PS and PSMME have been shown to inhibit Ca²⁺-influx through the TrpA1 ion channel in response to TrpA1 activators in mammalian cells (8, 9). The mammalian TrpA1 binding sites have been identified and predicted for two known TrpA1 antagonists, A-967079 and HC-030031, based on phylogenetic, mutational and modelling studies (24). In this study, we have used Drosophila as a model to study the effects of stilbenoid compounds on inflammatory responses. To be able to investigate if stilbenes can interact with the Drosophila TrpA1 (dTrpA1) channel, we used molecular modelling to computationally predict the antagonist binding sites at the receptor. In the absence of an experimentally defined structure of the dTrpA1 channel, we built a comparative homology model of dTrpA1 based on the experimental human TrpA1 (hTrpA1) structure. The dTrpA1 model with the best Discrete Optimized Protein Energy (DOPE) score (-317763.96875) and the root-mean-square deviation (RMSD) of 0.764 Å from the template was utilized for molecular modelling studies ( Figure 1A ). The two TrpA1 antagonists A-967079 and HC-030031 were used as reference compounds and were modelled into their respective binding sites described in the literature using molecular docking ( Figure 1A , zoom in). The A-967079 binding site has been described to locate within the channel pore between TM5, TM6 and pore helix 1 (PH1) from one subunit and TM6 from the adjacent chain (25). The residues forming this particular binding pocket in hTrpA1 are S873, T874, F877, F909 and M912 (25–27). The respective residues in dTrpA1 are V431, L432, F435, F468 and M471. The HC-030031 binding pocket is situated between the TM4-TM5 linker, preTM1 and TRP-like domains close to the membrane (28). The residues forming the HC-030031 pocket in hTrpA1 are W711, N855, Q979, H983, A971 and in dTrpA1 W268, Q413, Q538, H542 and A530.

To investigate possible binding modes and sites of stilbenes at dTrpA1, PS and PSMME were docked into both known antagonist binding sites. Based on the docking results, both PS and PSMME are predicted to form a hydrogen bond with S501 (of the adjacent TM6) and π-π interactions with F468 (of PH1) in the A-967079 binding site, while A-967079 itself is predicted to engage in hydrogen bonding with L428 backbone oxygen (in TM5) and π-π interactions with F468 ( Figure 1A , upper panel). In the HC-030031 binding pocket, PS and PSMME are predicted to form a hydrogen bond with Q538 (of the TRP-like domain) and π-π interactions with W268 (of the preTM1 domain), while HC-030031 itself is predicted to form hydrogen bonds with Q265 (of the preTM1) and D412 (of TM4-TM5 linker) ( Figure 1A , lower panel).

In addition to assessing the initial docking scores and favorable ligand-protein interactions to evaluate the predicted binding poses, we employed the Molecular Mechanics-Generalized Born Surface Area (MM-GBSA) method to estimate the free energy of binding (binding affinity) of the docked ligands. Moreover, the best docking poses of A-967079 and HC-030031 in their respective dTrpA1 binding sites and PS (as the representative of the stilbene compounds) in both the known binding pockets of the dTrpA1 model were subjected to molecular dynamics (MD) simulation to evaluate the stability of the predicted ligand-dTrpA1 complexes. After the MD simulation, the MM-GBSA binding affinities were recalculated to see if the simulation had improved or worsened the estimated binding affinity of the ligands. Whereas the MM-GBSA binding free energy of HC-030031 was significantly better after the 100-ns simulation (from -44.39 to -74.42 kcal/mol) at its corresponding binding site, it remained the same for A-967079 (-28 kcal/mol), suggesting a less favorable binding for A-967079 at its binding site. The MM-GBSA binding free energies of PS were also improved during the MD simulation (from -35.30 kcal/mol and -31.73 kcal/mol to -45.62 kcal/mol and -54.89 kcal/mol at the A-967079 and HC-030031 binding sites, respectively). PSMME (non-simulated) had very similar binding free energy values to those of PS (non-simulated) at both binding sites.

The predicted HC-030031-TrpA1 complex remained relatively stable during the simulations and HC-030031 engaged in water-mediated hydrogen bonding interactions with Q261 and L264 (of preTM1) (data not shown). However, the small reference ligand A-967079 was not stable in its predicted binding pose at the binding site during the simulation. It changed its position and orientation, and thus, also the interactions with the protein, e.g., the hydrogen bond with L428 (TM5) was exchanged to hydrogen bonds with S501 (of the adjacent TM6), L464 (of PH1) and S436 (of TM5) (data not shown). Although PS binding at the A-967079 site was relatively stable during the whole simulation via π-π interactions with F468 (of PH1), its hydrogen bond swapped from S501 (of the adjacent TM6) (in Figure 1B , left) to S436 (of TM5) and it could form an extra hydrogen bond through a water molecule with P461 (of PH1) (in Figure 1B , right). At the HC-030031 binding site, PS moved deeper into the cavity from its initial docking site, losing the initial interactions while forming a new water-mediated hydrogen bond with D412 (of the TM4-TM5 linker). This suggests that PS is too small for this site and that the A-967079 binding site is the preferred site for PS. Taken together, both PS and PSMME can bind to Drosophila TrpA1, according to the in silico studies.

To study the effect of stilbenoid compounds in vivo, we wanted to induce activation of the NF-κB-mediated inflammatory Imd/Relish pathway in the Drosophila intestine, without direct interference of the signaling mediators of the pathway, or by bacterial infection. In both murine and Drosophila models, oral administration with DSS has been used to chemically induce intestinal inflammation (29–31). DSS has been shown to disrupt the epithelia barrier and the intestinal homeostasis, inducing intestinal inflammation (31–35).

To assess the NF-κB-mediated inflammatory response induced by DSS in Drosophila, we fed foraging 3^rd instar larvae with fly food supplemented with DSS, whereafter, we used qPCR-based analysis to detect Relish-specific AMP target genes. We noticed that 5% w/v of 40 kDa DSS induced an increased gene expression of the NF-κB Relish target gene diptericin compared to control fed flies ( Figure 2A ), also in the absence of pathogenic infection. As Relish is involved in protecting the epithelial borders in the intestine from local insults, and thus maintains tissue homeostasis in situ also during non-pathogenic inflammation-inducing conditions (22, 36, 37), we wanted to specifically investigate if the local immune response in the gut is activated in response to DSS. As expected, we were able to detect local activation of the immune response in the Drosophila intestine by performing X-Gal staining on the dissected gut of diptericin-lacZ reporter flies fed with DSS ( Figure 2B ).

In addition, we analyzed if DSS treatment affects the intestinal microbiome in the fly. For this purpose, we performed 16S rRNA sequencing on control and DSS-fed larvae. The Drosophila gut harbors relatively few bacterial species, usually belonging to the families Enterococcaceae and Lactobacillaceae from the phylum Bacillota, and to the families Acetobacteraceae and Enterobacteriaceae from the phylum Pseudomonadota (38–40). We found the bacterial composition to be changed in DSS-treated larvae compared to control larvae ( Figure 2C ). The treatment with DSS leads to a decrease in the proportion of Bacillota to Pseudomonadota ( Figure 2C , Supplementary Figure S1 and Table S1 ), which is a marker for microbial instability and is associated with chronic inflammatory diseases also in humans (41, 42). Concomitantly, the Simpson index indicates a higher dominance and lower biodiversity in DSS treated larvae compared to control treated. Further supporting this notion, both the total number of observed families (Sobs) and the Shannon-wiener H index decreases in DSS treated larvae, indicating a decline in biodiversity ( Figure 2C ).

To assess if the fly commensal microbiome, alas dysbiotic, affects Relish activation upon DSS treatment, we reared flies under axenic conditions before treating them with DSS. Interestingly, DSS did not induce Relish activation in germ-free flies compared to their conventionally reared counterparts ( Figure 2D ). Similarly, the inducibility of diptericin is impaired in flies lacking the pattern-recognizing receptor (PRR) PGRP-LC ( Figure 2E ). This indicates that DSS-induced diptericin expression is not driven by disruption of the epithelial barrier, but rather mediated by receptor activation in response to a dysbiotic microbiome. In addition, this suggests that DSS-induced Relish target gene expression is mediated via activation of the Imd/Relish pathway. Finally, to assess if the DSS-induced expression of AMPs is indeed mediated by the NF-κB Relish, we used loss-of-function (LOF) mutants of Relish and confirmed that the inducibility of diptericin expression was Relish-dependent ( Figure 2E ). Taken together, DSS-treatment induces activation of Relish in flies and can be used to induce a modest inflammation in Drosophila.

Taking advantage of the model of DSS-induced dysbiosis and inflammation, we wanted to investigate the anti-inflammatory properties of stilbenoid compounds in flies. To relate our modelling studies of stilbenes and the reference TrpA1 antagonists with the experimental data, we first determined the anti-inflammatory properties of the reference compounds A-967079 and HC-030031 in DSS-treated flies. We fed the DSS-treated flies with A-967079 and HC-030031, and both TrpA1-inhibiting drugs alleviated DSS-induced inflammation 24 hours post DSS-treatment ( Figure 3A ), suggesting that the stilbenoid compounds may also exert their anti-inflammatory properties by inhibiting the TrpA1 ion channel also in Drosophila.

Before assessing the anti-inflammatory properties of the stilbenoids, we first fed flies with the stilbenoids to investigate if the treatment alone activates NF-κB in Drosophila. We analyzed four different stilbenoid compounds, the compounds modelled together with Drosophila TrpA1 pinosylvin (PS) and pinosylvin monomethyl ether (PSMME), as well as the stilbenoid glucosides isorhapontin (4,5’-dihydroxy-3-methoxy-3’-glucopyranosylstilbene) and astringin (3,4,3’,5’-tetrahydroxystilbene 3’-glucoside). When used at a concentration of 100 µM, none of the tested stilbenoids induced inflammation after 24 hours of feeding. On the contrary, treatment with 100 µM PS, PSMME and isorhapontin reduced basal Relish target gene expression ( Figure 3B ). Astringin, on the other hand, did not seem to influence basal Relish activity ( Figure 3B ). Similar results were obtained when using a higher, 500 µM concentration of PSMME, isorhapontin and astringin ( Figure 3B ). However, a higher concentration of PS resulted in an adverse spontaneous increase of Relish target gene expression ( Figure 3B ).

To assess the anti-inflammatory effect of stilbenes, we fed larvae with DSS for 3 hours, after which they were allowed to feed on control food or food supplemented with stilbenoids for 24 hours. While DSS-treated flies still expressed diptericin 24 hours post DSS-treatment in control conditions, PS and PSMME, were able to alleviate the DSS-induced inflammation ( Figure 3C ). However, isorhapontin had no alleviating effect on DSS-induced inflammation and astringin seemed to have an opposite effect ( Figure 3D ). As both isorhapontin and astringin were unable to reduce DSS-induced inflammation the compounds were excluded from further experiments.

To assess if the anti-inflammatory effect of PS and PSMME is indeed mediated via the TrpA1 channel in vivo, we investigated the ability of the stilbenoids to alleviate inflammation in TrpA1 LOF TrpA1¹ -mutant flies. We first ensured that the larvae of TrpA1 LOF flies responded to DSS-treatment by inducing expression of diptericin similarly as wildtype flies ( Supplementary Figure S2 ). At the same time this indicates that the DSS-induced Relish activation is TrpA1-independent. When we next treated the DSS-fed TrpA1-mutant larvae with stilbenoid compounds, PS and PSMME lost their anti-inflammatory properties observed in control larvae ( Figure 4A ). Finally, the DSS-induced diptericin expression could not be alleviated by feeding TrpA1 LOF flies with the known antagonists of mammalian TrpA1, A-967079 and HC-030031 ( Figure 4B ), hence, strengthening the functional role of TrpA1 in the immune response during DSS-induced intestinal inflammation ( Figure 4C ).

Drosophila melanogaster were maintained at 25°C with a 12 h light–dark cycle on Nutri-fly BF (Dutscher Scientific, Essex, UK). Canton^S wildtype flies and diptericin-lacZ reporter line and balancer lines fly lines were kindly provided by Prof. Pascal Meier and Dr. François Leulier (52, 53). The Drosophila fly lines TrpA1¹ (#36342), PGRP-LC^Δ5 (#36323), w:Rel^E20 (stock #9457) were obtained from the Bloomington stock center.

Early Drosophila 3^rd instar larvae were fed 40 kDa DSS (TdB Consultancy AB, Uppsala, Sweden) mixed in the fly food. 40 kDa DSS has been shown to induce the most severe colitis in mammalian model organism (54). For feeding with DSS 3-10 larvae were placed in 2 ml collection tubes with food (1-2 ml) containing DSS and sealed with pieces of Drosophila plugs (Genesee Scientific, California, USA). For qPCR experiments, more than 3 larvae were fed for 3 h with indicated DSS concentrations (1%, 5% or 10% w/v) and allowed to recover for at least 1 h before freezing. As DSS has been shown to compromise qPCR results by interfering with the activity of reverse transcriptase during cDNA synthesis and the polymerase during qPCR (55), we decided to use a one-hour recovery post DSS-treatment in all qPCR-experiments. For sequencing of the 16S rRNA gene larvae were fed with 10% w/v DSS for 3 h in room temperature with a 2 h recovery.

PS and PSMME were isolated from a mixture of Norway spruce and Scot pine knotwood using a batch reactor. In the first step of isolation, the wood materials were boiled in ethanol for about two hours. After cooling, the ethanol solution was filtrated and then condensed by distillation (56). Further, the ethanol extract was dissolved in Toluene and stirred in a dark place at room temperature to remove the polymers that could interfere with the purification process. After two days, it was filtrated, and the solid residue was returned to the flask. This procedure was repeated two more times. The final filtrate was evaporated with a rotary evaporator at 40℃. The extract was subjected to a normal phase column chromatography (Silica gel 60, 0.040–0.063 mm, Merck, Darmstadt, Germany) and eluted with petroleum ether (peth) and ethyl acetate (EtOAc) as a solvent system. PSMME was separated in a proportion of 80:20 (peth/EtOAc). The remaining obtained fractions contained PS, resin acids and hydroxymatairesinol (HMR) were subjected to the second normal phase column and eluted with 100% chloroform (CHCl3). The procedure continued by using a gradient of CHCl3/MeOH (98:2) and the fractions were collected based on TLC (Silica gel 60 F254, Merck, Darmstadt, Germany) profile. All the organic solvents in analytical grade (99.9%) were purchased from Sigma-Aldrich (St. Louis, MO, USA). The purity of the fractions containing PSMME and PS was analyzed by GC-MS after silylation ( Supplementary Figure S3, S4 ). The purity was determined to be >95% ( Supplementary Figure S5, S6 ). The silylation reagents hexamethyldisiloxane (HMDS) and trimethylchlorosilane (TMSCl) were purchased from Sigma-Aldrich (St. Louis, MO, USA) and pyridine was obtained from VWR (Fontenay-sous-Bois, France). The GC-MS instrument was Agilent 5975C TAD series GC/MSD system (Stevens Creek, Santa Clara, CA, USA). The stilbenoid glucosides (astringin and isorhapontin) were extracted from fresh inner bark of Norway spruce by acetone and further purified by column chromatography over 95% and analyzed by GC-MS and NMR described in detail in our previously reported method (57).

Similarly as with DSS feeding, 3-10 larvae were fed 24 h with indicated concentrations (100 µM or 500 µM) of stilbenoid compounds PS, PSMME, isorhapontin and astringin, or 200 µM of TrpA1 antagonists A-967079 (Sigma-Aldrich) and HC-030031 (Sigma-Aldrich) mixed in 1-2 ml fly food. For experiments with DSS treatment, larvae were moved from DSS-containing food to new tubes with stilbene/antagonist-containing food.

Drosophila larvae were homogenized using QIAshredder (QIAGEN) and total RNA was extracted with RNeasy Mini Kit (QIAGEN) and cDNA was synthesized with SensiFast cDNA synthesis kit (Bioline, London, UK) according to the manufacturers’ protocols. qPCR was performed using SensiFast SYBR Hi-ROX qPCR kit (Bioline). rp49 was used as a housekeeping gene for ΔΔCt calculations. The following gene-specific primers were used to amplify cDNA: diptericin (5’-ACCGCAGTACCCACTCAATC-3’, 5’-ACTTTCCAGCTCGGTTCTGA-3’), rp49 (5’-GACGCTTCAAGGGACAGTATCTG-3’, 5’-AAACGCGGTTCTGCATGAG-3’).

3^rd instar fly larvae were dissected in PBS and fixed for 15 minutes with PBS containing 0.4% glutaraldehyde (Sigma-Aldrich) and 1 mM MgCl₂ (Sigma-Aldrich). The samples were washed with PBS and incubated with a freshly prepared staining solution containing 5 mg/ml X-gal (5-Bromo-4-chloro-3-indolyl-β-D-galactopyranoside), 5 mM potassium ferrocyanide trihydrate (Sigma-Aldrich), 5 mM potassium ferrocyanide crystalline (Sigma-Aldrich) and 2 mM MgCl₂ in PBS at 37°C. After washing with PBS, the samples were mounted using Mowiol (Sigma) and imaged with brightfield microscopy (Leica, Wetzlar, Germany).

Genomic DNA was isolated from 40 3^rd instar larvae using a modified protocol for the QIAamp DNA mini kit (QIAGEN) (58). Larvae were surface sterilized by vortexing them twice in 2% active hypochlorite and sterile H₂O. The efficiency of the washes was confirmed by 16S PCR of water from the last wash step. Larvae were homogenized in lysis buffer containing 20 mM Tris, pH 8.0, 2 mM EDTA, 1.2% Triton X-100 and 20 mg/ml lysozyme and incubated 90 min at 37°C. 200 µl AL buffer (QIAamp DNA mini kit) with 20 µl proteinase K were added and the lysate was incubated 90 min at 56°C. Subsequent extraction was performed according to manufacturer’s protocol. Amplification and Illumina MiSeq sequencing of the V1-V3 region of the 16S rRNA gene, as well as selection of operational taxonomic units (OTUs) and taxonomy assignment of OTUs was done using Eurofins Genomics InView Microbiome Profiling 3.0 service. The data is deposited at NCBI, BioProject ID: PRJNA1005106. The proportion of Wolbachia species have been omitted from the bar graph in Figure 2C for easier comparison of bacterial species residing in the gut lumen. However, Wolbachia is included in Figure 2C , Supplementary Figure S1 and Supplementary Table S1 .

The fruit fly (Drosophila melanogaster) TrpA1 (dTrpA1) sequence was obtained from the UniProt knowledgebase (UniProtKB - Q7Z020). A Basic Local Alignment Search Tool (BLAST) (59) search for the sequence was run to find a suitable template in the Protein Data Bank (PDB) (60) for comparative modelling of its three-dimensional (3D) structure. Among protein structures with similar E-values (0.0) and Sequence Identity (35.81%), the human TrpA1 crystal structure (PDB ID: 6V9V) (61) was selected due to the highest resolution (2.60 Å) and the greatest sequence coverage (residues 1 – 1119). The 3D structure model of dTrpA1 was generated using Modeller (v. 9.24) (62). The modelling alignment was created with Clustal Omega (63) and manually curated. Out of the 20 generated alternative models, the one with the best Discrete Optimized Protein Energy (DOPE) score (64) was selected. The model was further evaluated by superimposition on the template structure using PyMOL (The PyMOL Molecular Graphics System, Version 4.6, Schrödinger, LLC) and the stereochemical quality of the model was verified with MolProbity (65).

The structures of natural stilbenes and the reference compounds were prepared using the LigPrep module of Maestro (Release 2020-2: Schrödinger, LLC, New York, NY, 2020) software and the Protein Preparation Wizard of Maestro (66) was used to minimize the dTrpA1 model using the OPLS3e force field (67) and the RMSD of 0.3 Å for heavy atoms as the convergence criteria. Two docking sites were defined based on the previously reported TrpA1 antagonist binding pockets (28) using the Receptor Grid Generation tool of Maestro: (i) A-967079 binding pocket around the amino acids V431, L432, F435, F468 and M471; and (ii) HC-030031 binding site around W268, Q413, Q538, H542 and A530. The stilbene compounds were docked at these alternative sites with the GLIDE docking tool (68–70) of Maestro, using the extra precision (XP) mode with flexible ligand sampling. A maximum of five poses per ligand were ranked based on the Glide XP docking score (XP Gscore) value. To allow for more accurate evaluation of the predicted docking poses, binding free energy calculations with the Prime/MM-GBSA module of Maestro (71) were carried out for the best-docked pose (according to the XP Gscore and the observed binding interactions) of each compound using the VSGB 2.0 solvation model (72) and the OPLS3e force field (67) and allowing the residues within 5 Å from the ligand to move.

The simulation systems (consisting of the membrane embedded dTrpA1 and the receptor-bound ligand in explicit solvent) were created with the System Builder tool of the Desmond module (Schrödinger Release 2020-2: Desmond Molecular Dynamics System, D. E. Shaw Research, New York, NY, USA, 2020. Maestro-Desmond Interoperability Tools, Schrödinger, New York, NY, USA, 2020) (73) using TIP3P water (74) as the solvent model and POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) as the membrane model. The systems were neutralized by adding Na⁺ counter ions. After the system relaxation, the production simulations were run for 100 ns at constant temperature (300 K) and pressure (1.01325 bar) according to our previously reported simulation protocol (75).

Results from qPCR were analyzed by ordinary one-way ANOVA and two-tailed Student’s t-test on the non-normalized -ΔCt values, the graphs depict relative fold induction of the target gene compared to a normalized sample (ΔΔCt). Statistical analyses were performed using GraphPad Prism version 9.5.0 for Windows (GraphPad Software, San Diego, California, USA). In figures, ns stands for p> 0.05, * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Error bars in figures specify SEM from the indicated number of independent experimental repeats with more than 3 larvae per treatment. Experiments were performed indicated number of times (n ≥ 3) and statistics were calculated for each individual experiment. For analysis of 16S rRNA sequencing data, Shannon-wiener H index was calculated according to H = -Σpi*ln(pi) and Simpson index was calculated according to D = Σni(ni-1)/N(N-1). In addition, the total number of observed families (Sobs) was calculated.