LPS depolarizes TRPA1-expressing nociceptors

Simultaneous monitoring of intracellular calcium levels and electrical activity in cell-attached mode revealed that calcium elevations caused by LPS in nodose neurons occurred during brief periods of action potential firing (Fig. 2a). Notably, excitatory effects of LPS on electrical activity were similar in WT and Tlr4 KO mice (Fig. 2b). During current-clamp recordings, a fraction (5/8) of cultured nodose neurons expressing TRPA1 channels depolarized rapidly (mean±s.e.m=16±4mV), firing trains of action potentials in response to LPS (Fig. 2c). In contrast, these responses were not observed (0/10) (P=0.006, Fisher’s exact test) in neurons insensitive to TRPA1 agonists (Fig. 2c,d).

Figure 2

E. coli LPS excites TRPA1-expressing sensory neurons.

To analyse whether LPS acts on the peripheral terminals of primary sensory neurons expressing TRPA1, we recorded the impulse activity of TG nerve afferents innervating the mouse tongue ex vivo. Application of LPS onto the receptive field of sensory units, identified as nociceptive by their response to heat or capsaicin, produced a significant increase in the firing frequency in 50% of them, precisely those that were also activated by the TRPA1 agonist allyl isothiocyanate (AITC; Supplementary Fig. 2). This indicates that LPS activates the fraction of peripheral nociceptor fibres expressing TRPA1.

Taken together, these results demonstrate that the chemosensitive cation channel TRPA1, but not the canonical LPS receptor TLR4, is critical for the acute excitatory effects of LPS on nociceptor sensory neurons.

LPS activates TRPA1 channels

Next, we determined whether LPS stimulates TRPA1 channels by measuring [Ca2+]i changes in Chinese hamster ovary (CHO) cells expressing recombinant mouse TRPA1 channels (mTRPA1 cells)13. LPS induced a dose-dependent and reversible increase of [Ca2+]i with an effective concentration of 3.3±0.5μgml−1 (mean±s.e.m for all values) (Fig. 3a,b). LPS did not induce Ca2+ responses in Ca2+-free (plus 1mM EGTA) solutions (ratio change=−0.0042±0.0008, n=46, P=1.0; paired t-test), or in non-transfected CHO cells (Fig. 3a), indicating that LPS induces Ca2+ influx through mTRPA1 channels in the plasma membrane. Importantly, LPS also stimulated human embryonic kidney (HEK)-293 cells transfected with human TRPA1 (Supplementary Fig. 3a), but was ineffective, at 3μgml−1, on three other human TRP channels known to be expressed in primary sensory neurons, namely, the vanilloid receptor TRPV1, the menthol receptor TRPM8 and the Δ9-tetrahydrocannabinol-activated receptor TRPV2 (Supplementary Fig. 3b–d).

Figure 3

LPS stimulates recombinant mouse TRPA1 channels.

Stimulation of mouse TRPA1 by LPS was confirmed in whole-cell patch-clamp recordings in mTRPA1 cells; extracellular application of LPS (20μgml−1) induced a significant increase in the amplitude of non-selective cation currents (reversal potential=−2.7±1.9mV, n=7; Fig. 3c,d). In contrast, LPS was ineffective in non-transfected CHO cells (Fig. 3d, Supplementary Fig. 4a,b). The maximal current increase induced by the potent TRPA1 agonist AITC (20μM) was approximately eightfold larger than that elicited by 20μgml−1 LPS (Fig. 3d), indicating that LPS has lower efficacy. Moreover, pre-application of the TRPA1 inhibitor HC-030031 fully prevented the activating effects of LPS (n=5). The inhibition of TRPA1 whole-cell current by HC-030031 was reversible on wash and could be reinstated on reapplication of the blocker (Supplementary Fig. 4c). At −75mV, current density averaged −8.2±1.7pApF−1 in HC-030031 plus LPS and −39.4±6.5pApF−1 after HC-030031 wash (n=7, P<0.001, paired t-test).

Next, we explored the mechanism(s) through which LPS stimulates TRPA1. LPS could potentially activate TRPA1 by modulating [Ca2+]i (refs 19,20). However, LPS was able to stimulate TRPA1 in Ca2+-free conditions (Fig. 3e, Fig. 4). Alternatively, TRPA1 could be activated by an endogenous electrophilic compound that could be eventually produced on LPS application. However, LPS induced a significant activation of the AITC-insensitive TRPA1-mutant C622S in HEK-293 cells (increase in Fura-2 ratio of 0.38±0.03, n=36, P<0.01, paired t-test). Analysis of the macroscopic TRPA1 currents activated by LPS in Ca2+-free conditions indicated a slowing of channel deactivation, an acceleration of channel opening (Fig. 3e, Supplementary Fig. 5) and a concomitant leftward shift in the voltage dependence of channel activation (Fig. 3f). These data demonstrate that, as occurs with other TRPA1 activators21,22, LPS acts as a gating modifier of TRPA1, destabilizing the closed state(s) and stabilizing the open conformation(s).

Figure 4

LPS activates TRPA1 channels in a membrane-delimited manner.

LPS activates TRPA1 channels in a membrane-delimited manner

To gain further insight into the mechanism of TRPA1 activation by LPS, we performed patch-clamp recordings in cell-free outside-out patches in the absence of intracellular and extracellular Ca2+. Under these conditions, extracellular application of LPS enhanced TRPA1 currents (6.1±1.6-fold increase at −75mV; P=0.018, n=5, paired t-test; Fig. 4a,b). The single-channel conductance, estimated from linear fits to currents evoked during voltage ramps, was nearly identical in control and in the presence of LPS (96±3pS and 94±2pS, respectively; n=5, Fig. 4b), and the TRPA1 antagonist HC-030031 induced a powerful inhibition of LPS-induced single-channel currents (97±2%, n=5, Fig. 4a,b). In outside-out patches held at +75mV, with Ca2+-free extracellular and intracellular solutions, the basal open probability was on average 0.014±0.004 (Fig. 4c), while mean closed time and mean open time were 2.0±1.2s and 8.1±1.0ms, respectively, (n=5). In these patches, LPS induced a 5.7±1.6-fold increase in the open probability (P=0.023), a 4±1-fold reduction of the mean closed time (P=0.0016) and a 1.8±0.4-fold increase in the mean open time (P=0.04; n=5, paired t-test; Fig. 4d). Of note, we recorded half-sized openings, which may reflect the incidence of subconductance open state(s).

LPS and endogenous TRPA1 agonists have synergistic effects

One fundamental aspect of TRPA1 function is to act as molecular integrator for multiple irritants and inflammatory mediators15,23. Thus, it was important to determine whether the effects of LPS on TRPA1 could be potentiated by other algesic compounds. At sites of infection and inflammation, various endogenous molecules are generated that could sensitize the response of TRPA1 to LPS. In particular, inflammation increases the local levels of 4-hydroxy-2-nonenal (HNE) and other lipid peroxidation products that are potent stimulators of TRPA1 (refs 24,25). Thus, we hypothesized that LPS and HNE may act synergistically on TRPA1. Indeed, the co-application of low concentrations of HNE and LPS produced a strong potentiation of TRPA1 activity, resulting in much larger currents than the sum of each agonist applied alone (Supplementary Fig. 6). As expected for a TRPA1-dependent effect, these currents were fully suppressed by HC-030031 (Supplementary Fig. 6).

The shape of lipid A determines TRPA1 activation by LPS

Suspecting that LPS insertion in the plasma membrane is responsible for its stimulatory action on TRPA1 channels, we tested the effect of lipid A, the membrane-anchored lipid moiety of LPS26. Application of 50μgml−1 diphosphoryl lipid A from E. coli induced [Ca2+]i responses in mTRPA1 cells, but not in control cells (Fig. 5a). Moreover, application of lipid A to nodose neurons produced robust Ca2+ elevations exclusively in a subpopulation (9/41, 22%) of CA-activated neurons (Fig. 5b). Interestingly, synthetic monophosphoryl lipid A also activated mTRPA1 cells (Supplementary Fig. 7) ruling out a biological contaminant during LPS purification as a possible source of the effect.

Figure 5

The shape of lipid A determines TRPA1 activation by LPS.

The intrinsic conformation of lipid A is critical for LPS bioactivity (for example, induction of cytokine production), with conical shapes conferring more activity than cylindrical ones27,28. This conformation is largely determined by the number of charged phosphate groups and acyl chains, and varies among different bacterial strains26. We found that LPSs with asymmetrical hexa-acyl lipid A (E. coli, Salmonella typhimurium and Klebsiella pneumoniae) induced the strongest TRPA1 responses (Fig. 5c,d). In comparison, LPSs with asymmetrical penta-acyl lipid A (Serratia marcescens, Pseudomonas aeruginosa) induced smaller effects on TRPA1. Finally, LPS from Neisseria meningitidis, with symmetrical hexa-acyl lipid A, and from S. minnesota, which is known to form symmetrical lamellar structures, had virtually no effect on TRPA1. Very similar results were obtained in cultured TG neurons: the percentage of neurons responding to LPS followed the sequence (E. coli>S. marcescens>S. minnesota) and the great majority of LPS responders had a nociceptive phenotype (Fig. 5e,f). These results strongly suggest that the capacity of a given LPS to activate TRPA1 correlates with its ability to induce mechanical alterations in the plasma membrane. We also found that trinitrophenol (TNP), a negatively charged amphiphatic compound inducing mechanical tension of the lipid membrane, and known to modulate the gating of mechanosensitive channels29, including TRPA1 (ref. 30) also potentiated LPS activity. Pretreatment with TNP strongly potentiated [Ca2+]i responses to low concentrations of LPS in mTRPA1 cells (Supplementary Fig. 8a–d).

LPS effects on TRPA1 are TLR4 independent

Generally, LPS effects on neurons have been linked to activation of the TLR4 signalling pathway10. However, as shown earlier, we found that the excitatory effects of LPS were still observed on sensory neurons isolated from Tlr4 KO mice. We also tested CLI-095, a novel specific inhibitor of TLR4 signalling31. As shown in Fig. 6a–c, pretreatment for 10min with 1μM CLI-095 did not affect the subsequent response to LPS in nodose neurons; neither the percentage of responding neurons nor the amplitude of the response was significantly affected. Similar results were obtained in TRPA1-expressing CHO cells (Supplementary Fig. 9). In addition, CHO cells transfected with human TLR4 did not show a TRPA1-like current when acutely exposed to LPS (Supplementary Fig. 10). Altogether, these results strongly support a TLR4-independent mechanism of action of LPS on TRPA1 channels.

Figure 6

The effects of LPS on TRPA1 channels are independent of TLR4 signalling and require lipid A activity.

We took additional steps to verify that activation of TRPA1 was truly mediated by LPS. To minimize the possible role of a contaminant, we tested ultrapure forms of LPS (see Methods). As shown in Supplementary Fig. 11, ultrapure LPS from E. coli (2μgml−1) also activated TRPA1-expressing cells. Moreover, the ultrapure form of S. minnesota R595, at a 10 times higher concentration, showed almost no response, similar to the results obtained from a different source (see Fig. 5c,d). Strikingly, an ultrapure LPS from Rhodobacter sphaeroides, which acts as a competitive antagonist of TLR4 (ref. 32), also activated TRPA1-expressing cells (Supplementary Fig. 11), strongly suggesting a TLR4-independent mechanisms of action. In contrast, the same preparations of LPS had the expected effects on human TLR4 signalling (Supplementary Fig. 11d).

Polymyxin B is a cationic cyclic polypeptide antibiotic that is widely used to neutralize the effects of endotoxin contamination in biological samples33. The action of polymyxin B is based on its binding, by electrostatic interaction, to the anionic lipid A. Incubation of LPS with polymyxin B completely eliminated the activation of TRPA1 by LPS (Fig. 6d,e). In contrast, activation of TRPA1 by AITC was minimally reduced (Fig. 6f). This result suggests that the inhibitory action of polymyxin B is not directly on the channel (for example, pore block) and further implicates lipid A in the effects observed.

Acute neuropeptide release by LPS depends on TRPA1 activity

TRPA1 participates in neurogenic inflammation and vasodilation34,35,36. To test whether activation of peptidergic sensory terminals by LPS leads to neuropeptide release and neurogenic inflammation14,37, we used an isolated mouse trachea preparation. The trachea is densely innervated with peptidergic vagal fibres that release calcitonin gene-related peptide (CGRP) in response to capsaicin38,39. Brief incubation of tracheal halves with acrolein, an electrophilic TRPA1 activator and product of oxidative stress, also present in fumes and as a toxic metabolite of chemotherapy agents15, produced a dose-dependent release of CGRP (Supplementary Fig. 12). Application of LPS (30–100μgml−1) to tracheal halves induced a modest but significant CGRP release (Fig. 7a). In addition, a brief pretreatment of the trachea with the membrane crenator TNP produced a massive potentiation of the CGRP release evoked by LPS (P<10−4) and by acrolein (P<0.01; Fig. 7a). Notably, responses to individual agonists, or potentiated response to LPS in the presence of TNP were fully abrogated in Trpa1 KO mouse tracheas (Fig. 7a), indicating that they depend strictly on activation of TRPA1 channels. We also observed a fourfold potentiation of CGRP release from the trachea by LPS in the presence of the endogenous TRPA1 agonist HNE (Fig. 7b), fully consistent with the synergistic action of both agonists on TRPA1 whole-cell currents as we have shown (Supplementary Fig. 6). Collectively, these results indicate that LPS, both alone and in combination with other endogenous inflammatory mediators, can activate TRPA1 in sensory nerve terminals leading to acute neurogenic inflammation.

Figure 7

Neurogenic inflammation and pain by LPS depend on TRPA1 activity.

Neurovascular responses to LPS involve activation of TRPA1

Systemic bacterial invasion can cause massive peripheral vasodilation, leading to a severe drop in blood pressure, a life-threatening condition known as septic shock40. Considering that chemical stimulation of TRPA1 in sensory fibres innervating resistance arteries induces potent vasodilation24,34,35,36, we tested whether this channel was involved in the vascular response to LPS. Using pressure myography ex vivo, we found that LPS (100μgml−1) induced a strong dilation in WT mouse mesenteric arteries precontracted with the alpha1 adrenergic agonist phenylephrine (2μM; Fig. 7c and Supplementary Fig. 13a). Notably, this effect was significantly reduced by the specific TRPA1 inhibitor HC-030031 (10μM) and by genetic ablation of TRPA1 (Fig. 7c and Supplementary Fig. 13b). Moreover, in Trpa1 KO animals, the residual LPS-induced dilation was completely insensitive to HC-030031 (Fig. 7c and Supplementary Fig. 13b). Thus, TRPA1 appears to participate in the vasodilation response to LPS, although other yet unidentified targets also appear to contribute to the arterial vasomotor responses.

Acute LPS-induced hypersensitivity engages TRPA1 but not TLR4

The typical symptoms of local Gram-negative bacterial infections include inflammation, pain and mechanical hypersensitivity41,42. They were readily reproduced experimentally with intraplantar injections of purified E. coli LPS in WT mice (Fig. 7d,e). Consistent with a major role of TRPA1 as a LPS effector, we found that such responses were severely attenuated in Trpa1 KO mice (Fig. 7d,e). A similar attenuation of mechanical hyperalgesia to LPS injection in Trpa1 KO mice was reported recently43. Moreover, the local inflammation caused by LPS molecules with different lipid A shape was strongly correlated with their ability to stimulate mTRPA1 in vitro (r=0.91, P<0.01; Fig. 7f) linear correlation coefficient (n=7). In contrast, acute pain responses, local tissue swelling and mechanical hyperalgesia developed normally in Tlr4 KO mice following LPS injection (Supplementary Fig. 14). Notably, mechanical hyperalgesia resolved more rapidly in Tlr4 KO mice compared with wild type, suggesting that immune response pathways are engaged by LPS in a slower timescale compared with TRPA1 activation.

Neuronal culture

Neuronal culturing methods were similar to those used previously65,66. TG and nodose ganglia were isolated from young adult male mice (postnatal months 1–3), WT and Trpa1 KO siblings64, Tlr4 KO (ref. 63) and Trpv1 KO (ref. 18), and incubated for 1h with 0.07% collagenase XI+0.3% Dispase at 37°C. Thereafter, ganglia were mechanically dissociated and cultured in minimal essential medium with Earle’s balanced salt solution and l-Glutamine (Invitrogen) supplemented with 10% fetal bovine serum (Invitrogen), 1% minimal essential medium vitamin solution (Invitrogen), 100μgml−1penicillin/streptomycin and 100ngml−1 nerve growth factor (NGF). Cells were plated on poly-l-lysine-coated glass coverslips and used after 1 day in culture.

Cell culture

CHO-K1 cells, from the American Type Culture Collection, were cultured in DMEM containing 10% of fetal bovine serum, 2% glutamax (Invitrogen), 1% non-essential amino acids (Invitrogen) and 200μgml−1penicillin/streptomycin. An inducible CHO-K1 cell line stably expressing mouse TRPA1 channels was kindly provided by Ardem Patapoutian (The Scripps Research Institute, USA)13 and cultured as detailed for the parental cell line. Expression of TRPA1 was induced by application of 0.25–0.5μgml−1tetracycline 4–18h before recording. HEK-293 cells, from the European Collection of Cell Cultures (Salisbury, UK), were seeded in 18mm glass coverslips coated with poly-l-lysine (0.1mgml−1) and grown in DMEM containing 10% (v/v) human serum, 2mM l-glutamine, 2Uml−1 penicillin and 2mgml−1streptomycin at 37°C in a humidity-controlled incubator with 10% CO2. Human TRPA1, TRPV1, TRPM8 and TRPV2 were cloned in the pAGGS-IRES-GFP vector and transiently transfected in HEK-293 cells using Trans-IT-293 reagents (Mirus, Madison, MI, USA). Transfected cells were identified by the expression of GFP (green fluorescent protein).

Calcium imaging

Calcium-imaging experiments were conducted with the fluorescent indicator Fura-2. Cells were incubated with 5μM Fura-2-acetoxymethylester (Invitrogen) for 60min at 37°C. Fluorescence measurements were made with a Zeiss Axioskop FS (Germany) upright microscope fitted with an ORCA ER charge-coupled device camera (Hamamatsu, Japan). Fura-2 was excited at 340 and 380nm with a rapid switching monochromator (TILL Photonics, Germany). Mean fluorescence intensity ratios (F340/F380) were displayed online with Metafluor software (Molecular Devices, Sunnyvale, USA). The control bath solution contained (in mM) 140 NaCl, 3 KCl, 2.4 CaCl2, 1.3 MgCl2, 10 HEPES and 10 glucose, and was adjusted to a pH of 7.4 with NaOH. Bath temperature was maintained at 34–36°C by a custom-built Peltier device.

Patch-clamp electrophysiology

Whole-cell voltage-clamp recordings were performed with standard patch pipettes (3–6MΩ resistance) made of borosilicate glass capillaries (Harvard Apparatus Ltd, UK) and contained (in mM) 140 NaCl, 5 CsCl, 10 EGTA, 10 HEPES, (305mOsmkg−1; pH 7.4, adjusted with NaOH). The control bath solution was kept at room temperature except for HC-030031 experiments (32±1°C). During calcium-free experiments, the bath solution was as follows (in mM): 140 NaCl, 3 KCl, 1.3 MgCl2, 10 HEPES, 1 EGTA, 10 glucose. Current signals were recorded with an Axopatch 200B patch-clamp amplifier (Molecular Devices). Stimulus delivery and data acquisition were performed using pClamp 9 software (Molecular Devices). Current development was monitored with repetitive (0.5Hz) injections of 500ms duration voltage ramps from −100 to +100mV, and the voltage steps consisted in a series of 400ms duration steps ranging from −150 to +175mV followed by an invariant pulse to −75mV (400ms). The holding potential was set to 0mV.

Outside-out patch recordings on mTRPA1 cells were performed in the absence of intracellular and extracellular Ca2+. Patch pipettes (3–6MΩ resistance) contained the standard internal solution (see above) and the calcium-free bath solution (see above). Currents were sampled at 20kHz and low-pass filtered at 2kHz. Current development was continuously monitored with repetitive (0.5Hz) injections of 500ms duration voltage ramps from −100 to +100mV from a holding potential of 0mV.

Patch-clamp data were analysed with WinASCD software written by Dr Guy Droogmans (ftp://ftp.cc.kuleuven.ac.be/pub/droogmans/winascd.zip) and Origin 7.0 (OriginLab Corporation).

Current-clamp recordings of cultured nodose neurons

Perforated-patch current-clamp recordings were performed on cultured nodose neurons from adult mice. Standard patch pipettes (2–5MΩ) were made of borosilicate glass capillaries (Harvard Apparatus Ltd) and contained (in mM) 105 K gluconate, 35 KCl, 10 NaCl, 10 HEPES, 0.1 EGTA and 1mgml−1 amphotericin B, pH 7.4. The external solution was identical to the one used in voltage-clamp recordings. Before electrical recordings, neurons expressing TRPA1 were identified with Ca2+ microfluorimetry (see above) by their response to 10μM nifedipine, a potent, reversible and non-electrophilic activator of TRPA1 channels22.

Recordings of lingual nerve afferents

The animals were killed by exposure to CO2 and quickly decapitated. The tongue was carefully dissected, split at the midline and placed in a recording chamber and continuously perfused (3mlmin−1) with physiological saline (in mM): NaCl 128, KCl 5, NaH2PO4 1, NaHCO3 26, CaCl2 2.4, MgCl2 1.3 and glucose 10. This solution was gassed with carbogen to pH 7.4. Temperature of the solution was kept around 34°C and could be modified with a water-cooled, home-designed, Peltier device. Fine filaments of the lingual nerve were split with fine forceps and placed on a platinum electrode for extracellular recording. Electrical signals were alternating current-amplified and high pass filtered. Data were captured and analysed using a CED 1401 interface and Spike 2 software (Cambridge Electronic Design, Cambridge, UK). Single units were detected online using a threshold-based criterion and sorted manually afterwards to remove artefacts.

Arterial pressure myography

Mice were killed by CO2 inhalation and decapitation. The mesenteric bed was removed and transferred to ice-cold solution (in mM): NaCl 124, NaHCO3 25, KCl 5.5, KH2PO4 1.2, MgSO4 1.2, glucose 5.5, CaCl22.5 (pH 7.4 when aerated with 95% O2–5% CO2). Second- and third-branch arteries were removed, dissected free of the adherent fat, cannulated and mounted in a pressure myograph (Danish Myotechnology, Aarhus, Denmark) for measurement of diameter. Before testing the effect of LPS or HC-030031 on arterial tone, arteries were allowed to equilibrate at 70mmHg and 37°C, and after 60min were precontracted with phenylephrine 2μM.

Assay of TLR4 activity

Activation of TLR4 by different forms of ultrapure LPS was tested in HEK-Blue-hTLR4 cells (InvivoGen, San Diego, USA). These HEK-293 cells stably express human TLR4, MD-2, CD14 and an NF-κB-inducible alkaline phosphatase reporter gene system. Expression of secreted alkaline phosphatase was detected with Quanti-Blue (InvivoGen) and quantified by reading the absorbance with a spectrophotometer at 620–655nm. Cells were maintained in complete DMEM with selective antibiotics, as described by the manufacturer’s instructions (InvivoGen). For testing, cells were plated on 96-well plates at 25,000 cells per well, and were incubated with different concentrations of LPS for 16h.

Analysis of IL-6 levels

WT and TRPA1 KO mice were injected i.p. with 0.75 or 1mgkg−1 E. coli LPS 055:B5 (Sigma-Aldrich) or saline and killed by decapitation 1h later. Serum and liver were quickly collected and immediately frozen and stored at −80°C. IL-6 serum levels were determined by enzyme-linked immunosorbent assay (R&D Systems, Minneapolis, USA).

Liver analysis of IL-6 messenger RNA levels was performed by reverse transcriptase PCR. Total RNA was extracted using Trizol Reagent (Invitrogen). Reverse transcriptase PCR was performed with a specific set of primers: forward: 5′-GCCACTCCTTCTGTGACTCCAGCT-3′. Reverse: 5′-ACCACTTCACAAGTCGGAGGCTT-3′. Hypoxanthine guanine phosphoribosyl-transferase was used as a housekeeping gene. The amplified products were resolved in a 2% agarose gel and quantified by densitometry using a digital imaging system (Molecular Analyst, Bio-Rad). The PCR products were digested with specific restriction enzymes to determine their specificity.

CGRP enzyme immunoassay and analysis

The CGRP content of the incubation fluid was measured using a commercial enzyme immunoassay kits with a detection threshold of 2pgml−1 (Bertin Pharma, Montigny-le-Bretonneux, France). For this purpose, 100μl fluid samples were stored on ice and mixed, immediately following the trachea incubation period, with 25μl of fivefold concentrated commercial CGRP enzyme immunoassay buffer that contained a cocktail of peptidase inhibitors. The CGRP contents were determined after the end of the experiment; the antibody reactions took place overnight. Enzyme immunoassay plates were scored photometrically using a microplate reader (Dynatech, UK). All results are presented as measured by the enzyme immunoassay in pg CGRP per ml incubation fluid. To reduce interindividual and day-to-day baseline variability, the data were normalized to the second baseline value (before any stimulation). This value was subtracted from all data points so that only the absolute change in CGRP release (Δpgml−1) is displayed in the figures.

Statistical comparisons were performed using the Statistica 7 software (Statsoft, Tulsa, USA). The baseline-normalized ΔCGRP values were entered into a one-way analysis of variance followed by Fisher’s least significant difference test, focusing on the peak values of stimulated CGRP release; *P<0.05 was considered significant. Data points represent means±s.e.m. of the given number (n) of experiments.

Behavioural testing

Behavioural studies were performed on littermates that were wild type or homozygous null at the Trpa1 or Tlr4 allele. All experiments were scored by a tester blind to genotype and to drugs injected. A total of 51 wild type, 25 Trpa1 KO and 6 Tlr4 KO male mice (2–3 months old) were used.

To measure acute nocifensive responses, mice were placed within transparent plastic cylinders and allowed to accommodate for 2h. LPS (5μgμl−1) was prepared in PBS and delivered via intraplantar injection in a volume of 10μl using a 30G needle coupled to a Hamilton syringe. Immediately after injection, nocifensive time was scored as the total time spent licking, lifting, guarding, shaking or biting of the injected hind paw over a period of 10min. A separate group of animals received injection of vehicle alone.

Mechanical thresholds were determined with Von Frey hairs using a modified version of Dixon’s up-down method67. Procedures and intraplantar dosis of LPS were identical to those used for nocifensive responses. Mice were placed in transparent plastic cylinders (13cm diameter × 8cm height) on a metal mesh platform and calibrated von Frey filaments were applied to the plantar surface of the hind paw for up to 2s with sufficient force to cause the filament to bow. Brisk withdrawal of the hind paw during or immediately after application was considered a positive response. The threshold force required to elicit withdrawal of the paw (50% hind paw withdrawal) was determined from two separate trials. Data presented are from both left and right hind paws averaged together, as there were no laterality effects.

To examine the effects of LPS on body temperature (Tb), animals (adult male mice) were adapted to a Plexiglass chamber (15 × 15 × 30cm3) in a room at 23±1°C. Baseline rectal temperature was recorded with a thermoprobe connected to a digital thermometer (Lendhermack, model IM902C). At time zero (10:00 day time), immediately after a second measure of Tb, vehicle (PBS) or 3mgkg−1 of E. coli LPS were injected i.p. Thereafter, Tb was monitored at 1h interval for 7h, and the following morning.

Chemical reagents

LPS from E. coli, S. typhimurium, S. minnesota and P. aeruginosa (purified by gel-filtration chromatography), and K. pneumoniae, S. marcescens (purified by phenol extraction) were purchased from Sigma-Aldrich. N. meningitidis LPS was purified by phenol extraction at the Center for Vaccines Research and Production (Finlay Institute, Havana, Cuba). The quality of this LPS (40.5 × 106EUml−1) was determined with a Kinetic-QCL Limulus amebocyte lysate, using the programme Wiea QCL and a Lonza kit. LPSs were dissolved as a stock solution in endotoxin-free distilled water, or PBS for behavioural experiments, at a concentration of 5mgml−1. Other ultrapure forms of LPS were purchased from InvivoGen (San Diego, USA), obtained by repeated phenol extractions and prepared as sterile stock solutions with sterile water and diluted right before use. These included LPS from E. coli strain 0111:B4, S. minnesota and R. sphaeroides. Table 1 summarizes the source and characteristics of the different LPS used.

Table 1

List of LPS reagents used in the study, purity and supplier.

Diphosphoryl lipid A (from E. coli F583 Rd mutant, Sigma) was dissolved in a glass tube at a concentration of 1mgml−1 in chloroform/methanol (3:1 v/v). The suspension was heated up to 40°C and solvent was dried by rotary evaporation under a stream of N2. The glass tubes were supplemented with extracellular solution and subjected to sonication for 40min. Synthetic monophosphoryl lipid A (InvivoGen) was prepared as a 1-mgml−1 solution in dimethyl sulphoxide (DMSO). Control experiments contained the same concentration of vehicle. CLI-095 (InvivoGen), also known as TAK-242, was prepared as a 3-mM stock solution in DMSO. Polymyxin B (InvivoGen) was prepared as a 50-mgml−1 stock in distilled water. AITC, CA, TNP and HNE were stored as stock solutions and diluted immediately before the experiments. HC-030031 was purchased from Tocris and stored in a DMSO solution at 100mM.

We thank E. Quintero, A Miralles, M. Tora, M. Benoit, S. Kerselaers and late A. Pérez Vegara for excellent technical assistance and S Ingham for illustrations. We are grateful to M. Valdeolmillos, E. Caparrós and A. Mälkiä for comments. The CHO cell line expressing murine TRPA1 was supplied by A. Patapoutian (The Scripps Research Institute). The mouse Trpa1 KO line was generously donated by K. Kwan and D. Corey (Harvard Medical School). We thank S. Akira (Osaka University), C. Guerri Centro de Investigación Príncipe Felipe (CIPF), H. Hammad and B. Lambrecht (VIB, Ghent, Belgium) for Tlr4 KO mice. O.F. and C.F.P. were predoctoral students of the Generalitat Valenciana. S.T., B.D., V.M. and J.A.M. were supported by predoctoral fellowships from the Spanish MINECO. Research was supported by Spanish public funds projects SAF2010-14990 and PROMETEO2010-046 to F.V., BFU2007-61524 to J.R.L.L., BFU2010-15898 to M.T.P.G., Instituto de Salud Carlos III PI12/00586 to R.S., BFU2005-08741 and CONSOLIDER-INGENIO 2010 CSD2007-00023 to C.B., ISCIII grants R006/009 (Red Heracles), the Spanish Fundación Marcelino Botín and Belgian Federal Government (IUAP P6/28 and P7/13), the Research Foundation-Flanders (F.W.O. G.0565.07, G.0686.09, G.A022.11N and G.0702.12), and the Research Council of the KU Leuven (GOA 2009/07, EF/95/010, PFV/10/006, OT/12/091 and GOA 14011).