activating PAR₂ on cells in the TME that subsequently release mediators that sensitize neurons or by proteases activating PAR₂ directly on neurons. PAR₂ has established roles in oral cancer progression and shaping of the TME, including modulating the immune landscape⁵⁶, mediating oral cancer cell proliferation and migration^57,58 and paracrine activation of PAR₂ on fibroblasts by ST14²⁶. Whether oral cancer nociception would be reduced by deletion of PAR₂ on other cell types in the TME has not yet been studied, but could be explored using approaches such as genetic epistasis analyses, as applied to investigate the role of PAR₂ in ST14-mediated oncogenesis²⁶.

Cell type expression of PAR₂ activating proteases may also result in local differences in the concentration of proteases that activate PAR₂ signaling pathways leading to release of mediators that sensitize neurons. We previously reported⁷ that the set of genes overexpressed in N+ cancers from patients reporting high levels of pain (“pain and metastasis genes”) overlaps with the meta-signature of the partial epithelial-to-mesenchymal transition (p-EMT). The p-EMT program is a transient state, which is characterized by increased expression of extracellular matrix genes, reduced expression of epithelial genes and lack of expression of classical EMT transcription factors⁵⁹. Cells expressing the p-EMT program are located at the leading edges of the cancer⁶⁰. MMP1 and PRSS23 are part of the p-EMT signature, suggesting that these proteases may be more highly expressed in cells at the leading edge of the cancer, favoring local biased agonism.

Mechanisms for pain production following activation of PAR₂ on neurons include PAR₂-mediated sensitization of the Transient Receptor Potential channels Ankyrin 1 and Vanilloid 1 and 4 (TRPA1, TRPV1 and TRPV4)^61–64, channels demonstrated to mediate (oral) cancer pain in rodent models^65,66. PAR₂ signaling via Gα_q and Gα_s pathways has been implicated in modulation of the activity of these receptors by activation of phospholipase C, protein kinase C and PKA. Additionally, we suggest that PAR₂ mediated increases in intracellular cAMP might directly lead to neuronal hyperexcitability. cAMP binds to the cyclic nucleotide binding domain on hyperpolarization-activated cyclic nucleotide-gated potassium and sodium (HCN) channels, promoting channel opening and subsequent firing in nociceptors⁶⁷. HCN2 has been linked to migraine, inflammatory and neuropathic pain⁶⁸ and could contribute to oral cancer pain.

There are limitations with this study. The proteases overexpressed in oral cancer were selected based on their mRNA expression levels. We did not measure protein expression of proteases in oral cancer tissues. We focused on two overexpressed proteases; future studies should include FURIN and MMP13, which are also overexpressed in oral cancers. MMP13 cleaves PAR₂ at the same site as MMP1 and is more abundant in N+ cancers. We anticipate that MMP13 will promote trafficking and signaling similar to that of MMP1, which will be favored in N+ patients experiencing pain and overexpressing MMP13.

Our study in HEK-293 cells reveals the potential PAR₂ trafficking and signaling elicited by proteases in the oral cancer TME. Although we focused our signaling studies on PAR₂ by overexpressing PAR₂ in HEK PAR₁-KO cells, HEK-293 cells express numerous GPCRs⁶⁹. We cannot rule out possible contributions from protease activation of other GPCRs. We note that HEK-293 cells differ in complexity and morphology from the cell types found in the TME. Yet, multiple studies have demonstrated that PAR₂ findings in HEK cells are recapitulated in dorsal root ganglia neurons, enteric neurons and intestinal epithelial cells^39,48,70. MMP1 and PRSS23 promote trafficking of PAR₂ to the cis-Golgi. In trigeminal neurons innervating oral cancer, however, the Golgi apparatus is located in the cell body in the trigeminal ganglion, distant from the site of PAR₂ activation in the oral cavity⁷¹. MMP1 and PRSS23 may promote neuronal PAR₂ trafficking to satellite Golgi organelles (Golgi outposts) located in the axons and dendrites^72,73, rather than the Golgi apparatus in the cell body.

We demonstrated that ST14, PRSS23 and MMP1 signaling is driven by intracellular PAR₂. Future work should include sensors selectively expressed in endosomes (ST14) and the cis-Golgi (PRSS23, MMP1) to validate that the signaling arises from these different intracellular locations. PAR₂ trafficking to the cis-Golgi promoted by MMP1 and PRSS23 was observed despite not promoting early endosomal trafficking. The route by which PAR₂ traffics to the cis-Golgi has not been demonstrated. Future work should investigate trafficking via routes distinct from those involving vesicles expressing Rab5, such as Rab22A, Rab22b or Rab31, also involved in early endosomes-Golgi transport⁷⁴.

Our findings were discrepant with previous publications reporting CTSS-mediated increases in cAMP levels³⁰ and lack of ERK phosphorylation^30,48. These differences could be attributed to the studied cell lines (HEK-293 and KNRK cells versus our study with HEK PAR₁-KO cells). HEK-293 cells express PAR₁. Studies suggest that CTSS can activate PAR₁⁷⁵ and there could be the possibility for unique signaling promoted by PAR₁-PAR₂ heterodimers⁷⁶. In addition, the source, potential post-translational modifications and activity of CTSS might contribute to the observed differences. Our study used active recombinant CTSS expressed in E. coli, a source that lacks the machinery to produce glycosylated proteins⁷⁷. By contrast, one of the previous studies³⁰ used activated human pro-CTSS expressed in Sf9 cells, which are capable of performing N-linked glycosylation in recombinant proteins⁷⁸.

Our findings have clinical implications. Opioids remain the most efficacious but imperfect analgesics for oral cancer pain. As tolerance develops, patients require progressively higher doses. Among cancers, oral cancer has the highest opioid escalation index⁷⁹. The severity and the profound impact that pain has on the quality of life in patients with oral cancer motivates our study of individual and combinations of proteases on PAR₂ trafficking and signaling.

We demonstrate the potential for proteases to promote PAR₂ trafficking to multiple cellular sites—plasma membrane, endosomes and cis-Golgi and to elicit multiple signaling events—Ca²⁺ mobilization, increased cAMP, PKA activation and phosphorylation of ERK. In the TME, all these PAR₂ trafficking and signaling events may occur simultaneously and contribute to the heterogeneity of oral cancer pain. We propose that the patient experience of pain and pain phenotype are linked to the concentration and types of proteases present in the TME.

While, to date, most intracellular GPCRs have been demonstrated to elicit pain from only one location, commonly endosomes^70,80–84, other receptors are known to signal from more than one location, such as the metabotropic glutamate receptor 5, which signals from the endoplasmic reticulum and the nucleus^82,83. More comprehensive investigation into receptor trafficking and signaling is warranted to improve therapeutic efficacy by designing new antagonists that can provide signaling bias to inhibit the cAMP/PKA pathway, demonstrated here to be activated by MMP1 and PRSS23. The use of drug delivery systems also affords the opportunity to deliver drugs in specific cellular locations, such as the Golgi or endosomes. Improved analgesia using drug delivery systems has been achieved by selective targeting of endosomal GPCRs^{10,70,81,85–89}, which involves delivering a single drug to a single compartment. In the case of PAR₂-mediated pain, greater relief might be provided by delivering a single drug to multiple compartments.

We conclude that proteases in the TME can elicit distinct PAR₂ trafficking and signaling. While the mixture of proteases does not represent the complexity of protease activity in the TME, our observations suggest that PAR₂ trafficking and signaling in cancers reflect the potency and efficacy of the individual proteases. The different expression profiles of oral cancer proteases can inform the potential PAR₂ cleavage and, thus, help predict the predominant trafficking and signaling of the receptor. Knowledge of the intra-cancer heterogeneity of PAR₂ trafficking and signaling could guide the design of targeted antagonists to provide relief for oral cancer patients.

Materials and methods

Reagents

Recombinant human proteases were purchased: MMP1 (Cat.# 420-01, PeproTech®, ThermoFisher Scientific), which contains the entire catalytic N-terminal and C-terminal domains, involved in substrate specificity, CTSS (Cat.#219343, MilliporeSigma), PRSS23 (Cat.# RPU53309, Biomatik) and ST14 (Matriptase/ST14 Catalytic Domain, CF, Cat.#3946-SEB, Bio-techne). Trypsin, obtained from porcine pancreas, was purchased from MilliporeSigma (Cat. #T0303). Sources of kits used to measure protease activities, reagents and plasmids are provided in Tables 1 and 2.

Cell lines

HEK-293 cells in which PAR₁ (F2R) had been deleted using CRISPR/Cas9 (HEK-PAR₁KO)⁹⁰ were a gift from Dr. Morley Hollenberg to the laboratory of Dr. Nigel Bunnett. Cells were cultured at 37 °C with 5% CO₂ in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with penicillin/streptomycin (50 U/mL) and 10% (vol/vol) fetal bovine serum (FBS). Loss of functional PAR₁ was confirmed by intracellular Ca²⁺ measurements, stimulating HEK-PAR₁KO cells with the PAR₁ agonist TFLLR-NH2 and PAR₂ agonist 2-Furoyl-LIGRLO-amide (2 F, Supplementary Fig. S2). For transfection, HEK-PAR₁KO cells were seeded into 10 mm Petri dishes (1.5 × 10⁶ cells) and transfected 24 h later using poly-ethyleneimine (PEI STAR™) at a 1:5 DNA: PEI ratio. Prior to transfection, culture media were changed to fresh media.

Intracellular Ca²⁺ measurements

HEK-PAR₁KO cells were transfected with human FLAG-PAR₂-HA (2.5 µg). The FLAG-PAR₂-HA construct includes an amino-terminal proopiomelanocortin signal peptide to allow expression at the cell surface in addition to the FLAG and HA epitopes fused to the N- and C-terminus, respectively⁵⁴. After 24 h, cells were transferred to 96-well clear plates and incubated for 24 h. On the day of the assay, cells were loaded with Fura-2 AM (1 mM) in calcium buffer (150 mM NaCl, 2.6 mM KCl, 0.1 mM CaCl₂, 1.18 mM MgCl₂, 10 mM D-glucose, 10 mM HEPES, 4 mM probenecid, 0.5% BSA, pH 7.4) for 30 min at 37 °C in a CO₂ free incubator. Cells were washed once with calcium buffer and left to equilibrate for 15 min.

Fura-2 AM fluorescence was measured with a FlexStation 3 (Molecular Devices) at 340 and 380 nm excitation and 530 nm emission. Fura-2 AM displays an excitation of 340 when bound to Ca ²⁺ and 380 nm in its free form. To measure Ca²⁺ mobilization, the ratio of Ca ²⁺-bound and free Fura-2 AM was used (340/380 nm). Fluorescence was measured every 5 s for 25 cycles⁹¹. Baseline was recorded for five cycles before adding proteases, vehicle (calcium buffer), or positive control (ionomycin, 1 µM). Cells were treated with graded concentrations of trypsin (1 µM–10 pM), ST14 (316 nM–1 nM), CTSS, PRSS23 and MMP1 (1 µM–1 nM) and the mixture of proteases (316 nM–1 nM). Antagonism of PAR₂ and inhibition of endocytosis was performed by pre-incubating cells for 60 min with a PAR₂ antagonist (AZ3451, 10 µM) and inhibitors of dynamin (Dyngo-4a, 30 µM) and clathrin (Pitstop 2, 30 µM). After 60 min, cells were washed once with calcium buffer.

BRET and FRET measurements

HEK-PAR₁KO cells were transfected with human PAR₂ variants (2.5 µg) and the BRET/FRET sensors (2.5 µg) described in Table 3.

After 24 h, cells were transferred to 96-well clear bottom black (FRET) and white (BRET) plates (catalog numbers #165305 and #165306, Nunc, Thermo Fisher) and incubated for a further 24 h. On the day of the assay, the media were changed to Hanks’ Balanced Salt Solution (HBSS), supplemented with HEPES (pH 7.4) and cells were left to equilibrate for 30 min at 37 °C in a CO₂-free incubator. Prior to luminescence measurements, cells were incubated for 10 min with Coelenterazine h (10 µM) for assays using Kras, Rab5 and CAMYEL and with Prolume purple (10 µM) for cis-Golgi assays. Additionally, for PKA measurements, the phosphodiesterase inhibitor 3-isobutyl-1-methylxanthine (IBMX) was added 5 min prior measurements to lower basal PKA activation. For nuclear and cytosolic ERK measurements, cells were serum-starved for 6–12 h before changing the media to HBSS.

Experiments were run in a CLARIOstar^® Plus plate reader (BMG Labtech) with optic module FI 430 530 480. Measurements were made every 90 s. Baseline was measured for five cycles, followed by stimulation with proteases, vehicle (HBSS) or positive controls (forskolin, PDBu), and measurements continued for 25 cycles. Trypsin was used as the positive control for trafficking experiments, while forskolin and PDBu were used for signaling experiments. Cells were treated with graded concentrations of trypsin (1 µM–10 pM), ST14 (316 nM–1 nM), CTSS, PRSS23, MMP1 (1 µM–1 nM) or a mixture of proteases (316 nM–1 nM). Antagonism of PAR₂ and inhibition of endocytosis was performed by pre-incubating cells for 60 min with AZ3451 or GB88 (10 µM) and 60 min with the endocytic inhibitors Dyngo-4a (30 µM) and Pitstop 2 (30 µM), respectively. Pre-treated cells were washed once with HBSS before starting the measurement protocol.

The PAR₂ antagonist AZ3451 (10 µM) was used as our first choice. The antagonist GB88 (10 µM) was used to examine the contribution of PAR₂ to cAMP production, as AZ3451 does not significantly inhibit cAMP production. AZ3451 blocks Ca²⁺ mobilization and phosphorylation of ERK in vitro, displaying more effective antagonism after PAR₂ activation by the synthetic agonist SLIGRL-NH2 than by trypsin^92,93. GB88 inhibits Ca²⁺ signaling in vitro, but also functions as a weak agonist of PAR₂ Gα_i/o and Gα_12/13 signaling^94,95.

Single cell RNA sequencing of human oral cancers

The filtered feature-barcode matrices containing detected cell-associated barcodes for CD45- oral cavity tumor samples were downloaded from the Gene Expression Omnibus database (accession ID: GSE164690). Further analysis, including quality filtering, identification of highly variable genes, and dimensionality reduction, was performed using Seurat⁹⁶.

Cells were filtered to only include those with >1000 UMIs, 200–5000 detected genes, and <10% of transcripts coming from mitochondrial genes. After removing unwanted cells from the dataset, 17,292 cells were retained. The UMI counts were normalized by the total number of UMIs per cell, multiplied by a scale factor of 10,000, and log-transformed. 50 principal components (PC) were calculated based on highly variable genes. To visualize the data, the dimensionality of the data matrix was further reduced to project the cells in two-dimensional space by uniform manifold approximation and projection (UMAP)⁹⁷. For visual data exploration, the Seurat object was converted to a 10x Genomics Loupe file using the loupeR package.

Statistics and reproducibility

Data were analyzed using the GraphPad Prism Software version 10. Inhibitor studies (Fig. 3) were evaluated using one-way ANOVA followed by Holm–Šídák multiple comparison tests. P < 0.05 was considered significant and p values are reported in the figures. Hierarchical clustering was performed using ClustVis⁹⁸. For the trafficking and signaling experiments, 6–15 independent experiments were performed for each individual signaling and trafficking experiment. The number of independent experiments is included in the legend of each figure. Each independent experiment was performed using 1–3 replicates, different cells, transfections and plates. The trafficking and signaling data were analyzed by subtracting the baseline (mean of 5 points prior to protease addition) from raw ratio values and subsequently the vehicle. This analysis was performed using MATLAB⁹⁹, version 23.2.0.2365128 (R2023b).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: P.D.R-G. and D D.G.A. Investigation: P.D.R-G., I.D., Z.D., R.L., L.A., N.H.T. and D.G.A. Visualization: P.D.R-G. and D.G.A. Funding acquisition: P.D.R.-G., B.L.S. and D.G.A. Project administration: P.D.R-G. and D.G.A. Supervision: D.G.A. Writing—original draft: P.D.R-G. and D.G.A. Writing—review & editing: P.D.R-G., I.G., Z.D., R.L., L.A., N.H.T., B.L.S. and D.G.A.

Peer review

Peer review information

Communications Biology thanks Moeno Kume and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Dr Mythreye Karthikeyan and Dr Ophelia Bu.

Data availability

Bulk RNA sequencing and single cell sequencing data can be accessed at GSE156178 and GSE164690, respectively. All data are available in the Supplementary Data Tables, as detailed in each figure legend.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-10215-x.