Anandamide applied for 20 seconds at a concentration range (1 μM, 3 μM, 10 μM and 30 μM) known to induce excitation³,¹⁶ of cultured rat PSN, produced concentration-dependent inward currents at −60 mV membrane potential with an EC₅₀ of 2.2 μM (Fig. 1A,B; Supplementary Table 1). The excitatory effect of anandamide was washed off within a few tens of seconds after stopping anandamide application (Fig. 1A,C). Anandamide at 30 μM reached its maximal excitatory effect (Fig. 1B).

All the anandamide-responding neurons that were tested for capsaicin-responsiveness (Supplementary Table 2) produced currents to capsaicin applied at around its EC₅₀ value (500 nM¹,⁴), 2 minutes after anandamide superfusion (n = 29; Fig. 1C,D). However, a sub-population of neurons, while not responding to anandamide responded only to capsaicin (n = 10; Fig. 1E,F; Table 1 and Supplementary Table 2). Hence, anandamide- and capsaicin-responsiveness defined two sub-populations of PSN: the “anandamide-and-capsaicin-responsive” (ACR) and the “capsaicin-only-responsive” (COR) neurons (Fig. 1C–F).

The proportion of COR neurons in the overall sample of capsaicin-responsive cells was 25.6% (10 of the 39 cells; Table 1; Supplementary Table 2) and independent of anandamide concentration up to 30 μM (Supplementary Table 2; p = between 1 and 0.6, Fisher’s exact test). The concentration of anandamide, applied before capsaicin application, did not have significant effect on the amplitude of capsaicin-evoked responses in ACR or COR neurons (data not shown; p = between 0.98 and 0.11, ANOVA). Therefore, we pooled the amplitudes of capsaicin-evoked responses in ACR and COR neurons, respectively. The pooled capsaicin-evoked amplitudes were significantly different (ACR: −3.19 ± 0.31 nA, n = 29; COR: −0.96 ± 0.27 nA, n = 10; p = 0.002, Student’s t-test).

Since the anandamide concentration had no effect on either the proportion of responsive cells or the response to subsequent capsaicin application, we used anandamide at 30 μM in the rest of the experiments.

TRPV1 is highly permeable to Ca^2+ 1,⁴. Therefore, to confirm the presence of ACR and COR neurons, we studied changes in the intracellular Ca²⁺ concentration ([Ca²⁺]_i) by anandamide (30 μM) and capsaicin (500 nM) in rat cultured PSN grown for 16–22 hours.

Of the 98 KCl-responding neurons, 57 responded to capsaicin. Only 40 of the capsaicin-responsive neurons responded also to anandamide (Supplementary Figure 1). The ACR:COR neuron ratio was not significantly different from that found using whole-cell recordings (Table 1). The amplitude of the capsaicin-evoked calcium transients was significantly greater in ACR than in COR neurons (1.01 ± 0.12 in ACR (n = 40) and 0.49 ± 0.13 (n = 17); Supplementary Figure 1; p = 0.015, Student’s t-test). The TRPV1 antagonist capsazepine (10 μM) blocked the anandamide-evoked Ca²⁺ transients indicating that the anandamide-evoked excitatory effect is indeed mediated through TRPV1 in PSN (Supplementary Figure 2).

Next we assessed 30 μM anandamide- and 500 nM capsaicin-evoked changes in [Ca²⁺]_i in PSN cultures prepared from WT-TRPV1 and TRPV1^−/− mouse DRG (Fig. 1G–I) to determine whether ACR and COR neurons are also present in another species and the excitatory effects by both anandamide and capsaicin are indeed mediated exclusively through TRPV1. To verify chemosensitivity of PSN isolated from TRPV1^−/− mice, we also administered mustard oil (50 μM) following capsaicin application to the cells.

Responses of 136 KCl-responding WT-TRPV1 cells were analysed. Of these, 54 neurons responded to capsaicin, of which 37 were ACR and 17 were COR neurons (Fig. 1G,H; Table 1). The ACR:COR ratio was not significantly different from that found with whole-cell recordings or Ca²⁺ imaging from rat cultured PSN (Table 1). The increase in [Ca²⁺]_i induced by capsaicin was significantly smaller in COR neurons than in ACR cells (ACR: 2.09 ± 0.13 (n = 37); COR: 0.81 ± 0.17 (n = 17); p < 0.0001, Student’s t-test).

Neither anandamide nor capsaicin increased the [Ca²⁺]_i in the 126 TRPV1^−/− neurons we assessed (Fig. 1I). However, 36 of the 126 neurons responded to mustard oil indicating that in general, chemosensitivity of the neurons remained intact (Fig. 1I).

These data show that anandamide- and capsaicin-responsiveness are partially segregated in rodent PSN. Although we cannot categorically exclude the possibility that compensatory changes in TRPV1^−/− mice contribute to the lack of anandamide responsiveness, we suggest that both the anandamide- and capsaicin-evoked excitation depend on TRPV1 in PSN. Considering that TRPV1 mediates the excitatory effect of both capsaicin and anandamide, the behaviour of ACR and COR neurons is underlain by differential responsiveness of TRPV1 to these agents in the two types of neurons.

The two main sub-populations of nociceptive PSN, the peptidergic and Bandeiraea simplicifolia lectin (IB4)-binding cells¹⁷, exhibit different capsaicin-evoked responses¹⁸,¹⁹. However, our data indicate that the presence or absence of neuropeptides do not contribute to the partial segregation of anandamide- and capsaicin-responsiveness in PSN (Supplementary Table 3).

The CB1 receptor is able to regulate TRPV1 responsiveness¹⁶,²⁰,²¹,²²,²³,²⁴,²⁵,²⁶,²⁷. Further, rimonabant at less than ~200 nM is a selective and specific CB1 receptor antagonist/inverse agonist²⁸,²⁹. Therefore, we used 200 nM rimonabant to determine whether the CB1 receptor contributes to the segregation of responses.

First we used whole-cell voltage-clamp recordings from cultured rat PSN to facilitate comparison with data obtained in the first experiment (Table 1). Rimonabant, which was continuously superfused, did not change the ACR:COR neuron ratio (Fig. 2A; Table 1). However, interpretation of this finding requires caution due to the relatively low number of neurons. Nevertheless, the amplitude of 30 μM anandamide-evoked currents was significantly reduced from −1.07 ± 0.33 nA (n = 7) to −0.59 ± 0.18 nA (n = 11; Fig. 2B; p = 0.006, Student’s t-test; Supplementary Table 2) with rimonabant. Rimonabant significantly reduced the amplitude of the capsaicin-evoked responses in ACR neurons (from −3.19 ± 0.31 nA (n = 29) to −1.82 ± 0.40 nA (n = 11); p < 0.001, Student’s t-test) but not in COR neurons (−0.96 ± 0.27 nA (n = 10) to −0.54 ± 0.31 nA (n = 7), p = 0.34, Student’s t-test; Fig. 2C). Because of the high variability of data on COR neurons, interpretation requires caution. Nevertheless, these findings did suggest that the CB1 receptor might have a sensitising effect on capsaicin-evoked responses in ACR neurons, which might not be present in COR neurons. Subsequently, we conducted experiments to confirm the CB1-mediated sensitising effect on TRPV1 and to study the underlying mechanisms.

First, to confirm the role of the CB1 receptor in sensitising TRPV1 in ACR neurons, we isolated PSN from Biozzi ABH (WT-CB1) and Biozzi ABH mice lacking the CB1 receptor (CB1^−/−). We have chosen these mice, because they exhibit a significantly altered endocannabinoid system in comparison to C57BL/6 mice, and we used Biozzi ABH mice in previous studies³⁰. Due to the significantly altered endocannabinoid system, we expected to find differences in TRPV1-mediated responses in C57BL/6 (background for TRPV1^−/− mice) and Biozzi ABH mice if our hypothesis on the sensitising effect of the CB1 receptor on TRPV1 were correct. To increase the efficacy of sampling, we again assessed changes in the [Ca²⁺]_i.

Anandamide increased the [Ca²⁺]_i in 113 of 230 KCl-responding neurons of WT-CB1 mice, all of which responded to capsaicin (Fig. 3A). In addition to the 113 ACR neurons, 11 of the 230 neurons exhibited responses only to capsaicin (Fig. 3A). The ACR:COR neuron ratio was significantly higher in WT-CB1 mice than in WT-TRPV1 mice or rats (Table 1). These data supports our view that the cannabinoid system contributes to regulating TRPV1 responsiveness. To get direct evidence for the involvement of the CB1 receptor, we next measured responses in PSN cultures prepared from CB1^−/− mice.

The proportion of neurons responding to anandamide was significantly reduced within the total number of capsaicin-responding cells in CB1^−/− mice (Fig. 3A; Table 1). At the same time, the proportion of COR neurons within the total number of capsaicin-responding neurons was significantly increased (Fig. 3A; Table 1). However, the overall proportion of neurons responding to 500 nM capsaicin in CB1^−/− mice (115 of 200) was not statistically different to that in WT-CB1 mice (124 of 230, p = 0.56, Fischer’s exact test).

Deleting the CB1 receptor significantly reduced the amplitude of both anandamide (from 0.42 ± 0.03, n = 113, to 0.30 ± 0.03, n = 74, p = 0.02, Student’s t-test; Fig. 3B) and capsaicin-evoked responses (from 0.56 ± 0.06, n = 113, to 0.39 ± 0.04, n = 74, p = 0.01, Student’s t-test; Fig. 3C) in ACR neurons. However, there was no significant difference in the amplitude of capsaicin-evoked responses between COR neurons isolated from WT-CB1 and CB1^−/− mice (from 0.45 ± 0.11, n = 11, to 0.38 ± 0.04, n = 41, p = 0.42, Student’s t-test; Fig. 3C). These findings confirm that the CB1 receptor executes a sensitising effect on TRPV1, which increases capsaicin- and anandamide-evoked responses in ACR neurons. Further, these findings also indicate that in addition to sensitising TRPV1, the CB1 receptor also contributes to the partial segregation of anandamide- and capsaicin-responsiveness in PSN.

CB1 receptor activity can be evoked by endogenous ligands, such as anandamide, synthesised by a major proportion of TRPV1-expressing (hence CB1 receptor-expressing¹¹,¹²) PSN²⁷,³¹,³²,³³. Alternatively, the CB1 receptor-mediated effects could be due to its constitutive activity²⁸,³⁴. To differentiate between these mechanisms, we next compared the effects on capsaicin-evoked responses of rimonabant, which blocks both evoked and constitutive activity, and NESS0327³⁵, which only inhibits ligand binding. To exclude the possibility of CB1 receptor activation by exogenous anandamide, capsaicin application was not preceded by anandamide superfusion. We assessed changes in the [Ca²⁺]_i to increase efficacy and we applied 100 nM capsaicin to diminish TRPV1 desensitisation³⁶.

Capsaicin was applied 3 times consecutively with 2 minutes intervals to identify capsaicin responsive cells. The further application of 30 μM anandamide revealed 161 ACR and 36 COR neurons in rat cultured PSN (Fig. 4; Table 1). This ratio was not significantly different from our previous findings with either whole-cell recordings or Ca²⁺ imaging from cultured rat PSN (Table 1). The average amplitudes of the first capsaicin-evoked responses in ACR neurons (1.07 ± 0.05, n = 161) was significantly greater than that in COR neurons (0.51 ± 0.07, n = 36; p < 0.0001, Student’s t-test; Fig. 4C,F).

100 nM capsaicin induced desensitisation in both ACR and COR neurons (Fig. 4A,C,D,F). Interestingly, the desensitisation was significantly greater in COR than in ACR neurons (Fig. 4A,C,D,F).

While rimonabant, applied immediately after the first capsaicin application until the end of the second capsaicin application (Fig. 4B,E), did not have any effect on the [Ca²⁺]_i per se, it significantly increased the rate of reduction in the average amplitude of the second capsaicin-evoked responses in ACR (Fig. 4B,C; p = 0.0001, Student’s t-test) but not in COR neurons (Fig. 4E,F; p = 0.62, Student’s t-test). Although, several ACR neurons exhibited partial recovery in their capsaicin-evoked responses after stopping rimonabant application (Fig. 4B), the rate of reduction in the average amplitude of the third capsaicin-evoked responses in these cells was still significantly greater when compared to that of the control experiment (Fig. 4C; p < 0.0001, Student’s t-test). No such differences were seen in COR neurons (Fig. 4F, p = 0.084; Student’s t-test). Notably, in addition to reducing the amplitude of the capsaicin-evoked responses in ACR cells, rimonabant also prolonged the recovery of responses in a group of both types of neurons (Supplementary Figure 3).

In contrast to rimonabant, NESS0327 at 100 nM per se substantially enhanced the duration of the capsaicin-evoked Ca²⁺ transients (Supplementary Figure 4A,B). This increase did not allow further assessment in the great majority of cells (Supplementary Figure 4A,B). The [Ca²⁺]_i-increasing effect was less pronounced at 1 nM, and when NESS0327 was applied 30 seconds after stopping capsaicin application an apparently capsaicin-independent increase in the [Ca²⁺]_i was produced (Supplementary Figure 4C). In this condition, 1 nM NESS0327 did not produce any effect on the amplitude of the second capsaicin application-evoked responses in COR neurons (n = 9; p = 0.74, Student’s t-test), whereas an increase was apparent in ACR neurons (n = 25; control: 94.66 ± 1.74% of the first response; NESS0327: 116.76 ± 9.52% of the first response; p = 0.0002, Student’s t-test; Supplementary Figure 4D,E). Although the excitatory effect of NESS0327 on PSN, the elucidation of which was not in the scope of the present study, makes interpretation of these findings very difficult, the lack of inhibitory effect may suggest that ligand-induced activation of the CB1 receptor does not contribute to TRPV1 sensitisation in PSN.

The finding that CB1 receptor inhibition or deletion has no effect on COR neurons may suggest that COR neurons might not express the CB1 receptor. To test this hypothesis, we next studied CB1 receptor expression in PSN by single-cell reverse transcription polymerase chain reaction (RT-PCR).

We collected 25 COR, 31 ACR and 20 cells, which did not respond to either of the TRPV1 activators (NR) from rat PSN cultures. The ACR:COR neuron ratio was not significantly different from the observation we made in whole-cell recordings from cultured rat PSN (Table 1). The amplitude of capsaicin-evoked responses was significantly larger (p < 0.0001, Student’s t-test) in ACR (−3.28 ± 0.35 nA, n = 31) than in COR (−0.67 ± 0.06 nA, n = 25) neurons. RNA of 10 COR, 11 ACR and 8 NR cells were processed successfully with single-cell RT-PCR (i.e. the sample showed GAPDH mRNA expression; Fig. 5A,B).

CB1 receptor primers produced amplicons of the predicted size, whereas no amplicons were present when no template was included in the PCR mixture (Fig. 5A; Supplementary Figure 5). All ACR neurons (n = 11), eight of ten COR neurons, and four of eight NR neurons exhibited detectable CB1 receptor mRNA expression (Fig. 5B). Immunolabelling of cultured rat PSN using an anti-CB1 receptor and anti-TRPV1 antibody confirmed the high proportion of TRPV1-expressing cells also expressing the CB1 receptor (149 of 150, Table 1; Supplementary Figure 6).

For the control of immunostaining and assessing possible differences between the intact DRG and cultured PSN, we also established the co-expression ratio of TRPV1 and the CB1 receptor in the intact DRG (Supplementary Figures 7 and 8). We found a slightly smaller proportion of TRPV1-expressing cells exhibiting obvious immunopositivity for the CB1 receptor in intact DRG than in PSN cultures (88.3 ± 3.8%, n = 3). This small difference could be due to culturing. Nevertheless, together, these findings indicate that the differing effects of rimonabant or deleting the CB1 receptor in ACR and COR neurons cannot be due to the exclusive expression of the CB1 receptor in ACR type neurons. Therefore, next we assessed the spatial distribution pattern of the CB1 receptor and TRPV1.

Immunocytochemistry on SDS-digested freeze-fracture replicas (SDS-FRL) is a highly effective method to reveal sub-cellular protein distribution³⁷. TRPV1-labelling, which served as a basis for selection of putative somatic membranes of PSN, was generally moderate to strong (Fig. 6A,B). All TRPV1-expressing membranes we examined exhibited immunolabelling for the CB1 receptor, which was also moderate to strong in general. Both TRPV1 and CB1 receptor immunoreactivities were specifically detected on the protoplasmic face (Fig. 6A,B). Virtually no labelling was observed on the exoplasmic face (Fig. 6A,B), which is in agreement with the epitope of both the anti-TRPV1- and anti-CB1 receptor antibodies raised against intracellular part of the respective molecules (see Material and Methods).

Membrane patches co-expressing TRPV1 and CB1 receptors differed in receptor densities and distribution (Fig. 6A,B). Some of the patches exhibited only a few isolated receptors (Fig. 6A). The majority of the membrane patches however exhibited a relatively high density of CB1 receptor expression, which was accompanied by moderate to high TRPV1 receptor density (Fig. 6B). It is unlikely that the differences in densities could be due to major regional variation of receptor distribution within a single cell, because no such differences were evident by visual assessment of about 100 TRPV1 and CB1 receptor co-expressing cultured PSN neurons or PSN in intact DRG (Supplementary Figure 6 and Supplementary Figure 8).

We found that two, three or sometimes four immunoparticles labelling TRPV1 in these membranes often formed clusters (Fig. 6B). The distance of the gold particles in these clusters was between ~10 and ~55 nm (Fig. 6B). Although the majority of the CB1 receptors appeared solitary, some of them seemed to be in close vicinity (<~55 nm) to TRPV1 (Fig. 6B).

We established morphometric data of 21 membrane patches (Table 2), which had an average size of 5.20 ± 0.39 μm² and exhibited 2788 gold particles of which 1793 (85.38 ± 16.03, n = 21, on average) and 995 (47.38 ± 6.78, n = 21, on average) labelled CB1 receptor and TRPV1, respectively. The densities for both receptors varied in a wide range (Table 2) and the CB1 receptor density was significantly higher than that of TRPV1 (p = 0.024, Student’s t-test). Correlations between some properties are shown in Supplementary Figure 9.

We also established the number of distances measured between each TRPV1 and its closest CB1 receptor neighbour as well as between each CB1 receptor and its closest TRPV1 neighbour, which were below 54.5 nm (Table 2). This threshold represents the maximum (“critical”) distance between the centres of the gold particles if the two receptors are positioned next to each other, and is based on the estimated size of the CB1 receptor, TRPV1, IgG and the gold particles³⁸,³⁹,⁴⁰,⁴¹ (Table 2; Supplementary Figure 10A). The number of distances between each TRPV1 and the closest TRPV1 neighbour shorter than 56 nm, which represents the estimated maximum distance if TRPV1 molecules are located next to each other (Supplementary Figure 10B), was also established. The number of critical distances was then divided by the number the appropriate receptor in each patch.

Principal component analysis (PCA), an unsupervised multivariate statistical tool, which reduces the number of variables and maximises the variability of data, suggested the presence of 2 major groups and 3 outliers (Supplementary Figure 11). PCA’s loading plot suggested that the number of CB1 receptor – TRPV1 critical distances does not contribute significantly to the variability of data (not shown). After removing the outliers and non-significant variables, PCA identified three principal components (PCs) that together contributed to more than 80% of the variability. The PCA score plot suggested the presence of 2–4 groups of cells (Supplementary Figure 11).

To study the putative groupings in more details, we built a partial least square discriminant analysis (PLS-DA) model, which again suggested the presence of 3 PCs (Fig. 7A). The PLS-DA score plot clearly showed the separation of three major groups (Fig. 7B). The loading plot (Fig. 7C) shows how, while the variable importance of projection plot (Fig. 7D) reveals the significance of, the various morphometric properties contributed to the separation of the membrane patches.

Neurons Membrane patches in Groups 1 (yellow in Fig. 7B) exhibited significantly smaller CB1 receptor densities, larger TRPV1 – CB1 receptor average distances, and fewer TRPV1 – CB1 receptor critical distances than patches in Group 2 (red in Fig. 7B) and Group 3 (purple in Fig. 7B,E,F; Table 2). Patches in Group 2 were different from patches in Group 3 mainly in TRPV1 distribution (Fig. 7E,F; Table 2). Together these data indicate that the distribution of the CB1 receptor and TRPV1 enables an intensive crosstalk in a large proportion of PSN (Group 2 and Group 3 neurons) and, due to the large distances between the two receptors, a limited crosstalk in a small group of neurons (Group 1). Importantly, the proportion of Group 1 – Group 2/Group 3 patches was not significantly different from ACR – COR neuron proportion found in rat or C57BL/6 mice (Table 1).

The presence of a significant proportion of CB1 receptors and TRPV1 closer than 54.5 nm to each other in PSN suggests that the two receptors, similarly to the cornea²⁶, could be engaged in protein-protein interaction. Therefore we next performed immunoprecipitation by either an anti-TRPV1 or an anti-CB1 receptor antibody followed by immunoblotting respectively with an anti-CB1 receptor or anti-TRPV1 antibody on protein samples isolated from rat DRG. We found in both experiments that the immunoprecipiated proteins contained the other molecule (Fig. 8).