Male and female hairless SKH1 mice (20–35 g, Charles Rivers, Wilmington, MA) were utilized for behavioral experiments. The hairless phenotype allows direct contact of the animal's skin with the thermal probes on the testing apparatus, thus eliminating the insulating effect of the fur. Additionally, shaving of furred mice leads to skin irritation that may induce hypersensitivity in the skin (14).

TRPM8^tm1Apat/J knockout mice (Jackson Labs, Bar Harbor, ME) were crossed in house with C57BL/6 mice (Charles Rivers, Wilmington, MA) to produce heterozygotes (TRPM8^EGFP−/+). The TRPM8^tm1Apat/J knockout mice express enhanced green fluorescent protein (eGFP) in place of TRPM8. Male and female heterozygotes (TRPM8^EGFP−/+) were utilized for electrophysiology and immunohistochemistry of TRPM8 expressing neurons.

The animals were housed with a 12-h light–dark cycle, and food and water were available ad libitum. Experiments were conducted between 09.00 and 18.00 h at a room temperature of ~22°C. All experiments were approved by the University of Florida Institutional Animal Care and Use Committee and were performed in compliance with the National Institutes of Health guidelines.

Hairless SKH1 mice of both sexes (Charles Rivers, Wilmington, MA, 20–35 g) were tested in Orofacial Pain Assessment Devices (OPAD, Stoelting Co., Wood Dale, IL). OPADs utilize a reward/conflict paradigm that requires the mice to place their faces onto temperature programmable Peltier bars to obtain a reward solution (14–27) (Figure 1A). Licking on the reward bottle is captured by the animal completing a circuit when they contact the reward bottle with their tongue (see example data in Figure 1A). The total number of licks on the bottle during the testing session was used as the dependent measure. The reward solution consisted of sweetened condensed milk diluted 1:2 with water. The animals' food was removed from their home cages 12–15 h prior to testing, but they received water ad libitum throughout the fasting period.

For experiments the mice were trained over a 2-week period at three sessions per week with the Peltier bars set at room temperature. Sessions were 10 min in length. The OPADs were then programed to test the animals at 18° or 42°C during 10-min sessions while recording licks on the reward bottle. Baseline responses at 18° and 42°C were collected prior to administering oxaliplatin or paclitaxel. The animals were tested three times per week for up to 80 days.

The oxaliplatin treated mice were given two injections of oxaliplatin (10 mg/kg, ip.) (Sigma Aldrich, St. Louis MO) dissolved in phosphate buffered saline (PBS). The injections were separated by 3 days. Vehicle control animals received two injections of an equal volume of PBS. The paclitaxel treated mice received four injections of paclitaxel (26 mg/kg, ip.) (Sigma Aldrich, St. Louis MO) spaced 3 days apart. The paclitaxel was dissolved in DMSO. Vehicle treated animals received four injections of an equal volume of DMSO.

TRPM8^EGFP−/+ mice were euthanized by isoflurane inhalation (5% in O₂) and the trigeminal ganglia were removed. The ganglia were then incubated for 2 h at 37° in Tyrode's buffer (mM: 140 NaCl, 4 KCl, 2 MgCl₂, 2 CaCl₂, 10 glucose, and 10 HEPES, adjusted to pH 7.4 with NaOH) containing collagenase (Sigma Aldrich, St. Louis, MO) (2 mg/ml). The ganglia were triturated with a plastic pipette, pelleted by centrifugation (100 X g), resuspended with fresh Tyrode's buffer, and plated onto 30 mm polystyrene plates. The cells were allowed to adhere to the plates for 1 h at room temperature prior to initiating the experiments (28–30). During the experiments the cells were superfused with Tyrode's buffer at 5 mls per minute at room temperature (~22°C). The bath's volume was maintained at ~1 ml. The TRPM8 expressing neurons were identified by the expression of eGFP using an inverted microscope equipped with fluorescence optics (Olympus IX70). EGFP labeled cells were whole cell patch clamped with 1.5 mm glass electrodes filled with electrode buffer consisting of (mM: 140 KCl, 1 CaCl₂, 10 EGTA, 10 HEPES, 2 MgCl₂). The pH was adjusted to 7.4 with KOH. The electrodes were pulled to resistances of 2–4 MΩ with a P-87 Flaming/Brown microelectrode puller (Sutter Instruments, Navato, CA). Data were collected using an Axopatch 200B amplifier, a Digidata 1200 analog to digital converter and PClamp8 software. Following establishment of the whole cell voltage clamp configuration the series resistance was compensated by 50–60% and cell capacitance was compensated using the settings on the amplifier.

Voltage protocols for voltage gated potassium channels (K_v), hyperpolarization-activated cyclic nucleotide-gated channels (HCN), and voltage gated sodium channels (Na_v) were as previously described (28, 30–32). Briefly, to activate K_v channels the cells were hyperpolarized to −100 mV for 500 ms and then stepped from −60 to 40 mV in 20 mV increments for 200 ms (Figure 2A). The HCN channels were activated by hyperpolarizing the cell membrane stepwise in 10 mV increments from −60 to −120 mV for 500 ms (Figure 2B). Na_v currents were activated by hyperpolarizing the cell membrane to −100 mV for 500 ms and then stepping the potential in 10 mV increments from −60 mV to 10 mV for 2 ms (Figure 2C).

To activate TRPM8 currents menthol (100 μM) was added to the Tyrode's buffer while holding the cells at −60 mV.

To calculate ion channel conductance the following equation was used:

where g is the conductance, I is the measured current, V is the command voltage, and V_r is the reversal potential of the current. To evaluate the probability of channel opening at a given membrane potential the conductance was normalized. Normalization was performed by taking the ratio of the measured conductance to the maximum conductance.

The stability of the membrane potential in neurons with pacemaker like properties is determined by the interaction of HCN channels and K_v channels (33). Therefore, the behavior of the TRPM8 neurons was evaluated by modeling the relationship between HCN and K_v channels using the modified Nernst equation:

Where V_m(t) is the membrane potential at time t, R is the gas constant, T is temperature in Kelvin, F is the Faraday constant, and z is the valence of the ions. The relative difference in the time constants (τ) for HCN and K_v channels was compensated for by using the variable n. At a temperature of 22°C the value of RT/zF is 24.89. P_h and P_k are the relative permeability of HCN channels and K_v channels, respectively. The membrane potentials at time t-n and t-1 were used to estimate ion permeability using lookup tables generated from non-linear curve fitting of conductance produced from the ion channels' current/voltage relationships (12). These permeabilities were used to calculate the membrane voltage at time t to generate a time series.

TRPM8^EGFP−/+ mice (2 oxaliplatin treated, 2 vehicle treated) were euthanized by isoflurane inhalation (5% in O₂) and perfused intracardially with PBS (pH 7.4) and subsequently 4% paraformaldehyde 6 weeks after oxaliplatin treatment. The skin from the whisker pads was shaved and removed. The tissue was fixed for 15 h in 4% paraformaldehyde, and then transferred to 30% sucrose in PBS for 24–30 h. The tissues were frozen and transverse sections cut at 40 μm. Five sections from each mouse were collected. The free-floating sections were incubated in PBS, 0.3% Triton X-100 and 5% normal goat serum (NGS–T) for 1 h before incubation in primary anti-green fluorescent protein antibody (Invitrogen, Waltham, MA) at a dilution of 1:50,000 in NGS-T for 18 h at room temperature. Preparations were then rinsed in NGS-T 3 times for 10 min each. The secondary antibody, Alexa Fluor 594 goat anti-rabbit IgG (Invitrogen, Waltham, MA), was diluted to 1:500 with NGS-T and the preparations incubated at room temperature for 1 h. The sections were then rinsed 3 times for 10 min each, mounted, cover slipped, and imaged. Images were taken of all fluorescent labeled basket like structures in the tissues. The images were then processed in ImageJ by outlining the labeled basket like structures and measuring the area of the structures.

Data on the OPADs was collected using AnyMaze (Stoelting Co., Wood Dale, IL). Data collected with PClamp8 were analyzed with Clampfit (Axon Instruments). Immunohistological data were analyzed with ImageJ. PRISM6 (GraphPad Software Inc., La Jolla, CA) was used for statistical analysis. The data were subjected to T-Tests, One-Way ANOVAs, Two-Way Repeated Measures ANOVAs, or non-linear regressions as appropriate. Dunnett's test or Bonferroni's multiple comparisons test were used for post-hoc analysis. Alpha was set to 0.05 for all experiments. All data are expressed as means ± SEM.

A common side effect of oxaliplatin chemotherapy is an acute hypersensitivity to cool temperatures, particularly in orofacial regions. The acute sensitivity usually resolves, but then a more chronic painful peripheral neuropathy may develop, which can become severe enough to necessitate curtailing therapy (1–5, 8, 11, 34). We evaluated the effects of oxaliplatin on cool temperature hypersensitivity in the face of hairless SKH1 mice using an OPAD assay (Figure 1A). Since paclitaxel is not associated with acute cool sensitivity but does produce a chronic painful peripheral neuropathy, we included paclitaxel as a comparator.

Treating SKH1 hairless mice with oxaliplatin did not produce hypersensitivity to an 18°C stimulus in the OPAD assay either acutely, tested within 5 h of the first treatment, or chronically [Figure 1B, 2-Way Repeated Measures ANOVA F_{(1, 40)} = 0.9409, P = 0.3379, N = 10 oxaliplatin, N = 32 vehicle]. Male and female mice did not differ in their responses so the data for male and female mice were combined [2-Way Repeated Measures ANOVA F_{(1, 18)} = 3.490e-005, P = 0.9954]. Similarly, paclitaxel did not produce hypersensitivity in the assay at 18°C [Figure 1C, 2-Way Repeated Measures ANOVA F_{(1, 38)} = 0.02492, P = 0.8754, N = 20 paclitaxel, N = 20 vehicle]. In contrast, paclitaxel treated mice demonstrated persistent hypersensitivity in the OPAD to 42°C [2-Way Repeated Measures ANOVA F_{(1, 34)} = 10.45, P = 0.0027, N = 16 paclitaxel, N = 20 vehicle], while oxaliplatin did not produce hypersensitivity to 42°C [2-Way Repeated Measures ANOVA F_{(1, 44)} = 0.1215, P = 0.7290, N = 10 oxaliplatin, N = 36 vehicle] (Supplementary Figure 1).

The lack of effect of both oxaliplatin and paclitaxel on nociception in the face of the mice at 18°C was unexpected given the literature on their effects in the limbs (5, 8–10, 34–44). Dorsal root ganglia (DRG) neurons that express TRPM8 are sensitized by oxaliplatin and believed to mediate the acute cool temperature hypersensitivity. The DRG TRPM8 neurons are also involved in oxaliplatin's chronic painful peripheral neuropathy (5, 8, 9). To determine if the chemotherapeutic agents also altered the function of TRPM8 expressing TRG neurons we treated TRPM8^EGFP−/+ mice with oxaliplatin or paclitaxel as described in the methods and then harvested TRG neurons 6–10 weeks following the last injection of the chemotherapeutic agent. The TRPM8^EGFP−/+ mice express eGFP in TRPM8 expressing neurons. The TRG neurons were dissociated, plated, and then TRPM8 eGFP expressing neurons were identified by their fluorescence (image in Figure 2A). The labeled neurons were whole cell voltage clamped. The standardized voltage protocols for K_v currents (Figure 2A), HCN currents (Figure 2B), and Na_v currents (Figure 2C) (12, 28) as described in the methods were used to characterize the long-term effects of the agents on the TRPM8 expressing TRG neurons.

After the whole cell voltage clamp was established the holding currents at −60mV were determined for the TRPM8 expressing neurons. The holding currents did not differ when comparing vehicle treated animals to either oxaliplatin or paclitaxel treated animals with currents of −44.95 ± 11.74 pA, −54.39 ± 12.20 pA, and −37.29 ± 12.87 pA, respectively [ANOVA F_{(2, 153)} = 0.2519, P = 0.7777, N = 91 neurons from 14 vehicle treated mice, N = 40 neurons from 9 oxaliplatin treated mice, and N = 25 neurons from 5 paclitaxel treated mice].

Figure 3 demonstrates that both oxaliplatin and paclitaxel had no effect on K_v currents [2-Way Repeated Measures ANOVA F_{(2, 180)} = 0.06814, P = 0.9342, N = 120 neurons from 14 vehicle treated mice, N = 34 neurons from 9 oxaliplatin treated mice, N = 29 neurons from 5 paclitaxel treated mice]. HCN currents, on the other hand, were significantly increased by oxaliplatin and paclitaxel treatment (Figure 4A) [2-Way Repeated Measures ANOVA F_{(2, 756)} = 30.94, P < 0.0001, N = 51 neurons from 14 vehicle treated mice, N = 39 neurons from 10 oxaliplatin treated mice, N = 21 neurons from 5 paclitaxel treated mice]. Normalizing the HCN conductance demonstrated that the activation threshold for the currents was not altered by the agents (Figure 4B) [2-Way Repeated Measures ANOVA F_{(2, 136)} = 2.227, P = 0.1118].

Na_v currents were also significantly increased by both agents (Figure 5A) [2-Way Repeated Measures ANOVA F_{(2, 616)} = 38.49, P < 0.0001, N = 27 neurons from 14 vehicle treated mice, N = 32 neurons from 9 oxaliplatin treated mice, N = 20 neurons from 5 paclitaxel treated mice]. The normalized Na_v conductance, however, demonstrated that the activation threshold shifted to more negative potentials (Figure 5B) [2-Way Repeated Measures ANOVA F_{(2, 75)} = 9.801, P = 0.0002]. The 50% activation potential for Na_v currents in TRG neurons isolated from vehicle, oxaliplatin, and paclitaxel treated mice were −38.07 ± 1.46 mV, −42.96 ± 0.87 mV, and −42.15 ± 0.87 mV, respectively.

HCN currents are responsible for pacemaker activity in neurons and cardiac cells (45–47), which suggests that the increase in HCN currents produced by the chemotherapeutic agents could decrease membrane stability and induce spontaneous firing. In pacemaker types of cells, the balance between HCN and K_v channels determines the relative firing rate of the cells (33). Therefore, to examine the relationship of the HCN currents to membrane voltage changes the HCN and K_v conductance/voltage relationships were curve fitted by non-linear regression. The curves were then used as lookup tables to model membrane voltages using equation 2 as previously described (12). Equation 2 estimates the interaction between HCN and K_v to model the stability of the resting membrane potential. Figure 6A illustrates the regressions for the conductances for both ion channels. The resultant time series for vehicle (Figure 6B), oxaliplatin (Figure 6C), and paclitaxel (Figure 6D) on predicted membrane potentials suggests that oxaliplatin or paclitaxel treated neurons were more unstable than the control neurons. The estimated variance in the membrane potential for vehicle, oxaliplatin and paclitaxel treated neurons was 43.41, 296.66, and 286.25, respectively. These models suggest that the chemotherapy treated TRPM8 expressing neurons were more likely to fire action potentials spontaneously. This concept was further supported by the lower threshold potentials for the Na_v currents from the oxaliplatin and paclitaxel treated mice when compared to vehicle treated mice (Figures 5A,B).

Previous work by others demonstrated that TRPM8 mediated responses were enhanced by oxaliplatin and that there is an increase in the expression of TRPM8 in DRG neurons (5, 9). Since our previous study found that TRPM8 was not activated by cooling these neurons (12) we activated TRPM8 mediated currents in TRG neurons by superfusing the cells with the TRPM8 agonist menthol (100 μM). As demonstrated in Figures 7A,B prior treatment of the mice with either oxaliplatin or paclitaxel increased the amplitude of menthol induced currents in the TRPM8 expressing TRG neurons [ANOVA F_{(2, 68)} = 3.482, P = 0.0363, N = 27 neurons from 12 vehicle treated mice, N = 29 neurons from 8 oxaliplatin treated mice, N = 15 neurons from 4 paclitaxel treated mice].

The incongruence of the behavior data with the electrophysiology data in oxaliplatin treated mice suggested that the TRPM8 expressing TRG neurons may have lost communication with the skin, which prevented the cool temperatures from activating the sensitized neurons. To address this issue, we sectioned whisker pads of TRPM8^EGFP−/+ mice that had been treated with ip. PBS or oxaliplatin 6 weeks prior to collection of the tissue. The sections were prepared as described in the methods section and examined under fluorescence microscopy. Axons from eGFP labeled neurons were identified in the whisker pads of PBS treated mice as illustrated in Figure 8. Many axons were observed to terminate near the surface of the skin in basket like structures. Work by Bouvier et al. demonstrated that these TRPM8 baskets wrap around Merkel cells and that Merkel cells respond to reduced temperatures (48). Dhaka et al., however, described these structures as bush/cluster free nerve endings (49). Following treatment with oxaliplatin the eGFP labeled axons were difficult to find and the basket structures were diffuse, smaller, and appeared damaged when compared to the vehicle treated baskets (Figure 8). The area of the baskets was quantified using ImageJ. The basket structures of vehicle treated animals had a mean area of 2,967.9 ± 592.8 μm² and oxaliplatin treated animals had a mean area of 1,146.8 ± 167.2 μm² (t-test, t = 2.740 df = 22, P = 0.012, N = 13 structures from 2 vehicle treated mice, N = 11 structures from 2 oxaliplatin treated mice). These findings indicated that the peripheral axons of TRPM8 expressing TRG neurons were damaged by oxaliplatin, likely within a few hours of the first treatment, resulting in a deafferentation of the neurons.