A ‘toothache tree’ alkylamide inhibits Aδ mechanonociceptors to alleviate mechanical pain


D. M. Bautista: Life Science Addition room 355, Berkeley, CA 94720-3200, USA. Email: dbautista@berkeley.edu; C. L. Stucky: 8701 Watertown Plank Road, Milwaukee, WI 53226, USA. Email: cstucky@mcw.edu

Key points

  • • Extracts from the toothache tree (Zanthoxylum) are used to treat inflammatory pain, such as toothache and arthritis.
  • • Hydroxy-α-sanshool (sanshool) is a major alkylamide in extracts from Zanthoxylum plants.
  • • Sanshool treatment in mice caused a selective attenuation of mechanical sensitivity under naïve and inflammatory conditions
  • • Sanshool inhibits Aδ mechanonociceptors that mediate both sharp acute pain and inflammatory pain, and inhibits the activity of multiple voltage-gated sodium channel subtypes, among which Nav1.7 is the most strongly affected.
  • • Our data implicate Nav1.7 as a key mediator of inflammatory mechanical pain in ‘fast pain’ mechanosensory neurons.

Abstract  In traditional medicine, the ‘toothache tree’ and other plants of the Zanthoxylum genus have been used to treat inflammatory pain conditions, such as toothache and rheumatoid arthritis. Here we examined the cellular and molecular mechanisms underlying the analgesic properties of hydroxy-α-sanshool, the active alkylamide produced by Zanthoxylum plants. Consistent with its analgesic effects in humans, sanshool treatment in mice caused a selective attenuation of mechanical sensitivity under naïve and inflammatory conditions, with no effect on thermal sensitivity. To elucidate the molecular mechanisms by which sanshool attenuates mechanical pain, we performed single fibre recordings, calcium imaging and whole-cell electrophysiology of cultured sensory neurons. We found that: (1) sanshool potently inhibits Aδ mechanonociceptors that mediate both sharp acute pain and inflammatory pain; (2) sanshool inhibits action potential firing by blocking voltage-gated sodium currents in a subset of somatosensory neurons, which express a unique combination of voltage-gated sodium channels; and (3) heterologously expressed Nav1.7 is most strongly inhibited by sanshool as compared to other sodium channels expressed in sensory neurons. These results suggest that sanshool targets voltage-gated sodium channels on Aδ mechanosensory nociceptors to dampen excitability and thus induce ‘fast pain’ analgesia.

AM fibre

Aδ mechanoreceptor fibre


action potential

D hair

down hair




dorsal root ganglion


local anaesthetic

Nav channel

voltage-gated sodium channel


quantitative PCR




Zanthoxylum plants (such as prickly ash) are also known as ‘toothache trees’ due to the analgesic properties of their bark and fruit. Native cultures from around the world, including Native American, African, South Asian and Chinese traditions have used prickly ash and related plants to treat toothache, arthritis and pain through topical application for hundreds of years (Foster & Duke, 1999; Pereira et al. 2010). The two most common sites of treatment by prickly ash, the teeth and joints, are richly innervated by pain fibres, suggesting that active chemicals within prickly ash may directly target peripheral pain neurons.

Distinct somatosensory neuron subtypes mediate sensitivity to touch and pain. They are classified as Aβ, Aδ and C fibres based on unique morphologies, mechanical response thresholds and electrophysiological properties (Iggo, 1960; Brown & Iggo, 1967; Harper & Lawson, 1985; Smith & Lewin, 2009; McGlone & Reilly, 2010; Lumpkin et al. 2010). Aβ fibres are heavily myelinated, have larger cell bodies, have terminals that interact with accessory cells in the skin, and function as innocuous touch receptors. Aβ fibres that exhibit a transient, phasic response to a constant stimulus are classified as ‘rapidly adapting’, whereas those that have a tonic response are called ‘slowly adapting’. C fibres are unmyelinated, have smaller cell bodies, terminate as free nerve endings in the skin, and many detect harmful mechanical, thermal and chemical stimuli. Aδ fibres are lightly myelinated and tend to have intermediate cell body sizes. Most slowly adapting Aδ fibres (AM fibres) are mechanoreceptors activated by intense (presumably noxious) mechanical stimuli, while rapidly adapting Aδ fibres (D hairs) detect innocuous tactile stimuli to hairy skin.

Underlying the morphological and physiological diversity of somatosensory neurons is a wide diversity of molecular components. In particular, recent studies have shown that somatosensory neurons express a wide array of voltage-gated sodium (Nav) channel subtypes that dictate excitability (Rush et al. 2007). Some Nav channel subtypes (such as Nav1.1, Nav1.2, Nav1.3 and Nav1.6) are expressed in both the central and peripheral nervous systems. In contrast, three Nav subtypes (Nav1.7, Nav1.8 and Nav1.9) are predominantly expressed in the somatosensory system, where these channels are thought to play critical functional roles (Rush et al. 2007). For example, C fibre nociceptors are enriched in Nav1.8 and Nav1.9, two Nav subtypes that are unique in their resistance to the potent sodium channel blocker tetrodotoxin (TTX). Nav1.8 plays a role in maintaining action potential (AP) transmission under cold, and contributes to lidocaine resistance in C fibres (Zimmermann et al. 2007; Kistner et al. 2010). Thus there is much interest in identifying subtype-specific sodium channel blockers that will attenuate excitability in specific subsets of somatosensory fibres and promote analgesia (Eijkelkamp et al. 2012).

Here we ask whether the toothache tree promotes analgesia by targeting distinct populations of somatosensory neurons. A relatively abundant component among several alkylamides found in prickly ash (Zanthoxylum piperitum) extracts is hydroxy-α-sanshool (sanshool), which induces tingling paraesthesia and numbing analgesia (Bryant & Mezine, 1999; Sugai et al. 2005a,b). Recent studies have shown that sanshool may cause tingling sensations by blocking two-pore potassium channels (Bautista et al. 2008) and inducing AP firing in a subset of rapidly adapting myelinated fibres that mediate light touch (Lennertz et al. 2010). However, the mechanism by which sanshool induces numbing analgesia is not understood. Here we used behavioural assays, cutaneous sensory fibre recordings, whole-cell patch clamp and calcium imaging to show that sanshool inhibits the excitability of Aδ nociceptive mechanosensory neurons that mediate fast pain by blocking voltage-gated sodium channels.



All behavioural protocols were approved by the UC Berkeley Institutional Animal Care and Use Committee, by the Institutional Animal Care and Use Committee at the Medical College of Wisconsin, and were performed in accordance with the guidelines put forth by the National Institutes of Health. Male C57/B6 mice between 8 and 14 weeks old were used. Mice were housed with unrestricted access to food and water and maintained on a 12:12 h light–dark cycle in a controlled environment. Animals were housed individually for all behavioural assays.

Behavioural assays

von Frey, radiant heat and hot plate assays were carried out essentially as described (Hargreaves et al. 1988; Chaplan et al. 1994; Cao et al. 1998). Briefly, animals were acclimated in the corresponding assay chambers starting 2 days before the day of the assay. To induce neurogenic inflammation, 5 μL of 10% allyl isothiocyanate (Sigma, St. Louis, MO, USA) in mineral oil was applied topically to the plantar and hairy surfaces of one hindpaw. For topical application to skin, sanshool was diluted in dimethylformamide (DMF) at 50 mm and 5 μl was applied to the plantar surface of the hindpaw. Sensory thresholds and latencies were tested 30–60 min after application of allyl isothiocyanate or sanshool. All behavioural assays were performed with the experimenter blind to treatment. Animals were killed by CO2 inhalation followed by cervical dislocation immediately after all experiments.

Teased nerve fibre recordings

Mice were briefly anaesthetized with isoflurane and killed via cervical dislocation. Subsequently, the skin from the hindpaw along with the saphenous nerve was dissected, placed corium side up and superfused with synthetic interstitial fluid as previously described (Reeh, 1986). The nerve was teased into thin filaments until extracellular recordings were obtained from isolated receptive fields. Data were collected using a PowerLab 4/sp, and were recorded and analysed using the Chart program (v5.5.6; ADInstruments, Colorado Springs, CO, USA). Receptive fields were identified with a mechanical probe. Fibres were characterized according to their conduction velocity, von Frey thresholds and adaptation to a suprathreshold mechanical stimulus as previously described (Aβ≥10 ms−1 > Aδ≥ 1.2 ms−1 > C) (Koltzenburg et al. 1997). Aβ fibres were characterized as slowly adapting if they responded throughout a sustained mechanical force or rapidly adapting if they respond predominantly at the onset or offset of force. Aδ fibres were classified as either slowly adapting AM or rapidly adapting Down hair (D hair) receptors. D hair receptors responded to very low mechanical forces (≪1 mN), whereas AM fibres exhibited mechanical thresholds ≥4 mN. Furthermore, D hair receptors had very large receptive fields (up to 8 mm in diameter) whereas AM fibres had very small (≤1 mm) discrete, sensitive receptive field spots. C fibres were all slowly adapting. The minimum AP amplitude for all fibre types was at least threefold greater than the noise level.

Fibres were exposed to sanshool or vehicle (0.2% DMF in synthetic interstitial fluid) for 5 min and stimulated with a 10 mN force for 10 s delivered by a feedback-controlled, computer driven mechanical probe, 0.8 mm in diameter with a smooth flat end (Kwan et al. 2009). Fibres were stimulated before exposure to sanshool or vehicle, at 1 min intervals during the exposure to compounds and then at 2 min intervals after the exposure. The same protocol was used for electrical stimulation, except a 10% suprathreshold stimulus was delivered to the receptive field. For concentration and mechanical stimulus dose–responses, AM fibres were exposed to different concentrations of sanshool for 2 min (20, 60, 100 or 200 μm sanshool). Subsequently, these fibres were stimulated with increasing mechanical forces (10, 40, 100 and 200 mN) at 1 min intervals in the presence of sanshool.

Cell and neuron cultures

Chinese hamster ovary cells stably expressing recombinant human Nav1.1, 1.2, 1.3, 1.6, 1.7 or 1.8 (Chantest, Cleveland, OH, USA) were cultured in Ham's F12 media (Life Technologies, Carlsbad, CA, USA) supplemented with 10% foetal bovine serum; Nav1.9-expressing cells were not available. Pilot studies were conducted on human embryonic kidney cells stably expressing recombinant human Nav1.7 (gift from T. Cummins). Dorsal root ganglion (DRG) neuron cultures were prepared as previously described (Bautista et al. 2008). Neurons were dissociated from dissected DRGs from adult (>8-week-old) animals. Culture media was minimal essential medium (Life Technologies) supplemented with glutamine (Thermo Scientific, Logan, UT, USA), MEM vitamin (Thermo Scientific) and penicillin-streptomycin (Thermo Scientific). Neurons were plated on round coverslips coated with poly-d-lysine and laminin, as previously described (Malin et al. 2007), were incubated overnight at 37°C/5% CO2, and assayed the following day.

Calcium imaging

Calcium imaging was carried out essentially as described (Lennertz et al. 2010). Assays were performed in extracellular Ringer's solution containing (in mm): 140 NaCl, 5 KCl, 10 Hepes, 10 glucose, 2 MgCl2, 2 CaCl2; pH 7.4. An increase in Fura-2 ratio of >20% over baseline was considered a response.

Quantitation of sodium channel mRNA expression

Between three and 10 cultured neurons classified based on their calcium response to sanshool and cell size were collected and pooled in lysis buffer using a pulled glass pipette. Lysis buffer contained (in mm): 75 KCl, 50 Tris-Cl (pH 8.3), 3 MgCl2 with 2 U μL−1 RNAsin (Promega, Madison, WI, USA). For a single biological replicate, cells from two to four experiments were pooled and treated with DNase (TURBO DNAfree, Life Technologies). First strand synthesis was carried out according to manufacturer's specifications (Superscript III, Life Technologies). Quantitative PCR (qPCR) was carried out on a StepOnePlus Real-time PCR System using SYBR GreenER reagents (Life Technologies) with primers for Nav channels described previously (Lopez-Santiago et al. 2006). qPCR for each channel subtype was carried out in triplicates. Expression of a channel subtype was defined as amplification of product and determination of a Ct value in two of three replicates within each sample. For quantitation of expression levels, all experiments were normalized to beta actin (forward: cacccgcgagcacagcttctttg, reverse: catgccggagccgttgtcgacg).


Whole-cell patch clamp recordings were made with an Axopatch 200B amplifier (Molecular Devices, Downingtown, PA, USA). Data were sampled at 10 kHz and filtered at 2 kHz. Data were acquired with PClamp10 (Molecular Devices, Natick, MA, USA) and analysed with custom scripts in MATLAB (Mathworks). Fire-polished electrodes (1–2 MΩ) were fabricated from borosilicate glass (Sutter, Instrument Company, Novato, CA, USA). Series resistance was compensated for by 90–95% and capacitance artefact was cancelled using the amplifier circuitry. For current clamp recordings, neurons that had resting membrane potential less than −50 mV were used.

Internal solution for voltage clamp recordings contained (in mm): 140 CsF, 10 NaCl, 5 Hepes, 1 EGTA, 2 MgCl2, 1 MgATP, 0.1 NaGTP 0.1 cAMP; pH 7.3. Internal solution for current clamp recordings contained (in mm): 140 potassium aspartate, 13.5 NaCl, 9 Hepes, 0.09 EGTA, 1.6 MgCl2, 1 MgATP, 0.1 NaGTP; pH 7.3. Standard Ringer's solution was used for current clamp recordings and Chinese hamster ovary cell voltage clamp recordings. For sodium current recordings in DRG neurons, extracellular solutions with reduced sodium contained: (in mm): 70/35/10 NaCl, 50/85/105 choline chloride, 20 TEACl, 5 KCl, 10 Hepes, 10 glucose, 2 MgCl2, 2 CaCl2, 0.1 CdCl2; pH 7.4. All solutions were adjusted to 305–320 mosmol kg−1 with 1 m mannitol.

A hydroxypropyl-β-cyclodextrin solution was added at 200 μm to extracellular solutions before addition of either DMF vehicle or sanshool stock. Solutions were subsequently sonicated in a water bath for 3–5 min. TTX was used at a final concentration of 300 nm.

Other reagents

Hydroxy-α-sanshool was purified from ethereal extracts of ground Szechuan peppercorns by Irvine Chemistry Laboratory (Anaheim, CA, USA) or prepared in-house. The purity of the compound was confirmed by H1-NMR as previously described (Bautista et al. 2008). Stock solution of sanshool was dissolved in DMF at 100 mm.

Data analysis

Fura-2 ratios were calculated using Metafluor (Molecular Devices). Cellular electrophysiology data was analysed in Clampfit 10.2 (Molecular Devices). Further analyses and statistical comparisons were carried out in MATLAB (Mathworks). For conductance–potential curves, whole-cell current (I) data were first converted to conductances (G) using the equation G=I/(VVrev), where V is the command potential and Vrev is the reversal potential. Subsequently, conductance was fit to the Boltzmann equation G=Gmax/(1 + exp((V1/2V)/k)), where Gmax is the maximum conductance, V1/2 is the half-maximal conductance potential and k is a slope factor. For steady-state inactivation curves, normalized current (Inorm) was fit to the equation Inorm= 1/(1 +exp ((VVh)/k)), where Vh is the potential for half-maximal reduction in current and k is a slope factor. The double Boltzmann equation, used to fit steady-state inactivation data in DRG neurons, was Inorm=f/(1 +exp ((VVh1)/k1)) + (1 −f)/(1 + exp((V− Vh2)/k2)), where f is the fraction of current contributed by one of two classes of Nav channels with either a higher or lower half-maximal potential for reduction in steady-state current. The value of f was chosen to numerically minimize the confidence intervals of both Vh1 and Vh2. All voltages were adjusted during data analysis according to experimentally measured junction potentials for each corresponding pair of solutions.


Behavioural responses to mechanical stimuli are reduced by sanshool

To examine the analgesic properties of sanshool, we measured the effect of topical application of sanshool on mouse somatosensory behaviours. Topical application of sanshool to the hindpaw plantar surface did not evoke any spontaneous licking, flinching or other obvious behaviours. However, topical sanshool significantly elevated the mechanical thresholds for paw withdrawal, as measured using the von Frey assay 60 min after sanshool application (Fig. 1A). In contrast, sanshool did not alter responsiveness to noxious thermal stimuli in the radiant heat paw withdrawal assay (Fig. 1B), or the hot plate test (Fig. 1C). These results suggest that sanshool specifically reduces baseline sensitivity to mechanical stimuli.

Figure 1.

Sanshool increases behavioural response thresholds for mechanical, but not thermal, stimuli 
A, individual von Frey response thresholds for naïve animals before and 60 min after topical treatment with vehicle or sanshool on the plantar surface of the hindpaw. Solid bar represents the mean threshold. n= 10. *P < 0.05 (Mann–Whitney U test). B, response latency to radiant heat stimuli applied to the hindpaw, before and after topical treatment with vehicle or sanshool. n= 6. C, response latency after being placed on a hot plate at 48°C. Either vehicle solution or sanshool was topically applied to the plantar surface of both hindpaws 50 min before placing the animal on the hot plate. n= 13 and 9 for vehicle and sanshool, respectively. D, von Frey response thresholds for individual animals before and after topical application of mustard oil (MO). After the baseline measurement, MO was applied topically to the hindpaw, and after 30 min, von Frey thresholds were measured. Animals were then either treated with vehicle or sanshool solution. Animals were tested again 60 min after sanshool/vehicle treatment. Solid bar represents the mean threshold. n= 11. *P < 0.05 (Kruskal–Wallis test, Tukey's HSD post hoc test). E, response latency to radiant heat stimuli applied to the hindpaw in mice before and after MO treatment. MO was applied topically to the hindpaw, and after 30 min, animals were either treated with vehicle or sanshool solution. Animals were tested for response to radiant heat stimuli 60 min after sanshool/vehicle treatment. n= 8, 7 for vehicle and sanshool, respectively. #P < 0.05 (two-way ANOVA, Tukey's HSD post hoc test). Error bars are s.e.m.

Because sanshool is used as a traditional medicine to treat inflammatory pain, we next measured the effect of sanshool on mechanical and heat sensitivity in a model of neurogenic inflammation. Topical application of mustard oil (10%) induced mechanical and heat hypersensitivity that was observed as a significant reduction in mechanical threshold for paw withdrawal and thermal paw withdrawal latency (Fig. 1D). Following induction of hypersensitivity by mustard oil, topical application of sanshool greatly attenuated the decrease in mechanical threshold while vehicle had no effect (Fig. 1D). Similar to its effect on naïve animals, sanshool did not alter hypersensitivity to noxious heat stimuli during mustard oil-induced neurogenic inflammation (Fig. 1E). Thus, sanshool selectively attenuates mechanical pain hypersensitivity.

Sanshool suppresses the mechanosensitivity of somatosensory fibres

We next used the ex-vivo skin–nerve preparation to determine if sanshool alters sensory transduction in the peripheral nervous system, and if so, whether sanshool targets specific subsets of somatosensory fibre types. As we have shown previously that sanshool activates rapidly adapting Aβ and Aδ fibres as well as a subset of C fibre nociceptors (Lennertz et al. 2010), we focused our study on fibres that display little or no sanshool-evoked excitation (Table 1). Nerve fibres were mechanically stimulated while the receptive fields were exposed to either vehicle (0.2% DMF), or sanshool (200 μm; Fig. 2A). Sanshool inhibited mechanically evoked AP firing in most AM fibres (90%) and a subset of slowly adapting Aβ fibres (38%) (Fig. 2A–C). In contrast, all mechanically sensitive C fibres continued to display mechanically evoked AP firing in the presence of sanshool (0%) (Fig. 2A–C). Fibre inhibition was concentration-dependent, with an IC50 of 70 ± 7 μm for AM fibres in response to 100 mN force (Fig. 2D). Sanshool inhibition was observed at a wide range of applied mechanical forces (Fig. 2E). A maximal inhibitory effect was observed after ∼4 min of 200 μm sanshool application; in some AM fibres, such inhibition persisted for more than 30 min following the wash out of sanshool (data not shown). These data demonstrate that sanshool strongly inhibits mechanical responses in slowly-adapting, myelinated fibres, especially the AM fibres, which mediate sharp acute mechanical pain as well as the hypersensitivity to mechanical pain after neurogenic inflammation (Lin et al. 2000; Magerl et al. 2001). These findings are consistent with the use of toothache tree extracts to treat inflammatory hypersensitivity in areas densely innervated by myelinated mechanosensitive pain fibres (e.g. tooth pulp or joints) (Heppelmann et al. 1988; Paik et al. 2010).

Table 1.  Distribution of cutaneous sensory fibres that are activated or inhibited by sanshool
Fibre type% activated (n)% inhibited (n)
  1. Fibres were stimulated with either a mechanical or an electrical stimulus. Data are from this study and Lennertz et al. (2010). ND, not determined.

Aδ mechanoreceptor2 (47)88  (17)
Down hair96  (23)ND
Rapidly adapting Aβ56  (27)11  (9)
Slowly adapting Aβ44  (34)33  (15)
C34  (68)0  (17)
Figure 2.

Sanshool suppresses mechanosensitivity of cutaneous somatosensory fibres 
A, example recordings demonstrate the inhibition of mechanically evoked action potentials in SA-Aβ and AM fibres in the presence of sanshool. B, percentage of nerve fibres that exhibit mechanical inhibition in sanshool. P < 0.001 (χ2 test). n= 8–10. C, number of action potentials fired at baseline and after 3 min in sanshool for each fibre tested. Solid bar represents the mean. ***P < 0.001, (two-way ANOVA). n= 8–10. D, sanshool concentration–inhibition curve for AM fibres using a 100 mN force stimulus. Data were fitted by a non-linear four-parameter regression model. E, dose-dependent inhibition of AM fibres by sanshool in response to increasing mechanical force. n= 9–14. AM, Aδ mechanoreceptor; AP, action potential; C, C fibre; n.s., not significant; SA, slowly adapting.

Electrical excitability of sensory fibres and dissociated sensory neurons is suppressed by sanshool

Sanshool may attenuate mechanically evoked AP firing by either inhibiting mechanosensitive channels that trigger excitability, or by directly inhibiting voltage-gated channels. Thus, we questioned whether sanshool inhibits AP firing using an electrical stimulus, which would bypass mechanical transduction. We found that sanshool attenuated the electrically evoked AP firing in all AM fibres and in a subset of slowly adapting Aβ fibres, but in contrast had no effect on C fibres (Fig. 3A and B). The time course of recovery of electrical excitability correlated well with the recovery of mechanical excitability (data not shown). This experiment suggests that sanshool inhibits the initiation or propagation of APs in slowly adapting, myelinated fibres, independently of its effect on mechanical transduction.

Figure 3.

Sanshool suppresses electrical excitability in myelinated nociceptors 
A, example recordings demonstrate the inhibition of electrically evoked action potentials in AM fibres but not in C fibres. B, percentage of nerve fibres that exhibit electrical inhibition in the presence of sanshool. P < 0.001 (χ2 test). n= 7–9. Error bars are s.e.m. C, representative voltage traces of APs evoked by current injection in dorsal root ganglion neurons, in the presence of vehicle (black) or 200 μm sanshool (green). Some neurons failed to generate an AP (top) while others remained excitable (bottom) in the presence of sanshool. D, percentage inhibition of AP amplitude relative to baseline in the presence of vehicle (filled) or 200 μm sanshool (open) solution. Neurons that fired spontaneous APs in the presence of sanshool were not included in the analysis. n= 31. E, thresholds for AP firing in neurons resistant or sensitive to the inhibitory effect of sanshool. Neurons were classified using a 50% cut-off for sanshool-induced reduction in AP amplitude. AP firing threshold was defined as the potential at which the rate of change in membrane potential (dV/ds) was greater than 15 mV ms−1. n= 22, 9 for sanshool-resistant and sanshool-sensitive neurons, respectively. **P < 0.01, ***P < 0.001 (Student's t test). Error bars are s.e.m. AM, Aδ mechanoreceptor; AP, action potential; SA, slowly adapting; San, sanshool.

Consistent with its effect on sensory fibres, sanshool also inhibited electrically evoked AP firing in a subset of cultured DRG neurons (Fig. 3C and D). Sanshool completely inhibited AP firing in 35% of neurons, and significantly reduced the AP amplitude in another 29% of sensory neurons following injection of suprathreshold current (n= 31; Fig. 3D). Another 36% were unaffected by sanshool. The sanshool-induced AP inhibition was concentration-dependent with an IC50 of 56 μm (n= 4, data not shown).

Sanshool-sensitive neurons had distinct physical and physiological characteristics compared to sanshool-resistant neurons. First, larger diameter neurons (>28 μm) were more likely to be inhibited by sanshool (nine of 16 large diameter vs. two of 15 small diameter; P < 0.05, Fisher's exact test). This result is consistent with the inhibition of myelinated fibres in our teased fibre recordings (Fig. 2). Small diameter neurons tend to express higher levels of Nav channels that are resistant to the specific sodium channel blocker TTX, such as Nav1.8, as compared to larger diameter neurons (Black et al. 1996; Fukuoka et al. 2008). Accordingly, the small diameter neurons that were insensitive to sanshool were also resistant to TTX (86%). Second, neurons inhibited by sanshool had a more hyperpolarized threshold for AP firing at baseline (Fig. 3E). Nav channels expressed in sensory neurons are activated at different membrane potentials depending on the channel subtype (Rush et al. 2005). Therefore, sanshool-sensitive neurons may express Nav channels that are activated at a more hyperpolarized membrane potential and are preferentially inhibited by sanshool.

Sanshool attenuates excitability by inhibiting voltage-gated sodium channels in dorsal root ganglion neurons

We next used whole-cell voltage clamp recording to examine the effects of sanshool on endogenous Nav channels in sensory neurons. The application of sanshool caused a rapid and reversible reduction in voltage-gated sodium current amplitude (Fig. 4A–D). Classification of neurons based on their sensitivity to TTX showed that TTX sensitivity is weakly correlated with sanshool sensitivity (Fig. 4E).

Figure 4.

Endogenous voltage-gated sodium currents are inhibited by sanshool in cultured sensory neurons 
A, representative voltage-gated sodium current traces from dissociated dorsal root ganglion (DRG) neurons, in the absence (left) and presence of sanshool (right, 200 μm). Inset shows the voltage protocol for the traces. B, representative time course of sanshool-induced voltage-gated sodium current inhibition. 30 ms step pulses from −110 mV to −20 mV were applied at 2 s intervals. Holding potential was −85 mV. Veh, vehicle; San, sanshool; Wash, washout. C, whole-cell voltage-gated sodium conductance of a representative DRG neuron shown in A. Data points were fit with a Boltzmann function. D, percentage inhibition of peak conductance relative to baseline in the presence of vehicle (filled) or 200 μm sanshool (open). 200 ms step pulses from −110 mV to −10 mV were applied. n= 16. E, percentage of suppression of total maximum conductance after stepping from −110 mV to −10 mV. Neurons were classified according to sensitivity to TTX. n= 16. #P < 0.05, ##P < 0.01, n.s. not significant (two-way ANOVA, Tukey's HSD post hoc test). F, representative voltage-dependent steady-state inactivation curve of a DRG neuron, in the absence and presence of sanshool (200 μm). Voltage protocol (inset) consisted of a 100 ms prepulse step to indicated potentials from a holding potential of −85 mV, followed by a 10 ms step to −10 mV. Data points were fit with a double Boltzmann function (see Methods). G, average shift in the midpoints of the steady-state inactivation curves derived from double Boltzmann fits. n= 9. **P < 0.01, ***P < 0.001 (Student's t test). TTX, tetrodotoxin.

Many Nav channel inhibitors act by preferentially binding to the inactivated state of the channel, which promotes a hyperpolarizing shift in the steady-state inactivation curve (Hille, 1977; Bean et al. 1983). We thus measured voltage-dependent steady-state inactivation of Nav channels in the absence and presence of sanshool (Fig. 4F). Sensory neurons express multiple sodium channel subtypes that differ in sensitivity to voltage-dependent inactivation. As such, our data were best fit with either a single or a double Boltzmann function to determine a single or two midpoint(s) for inactivation (Vh), respectively (see Materials and methods for fit assumptions) (Rush et al. 2005; Kistner et al. 2010). Sanshool caused a significant hyperpolarizing shift in both Vh1 and Vh2 (Fig. 4G), suggesting that sanshool may promote steady-state inactivation at or near the resting potential.

Sodium channel subtypes are selectively sensitive to sanshool

If sanshool dampens sensory neuron excitability by inhibiting Nav channels, sanshool sensitivity may correlate with the expression of sanshool-sensitive Nav channel subtypes. Our initial characterization showed that sensory neurons expressing TTX-sensitive Nav channels tend to be more strongly inhibited, while those expressing TTX-resistant Nav channels are less suppressed by sanshool (Fig. 4E). However, TTX sensitivity of AP firing did not correlate strictly with sanshool sensitivity, as many TTX-sensitive large diameter neurons were still resistant to the inhibitory effect of sanshool. We thus sought to determine the specific Nav channel subtypes expressed in neurons inhibited by sanshool.

Because molecular markers are not yet available to differentiate among somas of myelinated fibre subtypes in culture and soma size is a rough marker of functional subtypes, we used a ‘next best’ approach. We used calcium imaging to functionally classify neurons and then performed qPCR on subgroups of neurons to identify Nav channel subtype expression. Neurons were classified into three groups based on soma size and calcium response to sanshool (Fig. 5A). Group I neurons are larger in diameter and activated by sanshool, as previously described (Lennertz et al. 2010). Group I neurons likely correspond to myelinated light touch receptors, including rapidly adapting Aβ, some slowly adapting Aβ and rapidly adapting Aδ fibres. Group II neurons are also large diameter but are not activated by sanshool, and represent neurons that are more likely to be silenced (AM and another subset of slowly adapting Aβ fibres) (Fig. 5A). Group III small diameter neurons are activated by capsaicin and thus include a population of C fibre nociceptors. Neurons from each group were pooled and Nav channel mRNA expression levels were measured by qPCR.

Figure 5.

Sanshool preferentially targets Nav1.7 
A, schematic for classifying cultured sensory neurons based on calcium response properties to sanshool and capsaicin. Circle diameter represents relative cell body size, which in turn correlate with myelination. B, distribution of Nav channel subtype expression in groups of dissociated somatosensory neurons classified as in A. (Left) Prevalence of Nav channel subtype mRNA in different groups. Transcripts were detected in either >80% (++), between 80% and 50% (+), or <20% (–) of total samples tested. n= 6 samples. (Right) Summary of Nav channel mRNA distribution in group I (red), group II (blue) and group III (grey) DRG neurons. C, conductance–voltage relationship of human Nav1.7 expressed in Chinese hamster ovary cells, in the presence of vehicle or sanshool (200 μm). Data points were fit with a Boltzmann function. n= 5. D, percentage inhibition of current, relative to vehicle control, of the indicated human Nav (hNav) channels by 200 μm sanshool. Cells were held at −80 mV and stepped to 0 mV following a 20 ms step to −100 mV. Solid bar represents the mean percentage of inhibition. n= 4–10. ##P < 0.01 compared to all others, except indicated otherwise, ◊◊P < 0.01, n.s. not significant (one-way ANOVA, Tukey's HSD post hoc test). E, dose–response of sanshool-induced current inhibition in hNav1.2, hNav1.3 and hNav1.7. Data points for hNav1.7 and hNav1.3 were fit with a four-parameter sigmoidal function, with a mean IC50 value of 124 and 249 μm, respectively. n= 4–5. P < 0.001 between hNav1.2, hNav1.3 and hNav1.7 (repeated measures ANOVA). F, relationship between mean percentage of sanshool-induced hNav inhibition and average midpoint for steady-state inactivation at baseline. Percentage current inhibition for each cell expressing hNav subtype was derived as in D. For steady-state inactivation, cells were held at −80 mV and stepped to the indicated prepulse potentials for 100 ms, followed by a 10 ms step to −5 mV. Data points were fit with a Boltzmann function to obtain the inactivation midpoint potential, Vh (see Methods). n= 4–7. G, steady-state inactivation curves of hNav1.2 (black), hNav1.7 (red) and hNav1.8 (blue) in the presence or absence of sanshool (200 μm). n= 5–6. San, sanshool; Veh,vehicle. H, shift in Vh for hNav channels in the presence of sanshool (200 μm) relative to vehicle control. Solid bar represents the mean shift. n= 4–7. ##P < 0.01 compared to all others (one-way ANOVA, Tukey's HSD post hoc test). Error bars are s.e.m. AM, Aδ mechanoreceptor; RA, rapidly adapting; SA, slowly adapting.

The three groups of neurons displayed unique combinations of Nav channel subtype expression (Fig. 5B, Supplemental Fig. 1). Group I neurons expressed all subtypes tested except for Nav1.8. Similarly, group II neurons expressed most Nav channel subtypes, including Nav1.8, with the exception of Nav1.2. Group III neurons only expressed Nav1.7 and 1.8. These results are consistent with the previously described distribution of Nav channel subtypes based on relative cell size (Black et al. 1996; Fukuoka et al. 2008; Ho & O’Leary, 2011). Notably, larger diameter neurons that express both Nav1.7 and Nav1.8 probably include Aδ nociceptors (Djouhri et al. 2003a,b). These data suggest that Nav1.2 expression may underlie the resistance of larger diameter neurons to sanshool, and that Nav1.2 may be more resistant to sanshool than other subtypes. To test this, we characterized the degree of current block at 0 mV and the shift in steady-state inactivation induced by sanshool in Nav channel subtypes expressed heterologously.

All Nav channels tested displayed voltage-gated currents that were reduced, to varying degrees, by sanshool (Fig. 5C and D). Notably, Nav1.3 and 1.7 displayed significantly more block at 0 mV as compared to Nav1.1, 1.2, 1.6 and 1.8, with Nav1.2 displaying the least sensitivity (Fig. 5D and E). There are several classes of Nav channel inhibitors that have different mechanisms of action. Local anaesthetics, such as lidocaine, act by binding to the open or inactivated state of Nav channels (Hille, 1977; Ragsdale et al. 1994). Plotting the strength of sanshool-induced current inhibition against the mid-point for steady-state inactivation showed a clear correlation. The inhibition by sanshool tended to be stronger as the fraction of inactivated channels at resting potential increased (Fig. 5F). In DRG neurons, sanshool caused a hyperpolarizing shift in the steady-state inactivation curve of endogenous sodium channels (Fig. 4F and G). Interestingly, even though sanshool shifted the steady-state inactivation of all heterologously expressed Nav channels, sanshool caused the biggest shift in Nav1.7 (Fig. 5G and H). These results suggest that at resting potential, sanshool is likely to have the most profound inhibitory effect on the activity of Nav1.7 channels. Overall, sanshool inhibits Nav channel subtypes to varying degrees, and differences in the relative contribution of each channel subtype to excitability may dictate sanshool sensitivity among the different groups of neurons.


Extracts from the toothache tree have been used for decades to treat inflammatory pain. Yet, the molecular mechanisms that mediate the analgesia remained unknown. Here we demonstrate that topical application of sanshool to the skin blocks mechanical but not heat hypersensitivity following inflammation. Our results suggest that sanshool induces analgesia by completely silencing Aδ nociceptors through inhibition of Nav channels and blockade of AP firing.

Selective silencing of myelinated nociceptors

Sanshool is commonly used to treat toothache and rheumatoid arthritis. Structural and staining studies have shown that many fibres innervating the tooth pulp and knee joints are lightly myelinated and have free nerve endings, indicative of Aδ nociceptors (Heppelmann et al. 1988; Paik et al. 2010). In addition, recordings from fibres innervating the knee joint confirm that many of these fibres display functional properties of Aδ fibres (Guilbaud et al. 1985). In models of inflammatory pain, both myelinated and unmyelinated nociceptors have lower response thresholds, increased spontaneous activity and higher stimulus-evoked AP firing rate (Guilbaud et al. 1985; Schaible & Schmidt, 1985). Aδ fibres in particular may be sensitized during neurogenic inflammation via the release of inflammatory mediators such as substance P that then sensitizes Aδ terminals (Herbert & Schmidt, 2001; Magerl et al. 2001). Thus sanshool may be a highly effective analgesic for inflammatory pain because of its preferential silencing of Aδ mechanonociceptors.

Differential inhibition of Nav channel subtypes

The analgesic properties of sanshool resemble those of local anaesthetics (LAs) such as lidocaine, at both the physiological and molecular levels. LAs preferentially bind to the open and inactivated states of Nav channels by binding to specific residues in the S6 segments within the ion-conduction pore (Hille, 1977; Ragsdale et al. 1994). Structure–function studies for an insecticidal alkylamide that is structurally related to sanshool showed that the residues in S6 required for LA activity are also important for inhibition by alkylamides (Du et al. 2011). The leftward shift in the steady-state inactivation curve induced by sanshool also suggests that sanshool may bind to the open and/or inactivated state of Nav channels.

Like sanshool, LAs also display channel subtype specificity (Roy & Narahashi, 1992; Chevrier et al. 2004; Leffler et al. 2007; Kistner et al. 2010). For example, the efficacy of block by lidocaine is dependent on the fraction of inactivated channels present at a given membrane potential, and this fraction varies between channel subtypes (Wright et al. 1997; Herzog et al. 2003; Rush et al. 2005; Leffler et al. 2007; Kistner et al. 2010). The different subtypes of Nav channels expressed in DRG neurons also displayed varying degrees of sanshool-induced tonic inhibition at resting potentials (Fig. 5). Nav1.7 and 1.3 display the most hyperpolarized steady-state inactivation, indicating that more channels are inactivated at rest compared to other DRG neuron sodium channels, such as Nav1.2. This in turn would allow sanshool to inhibit Nav1.7 and 1.3 more potently than Nav1.2. Although both Nav1.7 and 1.3 voltage-gated currents were inhibited strongly by sanshool, Nav1.7 was unique in that its steady-state inactivation curve was shifted significantly more compared to other subtypes (Fig. 5). Thus the potent, dual action of sanshool on the current magnitude and steady-state inactivation properties of Nav1.7 may make this channel subtype particularly susceptible to sanshool.

Nav channel subtypes in sensory fibre function

Our Nav channel subtype expression and sanshool- sensitivity analyses (Fig. 5) suggest a model in which somatosensory neurons can be classified functionally based on their dependence on distinct but overlapping sets of Nav channel subtypes for maintaining excitability. Aδ nociceptors and slowly adapting Aβ fibres (represented by group II neurons) may be more likely to be inhibited by sanshool because they rely on Nav1.7 and/or Nav1.3. In contrast, rapidly adapting Aβ fibres (represented by group I neurons) are resistant to sanshool possibly because they rely less on Nav1.7 and/or Nav1.3 for excitability. C fibres (represented by group III neurons) are also resistant to sanshool because they rely on Nav1.8, which is more resistant to sanshool as compared to Nav1.7. These results support a model in which different Nav channels play specific roles in somatosenosory neuron function.

Unfortunately, we were unable to probe the sanshool sensitivity of Nav1.9 as there is no readily available and reliable source of Nav1.9 expressing cell lines. However, Nav1.9 is unlikely to play a key role in sanshool-evoked neuronal silencing because expression is limited to small diameter neurons and is unlikely to contribute to excitability of the sanshool-sensitive neurons (Fukuoka et al. 2008; Ho & O’Leary, 2011). In addition, Nav1.9 does not contribute to the rising phase of the AP (Herzog et al. 2001; Waxman & Estacion, 2008), thus it is unclear how sanshool sensitivity would affect AP firing.

The role of Nav1.3 in somatosensory neurons remains enigmatic. Expression of Nav1.3 is robust embryonically and after axotomy or nerve injury, but is otherwise weak (Waxman et al. 1994; Black et al. 1999; Kim et al. 2001). In addition, Nav1.3 expression is robust in dissociated DRG cultures whereas very little is detected in situ in tissue slices (Black et al. 1996). Likewise, our results showed robust expression of Nav1.3 in dissociated group I and II neurons. Thus, the relative contribution of Nav1.3 to sanshool sensitivity in vivo is unclear.

Previous studies of Nav channel subtype expression in DRG neurons suggest that Nav1.7 is expressed in a large population of neurons (Black et al. 1996; Fukuoka et al. 2008; Ho & O’Leary, 2011). Aδ nociceptors in particular express relatively high levels of Nav1.7 as well as Nav1.8 (Djouhri et al. 2003a,b). In addition, Nav1.7 expression is upregulated during inflammation of the tooth pulp and joints (Luo et al. 2008; Strickland et al. 2008). However, the functional contribution of each Nav subtype in Aδ nociceptors is not known. Our results implicate Nav1.7 as a critical mediator of Aδ nociceptor excitability, perhaps due to its proposed role in impulse initiation (Cummins et al. 1998). Consistent with this model, Nav1.7 expression has been observed in nerve endings of cutaneous myelinated nociceptors (Black et al. 2012). Future fibre recordings using Nav1.7 and Nav1.8 knockout mice, or Nav1.7-specific inhibitors (not commercially available), are needed to test this hypothesis.

Nav1.7 is also expressed in a subset of Aβ fibres that respond to gentle mechanical stimulation (Djouhri et al. 2003b). Our results suggest that Nav1.7 may also contribute to AP firing in a subset of slowly adapting Aβ fibres. The functional consequence of inhibiting Nav1.7 in these fibres is unclear. However, the reduced behavioural mechanosensitivity induced by sanshool in naïve animals may partly be due to sanshool's inhibitory effect on these tactile fibres.

Many unmyelinated nociceptors also express Nav1.7 and Nav1.8 (Black et al. 1996; Djouhri et al. 2003a,b; Fukuoka et al. 2008; Ho & O’Leary, 2011; Shields et al. 2012). However, C fibres remained largely resistant to the inhibitory effect of sanshool. While our PCR results suggest that Nav1.8 mRNA is expressed at similar levels between group II and group III neurons (Supplementary Fig. 1), there may be differential expression of Nav1.8 protein. In addition, electrophysiologically, unmyelinated nociceptors appear to rely more on Nav1.8 for AP generation (Blair & Bean, 2002). Indeed, the relative resistance of C fibres to lidocaine is thought to be due to high expression of Nav1.8 in these fibres (Kistner et al. 2010). Thus, sanshool resistance of C fibres may also be due to their reliance on Nav1.8, which is relatively insensitive to sanshool, for initiation of APs.

Acute, peripheral inhibition of mechanical pain by sanshool

We have previously shown that sanshool activates a subset of primary afferents, including Aβ fibres to induce tingling paraesthesia (Table 1) (Lennertz et al. 2010). Thus, it is possible sanshool-evoked analgesia may be indirect, as Aβ fibre activity can modulate nociceptor activity through central inhibition (Melzack & Wall, 1965; Woolf et al. 1980). However, central inhibition would be expected to alter both mechanical and thermal responsiveness (Chung et al. 1984; Paik et al. 1988) and cannot explain the activity of sanshool in the skin–nerve or in vivo experiments. Thus our data support a model whereby sanshool directly inhibits peripheral nociceptor terminals to mediate analgesia.

Studies on Nav channel knockout mice have shown significant roles for Nav1.7 in somatosensory behaviour. Nav1.7 is required for normal responses to radiant heat and noxious mechanical stimuli, but this channel is not required for responses to a noxious hot plate (Minett et al. 2012). As such, it has been proposed that neurons that predominantly depend on Nav1.7 are involved in acute thermal pain responses, while neurons that express both Nav1.7 and Nav1.8 mediate mechanical and inflammatory pain (Nassar et al. 2004; Minett et al. 2012). Humans with loss of function mutations in Nav1.7 do not experience pain in response to noxious mechanical or thermal stimuli (Cox et al. 2006). The lack of an effect of sanshool on thermal sensitivity was surprising given these studies. However, differences between acute inhibition of function and chronic genetic loss of Nav1.7 through development may account for such differences. Additionally, inhibition of other channel subtypes by sanshool may contribute to the specificity of sanshool action. One such possibility might suggest that the dual suppression of Nav1.7 and Nav1.3 by sanshool results in selective silencing of Aδ mechanoreceptors and subsequent inhibition of mechanical pain. Regardless, our data offer insights into the mechanism by which modality-specific analgesia is induced by sanshool. In addition, the development of synthetic sanshool derivatives (Wu et al. 2012) may be useful for treating painful inflammatory conditions that involve sensitization of myelinated mechanosensitive nociceptors.


Additional information

Competing interests


Author contributions

M.T. was responsible for the conception and design of experiments; collection, analysis and interpretation of data; drafting the article. R.C.L. was involved in the conception and design of experiments; collection, analysis and interpretation of data; drafting the article. D.V. dealt with the collection, analysis and interpretation of data. S.K. was involved in the collection, analysis and interpretation of data. C.L.S. contributed towards the conception and design of experiments and drafting of the article. D.M.B. was involved in the conception and design of experiments and drafting of the article. All authors approved the final version of the manuscript.

Behavioural assays, cellular electrophysiology and molecular biology were performed at UC Berkeley. Ex-vivo skin–nerve experiments and cellular electrophysiology were carried out at Medical College of Wisconsin.


This work was supported by the National Institutes of Health grants NS40538 and NS070711 to C.L.S. and AR059385 and DOD007123A to D.M.B.


We would like to thank Ted Cummins for generously providing the Nav1.7-expressing human embryonic kidney cell line and members of our labs for useful discussions.