Silent cold-sensing neurons contribute to cold allodynia in neuropathic pain

Abstract

Patients with neuropathic pain often experience innocuous cooling as excruciating pain. The cell and molecular basis of this cold allodynia is little understood. We used in vivo calcium imaging of sensory ganglia to investigate how the activity of peripheral cold-sensing neurons was altered in three mouse models of neuropathic pain: oxaliplatin-induced neuropathy, partial sciatic nerve ligation, and ciguatera poisoning. In control mice, cold-sensing neurons were few in number and small in size. In neuropathic animals with cold allodynia, a set of normally silent large diameter neurons became sensitive to cooling. Many of these silent cold-sensing neurons responded to noxious mechanical stimuli and expressed the nociceptor markers Nav1.8 and CGRPα. Ablating neurons expressing Nav1.8 resulted in diminished cold allodynia. The silent cold-sensing neurons could also be activated by cooling in control mice through blockade of Kv1 voltage-gated potassium channels. Thus, silent cold-sensing neurons are unmasked in diverse neuropathic pain states and cold allodynia results from peripheral sensitization caused by altered nociceptor excitability.

Keywords: cold allodynia; neuropathic pain; pain; potassium channels; sodium channels.

Figure 1

Silent cold-sensing neurons are activated by oxaliplatin. (A) Behavioural testing of the effect of intraplantar oxaliplatin injection on different sensory modalities (cold, mechanical and heat). n = 8 (five male and three female) for vehicle and n = 9 for oxaliplatin (five male and four female). Mean values before and after treatment were compared using repeated measures two-way ANOVA followed by post hoc Sidak’s test. Error bars denote 95% confidence interval. (B) Example images (i) and traces (ii) of cold-responding neurons in vehicle- and oxaliplatin-treated animals expressing GCaMP3. Cell 1 is a small diameter cold-sensing neuron in the vehicle condition, Cell 2 is a small diameter basal cold-sensing neuron after oxaliplatin and Cell 3 is a large diameter silent cold-sensing neuron unmasked by oxaliplatin that also responds to noxious mechanical stimuli. [C(i)] Histograms of the cross-sectional area of all neurons responding to any cold stimulus in the vehicle (top, blue, n = 82) and oxaliplatin (bottom, red, n = 179) groups. The distribution of areas for vehicle was fit by non-linear regression [least squares Gaussian; bin width is 80 µm²; mean = 214.9 µm², SD (σ) = 77.29 µm²]. This model is plotted over the oxaliplatin data to aid comparison with the dashed line denoting 3 SD from the mean. The difference in the distribution of areas between groups was assessed by Kolmogorov-Smirnov test (P < 0.001). [C(ii)] Bar plot of the percentage of responding neurons classed as silent cold-sensing neurons in the vehicle and oxaliplatin groups. Proportions were compared using a χ² test, and error bars denote 95% confidence intervals. (D) Relationship between the number of basal cold-sensing neurons and the drop in temperature can be fit by linear regression for both groups. For vehicle, y = −2.883x + 105.2, r² = 0.9809, n = 87. For basal cold-sensing neurons after oxaliplatin, y  = −3.443x + 105, r² = 0.9802, n = 39. The slopes are not significantly different (P = 0.12). (E) Quantification of the proportion of cold-sensing neurons responding to either heat (i) or mechanical (ii) stimuli in the vehicle and oxaliplatin groups. The proportion of polymodal neurons was compared using a χ² test, and error bars denote confidence intervals. Ice water: n_veh = 51, n_oxa = 81. Acetone: n_veh = 58, n_oxa = 145. [E(ii)] Cumulative probability plots showing mechano-cold neurons tend to have larger cross-sectional areas in the oxaliplatin group, compared using the Kolmogorov-Smirnov test. n_veh = 14, n_oxa = 62. For this experiment, 383 neurons responding to any stimulus were recorded in eight vehicle-treated mice (five males and three females) and 542 cells were recorded from nine oxaliplatin-treated animals (five males and four females).

Figure 2

Silent cold sensing neurons are activated after partial sciatic nerve ligation. (A) Behavioural testing of the effect of PNL on different sensory modalities. n = 3 (one male and two females) for sham and n = 6 (three males and three females) for PNL. For Von Frey and acetone tests, means over time were compared using repeated measures two-way ANOVA followed by post hoc Sidak’s test. Hot plate latencies at 4 weeks were compared using unpaired t-test. Error bars denote the SEM. (B) Example images (i) and traces (ii) of cold-responding neurons in sham- and PNL-operated animals expressing GCaMP3. Cell 1 is a small diameter cold-sensing neuron in the sham condition; Cell 2 is a small diameter basal cold-sensing neuron after PNL; and Cell 3 is a large diameter silent cold-sensing neuron unmasked by PNL. [C(i)] Histograms of the cross-sectional area of all neurons responding to any cold stimulus in the sham (top, blue, n = 113) and PNL (bottom, red, n = 109) groups. The distribution of areas for sham was fit by non-linear regression (least squares Gaussian; bin width is 80 µm²; mean = 222.7 µm², SD 60.9 µm²). This model is plotted over the PNL data to aid comparison with the dashed line denoting 3 SD from the mean. The difference in the distribution of areas between groups was assessed by Kolmogorov-Smirnov test (P < 0.001). [C(ii)] Bar plot of the percentage of responding neurons classed as silent cold-sensing neurons in the sham and PNL groups. Proportions were compared using a χ² test, and error bars denote confidence intervals. (D) Relationship between the number of basal cold-sensing neurons and the drop in temperature can be fit by linear regression for both groups. For sham, y = −3.603x + 101.6, r² = 0.9979, n = 51. For PNL, y = −3.875x + 107.8, r² = 0.9598, n = 40. The slopes are not significantly different (P = 0.66). (E) Quantification of the proportion of cold-sensing neurons responding to either heat (i) or mechanical (ii) stimuli in the sham and PNL groups. The proportion of polymodal neurons was compared using a χ² test, and error bars denote confidence intervals. Ice-water: n_sham = 64, n_PNL = 71. Acetone: n_sham = 95, n_PNL = 73. [E(iii)] Scatter plots showing mechano-cold neurons have both small and large cross-sectional areas in the PNL group. n_sham = 3, n_PNL = 19. For this experiment, 373 neurons responding to any stimulus were recorded in three sham-operated mice (one male and two females) and 297 cells were recorded from six PNL-operated animals (three males and three females).

Figure 3

Silent cold-sensing neurons are activated by ciguatoxin-2. (A) Behavioural testing of the effect of intraplantar injection of 100 nM ciguatoxin-2 (P-CTX-2) on cold sensitivity n = 6 for sham vehicle (three males and three females) and n = 6 for P-CTX-2 (three males and three females). Means were compared by repeated measures two-way ANOVA followed by post hoc Sidak’s test. Error bars denote 95% confidence interval. (B) Example images and traces of a large-diameter neuron (Cell 1) that is basally cold-insensitive but begins to respond to cooling after treatment with P-CTX-2. [C(i)] Heat map showing the effect of P-CTX-2 on the number of neurons responding to a cold ice-water stimulus. n = 48 for vehicle and n = 196 for P-CTX-2. The bar corresponds to 15 s. [C(ii)] Summary of the change in the number of sensory neurons responding to each modality after treatment with P-CTX-2. (D) Histograms of cross-sectional area of all neurons responding to any cold stimulus in the naïve state (left, blue, n = 91) and after P-CTX-2 (right). For P-CTX-2, blue denotes basally responsive neurons that maintained their response to cold (n = 70) and red denotes the silent cold-sensing neurons that were unmasked after treatment (n = 136). The distribution of areas in the naïve state was fit by non-linear regression (least squares Gaussian; bin width is 80 µm²; mean = 212.4 µm², SD 73.33 µm²). This model is plotted over the P-CTX-2 data to aid comparison with the dashed line denoting 3 SD from the mean. The different in the distribution of areas between groups was assessed by Kolmogorov-Smirnov test (P < 0.001). [E(i)] Box plot of the change in activation threshold of basally cold-sensitive neurons before and after treatment with vehicle (n = 35) or P-CTX (n = 8). [E(ii)] Box plot of the thermal activation threshold of all silent cold-sensing neurons unmasked by P-CTX-2 (n = 43) compared to all cold-sensing neurons recorded from naïve mice (n = 62). Medians were compared by Mann-Whitney test. (F) Quantification of the proportion of neurons responding ice-water that were also sensitive to either mechanical (i) or heat (ii) before and after treatment. Vehicle: n_pre = 36, n_post = 43. P-CTX-2: n_pre = 69, n_post = 174. [F(iii)] Comparison of the proportion of silent cold-sensing neurons that were responsive to other modalities before and after the induction of cold-sensitivity by P-CTX-2. n = 127. The proportion of polymodal neurons was compared using a χ² test, and error bars denote 95% confidence intervals. For this experiment, 615 neurons responding to any stimulus either before or after treatment were recorded in 10 P-CTX-2-injected mice (four males and six females) and 193 cells were recorded from three vehicle-injected animals (two males and one female).

Figure 4

Silent cold-sensing neurons express peptidergic nociceptor molecular markers Na_v1.8 and CGRPα. (A) Cartoon (left) of breeding strategy used to generate GCaMP3 reporter mice for each subset of interest. Bar plot (right) showing overlap of reporter expression for each marker with functionally defined cold-sensing neurons. TrkB-CreERT2 (Ntrk2): n_veh = 14 from two mice (one male and one female), n_oxa = 112 from three mice (two males and one female). Calb1-Cre (Calb1): n_veh = 7 from one mouse (one male), n_oxa = 15 from two mice (two females). Trpv1-Cre (Trpv1): n_veh = 4 from one mouse (one male), n_oxa = 87 from two mice (one male and one female). Na_v1.8-Cre (Scn10a): n_veh = 14 from four mice (two males and two females), n_oxa = 108 from six mice (four males and two females). Tmem45b-Cre (Tmem45b): n_veh = 40 from one mouse (one male), n_oxa = 36 from three mice (two males and one female). CGRPα-CreERT2 (Calca): n_veh = 45 from three mice (one male and two females), n_oxa = 122 from two mice (one male and one female). (B) Example images (left) and histograms (right) showing overlap of Na_v1.8-Cre-dependent tdTomato expression with cold-sensing neurons of different sizes in vehicle- and oxaliplatin-treated mice. Same data as in A. (C) Histogram (left) and bar plot (right) showing overlap of Na_v1.8-Cre-dependent tdTomato expression with different types of cold-sensing neurons in PNL-operated mice. n = 57 cells from two mice (one male and one female). (D) Histograms (left) and bar plot (right) showing overlap of Na_v1.8-Cre-dependent tdTomato expression with basally active and silent cold-sensing neurons in mice treated with P-CTX-2. n = 56 cells from four mice (one male and three females). (E) Example images (left) and histograms (right) showing overlap of CGRPα-CreERT2-dependent tdTomato expression with cold-sensing neurons of different sizes in vehicle- and oxaliplatin-treated mice. Same data as in A. (F) Histogram (left) and bar plot (right) showing overlap of CGRPα-CreERT2-dependent tdTomato expression with different types of cold-sensing neurons in PNL-operated mice. n = 16 cells from one mouse (one male). (G) Histograms (left) and bar plot (right) showing overlap of CGRPα-CreERT2-dependent tdTomato expression with basally active and silent cold-sensing neurons in mice treated with P-CTX-2. n = 56 cells from two mice (one male and one female). Error bars denote 95% confidence intervals. As these data were obtained as part of an exploratory screen, no statistical hypothesis testing was performed.

Figure 5

Diphtheria toxin-mediated ablation of Na_v1.8-positive nociceptors decreases oxaliplatin-induced cold allodynia. (A) Cartoon of diphtheria toxin-mediated ablation of Na_v1.8-positive neurons. [B(i)] Histogram of cross-sectional areas of all cold-sensing neurons imaged in Na_v1.8-Cre DTA mice treated with oxaliplatin. [B(ii)] Cumulative probability plot of cell areas in oxaliplatin-treated Na_v1.8-Cre (blue) and Na_v1.8-Cre DTA (red) mice, compared by Kolmogorov-Smirnov test. The distribution of cell areas in vehicle-treated Na_v1.8-Cre mice is shown for comparison. n = 108 cells from six oxaliplatin-treated Na_v1.8-Cre mice (four males and two females), n = 46 cells from two oxaliplatin-treated Na_v1.8-Cre DTA mice (one male and one female) and n = 14 cells from four vehicle-treated Na_v1.8-Cre mice (two males and two females). (C) Quantification of the number of nociceptive behaviours in 5 min on the 5°C cold plate in 10 control and 8 Na_v1.8-Cre DTA mice treated with oxaliplatin.

Figure 6

Voltage-gated sodium channel Na_v1.6 is required for excitability, but is not sufficient for cold sensitivity, of silent cold-sensing neurons. [A(i)] Histogram of cross-sectional areas of all cold-sensing neurons imaged in homozygous Na_v1.8-Cre tdTomato mice with oxaliplatin. [A(ii)] Cumulative probability plot of cell areas in oxaliplatin-treated heterozygous and homozygous Na_v1.8-Cre mice, compared by Kolmogorov-Smirnov test. [A(iii)] Bar plot showing the proportion of silent cold-sensing neurons expressing tdTomato in heterozygous and homozygous Na_v1.8-Cre mice, compared using a χ² test. n = 66 cells from three heterozygous (two males and one female) and n = 42 cells from three homozygous (one male and two females) Na_v1.8-Cre mice. (B) Cumulative probability plot of cell areas in oxaliplatin-treated wild-type and Na_v1.7 KO mice, compared by Kolmogorov-Smirnov test. n = 51 cells from five wild-type (one male and four females) and n = 18 cells from two Na_v1.7 knockout mice (two females). (C) Heat maps (i) and quantification (ii) showing the effect of intraplantar injection of different sodium channel blockers on the number of basal and silent cold-sensing neurons in mice pretreated with oxaliplatin. [C(iii)] Line plot showing the effect of blockers on median peak response, compared using Kruskall-Wallis test followed by Dunn’s multiple comparisons test. n = 35 cells from three mice for saline (one male and two females), n = 58 cells from four mice for TTX (three males and one female), n = 36 cells from two mice for 4,9-anhydrous-TTX (two females). [D(i)] Heat map showing the effect of sodium channel activation by intraplantar veratridine injection on the activity of cold-sensing neurons. [D(ii)] Box plot showing the size of cold-sensing cells is unaffected by veratridine, compared by Mann-Whitney test. [D(iii)] Bar plot showing veratridine reduces the number of cold-sensing neurons from 39 to 31. [D(iv)] Line plot showing veratridine reduces the response magnitude of cold-sensing cells, compared by Wilcoxon matched-pairs signed rank test. n = 53 cells from three mice (three males).

Figure 7

Blocking K_v1.1 voltage-gated potassium channels is sufficient to induce de novo cold sensitivity in silent cold-sensing neurons. (A) Examples images and traces showing that peripheral blockade of voltage-gated potassium channels induces novel cold-sensitivity in normally cold-insensitive sensory neurons (Cell 1). (B) Heat maps showing the effect of intraplantar injection of different potassium channel blockers on the peripheral representation of cold. The bar denotes 15 s. (C) Quantification showing the change in the number of cold-sensing neurons after treatment with different potassium channel blockers. (D) Violin plots showing the cross-sectional area of basal cold-sensing neurons in the naïve state (blue) and of silent cold-sensing neurons unmasked by potassium channel block (red). Medians were compared using Kruskall-Wallis test followed by Dunn’s multiple comparison test. (E) Bar plot of the percentage of polymodal cold-sensing neurons that also respond to noxious mechanical stimuli before (blue) and after (red) treatment with potassium channel blockers. Proportions were compared using a χ² test. Error bars denote 95% confidence intervals. (F) No change in the median response magnitude of neurons that responded to cold both before (blue) and after (red) treatment with potassium channel blockers, as determined by Kruskall-Wallis test followed by Dunn’s multiple comparison test. n = 42 from three saline-treated mice (one male and two females), n = 57 from six 4-AP-treated mice (five males and one female), n = 95 from three 4-AP-treated mice pre-injected with oxaliplatin (two males and one female), n = 101 from four α-dendrotoxin-treated mice (two males and two females), n = 48 from three k-dendrotoxin-treated mice (one male and two females), and n = 14 from three RIIIJ-treated mice (three females).

Figure 8

Proposed model of silent cold-sensing neuron activation during neuropathy to cause cold allodynia. In healthy mice, only small neurons respond to cold. The large-diameter silent cold-sensing neurons have high K_v1 activity, thus cold-induced terminal depolarization does not trigger action potential firing and there is no subsequent GCaMP signal at the level of the dorsal root ganglion. In neuropathic animals, we hypothesize that a functional reduction in K_v1 activity means that silent cold-sensing neurons are now sensitive to cold, increasing nociceptive input to the brain in response to cooling. Thus, both small and large neurons now show GCaMP signals to cold stimuli.