Endocannabinoids Have Opposing Effects On Behavioral Responses To Nociceptive And Non-nociceptive Stimuli

The endocannabinoid system is thought to modulate nociceptive signaling making it a potential therapeutic target for treating pain. However, there is evidence that endocannabinoids have both pro- and anti-nociceptive effects. In previous studies using Hirudo verbana (the medicinal leech), endocannabinoids were found to depress nociceptive synapses, but enhance non-nociceptive synapses. Here we examined whether endocannabinoids have similar bidirectional effects on behavioral responses to nociceptive vs. non-nociceptive stimuli in vivo. Hirudo were injected with either the 2-arachidonoylglycerol (2-AG) or anandamide and tested for changes in response to nociceptive and non-nociceptive stimuli. Both endocannabinoids enhanced responses to non-nociceptive stimuli and reduced responses to nociceptive stimuli. These pro- and anti-nociceptive effects were blocked by co-injection of a TRPV channel inhibitor, which are thought to function as an endocannabinoid receptor. In experiments to determine the effects of endocannabinoids on animals that had undergone injury-induced sensitization, 2-AG and anandamide diminished sensitization to nociceptive stimuli although the effects of 2-AG were longer lasting. Sensitized responses to non-nociceptive stimuli were unaffected 2-AG or anandamide. These results provide evidence that endocannabinoids can have opposing effects on nociceptive vs. non-nociceptive pathways and suggest that cannabinoid-based therapies may be more appropriate for treating pain disorders in which hyperalgesia and not allodynia is the primary symptom.

function of many neurons in each of these ganglia is known in considerable detail 20 . In terms of somatosensory signaling, the Hirudo CNS possesses three bilateral pairs of rapidly-adapting touch-sensitive neurons (T cells), two pairs of slow-adapting pressure-sensitive neurons (P cells) and two pairs of high-threshold nociceptive neurons (N cells) 21 . The N cells can be further divided into mechanical and polymodal nociceptors, with the latter being sensitive to noxious mechanical, thermal and chemical stimuli, e.g., H + , capsaicin and mustard oil [22][23][24][25] . P cell stimulation is capable of producing localized withdrawals from mechanical stimuli referred to as local bending and local shortening [26][27][28] . Hirudo are also capable of a whole body shortening reflex in which the entire animal withdraws from a noxious stimulus in a coordinated manner 29 . Whole-body shortening can be elicited by the P cells if multiple P cell receptive fields are activated, whereas this reflex can be activated by a single N cell 29,30 .
Previous studies in Hirudo have shown that 2-AG and anandamide elicit long-lasting (≥1 hr) depression in nociceptive N cell synapses and potentiation in non-nociceptive P cell synapses (summarized in Fig. 1A) [30][31][32][33][34][35] . These studies suggest that endocannabinoid effects on both synapses are mediated by a TRPV-like channel. The synaptic effects of 2-AG have also been observed at the behavioral level using semi-intact preparations in which it possible to monitor both physiological and behavioral changes. Specifically, N cell elicited whole-body shortening was reduced by 2-AG 30 . However, there has been no attempt to examine whether the effects observed in such reduce preparations can also be seen in intact animals. Therefore, the current study examined the effects of 2-AG and anandamide on behaviors elicited by non-nociceptive vs. nociceptive stimuli in vivo. Consistent with our earlier physiological studies, 2-AG and anandamide enhanced responses to non-nociceptive mechanical stimuli and reduced responses to nociceptive chemical stimuli. Furthermore, in animals that had undergone injury-induced sensitization, 2-AG and anandamide reversed the sensitized responses to nociceptive stimuli, but had no effect on sensitized responses to non-nociceptive stimuli. (A) Endocannabinoids directly depress the nociceptive (N) synapse via a presynaptic TRPV-like receptor that reduces neurotransmitter release 31,34,35 . Endocannabinoids potentiate the non-nociceptive synapse (P) via an indirect mechanism in which endocannabinoids depress inhibitory input from an unknown GABAergic interneuron (this is also TRPV-mediated). In N cells the Cl − gradient is dominated by the Cl − importer (NKCC1) while the P cell gradient is dominated by the Cl − exporter (KCC1) 36 . As a result of an elevated intracellular Cl − concentration, N cells are depolarized by GABA and consequently "protected" from disinhibition. (B) For tests of endocannabinoid effects in uninjured animals, Hirudo were given a pre-test to assess initial responsiveness to non-nociceptive and nociceptive stimuli 10 minutes prior to injection of either anandamide (100 μM) or 2-AG (75 μM), with or without 25 μM SB366791, subsequent tests of nociceptive and non-nociceptive responses were given 20, 60, and 120 mins post-injection. (C) For tests of endocannabinoid effects in animals that had undergone injury-induced sensitization, Hirudo responses to non-nociceptive and nociceptive stimuli were first assessed on day 0. On day 1, animals received a crush injury and responses to nonnociceptive and nociceptive stimuli were tested daily. On day 3, some animals received injections of 2-AG or anandamide.

Methods
Animals and behavioral protocols. Leeches (Hirudo verbana; 3 g) were obtained from commercial suppliers (Leeches USA, Westbury, NY or Niagara Leeches, Cheyenne, WY) and maintained in artificial pond water [0.52 g/L H 2 O Hirudo salt (Leeches USA Ltd.)] on a 12 hour light/dark cycle at 15 °C in a refrigerated incubator.
The techniques for testing Hirudo responses to nociceptive and non-nociceptive stimuli are based on previously published protocols 25,36 . Individual Hirudo were placed in a testing arena consisting of a plastic petri dish (145 mm diameter) lined with filter paper that had been saturated with pond water and maintained at room temperature. All animals were allowed to acclimate to the testing arena for 30 minutes prior to the start of the experiments. In experiments not involving injury-induced sensitization, each animal was initially tested (pre-test) for responses to nociceptive and non-nociceptive stimuli, followed by drug injection and then a post-test measurement for changes in responses to these stimuli at 20, 60 and 120 minutes post-injection (Fig. 1B).
For tests of responses to non-nociceptive stimulus an ascending range of von Frey filaments (0.008-2.0 g) was applied to the posterior sucker at 30-second inter-trial intervals. The threshold for a behavioral response was defined as the first von Frey fiber to elicit a localized shortening response that did not involve the sucker being picked up and withdrawn from the site of stimulation, which would correspond to a whole-body shortening response 37 . The range of von Frey fibers used to elicit this localized shortening behavior are well below the level necessary to elicit responses from the N cells 21,23,38 .
For a nociceptive stimuli, 800 μL of 250 μM allyl isothiocyanate (AITC, the active component of mustard oil) was applied to the external surface of the posterior sucker using a pipette similar to what we have previously reported 25 . AITC has been shown to elicit nocifensive responses in invertebrates possessing a TRPA1 channel homolog 39,40 . Previous studies in our lab have found that the Hirudo polymodal N cell does respond to peripheral application of AITC, which elicits withdrawal of the posterior sucker from the site of application 25 . The magnitude of this withdrawal reflex was noticeably greater than the responses elicited by the von Frey fibers and likely corresponds to a whole-body shortening response 29 . Responses to mustard oil were video recorded and subsequently analyzed using Noldus Observer XT software. Nocifensive behaviors were quantified in terms of latency to withdraw. The person performing these analyses was blind to the experimental conditions (i.e. drug treatment and/or injury status) of each animal. Animals were excluded from further analysis if during the pre-test they were observed to have begun to withdraw their sucker either immediately before or just as the AITC was applied to the animal (7 out of 109 animals tested). Following the nociceptive stimuli tests, all animals were rinsed off for 20 seconds with leech pond water and placed in a clean testing arena. The pre-injection behavioral measure (response threshold, response latency) were used to represent 100% for each animal tested. Consequently, the subsequent post-injection behavioral measure were normalized to the pre-injection level for each animal (subsequent statistical analyses only used the post-injection data). Prior to drug injections, animals were lightly anesthetized with ice-cold saline in an ice-lined dissecting dish and injected with 100 μL of either 100 μM anandamide, 100 μM of anandamide +25 μM of SB366791, 75 μM 2-AG, or 75 μM 2-AG +25 μM SB366791. Pilot studies examining the effects of SB366791 injections found that concentrations greater than 25 μM reduced responses to nociceptive stimuli. For vehicle control experiments, 100 μL of 0.01% DMSO were injected. As previously reported 36 , injections were made just anterior of the posterior sucker, a region where the dorsal and ventral sinuses that are part of the leech vascular system converge 19 . The leech CNS is contained within the ventral sinus so this method of injection is likely to be effective in delivering drugs to the CNS. Fig. 1C. On Day 0 (D0) pre-injury thresholds to non-nociceptive stimuli (von Frey fibers) and responses to nociceptive stimuli (AITC application) were measured as described previously. Injury-induced sensitization was delivered on Day 1 (D1). Each animal was initially anesthetized with ice-cold saline in an ice-lined dish for 20 seconds. Next, the posterior sucker was crushed for 20 seconds using a 13 cm long hemostat (crush dimensions were approximately 9 mm by 2.5 mm). This approach was used to produce an ethologically-relevant form of injury-induced sensitization that mimicked a potential injury produced by a predators bite. No obvious changes in sucker motor function were observed following application of this crush injury. Responses to mechanical and chemical stimuli were assessed one hour after injury and then each day for seven days. On day 3 (D3) each animal received an injection 100 μL of either 0.01% DMSO (vehicle control), anandamide (100, 75, or 25 μM) or 2-AG (75, 50, or 25 μM). Control, non-injured animals also receive a DMSO injection.

Injury-induced sensitization. The experimental protocol is summarized in
Statistical Analysis. Behavioral data were presented as mean ± standard error. Results were normalized to pre-test results for both nociceptive and non-nociceptive experiments. Statistical analyses using two-way analysis of variance (ANOVA) were performed to determine the main effects with Student-Newman-Keuls post-hoc to confirm the ANOVA results.

Results
Opposing effects of endocannabinoids are prevented by a TRPV inhibitor. First, the effects of endocannabinoid injections on Hirudo responses to mechanical non-nociceptive stimuli and chemical nociceptive stimuli were assessed. Mechanical non-nociceptive stimuli were delivered via von Frey fibers that apply force sufficient to activate the non-nociceptive sensory neurons in Hirudo (0.008 g-2.0 g), but were below the mechanical threshold for activating mechano-nociceptive cells (7.0 g) 21 . Chemical nociceptive stimuli were delivered via the application of 250 μM AITC (800 μL) to the posterior sucker as carried out in previous experiments with Hirudo 25,36 . The von Frey fibers elicit a localized shortening responses, whereas AITC elicits whole body shortening in which the animal picks up its sucker to withdraw it from the site of AITC application.
Next, the responses to a noxious chemical stimulus (AITC) topically applied to the posterior sucker were tested in endocannabinoid-treated animals. 2-AG-injection produced an increase in the response latency to noxious chemical stimuli over the 120 min testing period (Fig. 2C). Two-way ANOVA detected a statistically significant effect of drug treatment on withdrawal latency (F 3,86 = 19.09, p < 0.001), but no significant effect of . Although there appears to be an increase in latency in the 2-AG + SB366791 group at the 120 min post-injection, no statistically significant different was observed between the 2-AG + SB366791 group and the DMSO control group. No change in response latency was observed between in the Hirudo injected with SB366791 by itself ( Fig. 2C; p ≥ 0.05).

Effect of endocannabinoids on injury-induced sensitization.
Animals given a crush injury to the posterior sucker exhibited sensitization to non-nociceptive stimuli, expressed as a decrease in the response threshold to mechanical stimulation with the von Frey fibers. These animals also exhibited sensitization to nociceptive stimuli expressed as a reduced latency to respond to AITC application. Both types of sensitization were observed throughout the 7 day testing period (Fig. 3).
Sensitization to non-nociceptive stimuli was unaffected by 2-AG injection made on day 3 (Fig. 3A). A two-way ANOVA comparing non-injured control animals (N = 8), injured animals (N = 8) and injured animals receiving 25, 50 or 75 µM 2-AG (N = 4, 4, 7, respectively) detected a significant effect of treatment (F 4,391 = 530.36, p < 0.001), a significant effect of time (F 6,391 = 6.38, p < 0.001), but no significant treatment-time interaction effect (F 24,391 = 1.39, p > 0.05). Subsequent post-hoc analysis confirmed that all of the injured groups had a significant decrease in the threshold required to elicit localized withdrawal response compared to uninjured controls (p < 0.001). However, no there was no significant difference between injury-only group and the injury + 2-AG, regardless of the 2-AG concentration (p > 0.05 for all). Anandamide treatment also failed to alter sensitization to non-nociceptive stimuli (Fig. 3B). A two-way ANOVA comparing non-injured control animals (N = 8), injured  25 µM had no effect. Anandamide was also capable of reducing this injury-induced sensitization, but only at the highest concentration (75 µM).
Scientific RepoRts | 7: 5793 | DOI:10.1038/s41598-017-06114-1 animals (N = 8) and injured animals receiving 50, 75 or 100 µM anandamide (N = 3, 3, 5 respectively) detected a significant effect of treatment (F 4,321 = 640.30, p < 0.001), a significant effect of time (F 6,321 = 2.62, p < 0.05), but no significant treatment-time interaction effect (F 24,321 = 1.12, p > 0.05). Subsequent post-hoc analysis confirmed that all of the injured groups had a significant decrease in the threshold compared to the non-injured control group (p < 0.001). However, there was no significant difference between injury-only group and the injury + anandamide group, regardless of the anandamide concentration (p > 0.05 for all). Together these results indicate that 2-AG and anandamide had no effect on responses to non-nociceptive stimuli in Hirudo that have undergone injury-induced sensitization.
2-AG applied on day 3 did reduce sensitization to chemical nociceptive stimuli (Fig. 3C). A two-way ANOVA detected a significant effect of treatment (F 4,222 = 35.58, p < 0.001), a significant effect of time (F 6,222 = 3.79, p < 0.01) and a significant treatment-time interaction effect (F 24,222 = 2.63, p < 0.01). Subsequent post-hoc analysis showed that the response latency in injured animals was significantly lower compared non-injured controls (p < 0.001). The injury + 75μM 2-AG group was significantly different from all groups including the non-injured controls (p > 0.001), until days 6 and 7 when the 2-AG-treated group was no longer statistically different from the non-injured controls (although they remained statistically different from the injured group). 50 μM 2-AG reduced sensitization to nociceptive stimuli relative to the injured control group (p < 0.01; post-hoc of treatment effect), but was also different from the non-injured control group (p < 0.05), indicating this concentration of 2-AG was not as effective as 75 μM. 25 μM 2-AG had no effect on injury-induced sensitization to nociceptive stimuli (p ≥ 0.05).
Anandamide also reduced injury-induced sensitization to nociceptive stimuli, but not as effectively as 2-AG (Fig. 3D). A two-way ANOVA detected a significant effect of treatment (F 4,193 = 35.18, p < 0.001), a significant effect of time (F 6,193 = 2.16, p < 0.05), but not a significant treatment-time interaction effect (F 24,193 = 1.89, p ≥ 0.05). Subsequent post-hoc analysis showed that the response latency in injured animals was again significantly lower compared non-injured controls (p < 0.001). Only 100 μM anandamide was effective in reducing injury-induced sensitization based on a post-hoc comparison of the injury + 100 μM anandamide and injury groups (p < 0.05) and this effect was restricted to just the day of drug injection. 75 and 50 μM anandamide had no effect on injury-induced changes in response latency to nociceptive stimuli. In conclusion, while anandamide can ameliorate injury-induced sensitization, the effect is not as strong nor as long-lasting as the effect of 2-AG.

Discussion
This study presents behavioral evidence for opposing effects of endocannabinoids in responses to non-nociceptive mechanical vs. nociceptive chemical stimuli. Specifically, 2-AG and anandamide both enhanced responses to non-nociceptive stimuli and reduced responses to nociceptive stimuli. Both of these effects were blocked when the TRPV channel inhibitor SB366791 was co-injected. We also examined the effects of endocannabinoids on animals that had undergone an injury to the posterior sucker that produced persistent (at least 7 days) sensitization to both nociceptive and non-nociceptive stimuli. 2-AG and anandamide had no effect on responses to non-nociceptive stimuli in these injury-induced sensitized animals. This lack of effect is likely due to the fact that further decreases in the response threshold cannot occur because these animals already maximally sensitized due to the injury. However, it is also possible that endocannabinoids mediated the injury-induced sensitization itself and therefore this sensitization occludes any additional effects of injected 2-AG or anandamide. High frequency stimulation of the Hirudo nociceptive neurons does produce endocannabinoid-mediated potentiation of the pressure cell synapses that are likely to mediate responses to the von Frey fibers used in this experiment 33 . Future studies will examine the potential role of endocannabinoid signaling in mediating injury-induced sensitization to non-nociceptive injury.
2-AG and anandamide injections increased the response latency to chemical nociceptive stimuli (AITC), an anti-nociceptive effect. Injury-induced sensitization was not observed following injection of 100 µM 2-AG and there was an increased response latency relative to non-injured control animals for several days after injection. 50 µM 2-AG produced a smaller anti-nociceptive effect, but one that still lasted several days. 25 µM 2-AG was ineffective. By comparison anandamide was less effective with only the highest concentration (100 µM) producing an anti-nociceptive effect that only lasted one day. Both 2-AG and anandamide produce similar levels of depression in Hirudo nociceptive synapses although this was only measured for 1-2 hrs 31,33,35 . It is not clear at this time why 2-AG and anandamide are so different in terms of the duration of their behavioral effects.
These results are consistent with previous in vitro neurophysiological studies carried out in isolated ganglia and semi-intact preparation in which endocannabinoids depress nociceptive (N cell) synapses, but potentiate non-nociceptive (P cell) synapses (see Fig. 1A) [31][32][33] . Pharmacological inhibitors of TRPV1 blocked both endocannabinoid-mediated potentiation of P cell synapses and depression of the N cell synapses. Previous pharmacological studies have shown that Hirudo possesses a TRPV-like channel both peripherally and in the CNS that responds to capsaicin and TRPV1 antagonists such as SB366791 and capsazepine 24,31,33 . Depression of N cell synapses is due to activation of presynaptic TRPV-like channels that is thought to lead to a decrease in neurotransmitter release (see Fig. 1A) and is calcineurin-and transcription-dependent 30,34,35 . These mechanisms are similar to endocannabinoid/TRPV1-mediated depression in hippocampal synapses 41,42 . Although also TRPV-mediated, P cells lack TRPV-like channels and potentiation of P cell synapses is an indirect process that involves disinhibition of theses synapses (see Fig. 1A) 32,33 . Nociceptive synapses are "protected" from this disinhibition because they are depolarized by GABA due to elevated levels of intracellular Cl −33, 36 . At this time the GABAergic neurons in Hirudo that undergo this suggested endocannabinoid-mediated depression have not been identified.
An important caveat of these studies and much of the previous Hirudo work is that they are based on pharmacological manipulations. It is possible that the observed drug effects are due, at least in part, to off-target effects unique to invertebrates. For example, the platyhelminth Schistosoma mansoni, responds to capsaicin, but this effect is mediated by TRPA1 channels (which are also present in Hirudo) 25,40 . We have tried to minimize this potential confound in our past studies by utilizing multiple pharmacological agents 30,31,33 . However, this concern will remain until Hirudo versions of these proteins are isolated and directly examined.
The current experiments, combined with previous synaptic studies, have relevance to understanding endocannabinoid/TRPV-based modulation of nociception. Endocannabinoids have been reported to exert an anti-nociceptive effect due at least in part to depression of glutamatergic transmission at primary afferent synapses in the spinal cord 4,5,7,43,44 . However, stimulation of CB1 receptor can also enhance nociception due to depression of GABAergic/glycinergic inhibitory transmission within the spinal cord 8 . Interestingly, injury-induced allodynia due to TRPV1-mediated disinhibition in the spinal cord has been observed although it is not known whether endocannabinoids are activating the TRPV1 channel 45 . Finally, pro-nociceptive effects are also observed in animals with a genetic knock-out of fatty acid amide hydrolase (FAAH), which is responsible for anandamide metabolism, and these effects were mediated by both CB1 and TRPV1 9 . This capacity for endocannabinoids to have both pro-and anti-nociceptive effects, potentially through both CB1-and TRPV1-mediated signaling, may help to explain why some clinical studies of cannabinoid-based therapies to treat chronic pain can sometimes result in either no effect or a worsening of symptoms [10][11][12] . These pro-nociceptive effects may be due in part to endocannabinoids disinhibiting (and therefore enhancing) the nociceptive circuitry. Disinhibition is a critical mechanism that "opens the gate" for non-nociceptive afferents to have access to spinal nociceptive circuits [15][16][17]45 , but the mechanisms by which injury elicits disinhibition associated with sensitization of non-nociceptive pathways are not fully understood. Differences in Cl − gradients between nociceptive and non-nociceptive afferents may contribute to the latter being more sensitive to disinhibition 10, 33, 36, 46-48 . In Hirudo this disinhibition selectively affects non-nociceptive pathways, but may have effects on both non-nociceptive and nociceptive pathways in mammals possibly due to the more complex circuitry (in terms of multiple classes of inhibitory and excitatory interneurons) at the spinal cord level 14,15 .
Together, these findings demonstrate direct behavioral evidence for the opposing effects of endocannabinoids in both injured and non-injured animals. These results significantly contribute to understanding the potential role of endocannabinoids in both the induction and attenuation of pain conditions and demonstrate the need for more studies characterizing the specific mechanisms unique to sensitization of nociceptive sensory pathways versus sensitization of non-nociceptive pathways. From a clinical standpoint, endocannabinoid-based therapies may only have efficacy for certain types of pain conditions. Specifically, it is possible that cannabinoid-based therapies will be more effective for conditions that involve hyperalgesia, but either less effective for or perhaps even exacerbate conditions that include allodynia. Such considerations must be taken into account when designing future clinical studies that seek to use the endocannabinoid signaling system to treat chronic pain.