Different neuronal populations mediate inflammatory pain analgesia by exogenous and endogenous opioids

Mu-opioid receptors (MORs) are crucial for analgesia by both exogenous and endogenous opioids. However, the distinct mechanisms underlying these two types of opioid analgesia remain largely unknown. Here, we demonstrate that analgesic effects of exogenous and endogenous opioids on inflammatory pain are mediated by MORs expressed in distinct subpopulations of neurons in mice. We found that the exogenous opioid-induced analgesia of inflammatory pain is mediated by MORs in Vglut2+ glutamatergic but not GABAergic neurons. In contrast, analgesia by endogenous opioids is mediated by MORs in GABAergic rather than Vglut2+ glutamatergic neurons. Furthermore, MORs expressed at the spinal level is mainly involved in the analgesic effect of morphine in acute pain, but not in endogenous opioid analgesia during chronic inflammatory pain. Thus, our study revealed distinct mechanisms underlying analgesia by exogenous and endogenous opioids, and laid the foundation for further dissecting the circuit mechanism underlying opioid analgesia.


Introduction
Opium has been used for thousands of years, and opioid drugs still remain to be the most powerful analgesics used clinically. However, numerous adverse side-effects, including addiction, tolerance, hyperalgesia and respiratory depression, have limited its clinical use (Colvin et al., 2019;Streicher and Bilsky, 2018;Trang et al., 2015). The analgesic effect of morphine-like opioids is mainly mediated by mu-opioid receptor (MOR) (Kieffer, 1999;Matthes et al., 1996), which is encoded by the Oprm1 gene with many splicing isoforms (Chen et al., 1993;Pasternak, 2014;Wang et al., 1993). Besides their extensive distribution in the spinal cord and primary sensory neurons (Kemp et al., 1996;Scherrer et al., 2009), MORs are widely expressed in many pain-related brain regions, including the periaqueductal gray (PAG), thalamus, rostral ventromedial medulla (RVM), and anterior cingulate cortex (ACC) Erbs et al., 2015;Fields, 2004). In addition, MORs are highly expressed in the regions which are involved in reward or emotion such as the ventral tegmental area (VTA), nucleus accumbens (NAc), and amygdala (Fields and Margolis, 2015;Lutz and Kieffer, 2013).
The MORs expressed in different brain areas or different neuronal types might play distinct roles (Fields and Margolis, 2015;Kim et al., 2018). Activation of MOR by opioid drugs suppresses both sensory and emotional components of pain (Bushnell et al., 2013;Corder et al., 2018). Pharmacological and genetic approaches have previously been employed to examine the site of action for morphine analgesia (Maldonado et al., 2018). Injection of morphine in PAG, RVM and other brain areas evokes strong inhibition of nocifensive responses (Manning et al., 1994;Yaksh and Rudy, 1977), and activation of MOR in the PAG induces analgesia by descending modulation of the spinal cord via RVM (Basbaum and Fields, 1984). Moreover, opioids can differentially modulate subsets of the RVM neurons, which gate the spinal circuit for nociception via presynaptic mechanisms (Fields, 2004;Franç ois et al., 2017). MOR is also widely expressed in the dorsal root ganglion (DRG) and different neuronal populations at the spinal level (Kemp et al., 1996;Scherrer et al., 2009;Spike et al., 2002;Wang et al., 2018). Previous studies suggest that these MORs play important roles in mediating morphine-induced antinociception in both acute and inflammatory pain (Corder et al., 2017;Stein et al., 2009;Sun et al., 2019;Weibel et al., 2013). Thus, the site of action for analgesia by systemic morphine administration remains incompletely resolved.
Endogenous opioid system also plays a critical role in gating neural circuits underlying nociceptive information processing, as evidenced by elevated pain in response to the opioid antagonist (Sun et al., 2003). The important role of endogenous opioid system is also supported by the finding that persistent pain and placebo induced significant activation of opioid system (Wager et al., 2007;Zubieta et al., 2005), and placebo-induced analgesia was reversed by non-selective opioid antagonist naloxone (Benedetti et al., 1999). Analgesia by the endogenous opioid system is likely due to the fast release of endogenous opioid peptides in both spinal and supraspinal brain areas, leading to attenuation of sensory and pain-specific affective responses by activating the opioid receptors . In addition, opioid-independent constitutive MOR activation observed in chronic inflammatory pain represents another analgesic mechanism of the opioid system (Corder et al., 2013). Although brain imaging studies showed that exogenous and endogenous opioids may activate similar brain areas (LaGraize et al., 2006;Zubieta et al., 2001), it remains unknown whether their effects are mediated by the same population of neurons.
Elucidating the neural mechanism underlying opioid analgesia is essential for developing novel strategies for pain management. In this study, we determined the identity of MOR + neurons in multiple brain areas, and delineated the different mechanisms underlying the analgesic effects of exogenous and endogenous opioids, using a combination of genetic and pharmacological approaches.

Results
Characterization of MOR + neurons in the mouse brain MORs are widely expressed in the peripheral and central nervous system. Previous studies have examined the distribution of MORs in the nervous system with ligand binding-based autoradiography and knock-in reporter mouse lines (Erbs et al., 2015;Gardon et al., 2014;Mansour et al., 1988;Wang et al., 2018), or based on the distribution of Oprm1 mRNA, which encodes the MOR (George et al., 1994;Mansour et al., 1995). However, the identity of MOR + neurons in multiple brain regions is still not clear. Thus, we determined the identity of the MOR + neurons in the brain by examining the expression of glutamatergic and GABAergic neuronal markers in MOR + neurons. We performed triple fluorescence in situ hybridization (FISH) with Oprm1 (MOR), Slc17a6 (Vglut2), Slc32a1 (Vgat) probes in wild-type mice (Figure 1a-f, Figure 1-figure supplement 1a-b). Consistent with previous studies (Erbs et al., 2015;Mansour et al., 1995;Wang et al., 2018), Oprm1 was detected in many brain areas. By analyzing the co-localization between Oprm1 and Vglut2/Vgat, we found that in the central and medial part of thalamus and parabrachial nucleus (PBN), where neurons are mostly glutamatergic, nearly all Oprm1 + neurons were Vglut2 + (Figure 1a,b,d,e and g). In contrast, most Oprm1 + neurons were Vgat + in the striatum and central amygdala (CeA), which contain mostly GABAergic neurons (Figure 1a,c and g). For other brain areas, Oprm1 was expressed in both glutamatergic and GABAergic neurons, such as the periaqueductal gray (PAG), 22.5 ± 4.0% of Oprm1 + neurons were Vgat + and 73.3 ± 2.7% of Oprm1 + neurons were Vglut2 + (Figure 1d,f and g). For the cortex, which expresses low level of Vglut2, we used Slc17a7 (Vglut1) as a marker for glutamatergic neurons, and found that 78.8 ± 2.0% of Oprm1 + neurons were positive for Vglut1 and only 19.2 ± 2.2% of Oprm1 + neurons were positive for Vgat, suggesting that most cortical Oprm1 + neurons were glutamatergic (Figure 1-figure supplement 1c-f). In rare cases, such as dorsal raphe nucleus, Oprm1 + neurons were neither Vglut2 + nor Vgat + (Figure 1d). Taken together, these results demonstrate that MORs are widely expressed in both glutamatergic and GABAergic neurons in the brain.

Generation and verification of the Oprm1 KI/KI mouse line
To study the functional role of MORs expressed in distinct neuronal populations in opioid analgesia, we generated a mouse line with a strategy similar to the 'knockout first' approach (Skarnes et al., 2011). This mouse line allows for selectively expressing MORs in distinct neuronal populations. This was achieved by inserting one stop cassette flanked by two loxP sites between exon 1 and exon 2 of Oprm1 gene (Figure 2a). This mouse line, referred to as Oprm1 KI/KI hereafter, enables re-expression of MORs under the control of Cre recombinase. Immunostaining results showed that MOR expression was abolished in the brain, spinal cord and dorsal root ganglion (DRG) of Oprm1 KI/KI mice compared to wild-type mice (Figure 2b-c). Next, we determined whether Oprm1 KI/KI mice exhibit deficiency in MOR functions by testing the effects of morphine on pain and locomotion. We found that Oprm1 KI/KI mice exhibited no response to morphine in pain and locomotion tests, whereas wide-type littermates exhibited strong morphine-induced analgesia and hyper-locomotion (Figure 2d-h). MORs are also known to mediate analgesic effect of endogenous opioids in complete Figure 1. Distribution of Oprm1 in Vglut2 + and Vgat + neurons in the mouse brain. (a-f) Multiple in situ hybridization in wild-type mice shows the distribution of Oprm1 in glutamatergic and GABAergic neurons using RNAscope assay. Scale bars, 1 mm (a and d), 100 mm (b), (c), (e), and (f). (g) Percentage of glutamatergic and GABAergic neurons in MOR expressing neurons in representative brain regions (n = 3 mice). NAc, nucleus accumbens; MS, medial septal nucleus; HPF, hippocampal formation; AH/DMH, anterior hypothalamic area/dorsomedial hypothalamic nucleus; LH, lateral hypothalamic area; ZI, zona incerta; CeA, central amygdala; IPN, interpeduncular nucleus; SN, substantia nigra; PAG, periaqueductal gray; PBN, parabrachial nucleus; SC, superior colliculus; IC, inferior colliculus. Data are presented as mean ± SEM. The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Percentage of glutamatergic and GABAergic neurons in MOR expressing neurons in representative brain regions.  MOR expression in brain, spinal cord and DRG sections of wild-type and Oprm1 KI/KI mice. Scale bars, 2 mm (left), 100 mm (middle and right). (d and e) Effects of morphine (10 mg/kg, subcutaneous, s.c.) on locomotor activity in open field test for Oprm1 KI/KI mice (n = 9) compared with wild-type (n = 11) littermates. Two-way ANOVA (d), F 1,36 = 102.1, p<0.0001; (e), F 1,36 = 107.3, p<0.0001 with Bonferroni correction. (f and g) Effects of morphine (10 mg/kg, s.c.) on thermal pain tested with tail immersion (50˚C) and hot plate (52˚C) tests in Oprm1 KI/KI mice compared with wild-type littermates. n = 7-11 mice. Two-way ANOVA (f), F 10,132 = 36.01, p<0.0001; (g), F 1,36 = 145.0, p<0.0001 with Bonferroni correction. (h) Summary of the formalin-induced behavioral responses in phase I (0-10 min) and phase II (10-60 min) in Oprm1 KI/KI mice and wild-type littermates treated with saline or morphine (10 mg/kg, intraperitoneal, i.p.). n = 5-7 mice. One-way ANOVA (Phase I, F 3,20 = 10.10, p=0.0003; Phase II, F 3,20 = 8.085, p=0.0010) with Bonferroni correction. (i and j) Thermal and mechanical pain tests during complete Freund's adjuvant (CFA)-induced inflammatory pain in Oprm1 KI/KI (n = 6) and wild-type (n = 9) littermates. Two-way ANOVA (i), F 9,130 = 8.312, p<0.0001; (j), F 9,130 = 17.90, p<0.0001 with Bonferroni correction. (k) Schematic diagram for driving MOR re-expression. Nestin-Cre mice were crossed with Oprm1 KI/KI mice, and the stop cassette (red hexagon) was excised. (l) MOR expression in the brain, spinal cord and DRG sections of Nestin-Cre/Oprm1 KI/KI mice. Scale bars, 2 mm (left), 100 mm (middle and right). (m and n) Effects of morphine (10 mg/kg, s.c.) on moving distance and average speed of locomotion with open field test in Nestin-Cre/Oprm1 KI/KI mice (n = 10) compared to Oprm1 KI/KI (n = 9) littermates. Twoway ANOVA (m), F 1,34 = 61.95, p<0.0001; (n), F 1,34 = 51.10, p<0.0001 with Bonferroni correction. (o and p) Effects of morphine (10 mg/kg, s.c.) on thermal pain tested with tail immersion (50˚C) and hot plate (52˚C) tests in Nestin-Cre/Oprm1 KI/KI mice, compared to Oprm1 KI/KI littermates. n = 5-10 mice. Two-way ANOVA (o), F 10,99 = 16.66, p<0.0001; (p), F 1,34 = 8.616, p=0.0059 with Bonferroni correction. Data are presented as mean ± SEM, *p<0.05, **p<0.01, ***p<0.001. The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Raw data of the behavioral tests and MOR expression level in wild-type and Nestin-Cre/Oprm1-KI mice. Freund's adjuvant (CFA)-induced chronic inflammatory pain (Corder et al., 2013). Consistent with previous findings, we found that wild-type mice, but not Oprm1 KI/KI mice, recovered from CFAinduced hyperalgesia 3-4 weeks after CFA injection, suggesting that the endogenous opioid analgesia was abolished in Oprm1 KI/KI mice (Figure 2i-j). By contrast, Oprm1 KI/KI mice exhibited comparable basal locomotor activity and pain threshold with wild-type littermates (Figure 2d-j). These results demonstrate that insertion of a stop cassette in the first intron of Oprm1 gene abolished expression and function of MORs.
Next, we determined whether MORs could be re-expressed after removal of the stop cassette by Cre recombinase. We bred Oprm1 KI/KI mice with Nestin-Cre mice (Figure 2k), which express Cre recombinase in the nervous system under the control of Nestin promoter (Tronche et al., 1999). Immunostaining showed that the expression pattern of MOR in the brain, spinal cord and DRG of Nestin-Cre/Oprm1 KI/KI mice was comparable to that in wild-type mice (Figure 2b and l). We analyzed the expression level of MOR in sections of Nestin-Cre/Oprm1 KI/KI mice, and found the expression of MOR in Nestin-Cre/Oprm1 KI/KI mice was comparable to that in wild-type mice (Figure 2figure supplement 1). Whether re-expressing MOR in Nestin-Cre/Oprm1 KI/KI mice could restore MOR functions was further examined with behavioral experiments. We found that morphine significantly increased the response latency to noxious stimuli and locomotor activity in Nestin-Cre/ Oprm1 KI/KI but not in Oprm1 KI/KI mice (Figure 2m-p). These data confirmed that the Oprm1 KI/KI mouse line enabled restoration of functional MORs in the presence of Cre recombinase. This mouse line was used throughout this study to examine the role of MORs in different neuronal populations during opioid analgesia.

MORs in Vglut2 + glutamatergic neurons mediate exogenous opioid analgesia
Although MORs are highly expressed in the glutamatergic neurons of pain-related regions, the functional role of these MORs in opioid analgesia remains largely unknown. We examined the functional role of MORs expressed in glutamatergic neurons in opioid analgesia with Vglut2-Cre/Oprm1 KI/KI mice (Figure 3a), which selectively express MORs in Vglut2 + glutamatergic neurons. Immunostaining showed that the expression of MOR in several brain areas composed mostly by glutamatergic neurons, such as the habenula, thalamus and PBN, in Vglut2-Cre/Oprm1 KI/KI mice resembled that in wild-type mice (Figure 3b-c, Figure 3-figure supplement 1a-b). We next examined the functional role of MORs expressed in Vglut2 + glutamatergic neurons in opioid analgesia with Vglut2-Cre/ Oprm1 KI/KI mice. We found that systemic morphine injection induced significant analgesic effect in Vglut2-Cre/Oprm1 KI/KI mice during tail immersion, hot plate, and von Frey tests (Figure 3d-f). While, the basal pain thresholds were indistinguishable between Vglut2-Cre/Oprm1 KI/KI and Oprm1 KI/KI littermates (Figure 3-figure supplement 1c-e). Further, we determined the role of MORs on Vglut2 + glutamatergic neurons in mediating morphine analgesia for inflammatory pain. Although the nocifensive behaviors in formalin test were comparable between Vglut2-Cre/Oprm1 KI/KI and Oprm1 KI/KI mice (Figure 3-figure supplement 1f-g), Vglut2-Cre/Oprm1 KI/KI mice showed significantly less formalin-induced licking and flinching behaviors as compared with Oprm1 KI/KI mice after morphine treatment (Figure 3g-h), suggesting an important role of MORs expressed in Vglut2 + glutamatergic neurons in mediating morphine analgesia on inflammatory pain. Consistently, under chronic inflammatory pain condition induced by CFA, systemic morphine injection significantly increased the withdrawal latency in Hargreaves test and mechanical threshold in von Frey test in Vglut2-Cre/Oprm1 KI/ KI mice but not in Oprm1 KI/KI mice (Figure 3i-j), indicating that analgesic effects of morphine on CFA-induced chronic inflammatory pain was restored by selective expressing MOR in Vglut2 + glutamatergic neurons. Furthermore, locomotion analysis showed that morphine injection significantly increased the locomotor distance and speed in Vglut2-Cre/Oprm1 KI/KI mice but not in Oprm1 KI/KI mice after the morphine treatment (Figure 3-figure supplement 1h-i), suggesting the increase of latency in Hargreaves test and mechanical threshold in von Frey test was not due to deficit in motor function. Thus, activation of MORs in Vglut2 + glutamatergic neurons by systemic morphine suppressed both acute and chronic inflammatory pain.
We further confirmed these results by selectively deleting MORs from Vglut2 + glutamatergic neurons (

MORs in GABAergic neurons mediate endogenous opioid analgesia
MORs are also highly expressed in the GABAergic neurons. Although recent studies explored the role of MORs expressed in forebrain GABAergic neurons in morphine analgesia (Charbogne et al., 2017;Cui et al., 2014), the functional role of MORs in GABAergic neurons in opioid analgesia, especially endogenous opioid analgesia, remains largely unknown. We thus examined the role of MORs in GABAergic neurons using Vgat-Cre/Oprm1 KI/KI mice (Figure 4a), which selectively express MORs in GABAergic neurons. We found that Vgat-Cre/Oprm1 KI/KI mice showed strong expression of MORs in the striatum, hypothalamus, amygdala, interpeduncular nucleus (IPN) and ventral pallidum (VP) (Figure 4b and Figure 4-figure supplement 1a-b), consistent with the abundant Oprm1 expression in GABAergic neurons of these brain areas ( Figure 1g). Further analysis showed that the expression of MOR in multiple brain areas, such as the striatum, nucleus accumbens (NAc) and CeA, of Vgat-Cre/Oprm1 KI/KI mice resembled that in wild-type mice ( Figure 4c). Next, we performed behavioral experiments to examine the functional role of MORs expressed in GABAergic neurons with Vgat-Cre/Oprm1 KI/KI mice. The basal nociceptive responses showed no significant difference between Vgat-Cre/Oprm1 KI/KI and Oprm1 KI/KI mice ( Figure 4-figure supplement 1c-e). However, we found that morphine treatment significantly elevated the response latency of Vgat-Cre/Oprm1 KI/KI mice in the tail immersion test (Figure 4d), while systemic administration of morphine resulted in no significant analgesic effect on Vgat-Cre/Oprm1 KI/KI or Oprm1 KI/KI mice in the hot plate test (Figure 4e). We next examined the effect of morphine on mechanical pain using von Frey test, and found that systemic morphine injection increased the paw withdrawal frequency in response to mechanical stimulation, indicating that activation of   We further examined the role of MORs in GABAergic neurons in mediating the effect of morphine in inflammatory pain. For the formalin-induced inflammatory pain, we found that after systemic saline injection, the licking and flinching behaviors induced by formalin was comparable between Vgat-Cre/Oprm1 KI/KI and the Oprm1 KI/KI littermates (Figure 4-figure supplement 2g-h). Systemic morphine application shift the time course of formalin-evoked nocifensive behaviors rightward in Vgat-Cre/Oprm1 KI/KI mice compared to Oprm1 KI/KI mice, whereas the total durations were comparable in both phase I and phase II between the two groups ( Figure 4g-h). Furthermore, for CFAinduced chronic inflammatory pain, systemic morphine injection exhibited no significant effect on the responses evoked by thermal or mechanical stimuli, in either Oprm1 KI/KI or Vgat-Cre/Oprm1 KI/KI mice (Figure 4i-j), suggesting that the MORs in GABAergic neurons play insignificant role in mediating the analgesic effect of morphine in the CFA-induced inflammatory pain. In addition, the locomotion showed no difference between Vgat-Cre/Oprm1 KI/KI mice and Oprm1 KI/KI mice after morphine treatment (Figure 4-figure supplement 2i-j). Thus, MORs in GABAergic neurons are not involved in morphine analgesia under inflammatory pain conditions. Next, we asked whether MORs in GABAergic neurons are involved in analgesia by endogenous opioids under chronic inflammatory pain condition. We injected CFA in the hindpaw of both Vgat-Cre/Oprm1 KI/KI and Oprm1 KI/KI mice, and found both groups of mice developed hyperalgesia from 2 hr to 7 days after CFA injection. Surprisingly, the Vgat-Cre/Oprm1 KI/KI mice gradually recovered from hyperalgesia 3-4 weeks after CFA injection, but the Oprm1 KI/KI mice showed persistent hyperalgesia even 7 weeks after CFA application (Figure 4k-l), indicating that MORs in GABAergic neurons play a critical role in mediating the analgesia by endogenous opioids. Next, we confirmed that the recovery was mediated by MORs with the pharmacological approach, and found that intraperitoneal injection of b-FNA, a specific MOR antagonist, reduced the mechanical threshold in Vgat-Cre/ Oprm1 KI/KI mice, but not in Oprm1 KI/KI mice at 42 days after CFA injection ( Figure 4m). Further, we confirmed the role of MORs in GABAergic neurons in mediating analgesia by endogenous opioid with the mice in which the MORs were selectively deleted in GABAergic neurons ( Figure 4n). Consistently, we found that Vgat-Cre/Oprm1 fl/fl mice exhibited significantly slower recovery from hyperalgesia induced by CFA, as compared to Oprm1 fl/fl littermates (Figure 4o-p). Together, these results immersion (50˚C), hot plate (52˚C) and von Frey tests in Oprm1 KI/KI mice and Vgat-Cre/Oprm1 KI/KI mice. n = 5-8 mice. Two-way ANOVA (d), F 10,143 = 63.70, p<0.0001; (e), F 7,88 = 0.4163, p=0.8898; (f), F 5,78 = 7.448, p<0.0001 with Bonferroni correction. (g) Time course of formalin-induced nocifensive behaviors in Vgat-Cre/Oprm1 KI/KI mice compared to Oprm1 KI/KI mice with subcutaneous morphine injection (10 mg/kg). n = 8 mice for each group. (h) Summary of the nocifensive behaviors in phase I (0-10 min) and phase II (10-60 min) of formalin test in Vgat-Cre/Oprm1 KI/KI mice compared to Oprm1 KI/KI mice with subcutaneous morphine injection (10 mg/kg). n = 8 mice for each group. Student's unpaired t-test, t 14 = 1.248, p=0.2324 (Phase I); t 14 = 1.388, p=0.1868 (Phase II). (i and j) Time-course effects of morphine (10 mg/kg, s.c.) on thermal (i) and mechanical (j) sensitivities on CFA-induced pain responses in Oprm1 KI/KI mice (n = 8) and Vgat-Cre/Oprm1 KI/KI mice (n = 10) on d7 and d14, respectively. Two-way ANOVA (i), F 8,144 = 1.515, p=0.1567; (j), F 8,144 = 0.5818, p=0.7917 with Bonferroni correction. (k and l) Thermal and mechanical pain tests during CFA-induced inflammatory pain in Oprm1 KI/KI and Vgat-Cre/Oprm1 KI/KI mice. n = 8 mice for each group. Two-way ANOVA (k), F 11,168 = 2.848, p=0.0019; (l), F 9,140 = 6.034, p<0.0001 with Bonferroni correction. (m) Time-course effects of b-FNA (10 mg/kg, i.p.) on mechanical sensitivity on CFA-induced pain responses in Oprm1 KI/KI and Vgat-Cre/Oprm1 KI/KI mice on d42 when the pain response was mostly recovered (n = 8). RM one-way ANOVA (Oprm1 KI/KI , F 9,63 = 0.6393, p=0.7592; Vgat-Cre/Oprm1 KI/KI , F 9,63 = 5.992, p<0.0001) with Bonferroni correction. (n) Schematic diagram for conditional knockout MOR from GABAergic neurons. Vgat-Cre mice were crossed with Oprm1 fl/fl mice to generate Vgat-Cre/Oprm1 fl/fl mice. (o and p) Thermal and mechanical pain tests during CFA-induced inflammatory pain in Oprm1 fl/fl (n = 6) and Vgat-Cre/Oprm1 fl/fl mice (n = 7). Two-way ANOVA (o), F 11,132 = 5.428, p<0.0001; (p), F 11,132 = 4.943, p<0.0001 with Bonferroni correction. Data are presented as mean ± SEM, **p<0.01, ***p<0.001. The online version of this article includes the following source data and figure supplement(s) for figure 4: Source data 1. Raw data of the behavioral tests and MOR expression level in Vgat-Cre/Oprm1-KI and Vgat-Cre/Oprm1-fl groups of mice.  indicate that the analgesia by endogenous opioids under chronic inflammatory pain condition is mainly mediated by MORs expressed in GABAergic neurons.

MORs in the dorsal spinal cord mediate exogenous but not endogenous opioid analgesia
MORs are highly expressed in the dorsal spinal cord, and previous pharmacological studies indicate that these MORs play essential roles in opioid analgesia (Kemp et al., 1996;Wang et al., 2018). Given the potential non-specificity of pharmacological approaches, we further explored the functional role of MORs expressed in the dorsal spinal cord in exogenous and endogenous opioid analgesia with genetic approaches. We selectively expressed MORs in the dorsal spinal cord by crossing the Oprm1 KI/KI mice with Lbx1-Cre mice (Figure 5a), in which Cre is expressed in the dorsal spinal cord (Sieber et al., 2007). Immunostaining results showed that MORs were rescued in the dorsal spinal cord but not DRG in Lbx1-Cre/Oprm1 KI/KI compared to Oprm1 KI/KI mice, and the MOR signal in the superficial layer of spinal cord was restricted in layers labeled with IB4 (Figure 5b-c). We next tested the functional roles of MORs in the dorsal spinal cord in opioid analgesia with these animals. Although the basal pain thresholds of Lbx1-Cre/Oprm1 KI/KI and Oprm1 KI/KI littermates were comparable (  (Wang et al., 2018). We further determined the distribution of MOR + neurons in the spinal cord using in situ hybridization. Consistent with previous observations (Wang et al., 2018), we found that Oprm1 + neurons are widely distributed in the spinal cord, and that a significant proportion of the Oprm1 + neurons localized in superficial layers of dorsal spinal cord (Figure 6a-b). We next determined the identity of MOR + neurons by using in situ hybridization, and found that MORs are expressed in both GABAergic and glutamatergic dorsal spinal neurons with comparable percentage (Figure 6c-d).
We further determined the functional role of MORs expressed in excitatory neurons at the spinal level by using the Vglut2-Cre/Oprm1 KI/KI mice (Figure 3a). MORs were re-expressed in both spinal cord and DRG neurons in Vglut2-Cre/Oprm1 KI/KI mice (Figure 6e-f). We selectively activated MORs expressed in the glutamatergic neurons at the spinal level by intrathecal injecting of morphine, and (d-f) Effects of morphine (10 mg/kg, s. c.) on pain tests with tail immersion (50˚C), hot plate (52˚C) and von Frey tests in Oprm1 KI/KI (n = 12) and Lbx1-Cre/Oprm1 KI/KI (n = 11) mice. Two-way ANOVA (d), F 1,42 = 33.80, p<0.0001; (e), F 1,42 = 7.038, p=0.0112; (f), F 5,126 = 5.717, p<0.0001) with Bonferroni correction. (g) Time course of formalininduced nocifensive behaviors in Lbx1-Cre/Oprm1 KI/KI (n = 9) mice compared with Oprm1 KI/KI (n = 7) littermates with subcutaneous morphine injection (10 mg/kg). (h) Summary of the nocifensive behaviors in phase I (0-10 min) and phase II (10-60 min) of formalin test in Lbx1-Cre/Oprm1 KI/KI (n = 9) mice compared with Oprm1 KI/KI (n = 7) littermates with subcutaneous morphine injection (10 mg/kg). Student's unpaired t test, t 14 = 2.924, p=0.0111 (Phase I); t 14 = 0.1382, p=0.8921 (Phase II). (i and j) Time-course effects of morphine (10 mg/kg, s.c.) on thermal (i) and mechanical (j) sensitivities on CFAinduced pain responses in Oprm1 KI/KI mice (n = 8) and Lbx1-Cre/Oprm1 KI/KI mice (n = 5) on d7 and d14, respectively. Student's unpaired t-test, t 11 = 2.681, p=0.0214 (i), 60 min. (k and l) Thermal and mechanical pain tests during CFA-induced inflammatory pain in Oprm1 KI/KI (n = 8) and Lbx1-Cre/ Oprm1 KI/KI (n = 5) mice. Two-way ANOVA (k), F 9,110 = 0.6637, p=0.7400; (l), F 9,110 = 0.6157, p=0.7814 with Bonferroni correction. Data are presented as mean ± SEM, *p<0.05, ***p<0.001. The online version of this article includes the following source data and figure supplement(s) for figure 5: Source data 1. Raw data of the behavioral tests in Lbx1-Cre/Oprm1-KI and Lbx1-Cre/Oprm1-fl groups of mice.   Figure 6 continued on next page found morphine induced analgesic effect on both thermal and mechanical pain tests in Vglut2-Cre/ Oprm1 KI/KI mice but not Oprm1 KI/KI mice (Figure 6h-j). Since MORs in the DRG neurons are also located in the central terminals (Scherrer et al., 2009), the analgesic effect of intrathecal injection of morphine is also likely mediated by targeting at the MORs originating from DRG. To examine this possibility, we selectively expressed MORs in small DRG neurons with a DRG specific Cre mouse line (Agarwal et al., 2004), and found that MOR immunoreactivity was restricted to CGRP-positive lamina I and II outer layers in the dorsal spinal cord of SNS-Cre/Oprm1 KI/KI mice ( Figure 6-figure supplement 1a-b). In contrast, selective activation of MORs in the DRG neurons by using SNS-Cre/ Oprm1 KI/KI mice had no significant effect on the acute pain or formalin-induced nocifensive behaviors, and only slightly suppressed thermal pain in CFA-induced inflammatory pain ( Figure 6-figure supplement 1c-k). Taken together, our data indicate that MORs expressed in spinal glutamatergic neurons are the major target for morphine analgesia at the spinal level.
We next determined the functional role of MORs in GABAergic neurons at the spinal level using Vgat-Cre/Oprm1 KI/KI mice, in which expression of MORs was rescued in GABAergic neurons. We found that MORs were detected in the spinal lamina II inner layer labeled by IB4 and deeper layers ( Figure 6g). Selective activation of MORs expressed in spinal GABAergic neurons with Vgat-Cre/ Oprm1 KI/KI mice decreased both thermal sensitivity and mechanical thresholds in acute pain tests (Figure 6k-m). This manipulation also induced robust nocifensive responses including hindpaw licking, hindpaw flinching and body twisting behaviors, which could be blocked by a specific MOR antagonist, b-FNA (Figure 6n-o, Video 1). Consistent with the behavioral results, we found that activation of MORs expressed in spinal GABAergic neurons by intrathecal injection of morphine also significantly increased the number of neurons expressing c-Fos, a neuronal activity marker ( Figure 6p). Furthermore, we determined whether the activity of projection neurons was also increased after intrathecal morphine injection in Vgat-Cre/Oprm1 KI/KI mice. About 80% lamina I projection neurons project to PBN (Todd, 2010), and the spino-parabrachial pathway has been implicated in the processing of nociceptive signal (Han et al., 2015). We thus labeled the spinal projection neurons by  (    injection of retro-beads in the PBN (Figure 6q). We found that intrathecal morphine administration significantly increased the percentage of c-Fos + /beads + neurons in both superficial and deeper dorsal horn of Vgat-Cre/Oprm1 KI/KI mice (Figure 6r-s), suggesting that the PBN-projecting neurons were excited. These results further support the notion that activation of MORs in spinal GABAergic neurons induces hyperalgesia.

MORs in the PBN mediate morphine analgesia in inflammatory pain
Our results indicate that the MORs expressed in the glutamatergic neurons in the brain play a crucial role in morphine analgesia but not endogenous opioid analgesia. We further tested this hypothesis. Given that MORs are highly expressed in glutamatergic neurons in external lateral part of the PBN (Figure 1d-e) and that the PBN is involved in pain processing (Barik et al., 2018;Han et al., 2015;Huang et al., 2019;Rodriguez et al., 2017), we thus examined the role of MORs expressed in the PBN in opioid analgesia. We generated a MOR-iCreER T2 mouse line to label the MOR + neurons, and performed double in situ hybridization in MOR-iCreER T2 Â Ai9 mice to verify the efficiency and specificity of the MOR-iCreER T2 mouse line (Figure 7-figure supplement 1a-b). We found that 94.5 ± 1.8% tdTomato + neurons were positive for Oprm1, indicating high specificity, whereas the labeling efficiency varied among different brain areas, on average 46.8 ± 2.8% (range: 15.3 ± 3.1% to 85.6 ± 0.5%) of Oprm1 + neurons were positive for tdTomato (Figure 7-figure supplement 1cf). The labeling efficiency of MOR + neurons in PBN was 71.3 ± 5.2%. Next, we examined the response of MOR + neurons in the PBN to noxious stimuli by fiber photometry measurements of fluorescence signals of a calcium indicator GCaMP6s, which was specifically expressed in PBN MOR + neurons (Figure 7a). We observed elevated GCaMP6s fluorescent signals in the PBN during noxious mechanical stimuli, including tail and hindpaw pinch (Figure 7b). For the thermal stimuli, high temperature (52˚C) but not the warm temperature (40˚C) stimuli significantly increased the fluorescent signals of GCaMP6s (Figure 7c). The quantitative analysis of the fluorescent signal following stimuli onset showed that MOR + neurons in the PBN were activated by both noxious mechanical and thermal stimuli (Figure 7d). The increase of fluorescent signal during noxious stimulation was not due to movement artifact, because the change of fluorescent signal of EYFP was not significant (Figure 7d). Moreover, systemic morphine application significantly decreased the calcium response of MOR + neurons to tail pinch, and this effect largely diminished 5 hr after morphine injection (Figure 7e-g). These data supporting the idea that MORs expressed in the PBN are involved in modulating nociceptive information processing.
Next, we tested whether the MORs expressed in the PBN mediate morphine analgesia. We reexpressed MORs in PBN neurons by injecting an AAV expressing Cre-EGFP (AAV-hSyn-Cre-EGFP) bilaterally in the PBN of Oprm1 KI/KI mice (Figure 7h), and AAV-hSyn-EGFP was used as control. These mice are referred to as PBN-Cre/Oprm1 KI/KI and PBN-EGFP/Oprm1 KI/KI mice hereafter, respectively. We found that MOR expression in the PBN was largely restored after injection of AAV-Cre virus (Figure 7i), and the expression pattern was comparable to that in the wild-type mice (Figure 2-figure supplement 1a). Behavioral tests showed that there was no significant difference in basal pain thresholds or the locomotor activity between the two groups ( We further inquired whether MORs expressed in the PBN modulate inflammatory pain. We examined the effect of morphine on formalin-induced acute inflammatory pain in PBN-Cre/Oprm1 KI/KI and PBN-EGFP/Oprm1 KI/KI mice. These two groups of mice showed comparable nocifensive behavior in the formalin test (Figure 7-figure supplement 2g-h), while systemic morphine injection significantly reduced the nocifensive behavior in both phase I and phase II in PBN-Cre/Oprm1 KI/KI mice in comparison to PBN-EGFP/Oprm1 KI/KI mice (Figure 7j-k), indicating an important role of MORs in the PBN in morphine analgesia in acute inflammatory pain. We next explored the role of PBN MORs in CFA-induced chronic inflammatory pain. The mice in both groups showed hyperalgesia after CFA application, while systemic morphine injection significantly elevated the thermal and mechanical PBN-Cre/Oprm1 KI/KI mice (n = 12) with subcutaneous morphine injection (10 mg/kg). (k) Summary of the nocifensive behaviors in phase I (0-10 min) and phase II (10-60 min) of formalin test in control mice (n = 13) and PBN-Cre/Oprm1 KI/KI mice (n = 12) with subcutaneous morphine injection (10 mg/kg). Student's unpaired t-test, t 23 = 2.682, p=0.0133 (Phase I); t 23 = 2.253, p=0.0341 (Phase II). (l and m) Effects of morphine (10 mg/kg, s.c.) on thermal sensitivity at d7 (l) and mechanical sensitivity at d14 (m) on CFA-induced pain in control mice (n = 11) and PBN-Cre/Oprm1 KI/KI mice (n = 9). Two-way ANOVA (l), F 1,36 = 2.222, p=0.1448; (m), F 8,162 = 4.550, p<0.0001 with Bonferroni correction. (n and o) Pain tests for thermal and mechanical sensitivities during CFA-induced inflammatory pain in PBN-EGFP/Oprm1 KI/KI and PBN-Cre/Oprm1 KI/KI mice. n = 7 mice for each group. Two-way ANOVA (n), F 7,96 = 0.3247, p=0.9411; (o), F 7,96 = 0.8869, p=0.5200 with Bonferroni correction. Data are presented as mean ± SEM; *p<0.05, **p<0.01, ***p<0.001. The online version of this article includes the following source data and figure supplement(s) for figure 7: Source data 1. Raw data of the fiberphotometry recording in Oprm1-GCaMP6s mice and behavioral tests in PBN-Cre/Oprm1-KI mice.  thresholds in the PBN-Cre/Oprm1 KI/KI mice but not the PBN-EGFP/Oprm1 KI/KI mice (Figure 7l-m). However, both groups of mice exhibited comparable persistent pain and did not recover from hyperalgesia in long-term test after CFA application (Figure 7n-o), indicating that MORs expressed in the PBN are not involved in endogenous opioid analgesia. Taken together, our results indicate that the MORs expressed in PBN glutamatergic neurons play an important role in mediating morphine analgesia in both acute and chronic inflammatory pain, but are not involved in endogenous opioid analgesia.

Discussion
Our study revealed that MORs expressed in glutamatergic and GABAergic neurons play diverse roles in mediating opioid analgesia. We found that analgesia by systemic morphine is largely mediated by MORs expressed in Vglut2 + glutamatergic neurons. By contrast, the MORs expressed in GABAergic neurons are crucial for analgesia by endogenous opioids during chronic inflammatory pain. In addition, MORs expressed at the spinal level and in the PBN play an important role in the analgesic effect of morphine in acute pain and inflammatory pain, respectively, but not in endogenous opioid analgesia. These results demonstrated that analgesia by exogenous and endogenous opioids are mediated by MORs expressed in different neuronal populations.

Distribution of MORs in different neuronal populations
We determined the identity of the Oprm1 + neurons with triple FISH. We found that Oprm1 was widely expressed in different population of neurons, with the percentage of Oprm1 in different cell types varying across different brain areas and the spinal cord. For example, in the PAG, although it was thought that MOR is highly expressed in GABAergic neurons and that activation of MOR inhibits GABAergic synaptic transmission (Vaughan et al., 1997), we found that a significant proportion of Oprm1 + neurons were glutamatergic. In addition, we found that MORs expressed in both Vglut2 + excitatory neurons and GABAergic neurons in the dorsal spinal cord with comparable proportion, which is in contrast to a previous study in rat showing that MORs were mainly expressed in excitatory neurons in the spinal cord (Kemp et al., 1996). Complementing to the above results, we also revealed the distribution of MORs expressed in different types of neurons with immunohistochemistry using genetically modified mice (Figures 3b and 4b; Figure 3-figure supplement 2b; Figure 4-figure supplement 2b). This is important as MORs are highly expressed in presynaptic terminals and play important roles in gating the synaptic transmission presynaptically (Li et al., 2016;Vaughan et al., 1997). Together, these results provide a comprehensive view of MOR distribution in the central nervous system.

Neural mechanisms underlying morphine analgesia
Our results identified the neuronal population that is responsible for analgesia by systemic morphine. Although early studies have implicated MORs expressed in GABAergic neurons in opioid analgesia (Fields, 2004), the key neuronal population underlying the analgesic effect of systemic morphine is not known. Our study demonstrated that analgesia by systemic morphine is mainly mediated by MORs expressed in Vglut2 + glutamatergic neurons. We confirmed these results with different genetic approaches. Our results are in line with recent genetic studies, showing that MORs expressed in forebrain GABAergic neurons were not involved in morphine analgesia (Charbogne et al., 2017;Cui et al., 2014). However, previous pharmacological studies showed that local injection of morphine in some forebrain and midbrain areas enriched with GABAergic neurons produced analgesia (Cohen et al., 1984;Manning et al., 1994;Yaksh and Rudy, 1977). This could be explained by the possibility that local injection of morphine could activate not only MORs expressed in local neurons but also MORs expressed in presynaptic terminals that originate from other brain areas, including those with abundant glutamatergic neurons. This emphasizes the importance of dissecting the mechanism underlying morphine analgesia by combining genetic and pharmacological approaches. Glutamatergic neurons consist of several sub-populations, which could be marked by Vglut1, Vglut2 and Vglut3. Given that Vglut2 + glutamatergic neurons represent the largest population, here we only examined the functional role of MORs expressed in Vglut2 + glutamatergic neurons in morphine analgesia. It is likely that other sub-populations of glutamatergic neurons might also involve in analgesic effect of opioids, given MORs are also expressed in the Vglut1 + neurons.
MORs are widely expressed in the brain and at the spinal level. We found that MORs at different areas play diverse roles in morphine analgesia in acute and inflammatory pain. At the spinal level, MORs are expressed in the DRG and spinal cord (Scherrer et al., 2009;Wang et al., 2018). Our results indicate that MORs expressed in the spinal cord are important for morphine analgesia in acute pain, and are partially involved in analgesia during inflammatory pain. This is evidenced by the data showing that selective activation of MORs in the dorsal spinal cord evoked analgesic effect in phase I, but not the phase II in formalin test. Our results are different from that obtained in rats by ablation of MOR-expressing neurons in the spinal cord, in which the effect of morphine on phase II was also decreased (Kline and Wiley, 2008). This analgesic effect of morphine in the spinal cord is most likely mediated by MORs expressed in the glutamatergic neurons, as activation of MORs in spinal GABAergic neurons induced allodynia and hyperalgesia. Additionally, MORs in the DRG is only partially involved in morphine analgesia in thermal hyperalgesia during CFA-induced chronic pain, which is likely due to upregulation of MORs during inflammation (Ballet et al., 2003). A recent study showed that Vgat is expressed in primary sensory neurons (Du et al., 2017). However, we observed no significant MOR signal in the DRG of Vgat-Cre/Oprm1 KI/KI mice, suggesting that MORs are barely expressed in Vgat + DRG neurons. Our behavioral data are consistent with previous studies with mice lacking MOR in the DRG (Corder et al., 2017;Weibel et al., 2013). Together with the data obtained with Vglut2-Cre/Oprm1 KI/KI mice (Figure 3), these data indicate that the MORs expressed at the supraspinal level play a more important role in mediating the analgesic effect of morphine in chronic pain. This is further evidenced by the observation that selective activation of MORs in the PBN, which consists of mostly glutamatergic neurons, significantly reduced inflammatory pain but not acute pain. However, the PBN is most likely one of many key brain areas involved in analgesia by systemic morphine, as re-expressing MOR in the PBN only partially rescued the effect of systemic morphine.
Pain is a complex feeling that consists of both sensory and affective components. In addition to suppressing pain perception, opioid analgesics also play an important role in relieving the emotional discomfort of pain Price et al., 1985). For example, activation of MORs in the anterior cingulate cortex significantly alleviated the aversion caused by chronic pain (LaGraize et al., 2006). In the present study, we only examined the analgesic effect of morphine for the sensory component of pain. Further studies are warranted to examine the functional role of MORs expressed in different population of neurons in mediating the opioid analgesics in affective component of pain.

Analgesia by endogenous opioids in chronic inflammatory pain
Previous pharmacological studies suggest that endogenous opioids are involved in both acute and chronic pain modulation (Corder et al., 2013), although several genetic studies did not reach consensus about the role of MORs in endogenous pain control (Maldonado et al., 2018). Our study revealed that MORs expressed in GABAergic neurons, but not glutamatergic neurons, play a critical role in analgesia by endogenous opioids. This is evidenced by the data showing that mice with selective expression of MORs in GABAergic, but not Vglut2 + glutamatergic neurons, recovered from CFA-induced hyperalgesia, while animals that lacking MORs did not. This is in line with previous thought that activation of MOR in GABAergic neurons lead to pain suppression (Al-Hasani and Bruchas, 2011;Fields, 2004). Furthermore, our results indicate that MORs expressed in the brain are responsible for endogenous opioid analgesia observed in chronic inflammatory pain, as conditional knockout of MOR from the dorsal spinal cord did not affect endogenous opioid analgesia. However, our results are inconsistent with a previous study indicating that constitutive MOR activity in the spinal cord produces endogenous opioid analgesia (Corder et al., 2013). Given that MORs could be expressed presynaptically in descending fibers and that activation of supraspinal MORs leads to an increase of descending inhibitory component (Gogas et al., 1991), it remains possible that the antinociceptive effect of endogenous opioid in the spinal cord relies on the activation of MORs expressed presynaptically in the descending pathway. Future studies on the specific neuronal populations that mediate endogenous opioid analgesia are needed.
An unexpected finding in the present study is that the endogenous opioid analgesia is mediated by mechanism different from exogenous opioid analgesia. Enhancement of endogenous opioid peptide release and opioid-independent constitutive activation of MOR have both been implicated in endogenous opioid analgesia Maldonado et al., 2018;Stein, 2016). Thus, one possibility is that the endogenous opioid peptides released during inflammation targeted at MORs expressed in subpopulation of neurons other than those MORs activated by morphine. Consistently, several brain areas including amygdala, in which MORs are mostly expressed in GABAergic neurons, were shown to be activated during sustained muscle pain (Zubieta et al., 2001). The other possibility is that the MORs expressed in GABAergic neurons but not the glutamatergic neurons are constitutively activated during CFA-induced inflammation. The reason that exogenous opioid application did not cause analgesic effect in mice with MORs selectively expressed in GABAergic neurons could be that exogenous opioid activates more MORs than the endogenously active MOR populations, and some of MOR + neurons targeted by exogenous opioid play opposite roles in pain modulation. This is highly likely because our results showed that the activation of MORs in spinal GABAergic neurons evoked hyperalgesia. In addition, it is worth noting that delta and kappa opioid receptors also play important roles in opioid analgesia . Nevertheless, the distinct mechanism underlying endogenous opioid analgesia provides new opportunity for targeting the pain-gating circuits with new approaches.

Technical issues
We have developed a new genetic model for studying the functional role of MORs in different populations. This new genetic approach is achieved by inserting a stop cassette flanked by loxP between exon 1 and exon 2 of Oprm1 gene (Figure 2a), leading to loss of MOR. This design allows for selectively re-expressing MOR in distinct neuronal population in a Cre-dependent manner. We have verified this strategy at both expressional and functional levels. The key advantage of this strategy is able to maintain the original expression pattern and level of MOR during re-expression, since the reexpression of MORs is still driven by its own promoter. This strategy is different from other studies that employed promoter of other genes in order to selectively express MOR in certain brain areas, which might not restore MOR expression with endogenous pattern (Cui et al., 2014).
It is noteworthy that Oprm1 KI/KI line is not a MOR null mouse line due to the complexity of the Oprm1 gene. It is known that the Oprm1 gene has two promoters and produces 3 classes of splicing variants, 7-transmembrane (TM), truncated 6-TM and 1-TM variants (Pasternak, 2014). There is also evidence showing that some of them might form heterodimers (Xu et al., 2013). Moreover, the regional distribution patterns of these different splicing variants are distinct from one another (Abbadie et al., 2000a;Abbadie et al., 2000b;Abbadie et al., 2004;Xu et al., 2014). In generating the Oprm1 KI/KI mouse line, a stop cassette was inserted after exon 1. As some exon 11 promoter-driving MOR splice variants could skip exon 1 (Pan et al., 2001), these MOR splicing variants may still express in Oprm1 KI/KI mice. However, it has been previously shown that the exon 11-related MOR splicing variants are not involved in morphine analgesia (Pan et al., 2009;Schuller et al., 1999), consistent with our results showing that morphine analgesia was lost in Oprm1 KI/KI mice. Therefore, the residual exon 11-related MORs in Oprm1 KI/KI mice would not affect our conclusion.
In summary, our study deciphered the mechanism underlying opioid analgesia with pharmacological and genetic approaches. We revealed that analgesia by exogenous and endogenous opioids are mediated by MORs expressed in different population of neurons. Our results also demonstrate that MORs in glutamatergic and GABAergic neurons at different areas in the nervous system play distinct roles in modulating nociceptive information processing. This study provides new mechanistic insight into the neural mechanisms underlying pain modulation, paving the way for designing new strategy for pain management.
All mice were raised on a 12 hr light/dark cycle (lights on at 7:00 a.m.) with ad libitum food and water. Behavioral experiments were carried out in the light phase. All procedures were approved by the Animal Care and Use Committee of the Institute of Neuroscience, Chinese Academy of Sciences, Shanghai, China.

Tissue preparation
Mice were anesthetized with pentobarbital sodium (100 mg/kg) and perfused transcardially with saline followed by 4% paraformaldehyde (PFA, Sigma). The brain, DRG (lumber segments), and spinal cord (lumbar segments) were dissected and post-fixed overnight at 4˚C in 4% PFA, followed by dehydration in 30% sucrose dissolved in PBS at 4˚C. Sections (40 mm for brain, 30 mm for spinal cord, 10 mm for DRG, and 20 mm for in situ hybridization) were prepared with a cryostat (Leica CM 1950) for immunostaining or in situ hybridization.

Immunofluorescent staining
Immunofluorescent staining was performed as described previously (Gao et al., 2019). Tissue sections were blocked for 1 hr at room temperature in PBST (0.3% Triton X-100) with 5% normal donkey serum, followed by incubation with primary antibodies at 4˚C overnight and secondary antibodies at room temperature for 2 hr in PBST (0.3% Triton X-100) with 1% normal donkey serum. Primary antibodies used for immunofluorescent staining were anti-MOR (rabbit, 1:500 for brain and spinal cord,

DAB staining
Tissue sections were blocked in PBS with 0.3% H 2 O 2 for 30 min, and then blocked in PBST (0.3% Triton X-100) with 1% goat serum for 2 hr at room temperature. Next, sections were incubated with primary antibody of anti-MOR (rabbit, 1:500, ABCAM, ab134054) for 36 hr at 4˚C, followed by secondary antibody of Biotin-SP goat anti-rabbit IgG (1:200, Jackson ImmunoResearch Laboratories) for 2 hr at room temperature. After incubation with ABC (1:100, Vector, PK-4000) for 1 hr, sections were developed in DAB (3, 3-Diaminobenzidine)-ammonium nickel sulfate developing solution. In the end, ethyl alcohol and xylene were used for decoloration and sections were mounted with quickhardening mounting medium. For quantitative analysis of the expression level of MOR, sections from the test group were stained with those of wild-type control in one well to control the variables.

In situ hybridization
To investigate the identity of MOR-expressing neurons, multiple in situ hybridization experiments were performed using RNAscope Fluorescent Multiplex Assay with Oprm1, Slc17a6, Slc32a1, Slc17a7 probes (Advanced Cell Diagnostics), and the 3-Plex negative probe was used as control. Tissue sections were spread on slides and heated for 2 hr at 60˚C, and then kept at À80˚C before experiments. According to manufacturer's protocol, slides were pretreated with hydrogen peroxide for 20 min at room temperature and washed in DEPC-H 2 O for 1 min. Then slides were transferred to boiling retrieval regent for 7 min, and rinsed once in DEPC-H 2 O at room temperature. Protease digestion was performed for 15 min in 40˚C HybEZ oven. After washed in DEPC-PBS for 3 min and rinsed in DEPC-H 2 O, slides were treated with ethanol for 3 min twice, followed by air-dry at room temperature. Pre-warmed probes were mixed and slides were hybridized for 2 hr in 40˚C HybEZ oven. Signal amplification fluorescent labels are TSA-based. The specificity of MOR-iCreER T2 was validated using RNAscope assay in MOR-iCreER T2 Â Ai9 mice with Oprm1 and tdTomato probes with the same procedure.

Image acquisition and analysis
Images were taken using Olympus VS120 microscope, Olympus FV3000 confocal fluorescence microscope and Nikon Tie-A1 plus confocal fluorescence microscope. For co-localization analysis, confocal images taken by Olympus FV3000 were cropped in 400 mm Â 400 mm areas. The number of counting areas ranges from 1 to 6, which depends on the area of each brain region in one section. Cell counting was carried out manually using Fiji (NIH). For the calculating of MOR re-expression efficiency using MOR DAB staining, images were taken by Olympus VS120 microscope and processed in Fiji. Images were transformed to 8-bit and substrated the background, and then gray value in specific brain regions were measured in 800 mm Â 800 mm areas, except for the CeA that 600 mm Â 600 mm areas were used in Vgat-Cre/Oprm1 KI/KI group. The expression level of MOR was normalized to the mean gray value of MOR signal in each brain region of wild-type mice.

Pain behavior tests
To evaluate morphine-induced analgesia, morphine was diluted in saline and injected subcutaneously (10 mg/kg) or intrathecally (1.0 nmol, 5 ml), and behavioral tests were performed 40 min or 30 min later, respectively. For tail immersion test, mice were gently restrained in a cotton towel and one third of the tail from tip was then dipped into a water bath of 48 or 50˚C. And the tail flick latency was recorded, with a cut-off time of 20 s to avoid tissue damage. The hot plate test was performed by placing mice on the hot plate (Ugo Basile, Comerio, Italy) at 52˚C, and the first appearance of hind paw licking, hind paw lifting or jumping was recorded. A maximal cut-off of 45 s was set to avoid tissue damage.
For testing mechanical sensitivity, mouse hind paw was perpendicularly stimulated with a series of von Frey hairs with logarithmically incrementing stiffness (0.07-1.4 grams, Stoelting, Wood Dale, IL). One filament was applied five times in one test, with intervals more than 5 s, and the paw withdrawal times were recorded. Paw withdrawal frequency was calculated to present the mechanical threshold.

Formalin test
For the formalin test, the mice received intraplantar injection of formalin (Sigma, F1635, 5%, vol/vol diluted in saline, 10 ml) and then replaced in transparent plastic chambers. Morphine (10 mg/kg) or saline was injected subcutaneously or intraperitoneally 30 min before formalin injection. Total duration of the animal spending in licking and flinching behaviors of the injected paw was counted manually in 5 min interval for 1 hr. The first phase of nociceptive responses was calculated during 0-10 min, and the second phase of nociceptive responses was calculated during 10-60 min.

Complete Freund's adjuvant (CFA)-induced inflammatory pain
For the CFA-induced pain, the mice were subcutaneously injected of 50% CFA (20 ml) into the plantar of the right hind paws. The CFA (50%) was prepared by isometric CFA (Sigma, F5881) and saline, and the mixture was emulsified by intensive mixing. The mechanical threshold and thermal sensitivity on the ipsilateral paw during chronic pain were tested by von Frey test and plantar test, respectively. For the von Frey test, the modified Dixon's up-down method (Chaplan et al., 1994) was conducted by using a series of von Frey filaments to determine the 50% hind paw withdrawal threshold. For the plantar test, the Hargreaves apparatus (Ugo Basile, Comerio, Italy) was used to measure the paw withdrawal latency, a 20 s was set as the cut-off time to prevent injury. Both the thermal threshold and mechanical sensitivity were averaged from two trials, separated by 30 min intervals at each time point. To determine the morphine analgesia on CFA-induced inflammatory pain, morphine (10 mg/ kg) was subcutaneously administrated on day 7 and day 14 after CFA injection, and then the mechanical threshold and thermal sensitivity were measured on 30, 60, 90, 120, 180 and 240 min, respectively. Both the mechanical threshold and thermal sensitivity were measured once at each time point for morphine test. To investigate the endogenous opioid function mediated by MOR expressed on the GABAergic neurons, b-funaltrexamine hydrochloride (b-FNA, 10 mg/kg), a specific MOR agonist, was intraperitoneally injected on day 42 after the tests on pain sensitivity in Vgat-Cre/ Oprm1 KI/KI and Oprm1 KI/KI mice. The mechanical threshold was tested on the ipsilateral paw before and 15,30,45,60,90,120,150,180 and 210 min after the b-FNA injection.

Open field test
Locomotor activity of mice was evaluated by open field test. Mice were placed in the testing room for about 1 hr for habituation. For baseline test, mice were placed in the open field boxes (40 Â 40 Â 40 cm) and videotaped individually. Then mice were taken out from the boxes and subcutaneously injected with 10 mg/kg morphine. The open field test was performed 40 min after injection. Total travel distance and average speed were recorded in a 10 min period. The track was analyzed by LabState (AniLab).

Fiber photometry recording
For recording the neuronal response to acutely nociceptive tail pinch stimulation in freely-moving animals, mice were placed on an open field and habituated for 3 min while the implanted optic fiber was connected to an external patch cord. Unexpected tail pinch using a clip (~10 s) was delivered for 3-5 times with an interval of 60-120 s. The start and stop time of each stimulation was tagged by triggering TTL signal (Isolated Pulse Stimulator Model 2100, A-M SYSTEMS), which was synchronously output to the fiber photometry system, and recorded with calcium signal simultaneously at 1000 Hz using F-scope-G-2 (Biolink Optics). To examine the morphine's effect on PBN MOR + neurons during tail pinch, tail pinch was first applied for three times to record a basal response. Morphine (10 mg/kg) was injected intraperitoneally, then tail pinch and recording were performed 40 min later. Based on our data, morphine lost its analgesic effect in about 4 hr, thus we measured the tail pinch-induced response 5 hr post morphine application. For recording the neuronal response during nociceptive hindpaw pinch and hot water, anesthetized animals (1% pentobarbital sodium, 50 mg/kg) received right hindpaw pinch or tail immersion (40 or 52˚C) for three times for each stimulation (~10 s) with an interval of~120 s. The excitation laser power was controlled at~15 mW. The system noise was recorded after disconnecting the tip of photometry fiber from implanted fiber and blocking any optic input. Data were transformed into mat file and analyzed in MATLAB.

Fiber photometry data analysis
The onset of event stimulation was identified first. Fluorescence values were low-pass filtered at 2 Hz using a 4th order Butterworth filter with zero-phase distortion and then subtracted the noise signal of recording system. The resulting values were aligned to event onset and corresponding fluorescence values in each stimulation was derived. The dynamics of fluorescence in each stimulation was calculated by DF=F ¼ ðF À F 0 Þ=F 0 F stands for each fluorescence value and F 0 stands for the median of fluorescence values during baseline window (À5 to 0 s before stimulation). The averaged fluorescence change was visualized by heat plot using jet colormap in MATLAB. To quantify the change of fluorescence values across stimulation period, the event windows were defined (0-5 s after the onset of tail pinch). The averaged fluorescence changes in each baseline and event window was calculated and compared.

c-Fos double-staining with retro-beads
One week after red retro-beads injection in PBN, c-Fos experiment was performed in Vgat-Cre/ Oprm1 KI/KI mice. Mice were placed in the testing room for about 2 hr for habituation. After habituation, morphine (1.0 nmol/ 5 ml) or saline was intrathecally injected to Vgat-Cre/Oprm1 KI/KI mice, and these mice were sacrificed 2 hr after morphine injection for further staining. Then immunostaining of c-Fos (rabbit, 1:15000, Synaptic System, 226003) was performed with the method of immunofluorescent staining.

Morphine-induced nocifensive behaviors
The mice were placed in transparent plastic chambers on a glass board. Morphine (1 nmol/ 5 ml) was injected intrathecally and the performance of mice after morphine administration was recorded by camera. Time spent on grooming, hindpaw licking, flinching, and body twisting in the following 30 min was analyzed in Matlab R2009a manually.

Quantification and statistical analysis
Statistical analysis was performed using GraphPad Prism 6, MATLAB R2009a and MATLAB R2015b. All the experiments were performed in the blinded manner. For behavioral tests, experimenters were blinded to the mouse genotypes or virus information. The number of experiments performed with independent mice (n) is indicated in the legends. The data were analyzed using Repeated-measures two-way ANOVA, one-way ANOVA followed by Bonferroni post hoc analysis, and two-tailed un-paired student's t-test. The cut-off for significance was held at p=0.05. All data are presented as mean with SEM.