TRPA1 Mediates Aromatase Inhibitor–Evoked Pain by the Aromatase Substrate Androstenedione

Aromatase inhibitors (AI) are a mainstay in the treatment of estrogen-sensitive breast cancer in postmenopausal women (1). AIs block the activity of aromatase cytochrome P450, which transforms the androgens, androstenedione, and testosterone into the estrogens (estrone and 17β-estradiol, respectively; ref. 2), which are responsible for cancer cell replication and growth (3). Unfortunately, one-third of patients treated with AIs develop muscular and joint pain (AI-associated musculoskeletal symptoms, a condition that affects the quality of life of patients and limits adherence to AI therapy (4–6). The underlying mechanism of AIMSS is unknown and, accordingly, the treatment of AIMSS remains an unmet medical need.

Recently, we reported that the transient receptor potential ankyrin 1 (TRPA1), a cation channel highly expressed by a subpopulation of primary sensory neurons of the dorsal root ganglia (DRG; ref. 7), mediates AIMSS-like behaviors evoked by AIs in mice (8). TRPA1-expressing nociceptors contain the neuropeptides substance P (SP) and calcitonin gene–related peptide (CGRP), which mediate neurogenic inflammation (9). Exogenous compounds, including allyl isothiocyanate (AITC), and endogenously generated reactive oxygen species (ROS) and their derivatives, have been identified as TRPA1 agonists (9–12). Similar to other reactive agonists (7), highly electrophilic conjugated Michael acceptor groups of exemestane (13) and nitrile moieties of letrozole and anastrozole (14) react with the thiol groups of specific cysteine and lysine residues to trigger TRPA1 and activate nociceptors (8).

The ability to gate TRPA1 in vitro was confirmed in vivo by the observation that the pain-like behaviors evoked by AIs in mice are abrogated by genetic deletion or pharmacologic blockade of the channel (8). However, AI concentrations required for TRPA1 gating in vitro (8) are 1–2 order of magnitude higher than those found in patient plasma (15). In addition, an important proportion (30%–40%), but not all, of treated patients develop the painful condition (16, 17). These observations suggest that exposure to AIs is necessary, but not sufficient, to produce AIMSS, and that additional factors should cooperate with AIs to promote pain symptoms.

Aromatase inhibition, while reducing downstream production of estrogens, moderately increases upstream plasma concentrations of androgens, including androstenedione (ASD; ref. 18). Exemestane, a false aromatase substrate, blocks enzymatic activity by accommodating in the binding pocket that snugly encloses ASD (2). We reasoned that ASD, which retains some of the reactive chemical features of exemestane, such as the α,β-carbonyl moiety of the A ring and the ketone group at the 17 position, might target TRPA1 (Supplementary Fig. S1A). In this article, we report that ASD, unique among several steroid hormones upstream to aromatase, targets the recombinant and native TRPA1, thereby, alone or in combination with oxidative stress byproducts, contributing to dramatically lowering the concentrations/doses of letrozole required to engage TRPA1 in vitro and produce TRPA1-dependent AIMSS-like behaviors in mice.

In vivo experiments and tissue collection were carried out according to the European Union (EU) guidelines and Italian legislation (DLgs 26/2014, EU Directive application 2010/63/EU) for animal care procedures, and under University of Florence research permits #204/2012-B and #194/2015-PR. C57BL/6 mice (male, 20–25 g, 5 weeks; Envigo; N = 284), littermate wild-type (Trpa1^+/+, N = 22), and TRPA1-deficient (Trpa1^−/−; N = 22) mice (25–30 g, 5–8 weeks), generated by heterozygotes on a C57BL/6 background (B6.129P-Trpa1^tm1Kykw/J; Jackson Laboratories; ref. 19), TRPV1-deficient mice (Trpv1^−/−; B6.129 × 1-Trpv1^tm1Jul/J; N = 8) backcrossed with C57BL/6 mice (Trpv1^+/+; N = 8) for at least 10 generations (Jackson Laboratories; 25–30 g, 5–8 weeks) or Sprague–Dawley rats (male, 75–100 g, Envigo; N = 24) were used. Animals were housed in a temperature- and humidity-controlled vivarium (12-hour dark/light cycle, free access to food and water). Behavioral experiments were done in a quiet, temperature-controlled (20–22°C) room between 9 am and 5 pm, and were performed by an operator blinded to genotype and drug treatment. Animals were euthanized with inhaled CO₂ plus 10%–50% O₂.

The activating peptide of the human protease activated receptor 2 (hPAR2-AP) was synthesized by G. Cirino (University of Naples, Naples, Italy), and dissolved in distilled water. HC-030031 was synthesized as described previously (20). Letrozole, L-733,060, and CGRP_8-37 were from Tocris Bioscience. If not otherwise indicated, all other reagents were from Sigma-Aldrich. For in vitro experiments, all compounds were dissolved in 100% DMSO at 10 mmol/L concentration.

Naïve untransfected HEK293 cells (ATCCCRL-1573) were cultured according to the manufacturer's instructions. HEK293 cells were transiently transfected with the cDNAs (1 μg) codifying for wild-type (wt-hTRPA1) or mutant 3C/K-Q (C619S, C639S, C663S, K708Q; donated by D. Julius, University of California, San Francisco, CA; ref. 21) human TRPA1 (hTRPA13C/K-Q-HEK293) using the jetPRIME transfection reagent (Euroclone) according to the manufacturer's protocol. HEK293 cells stably transfected with cDNA for human TRPA1 (hTRPA1-HEK293, donated by A.H. Morice, University of Hull, Hull, United Kingdom), or with cDNA for human TRPV1 (hTRPV1-HEK293, donated by M.J. Gunthorpe, GlaxoSmithKline, Harlow, United Kingdom), or with cDNA for human TRPV4 (hTRPV4-HEK293, donated by N.W. Bunnett, Monash Institute of Pharmaceutical Sciences, Parkville, Australia), or with cDNA for both human TRPA1 and human TRPV1 (hTRPA1/V1-HEK293, hTRPA1/V1-HEK293, donated by A.H. Morice, University of Hull, Hull, United Kingdom; ref. 22) were cultured as described previously (22). Human lung fibroblasts (IMR90; ATCCCCL-186), which express the native TRPA1 channel, were cultured in DMEM supplemented with 10% FBS, 2 mmol/L glutamine, 100 U penicillin, and 100 μg/mL streptomycin. Cells were plated on glass-coated (poly-l-lysine, 8.3 μmol/L) coverslips and cultured for 2–3 days before being used for recordings. For all cell lines, the cells were used when received without further authentication. Primary DRG neurons were isolated from adult Sprague–Dawley rats and C57BL/6 or Trpa1^+/+ and Trpa1^−/− mice, and cultured as described previously (8).

Intracellular calcium was measured as reported previously (8). Results are expressed as the percentage of increase of ratio_340/380 (R_340/380) over the baseline normalized to the maximum effect induced by ionomycin (5 μmol/L; % change in R_340/380) or as the percentage of responding neurons, identified by KCl (50 mmol/L). Whole-cell patch-clamp recordings were performed as reported elsewhere (8). Peak currents were normalized to cell membrane capacitance and expressed as mean of the current density (pA/pF) in averaged results. Currents were evoked in the voltage-clamp mode at a holding potential of −60 mV; signals were sampled at 1 kHz and low-pass filtered at 10 kHz. Capsaicin (0.1 or 1 μmol/L) has been used to identify TRPV1-expressing nociceptors.

For behavioral experiments, after habituation, mice (C57BL/6, Trpa1^+/+ and Trpa1^−/−, Trpv1^+/+ and Trpv1^−/−) were randomized into treatment groups, consistent with experimental design. First, we evaluated the acute nocifensive response and mechanical allodynia evoked by ASD, H₂O₂ (20 μL, intraplantar, i.pl.), letrozole (0.5 mg/kg, intragastric, i.g.), or their vehicles [5% DMSO, isotonic saline or 0.5% carboxymethylcellulose (CMC)], respectively. α-lipoic acid (100 mg/kg, i.g.), or its vehicle (0.5% CMC, i.g.), was administered 2 hours after letrozole (i.g.). Mechanical allodynia evoked by the combination of different doses of letrozole (i.g.), ASD (20 μL i.pl.), H₂O₂ (20 μL, i.pl.), or their vehicles was studied in C57BL/6, Trpa1^+/+ and Trpa1^−/− mice, with the timing of the various pharmacologic interventions based on the timing of the responses to different stimuli. Mechanical allodynia and reduction in grip strength have been studied in C57BL/6 mice, Trpa1^+/+ and Trpa1^−/− mice treated with systemic D,L-buthionine sulfoximine (BSO; intraperitoneal, i.p.), ASD (i.p.), and letrozole (i.g.), or their vehicles (isotonic saline, 4% DMSO and 4% Tween 80 and 0.5% CMC, respectively), alone or in combination, with the timing of the various pharmacologic interventions based on the timing of the responses to different stimuli.

Immediately after intraplantar injection of the different compounds, C57BL/6, Trpa1^+/+ and Trpa1^−/− mice were placed in a Plexiglas chamber, and the total time spent licking and lifting the injected hind paw was recorded for 5 minutes, as described previously (23).

Mechanical threshold was measured in C57BL/6, Trpa1^+/+ and Trpa1^−/−, Trpv1^+/+ and Trpv1^−/− mice before (basal level threshold) and after the various intraplantar or systemic treatments by using the up- and down paradigm (24). The 50% mechanical paw withdrawal threshold response (in g) was then calculated as described previously (24, 25).

The grip strength test was performed with a grip strength meter (Ugo Basile), as reported previously (8, 26). The grip strength was measured in C57BL/6, Trpa1^+/+ and Trpa1^−/−, Trpv1^+/+ and Trpv1^−/− mice before and after the various treatments.

Sciatic nerves collected from euthanized C57BL/6 mice after various treatments were homogenized in 50 mmol/L phosphate buffer (pH 7.4) containing 5 mmol/L of sodium azide at 4°C for 60 seconds, centrifuged at 12,000 × g for 20 minutes at 4°C, and the supernatant was used to determine the H₂O₂ content. H₂O₂ levels were detected by using the phenol red-HRPO method (27, 28) corrected by protein content and expressed as μmol/mg of proteins (29).

C57BL/6 mice were anesthetized with a mixture (i.p.) of ketamine (90 mg/kg) and xylazine (3 mg/kg), and transcardially perfused with PBS followed by 4% paraformaldehyde. The sciatic nerve with the surrounding tissue was removed and embedded in paraffin. Immunofluorescence staining for 4-hydroxynonenal adducts and TRPA1 was performed as reported previously (30).

BSO was given 30 minutes before letrozole and ASD immediately after letrozole; mice were anesthetized 15 minutes after letrozole as described previously (10). HC-030031 (100 mg/kg, i.p.) or its vehicle (4% DMSO plus 4% Tween80, i.p.), or a combination of L-733,060 and CGRP_8-37, (NK₁/CGRP-RA; both 2 μmol/kg, i.v.) or its vehicle (isotonic saline), were administered 60 minutes or 15 minutes before ASD, respectively. The extravasated dye was extracted from synovial tissue of the knee joint by overnight incubation in formamide, and assayed by spectrophotometry at 620 nm, as reported previously (20).

The synovial fluid was collected from anesthetized mice treated with the combination of BSO (i.p.), letrozole (i.g.), and ASD (i.p.), or their vehicles (isotonic saline, 0.5% CMC and 4% DMSO plus 4% Tween80, respectively). BSO was given 30 minutes before letrozole and ASD immediately after letrozole. Synovial fluid was collected by instilling 3 times 0.1 mL of Hank buffer plus 10 mmol/L HEPES and 10 mmol/L EDTA in the knee 15 minutes after letrozole. The neutrophil count was performed using standard morphologic criteria on Diff-Quick stained cytospins. Data are expressed as total number of neutrophils in 100 μL of solution.

Slices (0.4 mm) of rat spinal cords were superfused with ASD or vehicle. Tissues were pre-exposed to capsaicin (10 μmol/L, 20 minutes), superfused with a calcium-free buffer containing EDTA (1 mmol/L), and pretreated with HC-030031 (50 μmol/L) or capsazepine (10 μmol/L). Superfusate fractions (4 mL) were collected at 10-minute intervals before, during, and after stimulus administration, freeze-dried, reconstituted with assay buffer, and analyzed for CGRP-like immunoreactivity (LI) as described previously (8). CGRP-LI was calculated by subtracting the mean prestimulus value from those obtained during or after stimulation. Results are expressed as femtomoles of peptide per gram of tissue. Stimuli did not cross-react with CGRP antiserum.

ASD levels were measured in mouse serum by using the Active ASD Radioimmunoassay (Beckman Coulter), a competitive RIA with a sensitivity of 0.1 nmol/L. Radiometric detection was performed using a 2470 WIZARD Automatic Gamma Counter (Perkin Elmer).

Letrozole levels were measured by LC/MS-MS. Briefly, plasma samples (50 μL) were obtained from the blood collected at different time points (1 and 3 hours) after i.g. administration of letrozole (0.1 and 0.5 mg/kg). At 50 μL of plasma sample, 200 pg of d4-letrozole was added. The sample was vortex-mixed for 10 seconds, then 50 μL of ZnSO₄ (90 mg/mL) diluted 1:4 with methanol was added, vortex-mixed for 30 seconds, and centrifuged at 12,000 rpm for 10 minutes. Supernatant (50 μL) was injected in On-line column-switching SPE (CS-SPE). The CS-SPE consists of two high-performance-liquid chromatography (HPLC) systems connected by a six-port switching valve. In the first step, analytes of interest are retained on column 1 (trapping column, SPE Strata C18 20 um, 20 × 2 mm, Phenomenex), whereas the matrix components can be washed off. In the second step, column 1 is switched in back-flush to column 2 (analytical column, LUNA, 3 um, C18 20 × 2 mm, Mercury MS Phenomenex). The mobile phases were the same for the trapping column and the analytical column, eluent A water with 0.1% formic acid and eluent B methanol. Samples were measured with a Perkin Elmer Sciex API 365 triple quadrupole mass spectrometer equipped with a Turbo IonSpray source, operating in positive ion mode, interfaced with a HPLC Perkin Elmer pump series 200. The capillary voltage was set to 5.5 kV. Heated turbo gas (450°C, air) at a flow rate of 10 L/minute was used. The ion transitions recorded in Multiple Reaction Monitoring (MRM) were m/z 286.2 → 217.2 for letrozole and 290.2 → 221.2 for d4-letrozole. A calibration curve was constructed for letrozole using the appropriate internal standard (d4-letrozole). Plasma samples (50 μL) from control mice were spiked with different concentrations of letrozole (from 2–16 ng/mL). A satisfying linearity was obtained for letrozole (r² = 0.994).

Data represent mean ± SEM. Statistical analysis was performed by the unpaired two-tailed Student t test for comparisons between two groups, and the ANOVA, followed by the Bonferroni post hoc test, for comparisons between multiple groups (GraphPadPrism version 5.00). Agonist potency was expressed as half maximal effective concentration (EC₅₀); that is, the molar concentration of an agonist producing 50% of the maximum measured effect. P < 0.05 was considered statistically significant.

In hTRPA1-HEK293 cells, but not in untransfected HEK293 cells, ASD evoked calcium responses in a concentration-dependent manner (EC₅₀, 49 μmol/L; Fig. 1A and Supplementary Fig. S1B). Responses to both ASD and the TRPA1 agonist, AITC, were abrogated by HC-030031 (Fig. 1B). Consistently, hTRPA1-HEK293 cells superfused with ASD elicited concentration-dependent inward currents (Supplementary Fig. S1E), an effect blocked by HC-030031 (Fig. 1C) and absent in untransfected HEK293 cells (Fig. 1D). In hTRPV1-HEK293 and hTRPV4-HEK293, activated by the selective TRPV1 agonist, capsaicin, or the selective TRPV4 agonist, GSK1016790A, respectively, ASD failed to evoke calcium responses or inward currents (Fig. 1E and F). Notably, hTRPA13C/K-Q-HEK293 did not respond to AITC or ASD, while they did respond to the nonelectrophilic TRPA1 agonist, menthol (Fig. 1G and Supplementary Fig. S1D; ref. 8).

In IMR90 cells, which constitutively express the TRPA1 channel (31) and do not respond to capsaicin, indicating the absence of a functional TRPV1 channel (Fig. 1H), ASD produced concentration-dependent (EC₅₀, 37 μmol/L) calcium responses that were fully and selectively inhibited by HC-030031 (Fig. 1H and Supplementary Fig. S1C). Similar results were obtained in electrophysiology experiments, where ASD activated TRPA1-mediated inward currents that were entirely and selectively abolished by HC-030031 (Fig. 1I). TRPA1 selectivity of HC-030031 in inhibiting ASD-evoked responses was supported by failure to affect responses produced by hPAR2-AP or KCl in hTRPA1-HEK293 or IMR90 (Fig. 1A–D, H, and I). Next, we wondered whether the other aromatase substrates, testosterone, or steroid hormones upstream to aromatase that maintain the α,β-carbonyl moiety of the A ring (progesterone, 17-hydroxy-progesterone), or the ketone group at the 17 position (dehydroepiandrostenedione), or other steroid hormones that retain the α,β-carbonyl moiety of the A ring (aldosterone, cortisol, corticosterone, deoxycorticosterone, 11-deoxycortisol), were able to activate hTRPA1-HEK293 cells. No hormone evoked a measurable response (Supplementary Fig. S1E).

ASD evoked concentration-dependent (EC₅₀ = 27 μmol/L) calcium responses in a subset of rat DRG neurons, identified by their ability to respond to KCl, AITC, and capsaicin as nociceptors (Fig. 2A and Supplementary Fig. S1F; ref. 23). Capsaicin-sensitive neurons that did not respond to AITC were also unresponsive to ASD (Fig. 2A). The percentages of ASD-responding and AITC-responding neurons out of the KCl-responding neurons were similar (Fig. 2B). Superimposable findings were obtained by electrophysiological recording (Fig. 3A). Surprisingly, the remarkable ASD selectivity for TRPA1 was challenged by experiments in rat DRG neurons. The selective TRPV1 antagonist, capsazepine, reduced both the calcium response and the inward currents evoked by ASD, and it abated the residual response observed in the presence of HC-030031 (Figs. 2C and D and 3A). The calcium response to ASD was, however, unaffected by the selective TRPV4 antagonist, HC-067047 (Fig. 2D). Results obtained in rat DRG neurons were replicated in mouse DRG neurons. Cells obtained from Trpa1^−/− mice exhibited a residual calcium response to ASD that, being consistently unaffected by HC-030031, was abated by capsazepine (Fig. 2E). The percentages of ASD-responding and AITC-responding neurons out of the KCl-responding neurons were similar (Fig. 2F) in DRG neurons isolated from Trpa1^+/+ mice. The percentage of neurons from Trpa1^−/− mice that exhibited a residual calcium response to ASD did not exceed the percentage obtained in neurons from Trpa1^+/+ mice (Fig. 2F).

To confirm the contribution of TRPV1 in the overall response to ASD, we used hTRPA1/V1-HEK293 cells (22), where the calcium response evoked by ASD was reduced by both HC-030031 and capsazepine, and was abated solely by the combination of the two antagonists, while responses to AITC and capsaicin were fully attenuated by respective antagonists (Fig. 2G). Thus, TRPV1, when coexpressed with TRPA1, as constitutively happens in DRG neurons, appears to contribute to the response to ASD.

CGRP release from slices of the rat dorsal spinal cord, an anatomical site enriched with terminals of TRPA1-positive peptidergic nociceptors, illustrates the ability of ASD to activate such neurons (8). The increased CGRP-LI outflow by ASD was abated by the removal of extracellular calcium, previous desensitization to capsaicin, or in the presence of HC-030031, but only partially reduced by capsazepine (Fig. 2H). Thus, ASD elicits CGRP release from a subset of TRPV1-positive neurons via a neurosecretory process, mediated by TRPA1.

To explore whether ASD cooperates with AIs and proinflammatory mediators to excite nociceptors via a TRPA1-dependent final common pathway, we used letrozole, the most prescribed AI in clinical practice (32). As a prototypical proinflammatory mediator, we selected the ROS H₂O₂, because oxidative stress is increased by breast cancer (33) and letrozole (34). In addition, the effect of letrozole (0.5 mg/kg, i.g.), at a dose that was previously shown to produce per se mechanical allodynia, was partially reduced by the antioxidant, α-lipoic acid (Supplementary Fig. S2). These observations suggest that the TRPA1-dependent letrozole-evoked mechanical hypersensitivity (8) is partially due to ROS generation, which cooperates with the anticancer drug to target TRPA1.

From the concentration–response curves of ASD, letrozole, and H₂O₂ (Fig. 3A–C), we selected subthreshold concentrations that were unable to elicit measurable inward currents in rat DRG neurons, and combined a per se inactive concentration of ASD with inactive concentrations of letrozole or H₂O₂ (ASD/letrozole or ASD/H₂O₂). Interestingly, we found that each combination evoked inward currents, which were abated by TRPA1 antagonism and reduced by TRPV1 antagonism (Fig. 3D). Finally, we identified an inactive combination of letrozole and H₂O₂ (Fig. 3E), and we found that adding an inactive ASD concentration to it triggered an inward current that was abated by HC-030031 and only partially reduced by capsazepine (Fig. 3E).

Previous in vitro findings were translated to an in vivo setting. Injection of ASD (1–10 nmol/paw) into the mouse paw did not evoke any acute nociceptive behavior (data not shown). However, 30 minutes after the injection, and for the following 2 hours, ASD produced a dose-dependent mechanical allodynia (Fig. 4A and B) that was partially and completely abrogated by capsazepine and HC-030031, respectively (Fig. 4B). H₂O₂ injected in the mouse paw produced a dose-dependent mechanical allodynia that was entirely dependent on TRPA1 (Fig. 4C). Similar to previous findings (8), systemic letrozole (0.1–0.5 mg/kg, i.g.) evoked a dose-dependent, delayed (1–6 hours) mechanical allodynia that was abated by HC-030031 and unaffected by capsazepine (Fig. 4D). The combined administration of allodynia-evoking doses of ASD and letrozole produced an exaggerated pain-like response (Supplementary Fig. S3A).

Next, we found doses of ASD and letrozole, or H₂O₂ and ASD, which, although per se ineffective, when given in combination, lowered the threshold for eliciting mechanical allodynia (Supplementary Fig. S3B and S3C). Finally, we identified ineffective combinations of ASD/letrozole, letrozole/H₂O₂, or H₂O₂/ASD that, when given simultaneously (letrozole/ASD/H₂O₂), caused mechanical allodynia (Fig. 4E). This response was partially reduced by capsazepine, completely reverted by HC-030031 (Fig. 4F), and absent in Trpa1^−/− (Fig. 4G).

In mice, letrozole (0.5 mg/kg, i.g.) has been reported to evoke TRPA1-dependent mechanical allodynia and a decrease in grip strength, two effects reminiscent of AIMSS (8). The same dose of letrozole (0.5 mg/kg, i.g.; ref. 8) increased H₂O₂ in the sciatic nerve tissue and slightly augmented ASD serum levels (Fig. 5A and C). BSO, by inhibiting γ-glutamylcysteine synthetase, causes systemic depletion of glutathione, and the ensuing increase in ROS (35). BSO (800 mg/kg, i.p.) increased mechanical allodynia and decreased forelimb grip strength through a TRPA1-dependent mechanism (Supplementary Fig. S4A), while, at the dose of 400 mg/kg, it slightly increased H₂O₂ in the sciatic nerve tissue (Fig. 5C) without affecting pain-like behaviors (Supplementary Fig. S4A). Finally, systemic administration of ASD (2 μg/kg, i.p.) caused, in mice, mechanical allodynia and reduced forelimb grip strength via TRPA1, with a partial contribution of TRPV1 (Supplementary Fig. S4B). ASD (2 μg/kg, i.p.) also increased H₂O₂ levels in the sciatic nerve (Fig. 5C).

To better understand the contribution of ASD and oxidative stress to the AIMSS-like behaviors, a low dose of ASD (0.2 μg/kg, i.p.) that failed to affect H₂O₂ generation (Fig. 5C), as well as mechanical allodynia and forelimb grip strength, (Supplementary Fig. S4B) was used. This same dose slightly increased hormone plasma concentration to levels comparable with those produced by a dose of letrozole (0.5 mg/kg, i.g.) that caused pain-like behaviors (Fig. 5A and Fig. 4D). In mice, 1 hour after oral administration of letrozole (0.1 and 0.5 mg/kg), drug concentrations in plasma were 13.5 ± 2.0 ng/mL (n = 4) and 55.3 ± 4.8 ng/mL (n = 4), respectively, whereas 3 hours after dosing, plasma concentrations were 8.41 ± 1.0 ng/mL (n = 4) and 45.4 ± 6.11 ng/mL (n = 4), respectively. Notably, the plasma concentration of 0.5 mg/kg letrozole measured at 1 hour after dosing was similar to that found previously (8). Systemic BSO (400 mg/kg, i.p.), letrozole (0.1 mg/kg, i.g.), and ASD (0.2 μg/kg, i.p.), which per se, or in combinations (letrozole/ASD, BSO/ASD or BSO/letrozole), did not affect behavioral responses, when given simultaneously caused remarkable mechanical allodynia and decreased forelimb grip strength (Fig. 5D and E). The triple combination of BSO, letrozole, and ASD increased H₂O₂ levels in the sciatic nerve (Fig. 5C). However, the increase equals that evoked by BSO (400 mg/kg) alone. Finally, the remarkable increase in 4-hydroxynonenal staining in the sciatic nerve allowed us to localize the oxidative stress generation within the neural structure (Fig. 5H). Behavioral responses evoked by the triple combination were partially and totally reverted by capsazepine and HC-030031, respectively (Fig. 5D and E), and were absent in Trpa1^−/− mice, but unaffected in Trpv1^−/− mice (Fig. 5F and G).

Finally, the combination of BSO (400 mg/kg, i.p.), letrozole (0.1 mg/kg, i.g.), and ASD (0.2 μg/kg, i.p.), which produced AIMSS-like behaviors, increased Evans blue dye extravasation in the synovial tissue and the number of neutrophils in the synovial fluid of mouse knee joint (Fig. 5I and J). Both responses were reduced by pretreatment with HC-030031 or a combination of L733,060 and CGRP_8-37 (both 2 μmol/kg, i.v.; Fig. 5I and J). These findings indicate that the letrozole/ASD/BSO combination via TRPA1 promotes two typical neurogenic inflammatory responses, such as plasma protein and neutrophil extravasation (9).

We found that ASD, unique among several steroid hormones, activates TRPA1, thereby promoting AIMSS-like responses. ASD behaves as a TRPA1 agonist across species, as it engages both the recombinant and native human channel and the rat and mouse channel. TRPA1 activation by ASD, similar to other electrophilic agonists (7), requires the presence of three cysteine (C619, C639, C663) and one lysine (K708) key residues. Furthermore, ASD exhibits a peculiar selectivity profile. Whereas in cells expressing only one channel (transfected HEK293 cells, IMR90 human fibroblasts; refs. 31, 36), ASD is a selective TRPA1 agonist, surprisingly, in rodent DRG neurons, which express multiple TRP channels, TRPV1 contributes to TRPA1-dependent ASD-evoked responses. Notably, DRG neurons from Trpa1^−/− mice maintained a residual responsiveness to ASD, which was abolished by capsazepine. One possible explanation for this unexpected finding is that the TRPA1 protein remaining after the homologous recombination, while lacking the domain required for channel activation by most agonists (37), including ASD itself, maintains the domain essential for TRPV1-dependent ASD activity. TRPV1 contribution to TRPA1-mediated responses evoked by ASD or other chemicals (22) was confirmed in cells expressing the recombinant forms of both the TRPA1 and TRPV1 (hTRPA1/V1-HEK293) channels. Formation of heterotetramers (38, 39), or intracellular calcium movement initiated by TRPA1 gating that results in a secondary TRPV1 activation (40), may explain how ASD heterologously interregulates channel activities (41, 42). Notwithstanding, this peculiar selectivity pattern was not replicated in vivo as TRPA1 inhibition was sufficient to totally prevent ASD actions.

Tenosynovitis and joint swelling are symptoms reported by patients treated with AIs (43). However, association with proinflammatory markers, including cytokines, such as IL6, has been excluded (44). The ability of ASD to release sensory neuropeptides or, in combination with letrozole and BSO, to provoke TRPA1-dependent SP/CGRP release, edema and neutrophil infiltration in the knee joint, suggests that neurogenic inflammation mediates the inflammatory component of AIMSS. The recent observation that, due to their electrophilic and reactive properties, exemestane, anastrozole, and letrozole target TRPA1, thus evoking pain-like responses and neurogenic inflammation, supported the hypothesis that channel activation in peptidergic nociceptors promotes AIMSS (8). However, while the three AIs target TRPA1 with remarkable selectivity, they exhibit low potency at both human and rodent channels (8). Notably, the peak plasma concentration of letrozole measured in mice in the current study after the 0.5 mg/kg dose (∼194 nmol/L) is close to the maximum concentration reported in human plasma after a single therapeutic dose (2.5 mg; ∼128 nmol/L; ref. 45). It must be underlined that the plasma concentrations found in mice and humans are 50–400 times lower than the threshold concentrations of letrozole able to gate in vitro the mouse, rat, or human TRPA1 (>10 μmol/L). This gap between plasma levels and threshold concentrations at TRPA1 argues against the hypothesis that AIs per se cause AIMSS. Elevated tissue concentrations due to high AI volumes of distribution (46) may still be insufficient for effective channel gating. Lack of evidence associating increased AI plasma levels with pain symptoms (45) further weakens the hypothesis that AIMSS results from an exclusive AI action. Similarly, attempts to associate serum ASD concentrations with AIMSS have failed (18). Thus, a direct cause-and-effect relationship between plasma levels of AIs or ASD and AIMSS has not been shown.

TRPA1 is amenable to sensitization by a variety of endogenous proinflammatory and proalgesic mediators (47), and we previously reported that H₂O₂ or PAR2 stimulation exaggerated TRPA1-dependent responses evoked by AIs (8). The present novel finding that the aromatase substrate, ASD, activates TRPA1 proposes a novel paradigm to explain AIMSS generation. Multiple factors, concomitantly occurring in patients with breast cancer treated with AIs, may cooperate to engage TRPA1, thus causing AIMSS-like behaviors. Letrozole (and probably other AIs) is the essential, although per se ineffective, initiating stimulus. Letrozole slightly augments ROS (34) and ASD concentrations (18). Breast cancer (33) or incidental inflammatory processes (48) may boost oxidative stress, thus increasing the possibility of the simultaneous presence of the three TRPA1 stimulants. However, the current hypothesis does not exclude that additional agents, able to activate or sensitize the TRPA1, may act along with AIs and ASD to reach the threshold for AIMSS generation. The current study, while showing that TRPV1 signaling negligibly contributes to ASD-evoked AIMSS-like behaviors, robustly underscores the paramount role of TRPA1. Thus, TRPA1 blockade by both new compounds currently under clinical scrutiny and old medicines recently identified as TRPA1 antagonists (49) may represent a new frontier to treat or prevent AIMSS.

No potential conflicts of interest were disclosed.

Conception and design: F. De Logu, S. Materazzi, T. Susini, P. Geppetti, S. Benemei

Development of methodology: F. De Logu, R. Tonello, S. Materazzi, C. Fusi, S. Li Puma, I.M. Marone, A.M. Morice, T. Susini, G. Moneti

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. De Logu, R. Tonello, S. Materazzi, R. Nassini, E. Coppi, S. Li Puma, I.M. Marone, L.R. Sadofsky, A.M. Morice, G. Moneti

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. De Logu, R. Tonello, S. Materazzi, R. Nassini, C. Fusi, G. Moneti, M. Di Tommaso, S. Benemei

Writing, review, and/or revision of the manuscript: R. Tonello, S. Materazzi, R. Nassini, L.R. Sadofsky, A.M. Morice, T. Susini, M. Di Tommaso, P. Geppetti, S. Benemei

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Tonello, A. Terreni

Study supervision: A.M. Morice, P. Geppetti, S. Benemei

The authors are grateful to M.J. Gunthorpe (GlaxoSmithKline) for the hTRPV1-HEK293 cells and N.W. Bunnett (Monash Institute of Pharmaceutical Sciences, Australia) for the hTRPV4-HEK293 cells, D. Julius (University of California, San Francisco, CA) for the human TRPA1 wild-type and human TRPA1 mutant (C619S, C639S, C663S, K708Q) cDNAs and G. Cirino (University of Naples, Italy) for providing hPAR2-AP, D. Preti (University of Ferrara, Italy) for providing HC-030031. They also acknowledge M. Lokken for her expert English revision.

This study was supported by Istituto Toscano Tumori (ITT), grant 2014, (P. Geppetti); Regione Toscana, grant Nutraceuticals 2014, “POFCADT” (P. Geppetti), and AIRC-Italy, grant 2012 (R. Nassini).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.