Reactive oxygen species (ROS) generation is stimulated by κ opioid receptor activation through phosphorylated c-Jun N-terminal kinase and inhibited by p38 mitogen-activated protein kinase (MAPK) activation

Activation of the mitogen-activated protein kinase (MAPK) c-Jun N-terminal kinase (JNK) by the Gi/o protein–coupled κ opioid receptor (KOR), μ opioid, and D2 dopamine receptors stimulates peroxiredoxin 6 (PRDX6)-mediated production of reactive oxygen species (ROS). ROS production by KOR-inactivating antagonists norbinaltorphimine (norBNI) and JDTic blocks Gαi protein activation, but the signaling mechanisms and consequences of JNK activation by KOR agonists remain uncharacterized. Binding of arrestins to KOR causes desensitization of G protein signaling and acts as a scaffold to initiate MAPK activation. Here, we found that the KOR agonists U50,488 and dynorphin B stimulated biphasic JNK activation with an early arrestin-independent phase, requiring the small G protein RAC family small GTPase 1 (RAC1) and protein kinase C (PKC), and a later arrestin-scaffolded phase, requiring RAC1 and Ras homolog family member (RHO) kinase. JNK activation by U50,488 and dynorphin B also stimulated PRDX6-dependent ROS production but with an inverted U-shaped dose-response relationship. KOR agonist-induced ROS generation resulted from the early arrestin-independent phase of JNK activation, and this ROS response was suppressed by arrestin-dependent activation of the MAPK p38. The apparent balance between p38 MAPK and JNK/ROS signaling has important physiological implications for understanding of dynorphin activities during the stress response. To visualize these activities, we monitored KOR agonist–mediated activation of ROS in transfected live cells by two fluorescent sensors, CellROX Green and HyPerRed. These findings establish an important aspect of opioid receptor signaling and suggest that ROS induction may be part of the physiological response to KOR activation.

Many of these signaling mechanisms at the receptor have been partially characterized (Fig. 1A): conventional KOR agonist binding results in G␣ i/o ⅐G␤␥ activation by stimulating guanine nucleotide (GDP-GTP) exchange typical of G protein-coupled receptors (GPCRs). Sustained KOR activation by high-efficacy agonists induces G␤␥-dependent G protein receptor kinase 3 (GRK3) phosphorylation of KOR and subsequent arrestin association (19,20). Arrestin binding to KOR sterically blocks G protein association and thereby causes KOR desensitization of that signaling pathway. KOR activation also stimulates other protein kinase cascades, including the mitogen-activated protein kinases (MAPKs). KOR activation of the MAPK ERK1/2 occurs in two phases: an arrestinindependent early phase and an arrestin-dependent late phase (21,22). In contrast, KOR activation of p38 MAPK occurs solely by the late-phase, arrestin-dependent mechanism (22,23). Arrestin associated with KOR acts as a kinase scaffold that increases both phospho-ERK (pERK) and phospho-p38 MAPK (pp38). KOR-dependent aversion results from pp38-induced activation of serotonin transport and direct effects of pp38 on ventral tegmental area dop-amine neuron physiology (8,24,25). Additionally, analgesia induced by KOR agonists in females is inhibited by estradiol, which stimulates GRK2 to sequester G␤␥ and block arrestin-independent signaling while leaving arrestin-dependent signaling and aversive behaviors largely intact (26).
Recent studies have also demonstrated that receptor binding of either KOR agonists or antagonists activates a third MAPK, c-Jun N-terminal kinase (JNK) by stimulating phosphorylation at Thr-183 and Tyr-185 of JNK-1 (27)(28)(29). The selective KOR antagonists norbinaltorphimine (norBNI) and JDTic produce their long-lasting inhibition of KOR signaling following JNK-1 isozyme activation (30), and the molecular mechanisms involved have recently been identified (29). Activation of phospho-JNK by norBNI stimulates peroxiredoxin 6 (PRDX6), which in turn stimulates NADPH oxidase to generate reactive oxygen species (ROS). Local production of ROS oxidizes the cysteine sulfhydryl of G␣ i required for reversible palmitoylation. The depalmitoylated G␣ i binds tightly to KOR but in a conformation that prevents nucleotide exchange by the KOR signaling complex (29). The tightly bound, depalmitoylated KOR-G␣ i complex is permanently inactivated, and recovery presumably requires new receptor synthesis to restore KOR function. Thus, these receptor-inactivating KOR antagonists are actually functional (JNK-biased) agonists that have long durations of effect through a JNK-dependent mechanism rather than through slow drug dissociation (27)(28)(29)(30)(31). This form of receptor regulation is likely to be a general property of G␣ i/o signaling as a similar mechanism was found to mediate morphine tolerance at the opioid receptor and quinpirole tolerance at the D2 dopamine receptor (29). KOR agonists have also been shown to increase phospho-JNK (27,32,33), but how this process differs from KOR antagonist mechanisms is unclear, and whether KOR agonist activation of pJNK signaling has other physiological consequences is not known. To address these questions, we have focused on how KOR activation by conventional agonists stimulates JNK signaling. KOR agonists were found to increase phospho-JNK by both arrestin-independent and arrestin-dependent mechanisms. JNK activation by KOR agonists stimulated the production of ROS that could be detected by the fluorescent sensors CellROX Green and HyPerRed in transfected HEK293 cells.

Activation of JNK by the KOR agonist U50,488
HEK293 cells stably transfected with KOR respond to the selective agonist U50,488 (10 M) with a biphasic increase in phospho-JNK (Fig. 1, B and C). Phospho-JNK immunoreactivity (IR) was significantly increased in an early phase at 15 min (170 Ϯ 21% of basal, p Ͻ 0.05) and significantly increased in a late phase with a peak at 60 min (160 Ϯ 21% of basal, p Ͻ 0.05). To confirm that the increase in JNK phosphorylation was mediated through KOR activation of G␣ i , cells were pretreated with pertussis toxin (PTX; 10 ng/ml) 3 h prior to treatment with 10 M U50,488 for 15 or 60 min, and cell lysates were immunoblotted for phospho-JNK IR (Fig. 1, D and E). PTX significantly inhibited U50,488-induced phospho-JNK IR compared with vehicle pretreatment at both 15 min (104 Ϯ 4 and 138 Ϯ 12%, respectively; p Ͻ 0.05) and 60 min (81 Ϯ 12 and 126 Ϯ 7%, respectively; p Ͻ 0.01). To further confirm agonist-induced JNK activity, we looked at U50,488-mediated phosphorylation of the JNK substrate c-Jun (Fig. 1, F and G). U50,488 treatment resulted in a monophasic increase in phospho-c-Jun IR with a significant increase at 60 min (172 Ϯ 18% of basal, p Ͻ 0.01).
We next sought to replicate these findings with dynorphin B, an endogenous peptide ligand for KOR (34,35). KOR-expressing HEK293 cells were treated for 5-60 min with 1 M dynorphin B, and cell lysates were then immunoblotted for phospho-JNK (Fig. 1, H and I). Similar to U50,488, dynorphin B stimulated a biphasic increase in phospho-JNK IR with significantly increased phospho-JNK IR at 30 min (136 Ϯ 6% of basal, p Ͻ 0.01) and 120 min (133 Ϯ 12% of basal, p Ͻ 0.05), indicating that JNK activation by dynorphin is qualitatively similar to U50,488 but with slightly delayed kinetics. The reason for the slower kinetics of JNK phosphorylation following dynorphin is unclear as no difference in kinetics is observed in dynorphininduced ERK1/2 activation (Fig. 1, J and K).
We next sought to characterize the mechanisms underlying KOR-mediated JNK activation. A role for the small G protein RAC, but not RHO, in KOR-stimulated JNK phosphorylation has been described previously (32). Several studies have also indicated that KOR is capable of stimulating PKC activity (36 -40), and PKC activity has been implicated in morphine-stimulated JNK activation (41). Based on these studies, we pretreated KOR-expressing HEK293 cells with 10 M Y27632 (a selective inhibitor of RHO kinase), 100 M NSC23766 (a selective inhibitor of RAC1), 5 M Gö6976 (a PKC inhibitor), or vehicle prior to treatment with 10 M U50,488 for 15 or 60 min and immunoblotting cell lysates for phospho-JNK IR. NSC23766 and Gö6976 significantly inhibited U50,488-stimulated phospho-JNK IR at 15 min (95 Ϯ 8% of basal, p Ͻ 0.01, and 90 Ϯ 12% of basal, p Ͻ 0.01, respec-

ROS activated by opioid receptors through c-Jun kinase
arrestin-dependent JNK activation, in agreement with our similar studies on the opioid receptor (41).

Generation of reactive oxygen species is induced by KOR agonists
JNK activation by long-acting KOR antagonists results in receptor inactivation through a PRDX6 -ROS-mediated depalmitoylation mechanism (29), but it is unclear whether JNK activation by conventional KOR agonists also stimulates ROS production. Using CellROX Green (a cell-permeable dye that fluoresces upon oxidation-induced nucleic acid binding, which is not readily reversible) to detect ROS production, KOR-expressing HEK293 cells or untransfected control HEK293 cells were incubated with CellROX Green and U50,488 (concentration range, 1 nM to 10 M), fixed, and imaged for CellROX Green fluorescence (Fig. 3, A and B). U50,488 treatment resulted in a significant increase of CellROX Green fluorescence in the first 30 min in KOR-expressing cells but not in untransfected HEK293 cells (p Ͻ 0.01). In contrast, no increase in fluorescence was observed in KOR-expressing cells that were incubated with CellROX Green from 30 to 60 min after U50,488 (Fig. 3B). Interestingly, U50,488 stimulated ROS generation with an inverted U-shaped dose response with peak fluorescence at 100 nM (3.5 Ϯ 1.6-fold over vehicle, p Ͻ 0.05 versus untransfected) and no change in fluorescence intensity after treatment with 10 M U50,488 (1.0 Ϯ 0.6-fold over vehicle). No changes in CellROX Green fluorescence were observed at any concentration of U50,488 in untransfected HEK293 cells, demonstrating that this inverted U-shaped dose-response relationship was not a result of off-target effects. To confirm that the changes in CellROX Green fluorescence were due to increases in ROS, KOR-expressing HEK293 cells were pretreated with 10 M N-acetylcysteine (NAC; a nonselective antioxidant) or vehicle prior to treating cells for 30 min with 100 nM U50,488 in the presence of CellROX Green (Fig. 3, C and D). Based on these results, we predicted that treatment with the endogenous KOR ligand dynorphin B would also induce ROS generation. KOR-expressing HEK293 cells were treated with dynorphin B (concentration range, 100 pM to 10 M) for 30 min in the presence of CellROX Green; cells were then fixed and imaged for CellROX Green fluorescence (Fig. 3, E and F). As

ROS activated by opioid receptors through c-Jun kinase
observed with U50,488, dynorphin B stimulated ROS generation with an inverted U-shaped dose response. Peak fluorescence was observed at 100 nM (2.9 Ϯ 0.6-fold over vehicle, p Ͻ 0.05 versus 0.1 nM), and there was no change in fluorescence after treatment with 10 M dynorphin B (1.3 Ϯ 0.2-fold over vehicle).

Mechanism of reactive oxygen species stimulation by KOR
To determine whether JNK and the phospholipase A 2 (PLA 2 ) activity of PRDX6 are also required for U50,488stimulated ROS production, we pretreated KOR-expressing HEK293 cells with 10 M MJ33 (an inhibitor of PRDX6 PLA 2 activity) (42-44), 1 M JNK-IN-8 (a JNK inhibitor) (45), or vehicle prior to treating cells for 30 min with 100 nM U50,488 in the presence of CellROX Green. Cells were then fixed and imaged for CellROX Green fluorescence (Fig. 4, A and B).  . These results demonstrate that U50,488 stimulated generation of ROS via JNK and PRDX6, similar to nor-BNI and morphine (29). Naloxone also blocked HyPerRed fluorescence induced by U50,488 (1.0 Ϯ 0.1-fold of vehicletreated, p Ͻ 0.01), further confirming that ROS generation was KOR-mediated.
Because U50,488-stimulated ROS was JNK-dependent, we tested whether JNK phosphorylation induced by U50,488 also followed an inverted U-shaped response by treating KOR-expressing HEK293 cells with two different concentrations of U50,488 for 15 min (Fig. 4, E and G). Both 100 nM and 10 M U50,488 resulted in significant increases in phospho-JNK IR at 15 min (125 Ϯ 5% of vehicle, p Ͻ 0.05, and 145 Ϯ 8% of vehicle, p Ͻ 0.001, respectively) with significantly greater IR at 10 M as compared with 100 nM U50,488 treatment (p Ͻ 0.05). Furthermore, a similar concentration-dependent increase in phospho-JNK IR following 60-min treatment with 100 nM and 10 M U50,488 (123 Ϯ 4% of vehicle, p Ͻ 0.05, and 158 Ϯ 14% of vehicle, p Ͻ 0.001, respectively) with significantly greater IR at 10 M as compared with 100 nM U50,488 treatment (p Ͻ 0.05) (Fig. 4, F and H) was also observed. These results suggest that the U-shaped dose response of ROS activation by KOR was not due to decreased JNK activation at higher agonist concentrations.

Reactive oxygen species stimulation by KOR is antagonized by the GRK pathway
Based on the time course of ROS generation, we hypothesized that ROS generation was mediated through the earlier, arrestin-independent phase of JNK activation. To test this, HEK293 cells were transiently transfected with KOR or KOR(S369A), a receptor mutant form that cannot be phosphorylated by GRK3 or recruit arrestin (47,48), and then treated with U50,488 (1 nM-10 M) for 30 min in the presence of Cell-ROX Green (Fig. 5, A and B). As observed in stable KOR-expressing cells, transiently transfected KOR HEK293 cells responded to U50,488 with an increase in CellROX Green fluorescence. The U50,488 dose response also had an inverted U-shaped dose response peaking at 100 nM (1.9 Ϯ 0.4-fold of vehicle-treated). In contrast, U50,488 did not produce a U-shaped dose response in HEK293 cells transiently transfected with KOR(S369A) cells but instead showed a linear increase in response with increasing dose (Fig. 5B). Higher concentrations (10 M) of U50,488 produced greater CellROX Green fluorescence (2.8 Ϯ 0.8-fold over vehicle, p Ͻ 0.01) compared with KOR-expressing cells. These results support the hypothesis that ROS generation was mediated by the arrestinindependent phase of JNK activation as KOR(S369A) solely activates arrestin-independent signaling.
It was not clear why arrestin-dependent JNK activation was unable to activate ROS as the late phase-generated phospho-JNK should be able to readily diffuse away from the receptor complex and interact with PRDX6 or why high concentrations of drug resulted in a complete loss of ROS activation, which we have not observed for other KOR signaling pathways (10,22,23,27). Furthermore, classic arrestin-mediated GPCR desensitization mechanisms did not clearly explain the loss of ROS stimulation as upstream JNK activation was maintained. An alternative hypothesis is that, following recruitment to KOR, arrestin also acts as a scaffold for activation of p38␣ MAPK (23,49), and phospho-p38 has been shown to inhibit JNK in other systems (50). To test the role of phospho-p38 in the ROS response to U50,488, KOR-expressing cells were pretreated with the p38 inhibitor SB203580 (10 M) or vehicle prior to treating cells for 30 min with 100 nM or 10 M U50,488 (Fig. 5, C and D). SB203580 had no effect on the CellROX Green response to 100 nM U50,488 (2.1 Ϯ 0.3-and 2.5 Ϯ 0.4-fold over vehicle, respectively). However, a significant increase in 10 M U50,488stimulated CellROX Green fluorescence was observed in cells pretreated with SB203580 instead of vehicle (2.1 Ϯ 0.3-and 0.75 Ϯ 0.1-fold of vehicle, respectively; p Ͻ 0.05). Consistent with this explanation, KOR(S369A) does not activate p38 MAPK and hence does not produce an inverted U-shaped dose response. These results suggest that GRK-dependent activation of p38, rather than receptor phosphorylation or arrestin recruitment per se, inhibits ROS generation at higher concentrations of agonist. In contrast, no effect of SB203580 was observed on CellROX Green fluorescence stimulated by the KOR ligand norBNI, which has functional selectivity for the JNK-ROS pathway and does not activate arrestin-dependent p38 signaling (Fig. 5, E and F) (27)(28)(29)(30).

Real-time measurement of reactive oxygen species following KOR agonist
The results described above suggested that ROS generation by KOR agonists may have physiological effects and that livecell imaging could reveal the kinetics of KOR signaling. To assess this possibility, we next used live-cell imaging of HyPer-Red in KOR-expressing HEK293 cells. Live-cell imaging showed that 100 nM U50,488 elevated ROS at 30 min, and ROS remained elevated at 45 min post-drug application (Fig. 6, A  and C). Live imaging also revealed that U50,488 induced a transient decrease in HyPerRed fluorescence peaking at ϳ2-3 min, an effect that was not resolved in the initial studies performed in fixed cells. Both the transient decrease and longer-lasting elevation in ROS, as measured by HyPerRed fluorescence, were blocked by pretreatment with 10 M naloxone (Fig. 6, B and D). In addition, treatment with 10 M naloxone 30 min after U50,488 activation of HyPerRed fluorescence completely reversed the increase within 15 min after the addition of naloxone (Fig. 6D). This result indicates that the increase in HyPerRed fluorescence required continuing KOR-stimulated ROS production.

Discussion
The key finding of this study is that conventional KOR agonists stimulate ROS generation through a JNK-PRDX6 pathway. This study identified biphasic activation of JNK by KOR agonists through two distinct G protein-dependent mechanisms: an early arrestin-independent pathway and a slower

ROS activated by opioid receptors through c-Jun kinase
arrestin-scaffolded pathway. Induction of ROS generation was mediated by arrestin-independent JNK activation and inhibited by GRK/arrestin-dependent stimulation of p38 activity. JNKdependent ROS activation is likely to be a general mechanism of G␣ i/o receptor signaling, representing another aspect of GPCR signaling through G␣ i , as other members of this receptor family (e.g. opioid, D2 dopamine, and CB1 cannabinoid receptors) show similar activation of ROS (29). 3 The initial arrestin-independent JNK phosphorylation required PKC and the small G protein RAC but not RHO. This pathway likely corresponds to the Src and RAC/CDC42-dependent mechanism reported by Kam et al. (32). Arrestin-inde-pendent, PKC-dependent JNK phosphorylation is also stimulated by morphine-like agonists of MOR (41). Previous investigators reported PKC-mediated KOR signaling and desensitization (36 -40), and our data link these two pathways. In contrast, arrestin-dependent activation of JNK by U50,488 required both RAC and the RHO effector ROCK1. Unlike the early-phase JNK phosphorylation but similar to arrestin-dependent JNK phosphorylation by the MOR agonist fentanyl (41), PKC was not required. Rather, inhibition of PKC enhanced arrestin-dependent JNK phosphorylation, suggesting that PKC inhibits arrestin-mediated JNK phosphorylation by receptor desensitization or another mechanism.
U50,488 was also found to stimulate increased ROS in a PRDX6 PLA 2 activity-dependent manner. These results are 3 S. S. Schattauer, and C. Chavkin, unpublished observations.

ROS activated by opioid receptors through c-Jun kinase
analogous to the previously published findings of JNK/PRDX6induced ROS by the MOR agonist morphine, which stimulates JNK phosphorylation by an arrestin-independent mechanism (29,41). Unlike morphine, U50,488 and dynorphin B both stimulated ROS generation with a complex concentration response. Maximal ROS generation was observed at moderate concentrations with a loss of ROS stimulation at receptor-saturating concentrations as a result of receptor phosphorylation by GRK and downstream p38 activity. Inhibition of JNK by p38 explains the lack of ROS generation by the arrestin-dependent JNK pathway observed in this study and previously (29). Based on this, we would predict that partial or G protein-biased agonists might be more efficacious at promoting ROS generation. We would further predict that system biases toward or against GRK-arrestin pathways would change the extent to which JNK-ROS signaling is activated. GPCR signaling has been broadly divided into two branches: arrestin-independent signaling, historically referred to as G protein-dependent signaling (51) (including KOR-induced ERK1/2 MAPK phosphorylation and ion-channel modulation), and arrestin-dependent signaling, which requires G proteins as well but is also dependent on arrestin scaffolding of signaling complexes to facilitate signal transduction (52) (including KOR-induced p38 MAPK phosphorylation). Similar to other arrestin-independent G protein-initiated signaling cascades that are desensitized by arrestin recruitment, this JNK-ROS cascade is also inhibited by arrestin-mediated signaling but by a different arrestin-dependent mechanism. The opposing bal- Fluorescence was significantly increased after U50,488 treatment in mycKORexpressing HEK293 cells that were not pretreated with naloxone (repeated-measures two-way ANOVA (significant interaction between time and drug, F 51,306 ϭ 3.475, p Ͻ 0.0001; effect of time, F 51,306 ϭ 0.9146, p ϭ 0.6409; effect of drug, F 1,6 ϭ 0.8464, p ϭ 0.3931; n ϭ 7) with Holm-Šidák post hoc comparison against naloxone pretreatment (*, p Ͻ 0.05; ns, not significant)). E, schematic of JNK signaling by KOR agonists. KOR activation promotes analgesia and cognitive disruptions through G protein-mediated mechanisms, including regulation of various ion channels, adenylyl cyclase, and ERK1/2 MAPK. G proteindependent GRK3 phosphorylation of KOR at Ser-369 promotes arrestin (arr) recruitment, resulting in arrestin-scaffolded p38 MAPK activation and aversion in addition to receptor desensitization. This study describes another aspect of signaling in which ␤␥ subunits also promote JNK activation through G proteinmediated activation of RAC1 and PKC, resulting in increased oxidative signaling via PRDX6 PLA 2 activity. This pathway is directly inhibited by G proteinmediated, arrestin-scaffolded activation of p38 MAPK.

ROS activated by opioid receptors through c-Jun kinase
ance between JNK-ROS signaling and GRK-arrestin-p38 described here may contribute to differential effects of GPCR activation in different cell types and cellular compartments (53)(54)(55)(56), dependent on signalosome composition and local cellular environment.
JNK and oxidative pathways both have established roles in cellular stress pathways (57)(58)(59). In addition to their role in oxidative stress, ROS can act as signaling molecules, regulating tyrosine phosphorylation, small G protein activity, and transcription factor complexes (60,61). ROS have also recently been shown to promote desensitization and tolerance for GPCRs, including MOR and KOR (29,(62)(63)(64), and adversely affect pain management by opioids. The insights from this study add further clarity to our understanding of functional selectivity at opioid receptors and how the dynorphin-KOR system may contribute to the stress-induced vulnerability to neurodegeneration, mood disorders, and cardiovascular disease (65)(66)(67)(68). Markers of oxidative stress are altered in clinical studies and rodent models of psychiatric stress, and ROS have been linked to neurodegenerative disease, pathological pain, and psychiatric disease (67)(68)(69); this study suggests that the dynorphin-KOR axis may play an important role in these effects. Further studies will be needed to understand under what conditions ROS is generated by KOR and what roles KORinduced ROS play in the stress response, aided by the signaling insights and tools validated in this study.
We confirmed that the two different methods of receptor tagging used in this study (N-terminal myc and C-terminal GFP) did not impact JNK activation by treating mycKOR-and KORGFP-expressing HEK293 with 100 nM and 10 M U50,488 for 15 or 60 min and immunoblotting for phospho-JNK. Significantly more phospho-JNK IR was observed following 15-min 10 M U50,488 than 100 nM U50,488 (145 Ϯ 8 and 125 Ϯ 5%, respectively) (Fig. 4, E and G). Similarly, more phospho-JNK IR was observed following 60-min 10 M U50,488 than 100 nM U50,488 (152 Ϯ 7 and 122 Ϯ 7%, respectively) (Fig. 4, F and H). At neither time point was any effect of receptor tagging observed. Based on this, we used mycKOR for all experiments after studies using CellROX Green were initiated. 2) overnight at 4°C. Relative fluorescent band intensity was measured using Odyssey software and expressed as arrestin 2 IR or arrestin 3 IR intensity over actin IR intensity and then expressed as normalized to scrambled siRNA treatment.

Western blot analysis
HEK293 cells stably expressing the indicated constructs were serum-starved for 6 h prior to drug treatment. Cells were treated as described in figure legends and then lysed in lysis buffer (50 mM Tris-HCl, pH 7.5, 300 mM NaCl, 1 mM EDTA, 1 mM Na 3 VO 4 , 1 mM NaF, 10% glycerol, phosphatase and protease inhibitors). For JNK and c-Jun experiments, buffer was supplemented with 1% Triton X-100. Lysates were sonicated and centrifuged (15,000 ϫ g, 25 min, 4°C), and the supernatant was stored at Ϫ20°C. Total protein concentration was determined by BCA assay (Pierce) with BSA standards before loading 40 g (JNK), 30 g (c-Jun), 20 g (ERK1/2), or 10 g (arrestin) onto 10% Bis-Tris precast gels (Life Technologies) and running at 120 V for 1.5-2 h. Blots were transferred to nitrocellulose (Whatman) for 1.5 h at 30 V. Nitrocellulose was blocked with 5% BSA-TBST for 1 h at room temperature and stained overnight for c-Jun and phospho-c-Jun, JNK and phospho-JNK, ERK1/2 and phospho-ERK1/2, or arrestin 2 or 3 and actin. Blots were incubated in IRDye secondary antibody (LI-COR Biosciences, catalog number 926-68070, lot C50721-05 and catalog number 926-32211, lot C506602-05; 1:10,000) in 1:1 Odyssey buffer (LI-COR Biosciences) and 5% milk-TBST for 1 h at room temperature and then scanned on the Odyssey IR Imaging System (LI-COR Biosciences). Band intensity with background subtraction was measured using Odyssey software (Image Studio 3.1 and Image Studio Lite 4.0, LI-COR Biosciences). Phosphoprotein band intensity was normalized to total (ERK1/2 and c-Jun) or actin (JNK) band intensity. Data were normalized to

ROS activated by opioid receptors through c-Jun kinase
percentage of control sample (percent vehicle or percent basal) and plotted using GraphPad Prism 6.07 (GraphPad Software, Inc.). Statistical significance (p Ͻ 0.05) was determined by analysis of variance (ANOVA) followed by Holm-Šidák post hoc test.

Reactive oxygen species measured by CellROX Green fluorescence
Untransfected HEK293 cells or cells stably expressing mycKOR were grown on poly-D-lysine-treated coverslips the day prior to the experiment. For transient mycKOR and mycKOR(S369A) experiments, HEK293 cells were plated on coverslips 48 h prior to the experiment and transiently transfected with mycKOR or mycKOR(S369A) 24 h prior. Cells were serum-starved for 5 h and then treated as described. CellROX Green (10 M; Molecular Probes) was added during the last 30 min of treatment. Cells were rinsed in PBS and fixed for 15 min with 4% paraformaldehyde. Cells were mounted on glass slides with VectaShield HardSet with DAPI (Vector Laboratories) and imaged within 12 h. Coverslips were imaged on a Nikon upright fluorescence microscope with Nikon Elements AR v3.1 software (Nikon Instruments). To prevent photoactivation, exposure to light during sample generation and imaging was minimized. Two representative fields from each coverslip were imaged for CellROX (488 nm) and DAPI. Exposure times were held constant for every fluorophore throughout the entire experiment. Image intensities for between 7 and 20 cells per image were quantified using ImageJ v 1.42q (National Institutes of Health), and these were averaged to give an average field intensity value. This was done for the two coverslip images and averaged to make one sample (n).

Reactive oxygen species measured by HyPerRed fluorescence
mycKOR-expressing or untransfected HEK293 cells grown on poly-D-lysinecoated coverslips were transiently transfected with HyPerRed using FuGENE HD. The following day, cells were serum-starved for 6 h and treated as described. Cells were blocked for 30 min with 1% BSA in PBS, 0.025% Triton X-100; incubated in rabbit anti-KOR antibody (70) at 1 g/ml overnight at 4°C; washed; and incubated with Alexa Fluor 488 anti-rabbit secondary (Life Technologies) for 1 h. Cells were washed, mounted on glass slides with VectaShield HardSet with DAPI, and stored at 4°C until imaging. Coverslips were imaged on a Nikon upright fluorescence microscope at 10ϫ. Repre-sentative fields from each coverslip were imaged for KOR, HyPerRed (555 nm), and DAPI. Exposure times were held constant for every fluorophore throughout the experiment. Images were adjusted to a standard threshold value. Image intensities for between 20 and 40 cells per image were quantified using ImageJ v 1.42q (National Institutes of Health), and these were averaged to give an average field intensity value.

Live-cell imaging of HyPerRed fluorescence
HEK293 cells or cells stably expressing mycKOR were grown on chambered coverslips (Ibidi) 18 h prior to transient transfection with HyPerRed using FuGENE HD, and media were changed after 24 h. Cells were serum-starved for 6 h prior to the experiment, which took place 36 -48 h after transfection. Cells were imaged on a Leica SP8X confocal microscope using LASX software at 20ϫ with a 2ϫ digital zoom, maintaining temperature in the imaged chambers at 34.9°C (S.D., Ϯ0.2°C). Two treatment conditions were imaged in parallel for each experiment, tracking two fields per condition using the "hold focus" and "mark and track" functions, with scanning intervals of 60 s. Each replicate represents the average of the two positions in the well from each experiment. HyPerRed (excitation with WLL 577 laser at 10%, emission with HyD sensor at 584 -692 with 0.3% gating and smart gain of 33.1%) and brightfield were imaged with 512 ϫ 512 resolution, 600 speed with bidirectional scanning, and live averaging of 2. Ten frames were measured before the first drug treatment; cells were treated as described and treated with 200 M H 2 O 2 at the end of all experiments to confirm response with a positive control. MATLAB R2017a software was used for live-cell image processing. Fluorescence was quantified by averaging the pixel intensity across the whole frame. Background intensity was consistent with imaging noise and thus negligibly affected the average whole-frame intensity. A two-frame-window moving average was used to adjust for frame-to-frame fluorescence noise fluctuations in the timecourse data.

Quantification and statistical analysis
GraphPad Prism 7.03 software was used for data analysis. Data are represented as mean Ϯ S.E. When data are presented as percent baseline, percent vehicle, or -fold change, data were normalized to a within-replicate control set at 100%, which is represented by a dashed line. Statistical analyses are described in the corresponding figure legends.