Activation of mesolimbic reward system via laterodorsal tegmental nucleus and hypothalamus in exercise-induced hypoalgesia

Ventral tegmental area (VTA) dopamine (DA) neurons are the primary source of dopamine in target structures that constitute the mesolimbic reward system. Previous studies demonstrated that voluntary wheel running (VWR) by neuropathic pain (NPP) model mice produces exercise-induced hypoalgesia (EIH), and that activation of mesolimbic reward system may lead to EIH. However, the neuronal mechanism by which the mesolimbic reward system is activated by VWR is unknown. Here, we found that VWR produces EIH effects and reverses the marked reduction in activated lateral VTA (lVTA)-DA neurons induced by NPP. The proportions of activated laterodorsal tegmental nucleus (LDT)-cholinergic and lateral hypothalamus-orexin neurons were significantly enhanced by VWR. Retrograde tracing and dual immunostaining revealed that VWR activates lVTA-projecting LDT-cholinergic/non-cholinergic and lateral hypothalamic area (LHA)-orexin/non-orexin neurons. Therefore, EIH effects may be produced, at least in part, by activation of the mesolimbic reward system via activation of LDT and LHA neurons.


Changes in running distances and pain behaviors. The manner of exercise by experimental animals
is classified into forced exercise, such as treadmill running and swimming, and voluntary exercise such as free running on a running wheel. The effectiveness of forced exercise in producing EIH effects has been demonstrated in many studies 3 . However, forced exercise induces several physiological adaptations similar with chronic stress 21 , whereas VWR produces not only EIH effects in NPP, inflammatory pain and chronic muscle pain model animals, but also improves symptoms of chronic stress such as depression and anxiety-like behaviors [22][23][24][25][26][27][28] . In addition, VWR can be performed in the active phase during night without perturbing the rest phase during daytime. Thus, VWR is a stress-free exercise style, which may be more effective in producing EIH effects. Therefore, the present study used VWR and investigated EIH effects. Figure 1B shows changes in running distances per day by Naive, Sham, and PSL-Runner mice. The mean running distance in the three groups was approximately 3000 m/day on day 1 after the start of the experiment (Pre14 d), and running distances markedly increased by day 14 (Pre1 d) (Naive-runner: 12948.03 ± 943.77 m/ day; Sham-Runner: 12314.39 ± 1304.08 m/day; PSL-Runner: 12766.73 ± 886.59 m/day). The running distance by Sham-Runner mice decreased to 5591.72 ± 1147.08 m/day on day 1 post-surgery (1 d), but returned to nearly pre-surgery levels by day 4 post-surgery (11503.76 ± 1887.23 m/day: 93.4%). On the other hand, the running distance by PSL-Runner mice markedly decreased on day 1 post-surgery (1337.68 ± 286.54 m/day: 10.5%). However, these values gradually increased thereafter. The running distance by PSL-Runner mice recovered to approximately 72% (9250.59 ± 488.29 m/day) of the pre-surgery level by day 14 post-surgery (14 d). These changes were consistent with our preliminary results 19 . We then examined the diurnal variation in running distances (m/hour) by PSL-Runner mice (Fig. 1C). The peak running distance by Runner mice at day 1 pre-surgery (Pre1 d) was approximately 2449.52 ± 167.72 m/hour, but at day 1 post-surgery (1 d), it markedly decreased to 541.03 ± 238.95 m/ hour. Then, the running distances recovered day by day after surgery, and at day 14 post-surgery (14 d), the peak running distance recovered to approximately 84.3% (2066.96 ± 66.89 m/hour) of the pre-surgical level.
We performed von Frey and plantar tests to evaluate the effects of VWR on mechanical allodynia and heat hyperalgesia (Fig. 1D,E). No marked differences were noted in the pre-operative data of withdrawal thresholds or withdrawal latencies among mice assigned to the six experimental groups. Furthermore, Runner mice in the Naive and Sham groups exhibited no significant alterations in withdrawal thresholds or withdrawal latencies throughout the experimental period, indicating that VWR had no effects on sensory sensitivity in mice without PSL. PSL-Sedentary mice had markedly lower withdrawal thresholds (0.227 ± 0.035 g) and shorter withdrawal latency (4.17 ± 0.137 sec) at 2 days after PSL than at pre-surgery (withdrawal threshold: 1.16 ± 0.052 g; withdrawal latency: 12.9 ± 0.307 sec), and these decreased thresholds and latencies were maintained throughout the experimental period. Withdrawal thresholds and latencies were significantly higher and longer on day 7 in PSL-Runner mice (0.402 ± 0.044 g, p < 0.01; 6.65 ± 0.342 sec, p < 0.01) than in PSL-Sedentary mice (7 days = 0.192 ± 0.017 g; 7 days = 4.4 ± 0.245 sec). Furthermore, significant increases in withdrawal thresholds and latencies were observed until day 13 after PSL in PSL-Runner mice compared with those in PSL-Sedentary mice (p < 0.01). Taken together with our previous findings 19 , the present results confirmed that the pain behaviors in NPP model mice were significantly improved by the VWR protocol employed in the present study. We next examined the relationships between the threshold of final pain on behavioral tests (13 d) and total running distances during the 13-day period after PSL in PSL-Runner mice. We found a significant positive correlation between total running distance, and withdrawal threshold ( Fig. 1F; R = 0.934, p < 0.001, n = 9) and withdrawal latency ( Fig. 1G; R = 0.929, p < 0.001, n = 8). These results also confirmed our previous findings 19 . Scientific  VWR/PSL-induced changes in lVTA-dopamine neurons. As TH is a rate-limiting enzyme for dopamine synthesis, immunostaining with a TH antibody enabled identification of DA neurons in brain sections. In addition, immunoreactivity of FosBΔFosB was used as a marker of activated neurons because FosB protein gradually accumulates in response to chronic stimuli, including wheel running, and persists for long periods of time because of its high stability 16 . The VTA, which is composed of anatomically and functionally heterogeneous DA neuron subpopulations with different axonal projections, is divided into two parts: the lateral VTA (lVTA) and medial VTA (mVTA) 5   Naive-, Sham-and PSL-Runner mice were allowed to run freely on the running wheel, whereas Sedentary mice were reared in cages with a locked running wheel. Naive-Runner, Sham-Runner and PSL-Runner mice were placed in individual cages equipped with a low-profile wireless mouse running wheel. (B) Changes in running distances by Runner mice. (C) Changes in diurnal variation in running distances by PSL-Runner mice (n = 7). (D) von Frey and (E) plantar tests were performed for all experimental groups. Mechanical withdrawal thresholds and thermal withdrawal latencies were significantly higher for PSL-Runner mice than for PSL-Sedentary mice (*p < 0.01). Significant positive correlations were observed between total running distances during the 13-day period after PSL, and (F) the withdrawal threshold (R = 0.934, p < 0.001, n = 9) and (G) withdrawal latency (R = 0.929, p < 0.001, n = 8).
the primary subpopulation of DA neurons that are preferentially activated by aversive stimuli, including chronic pain, whereas lVTA-DA neurons projecting to the NAc lateral shell may primarily signal reward and salience 29 . Therefore, we focused on lVTA-DA neurons, and examined the effects of PSL and VWR on these neurons. Figure 2A shows the position of lVTA on photomicrographs, which confirmed the report of Lammel et al. 5 . Figure 2B-M show typical immunostained images of activated lVTA-DA neurons in each experimental group, and the proportions of activated DA neurons in the lVTA on the ipsilateral and contralateral sides were compared among each group (Fig. 2N). The square in Fig. 2A indicates the area of lVTA in which the quantitative analysis was made. In agreement with the photomicrographs, the proportions of activated lVTA-DA neurons (TH + FosBΔFosB + /TH + ) were significantly higher in Naive-, Sham-and PSL-Runner mice than those in Naive-and Sham-Sedentary mice (p < 0.01). In contrast, the proportion of activated lVTA-DA neurons in PSL-Sedentary mice were significantly lower compared with those in the other groups (p < 0.01). There were no significant differences between the ipsilateral and contralateral sides in all groups. VWR/PSL-induced changes in LDT-cholinergic neurons. The LDT contains three distinct neuronal types, cholinergic, GABAergic and glutamatergic neurons 30 . In the present study, immunostaining with a nNOS antibody was used to identify cholinergic neurons in brain sections because nNOS and choline acetyltransferase are useful target proteins for the immunohistochemical identification of cholinergic neurons in the mouse LDT 31 . Using dual immunostaining with nNOS and FosBΔFosB antibodies, we found many nNOS + neurons without FosBΔFosB + immunoreactive nuclei on both the ipsilateral and contralateral sides of the LDT in Sham-Sedentary mice (Fig. 3A). However, robust immunoreactivity of FosBΔFosB was bilaterally detected in nNOS + neurons located in the LDT in Sham-Runner, PSL-Sedentary and PSL-Runner mice ( Fig. 3B-D). The proportions of activated cholinergic (nNOS + FosBΔFosB + /nNOS + ) neurons on the ipsilateral and contralateral sides in the LDT were compared in Fig. 3F. Consistent with the immunostaining images, the proportions were significantly higher in all Runner mice and PSL-Sedentary mice than in Naive-and Sham-Sedentary mice (p < 0.01). No significant differences between the ipsilateral and contralateral sides were observed in all groups.
When the relationship between the total running distance during the 14-day period after surgery in PSL-Runner mice and the proportion of activated cholinergic neurons (sum of both the ipsilateral and contralateral sides) in PSL-Runner mice was examined, a significant positive correlation was noted (Fig. 3G, R = 0.939, n = 8, p < 0.01).
lVTA-projecting LDT cholinergic neurons in Runner mice. Using retrograde tracing and dual immunostaining, we investigated whether lVTA-projecting LDT-cholinergic neurons are activated by VWR. Mice were placed in individual cages with a running wheel at the start of the experiment, and after 2 weeks (Pre14 d), the running distance increased to 14999.47 ± 1124.61 m/day, as shown in Fig. 4B. A retrograde tracer, Retrobeads Red (RBR), was unilaterally injected into the left side of the lVTA, as described in Methods, and the RBR-injected mice then continued VWR for 2 weeks (Fig. 4A,B). Although the injection of RBR markedly decreased the running distance at day 1 post-injection (1 d: 4214.88 ± 1021.74), the running distances gradually increased thereafter until day 14 post-injection (14 d: 20193.53 ± 2633.39 m/day) (Fig. 4B). The brainstems in six mice in which RBR was precisely injected into the lVTA were used for dual immunostaining.
Although RBR-incorporated neurons were bilaterally detected in the LDT (Fig. 4C), more RBR + nNOS + neurons (arrows in Fig. 4D) were detected on the ipsilateral side (13.08 ± 0.98) than on the contralateral side (7.91 ± 1.45) of the injection (Fig. 4F). These results were consistent with previous findings, which demonstrated that following the unilateral injection of Fluoro-Gold or RBR into the VTA, the positive labeling of tracers was bilaterally detected in LDT-cholinergic neurons; however, the amount on the ipsilateral side was higher than that on the contralateral side 32,33 . On dual immunostaining with nNOS and FosBΔFosB antibodies, we found two distinct neuronal types, namely, lVTA-projecting activated cholinergic (nNOS + RBR + FosBΔFosB + ) and lVTA-projecting activated non-cholinergic (nNOS − RBR + FosBΔFosB + ) neurons (Fig. 4E). In the quantitative analysis, many lVTA-projecting activated non-cholinergic and lVTA-projecting activated cholinergic neurons were detected on the ipsilateral and contralateral sides, and the number of lVTA-projecting activated non-cholinergic neurons was approximately 2-fold that of the lVTA-projecting activated cholinergic neurons on the ipsilateral side (Fig. 4F).

VWR/PSL-induced changes in LH-orexin neurons.
Orexins are hypothalamic neuropeptides comprising orexin A and orexin B that are involved in multiple physiological processes such as the regulation of the sleep/ wakefulness state, energy homeostasis, analgesia and reward seeking. Orexins bind to the orexin receptor type 1 (OR1) and orexin receptor type 2 (OR2), which are expressed on VTA neurons 34 . Orexin neurons are localized exclusively in the lateral hypothalamus (LH), including the lateral hypothalamic area (LHA), perifornical area (PFA) and dorsomedial hypothalamus (DMH), but their projections are widely distributed over many brain areas 34 . As shown in Fig. 5A, the LH was divided into two regions, the DMH and PFA/LHA, according to a criterion described in the Methods section. Orexin neurons arising in the DMH and PFA project to brainstem nuclei and are involved in the control of arousal and modulation of stress responses, whereas neurons in the LHA project to brain areas, including the VTA and NAc, and modulate reward and addiction processes 35,36 . However, as the PFA is anatomically small and disappears in the rostral part, we considered PFA/LHA-orexin neurons to mainly regulate reward processing in this study.
Using dual immunostaining with orexin A and FosBΔFosB antibodies, we found many orexin A + neurons with strongly FosBΔFosB + immunoreactive nuclei in both the DMH and PFA/LHA in Naive-and PSL-Runner mice (Fig. 5D,E,H,I), whereas many orexin A + neurons lacking FosBΔFosB + immunoreactivity were detected in both regions in Naive-and PSL-Sedentary mice (Fig. 5B,C,F,G). Furthermore, the intensity of orexin A immunoreactivity was very weak in Sedentary (Fig. 5B,C,F,G) compared with that in Runner mice (Fig. 5D,E,H,I). Based on the photomicrographs, the proportion of activated orexin neurons (orexin A + FosBΔFosB + /orexin A + ) was significantly higher in Naive-, Sham-and PSL-Runner mice than that in Sedentary mice (n = 6, p < 0.01) (Fig. 5J). No significant differences between the ipsilateral and contralateral sides were observed in all groups. When the relationships between total running distances during the 14-day period after PSL and the proportions of activated orexin neurons in PSL-Runner mice (sum of both the ipsilateral and contralateral sides) were examined, a significant positive correlation was detected between total running distance and the proportion of activated orexin neurons in the PFA/LHA (R = 0.846, n = 8, p < 0.01) (Fig. 5L) but not in the DMH (R = 0.585, n = 8, NS) (Fig. 5K). lVTA-projecting orexin neurons in Runner mice. We next investigated whether orexin neurons projecting to the lVTA are activated by VWR. The VWR protocol and retrograde tracing were the same for the LDT, as shown in Fig. 4A and B. Almost all RBR + cells in the LH were detected on the ipsilateral side of the injection, although RBR + cells were occasionally observed on the contralateral side (Fig. 6A). In the present study, structures in which at least five RBR particles were arranged in either a ring-or disk-like shape were considered to be RBR + cells (e.g., arrows and arrowheads in Figs 4D and 6C). Most RBR + cells (approximately 80%) were distributed in the PFA/LHA and the rest were found in the DMH (Fig. 6B), indicating that PFA/LHA-orexin neurons abundantly project to the lVTA. On dual immunostaining with orexin A and FosBΔFosB antibodies, we found two distinct neuronal types, namely, lVTA-projecting activated orexin (orexin A + RBR + FosBΔFosB + ) and lVTA-projecting activated non-orexin (orexin A − RBR + FosBΔFosB + ) neurons (Fig. 6C,D). Quantitative analysis indicated that the VWR activated approximately 42% of the lVTA-projecting orexin neurons, and the number of lVTA-projecting activated non-orexin neurons was approximately 1.8-fold greater than that of lVTA-projecting activated orexin neurons (Fig. 6E).

Discussion
The ability of the VWR to produce EIH effects has been confirmed by recent experimental studies [22][23][24][37][38][39][40] . Consistent with these studies, our results demonstrated that VWR for 2 weeks before and after surgery significantly improves pain behaviors. Thus, VWR is a reliable exercise style to produce EIH effects, although the VWR protocols and possible mechanisms of EIH differ among studies. We found a significant positive correlation between the total running distance and threshold of pain on behavioral tests, implying that the amount of VWR is a critical factor governing the analgesic level of EIH effects. On the contrary, Grace et al. 23 reported that VWR prior to chronic constriction injury in rats completely prevented the development of allodynia, but the running distance and pain threshold were not correlated. Similarly, Pitcher et al. 24 found that VWR by inflammatory pain model rats improved pain behaviors, but the amount of VWR did not correlate with analgesic levels. These discrepancies may be due to the differences in experimental designs and animal species used. We used mice as experimental animals, whereas they used rats. The physical activity level of mice is much higher than that of rats, and more active VWR may lead to more effective EIH. In accordance with our preliminary results 19 , the present study confirmed that the decreased running distances in VWR by PSL-Runner mice markedly recover day by day. In addition, the diurnal variation in running distances also recovered to almost pre-surgical levels. VWR has been well-reported as a natural reward that can activate VTA-DA neurons and the mesolimbic reward system [14][15][16][17] . Therefore, the gradual increase of running distance in VWR may cause higher activation of the mesolimbic reward system, which may help reduce NPP-induced aversion.
Our results revealed that DA neurons in the lVTA are significantly inactivated in NPP model mice, which agrees with a recent finding that DA efflux in the NAc is reduced by painful stimulation 6 . It is well-documented that VTA-DA neurons are inactivated in painful conditions. Chronic pain or other aversive stimuli activate glutamate-containing LHb neurons, which then project to GABA neurons in the VTA and the more caudally located rostromedial tegmental nucleus (RMTg) to inhibit DA neurons in the VTA 5,41 . In addition to this LHb → RMTg/VTA (GABA) pathway, we cannot exclude the possibility that the direct LDT → VTA (GABA) pathway 42-45 may also be involved in the inhibition of VTA-DA neurons and maintenance of NPP (Fig. 7A). Further studies are needed to elucidate the precise roles of this GABA neuron-targeted input from the LDT in the control of the reward system. On the other hand, we found that VWR significantly activates lVTA-DA neurons, and simultaneously enhances EIH effects in PSL-Runner mice. This is reasonable because many studies have demonstrated that VWR activates the mesolimbic reward system via VTA-DA neurons, and activation of the mesolimbic reward system induces analgesic effects 6,12,14-17,46 . Furthermore, Wakaizumi et al. 18 demonstrated that specific inactivation of VTA-DA neurons attenuates the EIH effects induced by treadmill running in NPP model mice. In addition, based on our previous results, VWR significantly enhances TH immunoreactivity and pCREB expression in DA neurons in the lVTA 19 . Thus, activation of the mesolimbic reward system via activation of lVTA-DA neurons following VWR likely plays an important role in producing EIH effects.
We found that VWR significantly increases the proportion of activated LDT-cholinergic neurons. The lVTA-DA neurons receive direct cholinergic innervation from the LDT of the mesopontine tegmentum 33,47 , which regulates the activity of VTA-DA neurons. Efflux of DA in the NAc is stimulated by intra-VTA neostigmine, a cholinergic agonist, and this effect is blocked by excitotoxic lesions of the LDT 48 . Electrical stimulation of the LDT also increases NAc DA efflux through activation of nicotinic ACh and ionotropic glutamate receptors in the VTA 49 . Among the pontomesencephalic cholinergic neurons, VTA-projecting neurons are mostly concentrated in the LDT 33 and these cholinergic neurons are noci-responsive 50 . It has been demonstrated that approximately 40% of them exhibit bursts of neuronal activity in response to noxious heat stimuli 50 . In the present study, approximately 60% of the LDT-cholinergic neurons on both sides were activated in PSL-Sedentary mice. On the other hand, 70∼80% of cholinergic neurons in the LDT were activated in Runner mice. Of note, similar proportions of LDT-cholinergic neurons (70%) were activated in PSL-Runner mice. As no additive effects were observed by concomitant PSL-surgery and VWR, the same neuronal group may respond to both PSL and VWR. Therefore, we speculate that these cholinergic neurons respond to arousal-related stimuli, such as pain, stress and exercise, and increase activity by controlling VTA-DA neurons for adaptation to new circumstances.
It has been reported that forced running by Parkinson's disease model rats activates LDT neurons 51 . Using retrograde tracing and dual immunostaining, we found that VWR activated approximately 60% of the LDT-cholinergic neurons projecting to the lVTA on both sides in the present study. Therefore, based on the critical role of cholinergic innervation from the LDT for VTA-DA neurons 5,48 , our results suggest that VWR bilaterally activates LDT-cholinergic neurons, thereby activating lVTA-DA neurons. On the other hand, VWR also activated LDT-non-cholinergic neurons projecting to the lVTA, i.e., GABAergic and/or glutamatergic neurons. A previous study revealed that both LDT-cholinergic and LDT-glutamatergic inputs to the VTA contribute to rewarding effects, but they do so in different ways 52 . Therefore, both cholinergic and glutamatergic LDT neurons activated by VWR may produce EIH effects in different manners. However, the precise mechanisms need to be elucidated in more detail in future studies.
The present study demonstrated that VWR significantly increased the proportion of activated orexin neurons. Physical exercise has been reported to activate orexin neurons and increase orexin projection levels [53][54][55] . Upon investigation of the relationship between the proportion of activated orexin neurons and the total running distances after PSL, a significant positive correlation was detected for PFA/LHA-orexin neurons, but not for DMH. DMH-orexin neurons project preferentially to the locus ceruleus and are involved in arousal, whereas LHA-orexin neurons are mainly involved in reward and reinforcement of VTA-DA neurons 56 . Therefore, our results suggest that DMH-orexin neurons are activated in response to exercise irrespective of amount, and contribute to wakefulness, whereas LHA-orexin neurons enhance their activation levels in a VWR-dependent manner, which may activate lVTA-DA neurons and EIH.
According to retrograde tracing and dual immunostaining, the VWR activates lVTA-projecting orexin neurons. OR1 and OR2 are present in the VTA, and VTA-DA neurons receive excitatory inputs from LHA-orexin neurons 20,34 . Therefore, activation of lVTA-projecting LHA-orexin neurons following VWR is sufficient to activate lVTA-DA neurons. In addition, we found that VWR also activates lVTA-projecting non-orexin neurons. The LHA contains not only orexins and MCH, but also GABA neurons 57,58 , and optogenetic stimulation of LHA-GABA neurons projecting to the VTA produces reward-related behaviors 59,60 . In our preliminary experiment, we found that VWR does not induce expression of FosBΔFosB in MCH neurons (data not shown), suggesting that VWR-activated non-orexin neurons correspond to LHA-GABA neurons. Therefore, VWR may also activate LHA-GABA neurons in addition to LHA-orexin neurons, both of which project to the lVTA. Yazdi-Ravandi et al. 61 demonstrated that the injection of orexin A into the rat VTA produced analgesic effects, but this effect was blocked by the injection of D1 and D2 receptor antagonists into the NAc. In inflammatory pain model rats, pharmacological activation of orexin neurons induces analgesia, and inhibition of orexin receptors in the VTA can abolish these analgesic effects 62 . Furthermore, selective ablation of orexin neurons produces pain-related behaviors, whereas pharmacological activation of orexin neurons induces analgesia 63 . Thus, the analgesic effects of orexin may be mediated via activation of the mesolimbic reward system through activation of VTA-DA neurons. Therefore, activation of lVTA-DA neurons via LHA-orexin neurons, and possibly LHA-GABA neurons, following VWR may produce EIH effects.
Based on the present study, we propose two potential neuronal mechanisms of lVTA-DA neuron activation following VWR. One is by cholinergic/glutamatergic neurons in the LDT and the other is by orexin neurons in the LHA. The former neurons can be activated by both VWR and PSL, whereas the latter neurons are activated by VWR but not by PSL. Furthermore, many of these neurons project to the VTA. Therefore, we conclude that both LDT (cholinergic/glutamatergic) → lVTA (DA) and LHA (orexin) → lVTA (DA) pathways are involved in the activation of the mesolimbic reward system, which may produce EIH effects. The creation of positive emotion and pleasure following attenuation of pain perception reinforces the motivation to perform physical exercise, which may further potentiate EIH effects (Fig. 7B). This virtuous cycle may help to maintain, improve and establish an active lifestyle, which further relieves chronic pain and improves QOL. Therefore, further studies focused on brain neuronal systems, including the mesolimbic reward system, will enhance our understanding of the neuronal mechanisms producing EIH effects, and lead to the development of novel therapeutic strategies to improve and prevent chronic pain.

Methods
Preparation of NPP model mice. Adult C57BL/6J mice were used in this study. NPP model mice were prepared by PSL 64 . Under deep anesthesia, approximately one-third to one-half of the sciatic nerve at the right mid-thigh level was tightly ligated with 8-0 silk sutures. Sham operations were performed using the same procedures described above excluding PSL. All experiments in this study were approved by the Animal Care Committee  (6) PSL-Runner mice (n = 6~10). All Runner and Sedentary mice were placed in individual cages equipped with a running wheel or locked running wheel under a 12/12-hour light-dark cycle, and Naive-, Sham-and PSL-Runner mice were allowed to run freely on the running wheel. The VWR protocol is shown in Fig. 1A. All mice, except for Naive mice, performed VWR for 2 weeks prior to the sham operation or PSL. PSL and Sham mice were returned to individual cages immediately after surgery, and running distances were measured during the 14-day period after the surgery. The rotations ran on the running wheel/ hour were monitored using a magnetic reed switch attached to a computerized exercise-monitoring system (SOF-860 wheel manager software, MED Associates, Inc.), and daily running distances (m/day) and diurnal variation in running distance (m/hour) were calculated using the number of rotations on the running wheel.
Evaluation of pain behaviors. In order to abolish diurnal variation in pain behaviors, von Frey and plantar tests were performed at 8:15 a.m. Prior to both tests, mice were placed in an acrylic glass enclosure with a wire mesh bottom (8.3 × 8.3 × 8.0 cm) and allowed to acclimate for 30 min. Mechanical allodynia was evaluated using withdrawal threshold upon von Frey monofilament stimulation to the plantar surface of the hindpaw (0.023, 0.0275, 0.068, 0.166, 0.407, 0.692, 1.202 and 1.479 g). The up-down method of von Frey monofilament testing was used in the present study, and the minimum pressure (g) that evoked a quick withdrawal response by stimulation with a thin von Frey monofilament was taken as the withdrawal threshold 65  Heat hyperalgesia was evaluated by the plantar test, which measures withdrawal latency (sec) to a radiant thermal stimulus delivered from beneath the glass floor to the plantar surface of the hindpaw (Plantar Test, Model7371, Ugo Basile, Comerio, Italy). The heat stimulus by the projector lamp bulb was focused on the plantar surface of the ipsilateral hindpaw, and the minimum time (sec) needed to evoke quick withdrawal of the paw by heat stimulation was taken as the withdrawal latency 66 . The von Frey and plantar tests were performed three times for each mouse, and mean values were considered as the withdrawal threshold or withdrawal latency, respectively. In all behavioral evaluations, the investigator was blinded to the experimental groups in order to avoid any bias.
Mice were perfused transcardially with 4% paraformaldehyde in 0.1 M phosphate-buffered saline (PBS), and the brain was post-fixed. After cryoprotection, the brain was frozen using dry ice-cooled hexane. Serial 25-μm thick brain sections including the VTA (Bregma −3.52 mm~−3.08 mm), LDT (Bregma −5.02 mm~−4.84 mm) or LHA (Bregma −2.06 mm~−1.34 mm) in the coronal plane were mounted serially on slides. After blocking for non-specific staining, sections for dual immunofluorescence staining were incubated simultaneously with two kinds of primary antibodies diluted in 0.1 M PBS containing 5% normal donkey serum and 0.3% Triton X-100 at 4 °C for 48 hours. Then, sections were incubated with the secondary antibodies diluted in 0.1 M PBS containing 5% normal donkey serum and 0.1% Triton X-100 during overnight at 4 °C. The secondary antibodies used for the detection of TH, nNOS and orexin A immunoreactivity were Alexa Fluor 594-labeled donkey anti-sheep (1:500, Jackson ImmunoResearch Laboratories, Inc., Coad Number: 713-585-147) or Alexa Fluor 594-labeled donkey anti-goat IgG (1:500, Abcam, ab150136). The secondary antibody used for the detection of FosBΔFosB immunoreactivity was Alexa Fluor 488-labeled donkey anti-rabbit IgG (1:500, Abcam, ab150061). Sections were washed in 0.1 M PBS and mounted in Vectashield mounting medium with DAPI (H-1200, Vector Labs., Burlingame, CA). Fluorescence signals were detected using a confocal microscope (LSM700, Carl Zeiss, Oberkochen, Germany) equipped with an argon-helium laser. Negative control sections, which were processed without primary antibodies, had no significant positive immunoreactivity.
Quantitative analysis of immunopositive neurons. Immunofluorescence images of lVTA or LDT on the ipsilateral and contralateral sides were acquired at a magnification of 20×. A square of 200 μm × 200 μm in size, as shown in Fig. 2A for lVTA and Fig. 3A for LDT, was placed on the microscope images using ImageJ software (version 1.48, National Institutes of Health, USA), and the number of immunopositive cells within was counted. In a previous study, the LH was further divided into three regions: DMH, PFA and LHA 30 . In this study, the LH was divided into two regions, DMH and PFA/LHA, by a pink vertical line along the medial edge of the fornix, as shown Fig. 5A. Immunofluorescence images of the DMH and PFA/LHA on the ipsilateral and contralateral sides were acquired at a magnification of 10×. The number of immunopositive cells within the yellow square (DMH) or green pentagon (PFA/LHA) on microscopy images was counted. The mean value of three randomly selected sections from six sections per brain was regarded as the value of immunopositive neurons per mouse. In order to avoid any bias in the results of the quantitative analysis, investigators were blinded to all experimental groups throughout quantitative analyses.
Injection of retrograde tracer into the lVTA and dual immunofluorescence staining. Mice performed VWR for 2 weeks prior to injection of the retrograde tracer, and were then placed in a stereotaxic apparatus. Unilateral injection of Retrobeads Red (RBR) (50 nl, Lumafluor Inc., Naples, FL) was performed by inserting a microsyringe into the left side of the lVTA (Bregma: anteroposterior, −3.1 mm; mediolateral, 0.5 mm; and dorsoventral, −4.5 mm). RBR-injected mice were returned to their individual cages, and the running distance was recorded for 2 weeks before and after injection. The brains of RBR-injected Runner mice were processed in the same manner as described for the Immunofluorescence analysis. To evaluate RBR-injection sites, 100-μm thick brain sections including the lVTA were mounted serially, and after briefly rinsing with PBS, slides were mounted in Vectashield mounting medium with DAPI.
Except for secondary antibodies to detect nNOS and orexin A immunoreactivity, the same immunostaining procedure and quantitative analysis were adapted for RBR-injected brain sections. The secondary antibody used to detect specific signals for nNOS and orexin A was DyLight 405-labeled donkey anti-goat IgG (1:500, Jackson Immuno Research Laboratories, Inc.). The total number of cells in three randomly selected sections from six sections per RBR-injected Runner mouse was regarded as the value of immunopositive cells. In order to avoid any bias in the results of the quantitative analysis, investigators were blinded to all experimental groups throughout quantitative analyses.
Statistical analysis. Quantitative data are presented as the mean ± standard error of the mean (SEM). A repeated measures ANOVA followed by Bonferroni's post hoc test was used to compare withdrawal threshold and withdrawal latency among the experimental groups. Multiple comparisons of the number of immunopositive cells among experimental groups were performed using a one-way ANOVA followed by Bonferroni's post hoc test. Differences were considered to be significant at p < 0.05. Data availability. All data generated or analysed during this study are included in this published article.