All animal experiments were approved by the Institutional Animal Care and Use Committee of College of Medicine, National Taiwan University following ARRIVE guidelines. Male C57BL/6 mice of 8–12 weeks were housed in groups of 10 in plastic cages and maintained in a holding room with a 12 h light-dark cycle with free access to food and water ad libitum. On the experimental day, mice were moved in their home cages to the behaviour room and acclimated there for at least 1 h before testing.

The hot-plate test in mice was performed as reported previously [12]. Briefly, the mouse was placed on a hot-plate maintained at 50 °C and paw withdrawal latency was recorded with a cut-off time of 60 s to prevent tissue damage. The antinociceptive effect in each mouse at each time point was calculated as the % of the maximal possible effect (MPE) by the equation: %MPE = 100 x (Latency_{after treatment} - Latency_{before treatment}) / (60s - Latency_{before treatment}). The AUC of withdrawal latencies during the 60 min-recording period was calculated as the total antinociceptive effect in each mouse.

To induce SIA, mice were restrained in a 50-ml centrifugal tube with several small holes for 30 min as reported previously [12]. The control non-stress group of mice remained at their home cages for the same 30 min before being subjected to the hot-plate test.

Locomotor activity was assessed by an open-field test in a 48 × 48 × 40 cm³ acrylic chamber with the arena floor divided into 36 squares, as described previously [43]. The mouse was placed in the center of the chamber, and the number of squares the mouse transpassed with forepaws (number of crossing) and the number of times the mouse stood up with forepaws on the floor (number of rearing) were counted for 5 min.

When drugs were given by i.pag. or i.c.v. microinjection, mice received the i.pag. or i.c.v. cannulation surgery 1 week before the microinjection experiment, as reported previously [12]. Briefly, under anesthesia with 50 mg/kg Zoletil® 50 (a mixture of tiletamine and zolazepam) and xylazine (10 mg/kg), the mouse was placed in a stereotaxic apparatus and a 24-gauge, 10 mm stainless-steel guide cannula was implanted into the right vlPAG (− 4.8 mm caudal, − 0.5 mm lateral, − 2.8 mm ventral from bregma, Additional file 1: Figure S1A) or the right ventricle (− 0.5 mm caudal, − 1.0 mm lateral, − 2.2 mm ventral from bregma, Additional file 1: Figure S1B), according to stereotaxic coordinates of mice [44]. On the day of experiments, i.pag. or i.c.v. microinjection was performed through a 30-gauge injection needle (10 mm) connected to a Hamilton syringe (1.0 μl) on a microinfusion pump (KDS311, KD Scientific Inc., Holliston, MA, USA). The drug solution (0.1 μl) was delivered in 60 s, followed by a 240 s-residual time to avoid back-flow of drug solution. Nociceptive responses were measured 5 mins before as well as 5, 10, 20, 30 and 40 mins after i.pag. or i.c.v. microinjections. For the mice that underwent restraint stress, i.pag or i.c.v. microinjections of the antagonists were performed 5 mins before stress, and nociceptive responses were measured immediately, 5, 10, 20, 30 and 40 mins after stress. After the final behavioural evaluation, the animals were microinjected with 0.5 μl of 0.4% trypan blue solution (Sigma-Aldrich, St. Louis, MO, USA) through the guide cannula to verify the injection tract location. Animals were then sacrificed by decapitation, coronal brain sections (300 μm) were prepared on a vibratome (DSK microslicer DTK-1000, Dosaka, Japan). The injection site was identified by the presence of trypan blue stain diffusion in the vlPAG tissue. Only animals with the cannula correctly targeting the ventricle or vlPAG were included in data analysis.

The preparation of the mouse vlPAG and LH homogenate is the same as previously reported [12]. Briefly, immediately after restraint stress, the mouse was sacrificed. Its brain was removed, placed in a pre-cooled adult mouse brain slicer matrix (Roboz Surgical Instrument, Gaithersburg, MD, USA), and sliced into 1 mm-thick coronal sections. vlPAG or LH brain tissues were bilaterally punched out with a 0.5 mm-tip according to a mouse brain atlas [44]. Each vlPAG sample was collected from one mouse brain, whereas each LH sample was from two mouse brains. After ultrasonication in lysis buffer, the lysates were homogenized and centrifuged (1900 g,14,000 rpm, 15 min) and supernatants collected. The protein concentration in the supernatant was measured by the Bradford method [45].

The substance P level in the vlPAG homogenate was measured with an EIA kit (Cat. No. 583751, Cayman Chemical. Ann Arbor, MI, USA) with a detection range of 3.9–500 pg/ml. The NPS level in the LH homogenates was measured with an ELISA kit. (Cat. No. CSB-EL016026MO, Cusabio, College Park, MD, USA) with a detection range of 4.69–300 pg/ml.

NPS and [tBu-D-Gly⁵] NPS were synthesised and purified as previously described [46]. N-(2-methyl-6-benzoxazolyl)-N-1,5-naphthyridin-4-yl-urea (SB-334867, a selective OX₁R antagonist), 6-methyl-2-(phenylethynyl) pyridine hydrochloride (MPEP, a selective mGlu₅R antagonist) and orexin-A were purchased from Tocris Bioscience (Bristol, UK). Substance P, 1-(2,4-dichlorophenyl)-5-(4-iodophenyl)-4-methyl-N-1-piperidinyl-1H-pyrazole-3-carboxamide (AM251, a CB₁R antagonist), and cis-2-(Diphenylmethyl)-N-[(2-iodophenyl)methyl]-1-azabicyclo [2.2.2] octan-3-amine oxalate salt (L-703,606, a selective NK₁R antagonist) were purchased from Sigma-Aldrich. NPS and [tBu-D-Gly⁵] NPS were dissolved in 0.9% normal saline. Substance P was dissolved in 0.1 M acetic acid. SB-334867, L-703,606, MPEP and AM251 were dissolved in dimethyl sulfoxide (DMSO). All drugs were prepared at the working concentration for the intended i.pag. or i.c.v. injection doses.

Data are expressed as the mean ± S.E.M. and the “n” indicates the number of mice tested in each group. In the hotplate test, two-way ANOVA with post hoc Bonferroni test was used to analyse time courses of antinociceptive effects among different groups. The antinociceptive effect was also assessed by the area under the curve (AUC) of the time courses of quantification of the line graph from baseline to the last time-point of the experiment. Each AUC bar graph was calculated by one-way ANOVA followed by Tukey’s multiple comparison test. Student’s T-test was employed to analyse the results obtained in EIA and ELISA tests. Differences were considered significant if p < 0.05.

NPS when given by i.c.v. injection at the doses 0.3 and 1.0 nmol, which did not affect the spontaneous locomotor activity of mice (Additional file 2: Figure S2), significantly prolonged the latency of nociceptive response in the hot-plate test (Fig. 2a and b) in a time- (F_6,38 = 5.696, p < 0.001, two-way ANOVA, Fig. 2a) and treatment-dependent (F_4,23 = 10.25, p < 0.001, two-way ANOVA, Fig. 2a) manner. However, when NPS was given by i.pag. microinjection, it did not produce significant antinociceptive effect at either 0.3 or 1.0 nmol (Fig. 2a and b). This suggests that the site of action for NPS-induced supraspinal antinociception is brain region(s) other than the vlPAG.

To investigate if the central antinociceptive effect of NPS is mediated via NPSR, we co-administered [tBu-D-Gly⁵] NPS (10 nmol, i.c.v.), a selective and potent NPSR antagonist [47], along with NPS (0.3 or 1.0 nmol, i.c.v.) to mice before the hot-plate test. [tBu-D-Gly⁵] NPS at 10 nmol (i.c.v.) did not affect the nociceptive response in naïve mice, but completely blocked the antinociceptive effect of i.c.v. NPS at the doses of 0.3 and 1.0 nmol (Fig. 2c and d) The overall comparison of the time course of the antinociceptive effect showed a significant difference between time and treatment (F_30,222 = 1.872, p = 0.0057, two-way ANOVA, Fig. 2c). This suggests the central antinociceptive effect of NPS is mediated by NPSRs in the brain.

To substantiate our hypothesis that a cascade mediated by NPS, orexins and substance P sequentially is involved in SIA, we evaluated equipotent doses of these three neuropeptides in a concurrent behavioural assay. As shown in Fig. 2e and f, NPS (0.3 nmol, i.c.v.) produced a antinociceptive effect in the mouse hot-plate test that was comparable to the effects induced by i.pag. microinjection of orexin-A at 1 nmol and i.pag. substance P at 5 nmol, respectively, with a significant difference between time and treatment (F_18,120 = 1.924, p = 0.0198, two-way ANOVA, Fig. 2e).

To ascertain if the NK₁R-mGlu₅R-CB₁R pathway revealed by the electrophysiological study of Drew et al. (2009) [36] is involved in i.pag. substance P-induced antinociceptive effect, we challenged the antinociceptive effect of substance P with selective antagonists of NK₁Rs (L-703,606), mGlu₅Rs (MPEP) and CB₁Rs (AM251), respectively, in a treatment-dependent manner (F_3,18 = 5.316, p = 0.0084, two-way ANOVA, Fig. 3a; F_3,18 = 10.97, p = 0.0003, two-way ANOVA, Fig. 3b; F_3,17 = 5.929, p = 0.0059, two-way ANOVA, Fig. 3c). Indeed, i.pag. co-administration of L-703,606 (10 nmol), MPEP (30 nmol) or AM251 (30 nmol) with substance P (5 nmol) significantly antagonized the antinociceptive effect of i.pag. substance P (Fig. 3).

Our previous findings that i.pag. orexin-A induced antinociception through an OX₁R-initiated eCB signalling [11], where the downstream mechanism was highly similar to i.pag. substance P induced antinociception as shown in the subsection above. In order to ascertain the interaction between orexin-A and substance P in the vlPAG, we challenged orexin-A-induced antinociception with i.pag. NK₁R and mGlu₅R antagonists, respectively. Co-administration of L-703,606 (10 nmol, i.pag.) or MPEP (30 nmol, i.pag.) significantly antagonized i.pag. orexin-A (1 nmol)-induced antinociception (Fig. 4). The overall comparison of time courses of the antinociceptive effect showed a significant difference between time and treatment (F_18,108 = 3.841, p < 0.001, two-way ANOVA, Fig. 4a; F_18,108 = 4.597, p < 0.001, two-way ANOVA, Fig. 4b). These results in combination with our previous findings suggest that orexin-A-induced antinociception is mediated by OX₁Rs, NK₁Rs, mGlu₅Rs and CB₁Rs sequentially in the vlPAG.

Next, we investigated whether the now established OX₁R-NK₁R-mGlu₅R-CB₁R cascade in the vlPAG is involved in the supraspinal antinociceptive effect of NPS. Co-administration of the respective antagonists of OX₁Rs (SB-334867, 15 nmol, i.pag.), NK₁Rs (L-703,606, 10 nmol, i.pag.), mGlu₅Rs (MPEP, 30 nmol, i.pag.) or CB₁Rs (AM251, 30 nmol, i.pag.), significantly supressed the antinociceptive effect of i.c.v. NPS (0.3 nmol) (Fig. 5), in a time- (F_6,114 = 3.252, p = 0.0055, two-way ANOVA, Fig. 5a; F_6,114 = 2.936, p = 0.0106, two-way ANOVA, Fig. 5b; F_6,114 = 2.603, p = 0.211, two-way ANOVA, Fig. 5c; F_6,114 = 2.2, p = 0.0479, two-way ANOVA, Fig. 5d) and treatment- (F_3,19 = 36.96, p < 0.001, two-way ANOVA, Fig. 5a; F_3,19 = 28.58, p < 0.001, two-way ANOVA, Fig. 5b; F_3,19 = 67.33, p < 0.001, two-way ANOVA, Fig. 5c; F_3,19 = 23.44, p < 0.001, two-way ANOVA, Fig. 5d) dependent manner. Thus, i.c.v. NPS-induced analgesia is mediated by the OX₁Rs, NK₁Rs, mGlu₅Rs and CB₁Rs in the vlPAG.

We have previously demonstrated that the vlPAG OX₁R-CB₁R pathway is involved in the SIA induced by an acute restraint stress in mice [12]. We subsequently examined whether the now established NPSR-OX₁R-NK₁R-mGlu5R-CB₁R cascade is involved in the SIA induced by the same restraint stress protocol. Mice receiving an acute restraint stress for 30 min exhibited significant reduced paw withdrawal response in the hot-plate test. This SIA diminished within 20 min (Fig. 6a) [12] and was significantly prevented in mice i.c.v. pre-treated with an NPSR antagonist ([tBu-D-Gly⁵] NPS, 10 nmol) (Fig. 6a and d) or with i.pag. Pretreated with NK₁R (L-703,606, 10 nmol) (Fig. 6b and d) and mGlu₅R (MPEP, 30 nmol) antagonists (Fig. 6c and d), respectively. The overall comparison of the time course of the antinociceptive effect showed a significant difference between time and treatment (F_18,114 = 4.317, p < 0.001, two-way ANOVA, Fig. 6a; F_18,108 = 3.780, p < 0.001, two-way ANOVA, Fig. 6b; F_18,108 = 3.501, p < 0.001, two-way ANOVA, Fig. 6c). Take together with our previous findings that SB-334867 (15 nmol, i.pag.) and AM251 (30 nmol, i.pag.) prevented SIA [12], it is suggested that SIA is mediated via the NPSR-evoked OX₁R-NK₁R-mGlu₅R-CB₁R cascade in the vlPAG.

All the tested antagonists at the administered doses that attenuated SIA had no effect on the number of crossing and rearing, including the NPSR antagonist ([tBu-D-Gly⁵] NPS, 10 nmol, i.c.v.) (Additional file 3: Figure S3a and b), the NK₁R antagonist (L-703,606, 10 nmol, i.pag.) (Additional file 3: Figure S3, C and D) and the mGlu₅R antagonist (MPEP, 30 nmol, i.pag.) (Additional file 3: Figure S3, E and F), similar to the OX₁R antagonist (SB-334867, 15 nmol, i.pag.) and the CB₁R antagonist (AM251, 30 nmol, i.pag.) which we have reported previously [12]. This supports that these antagonists attenuate SIA by blocking their respective endogenous ligands.

Measurement of the neuropeptide content in the brain homogenate revealed that restrain stress significantly elevated the NPS level in the LH (df = 12, t = 2.987, p < 0.05, Student’s t-test, Fig. 7a) and the substance P level in the vlPAG (df = 9, t = 2.72, p < 0.05, Student’s t-test, Fig. 7b). Similar elevations in orexin-A levels were observed in the vlPAG of restrained mice as we previously reported [12].

Drew et al. [36], using an electrophysiological approach, has shown that, in the vlPAG, substance P can facilitate the release of glutamate that subsequently activates postsynaptic mGlu₅Rs located at perisynaptic sites, leading to the synthesis of 2-AG that retrogradely inhibits presynaptic GABA release via CB₁Rs. Gregg et al. [40], using a behavioural approach, also demonstrated that activating mGlu₅Rs in the PAG can induce an antinociceptive effect mediated by 2-AG and CB₁Rs. Here, we further demonstrated that this substance P-initiated and mGlu₅R-mediated eCB retrograde signalling contributes to the antinociceptive effect of substance P in the vlPAG since i.pag. substance P-induced antinociception was antagonized by i.pag. blockade of NK₁Rs, mGlu₅Rs or CB₁Rs. The present study also supports that substance P is antinociceptive at the supraspinal level and the vlPAG is one of the sites of action.

The finding that the NK₁R antagonist attenuated orexin-induced antinociception in the vlPAG (Fig. 4) suggests that substance P acts downstream of orexin-induced antinociception. This finding is in agreement with a recent study that reported that the level of substance P in the vlPAG was increased following i.pag. orexin-A administration in rats [48]. Previously, we have shown that orexin via OX1Rs induces analgesia through a GqPCR-PLC-DAGL-2-AG-CB₁R retrograde disinhibition mechanism in the vlPAG [11]. Given that mGlu₅R, a GPCR, is also coupled to Gq proteins and mediates the antinociceptive effect via the same 2-AG-dependent disinhibition mechanism in the PAG [40] as the OX₁R [11], it is reasonable to suggest that the mGlu₅R is a downstream target after OX₁R-NK₁R activation. That is, orexin may induce analgesia via a cascade mediated by the OX₁R-substance P-NK₁R-glutamate-mGlu₅R-PLC-DAGL-2-AG-CB₁R signalling sequentially in the PAG (Fig. 8).

This sequential cascade may be able to explain the previous finding that i.pag. blockade of NK₁Rs attenuated LH stimulation-induced antinociception [49]. It is likely that orexin is the mediator released from the LH to induce antinociception indirectly via the NK₁R in the PAG. Additionally, involvement of substance P in the antinociceptive effect of orexin may also explain our previous electrophysiological finding that, in certain recorded vlPAG neurons, orexin-A did not induce postsynaptic depolarization but attenuated GABA release via presynaptic CB₁Rs [11]. In addition to the 2-AG spill-over hypothesis, orexin-A may activate neurokinin neurons to release substance P that indirectly inhibits GABA release via mGlu₅R-eCB signalling in those neurons that were not depolarized by orexin-A.

In agreement with previous studies that i.c.v. NPS was antinociceptive in swiss mice [19–21], we also found i.c.v. NPS reduced the hot-plate nociceptive response in C57BL/6JNarl mice. Peng et al. [20] suggested that the PAG is likely the site of action of NPS since i.c.v. NPS increased the c-Fos expression in the PAG where the NPSR mRNA is abundant [23]. However, our findings that direct i.pag. microinjection of NPS failed to induce antinociception and that i.c.v., but not i.pag., blockade of NPSRs antagonized i.c.v. NPS-induced antinociception indicate that i.c.v. NPS may act in brain regions other than the PAG to exert its antinociceptive effect.

The findings that i.pag. blockade of OX₁Rs, NK₁Rs, mGlu₅Rs and CB₁Rs prevented i.c.v. NPS-induced antinociception suggest the involvement of the OX₁R-NK₁R-mGlu₅R-CB₁R signalling in the vlPAG in the supraspinal antinociceptive action of NPS. The site of action is likely in the hypothalamic areas where orexin neurons are located, especially the LH that is involved in pain regulation. Ideally, it would be more precise to study the action of NPS and its antagonist on orexin neurons in the LH via intra-LH microinjection. However, due to the difficulty of performing both intra-LH and i.pag. cannulations in mice, i.c.v. and i.pag. microinjections were employed (Figs. 5 and 6). Nevertheless, several studies have suggested an interaction between NPS and orexin systems. Anatomical and functional studies suggest that NPS can activate orexin neurons and may modulate biological functions indirectly via released orexins. First, the hypothalamic regions where orexin neurons are located, including the LH, PFA, and DMH, are enriched with NPSRs [23]. Second, after i.c.v. injection of NPS in rats, fos-immunoreactive cells in the hypothalamus, especially in the LH, were orexin-A-positive [27, 28]. Third, NPS has been reported to be an upstream activator of the orexin system in feeding [27] and addiction [28] behaviours acting in the hypothalamus. Therefore, it is likely that NPS activates the orexin neurons at the LH, releasing orexins in the vlPAG to induce antinociception.

Previously, we have demonstrated that SIA is mediated by orexins released from the LH, an important region for SIA [13], via an OX₁R-initiated 2-AG-dependent disinhibition mechanism in the vlPAG [12]. Here, we extend the findings in this study to suggest that NPS activates hypothalamic orexin neurons and add substance P as downstream of vlPAG OX₁R activation in this SIA mechanism. That is, during stress, hypothalamic orexin neurons are activated by NPS, which is released possibly from the peri-LC and/or the KF-PBN in mice [24], releasing orexins that activate the OX₁Rs on neurokinin neurons in the vlPAG. Then, substance P is released and activates the NK₁R-containing glutamate neurons, yielding massive glutamate release that in turn activates perisynaptic mGlu₅Rs to initiate GqPCR signalling and generation of 2-AG. This eCB then retrogradely activates presynaptic CB₁Rs to inhibit GABA release in the vlPAG, ultimately leading to analgesia (Figs. 1 and 8). This conclusion is based on the following findings, which may fill the gaps [★] in the schema depicted in Fig. 1, that (1) stress increased NPS levels in the LH (Fig. 7a) and SIA was reduced by blocking NPSRs (Fig. 6a); (2) stress increased orexin levels in the vlPAG and SIA was reduced by blocking OX₁Rs in the vlPAG [12]; (3) stress increased substance P levels (Fig. 7b) and SIA was reduced by blocking NK₁Rs in the vlPAG (Fig. 6b); (4) SIA was reduced by blocking either mGlu₅Rs (Fig. 6c), CB₁Rs or DAGL in the vlPAG [12]. The antagonist/inhibitor of NPSRs (BuG-NPS, Fig. 6a), OX₁Rs (SB-334867) [12], NK₁Rs (L-703,606, Fig. 6b), mGlu₅Rs (MPEP, Fig. 6c), CB₁Rs (AM251) or DAGL (tetrahydrolipstatin) [12] employed at the dose blocking SIA, per se, did not affect nociceptive threshold in unrestrained normal mice, suggesting no non-specific effects of these antagonists employed at the concentrations used in this study.

Since 1990s, substance P has been reported to play a role in SIA while the site(s) of action remain unidentified. Rosen et al. [38] reported that substance P was released from the PAG of animals in response to a behavioural stress, suggesting that endogenous substance P contributes to SIA originated from the PAG-mediated descending pain inhibition. The finding that the antinociceptive effect induced by stimulating the LH was abolished by i.pag. L-703,606 [39], suggesting that stimulating the LH can release substance P to induce antinociception via the NK₁Rs in the PAG. Here, we provided direct evidence supporting that SIA is mediated by elevated substance P in the PAG.

Several lines of evidence have indicated the involvement of NPS in stress-induced responses. NPSRs are enriched in the amygdala and hypothalamus [22], stress-related brain regions. The number of c-fos-containing NPS neurons in the peri-LC and KF-PBN was increased after a short-term forced swim stress or restraint stress [24]. The current finding that the acute restraint stress that induces analgesia can increase the NPS level in the LH directly supports that NPS is released during stress and contributes to SIA.

Several reports have indicated a cross-modulatory relationship between NPS and the corticotrophin releasing factor (CRF) system in stress-related responses. Paneda et al. [50] reported that CRF₁ receptor may mediate NPS-induced cocaine reinstatement in mice. Conversely, Jungling et al. [4]. demonstrated that CRF can modulate NPS neurons in the LC of mice following acute stress. It remains to be elucidated if interactions between the CRF system and the NPSR-OX₁R-NK₁R-mGlu₅R-CB₁R-mediated sequential cascade in SIA occur.

In the current study, we found that NPS (i.c.v.) at 0.3 and 1 nmol in C57BL/6 did not induce significant hyperlocomotion (Additional file 2: Figure S2). This is different with previous studies, where i.c.v. NPS at doses of 0.1 and 1.0 nmol induced hyperlocomotion in C57BL/6 [50] and Swiss mice [4]. However, Rizzi et al. [51], Castro et al. [52] and Boeck et al. [53] consistently demonstrated that i.c.v. NPS, only at the dose of 0.1 nmol, but not 0.01 and 1.0 nmol, exhibited significant hyperlocomotion in CF-1 mice. Furthemore, Holanda et al. [21] reported that i.c.v. NPS at 0.1 nmol did not increase locomotor activity in CF-1 mice. The discrepancy among studies is unclear. It may be that the i.c.v. NPS doses employed under the conditions (mouse strain and the motor activity assessment) in the present study fall outside the optimal dose for inducing hyperlocomotion.

All data generated or analyzed during this study are included in this published article and its supplementary information files.