Research ArticlePain

Anti–PD-1 treatment impairs opioid antinociception in rodents and nonhuman primates

See allHide authors and affiliations

Science Translational Medicine  19 Feb 2020:
Vol. 12, Issue 531, eaaw6471
DOI: 10.1126/scitranslmed.aaw6471

A checkpoint for pain

Immune checkpoint blockade therapies using antibodies against programmed cell death protein–1 (PD-1) have been shown encouraging results for treating several tumors. Opioid treatment is frequently administered to patients with cancer for controlling disease-associated pain. PD-1 is expressed in sensory neurons and mediates pain sensitivity in mice. Now, Wang et al. studied the relationship between PD-1 expression and opioid effects in mice and nonhuman primates (NHPs). Morphine analgesic effect was diminished after PD-1 inhibition, and PD-1 blockade increased opioid-induced hyperalgesia in mice. In NHPs, immunotherapy targeting PD-1 inhibited morphine-induced antinociception. The results suggest that immunotherapies targeting PD-1 might reduce the sensitivity to opioids.


Emerging immunotherapies with monoclonal antibodies against programmed cell death protein–1 (PD-1) have shown success in treating cancers. However, PD-1 signaling in neurons is largely unknown. We recently reported that dorsal root ganglion (DRG) primary sensory neurons express PD-1 and activation of PD-1 inhibits neuronal excitability and pain. Opioids are mainstay treatments for cancer pain, and morphine produces antinociception via mu opioid receptor (MOR). Here, we report that morphine antinociception and MOR signaling require neuronal PD-1. Morphine-induced antinociception after systemic or intrathecal injection was compromised in Pd1−/− mice. Morphine antinociception was also diminished in wild-type mice after intravenous or intrathecal administration of nivolumab, a clinically used anti–PD-1 monoclonal antibody. In mouse models of inflammatory, neuropathic, and cancer pain, spinal morphine antinociception was compromised in Pd1−/− mice. MOR and PD-1 are coexpressed in sensory neurons and their axons in mouse and human DRG tissues. Morphine produced antinociception by (i) suppressing calcium currents in DRG neurons, (ii) suppressing excitatory synaptic transmission, and (iii) inducing outward currents in spinal cord neurons; all of these actions were impaired by PD-1 blockade in mice. Loss of PD-1 also enhanced opioid-induced hyperalgesia and tolerance and potentiates opioid-induced microgliosis and long-term potentiation in the spinal cord in mice. Last, intrathecal infusion of nivolumab inhibited intrathecal morphine-induced antinociception in nonhuman primates. Our findings demonstrate that PD-1 regulates opioid receptor signaling in nociceptive neurons, leading to altered opioid-induced antinociception in rodents and nonhuman primates.


Programmed cell death protein–1 (PD-1) is a member of the immunoglobulin gene superfamily, initially identified in T cells upon programmed cell death, with Pd1 mRNA expression restricted to the thymus (1). Mice lacking Pd1 develop autoimmune disease, suggesting an immunosuppressive role of PD-1 (2). PD-L1, the ligand for PD-1, is highly expressed in many cancers and associated with mortality in patients with cancer (3, 4), establishing a role of PD-1 in cancer-induced immune suppression. Inhibition of the interaction between PD-1 and PD-L1, known as an immune checkpoint blockade, enhances T cell responses to produce antitumor activity (5). Emerging immune therapies using anti–PD-1 monoclonal antibodies have shown success in treating various cancers such as melanoma (4, 5). PD-1 is also expressed by melanoma cells, promoting tumor growth (6).

Despite extensive studies of PD-1 in non-neuronal cells, the nature of PD-1 signaling in neurons is largely unknown. Is PD-1 a neuromodulator or a neuro-checkpoint inhibitor? We recently showed that primary sensory neurons of dorsal root ganglion (DRG) also express functional PD-1 receptor and that activation of PD-1 by PD-L1 inhibits neuronal excitability and pain in mice (7). PD-L1 is produced by nonmalignant tissues including DRG and spinal cord (7), implicating a physiological role of PD-L1. Furthermore, Pd1-deficient mice exhibit increased excitability in DRG neurons and increased pain sensitivity. In a mouse model of melanoma, cancer pain is masked by PD-L1, produced from melanoma cells and acting on skin nerve terminals (7).

Opioids are mainstay treatments for cancer pain (810). Morphine produces antinociception via mu opioid receptor (MOR) (11, 12), which is expressed in the central nervous system (CNS; brain and spinal cord) and peripheral nervous system (PNS; DRG and nerve) (11, 13). MOR mediates both the beneficial effects (such as antinociception) and the unwanted side effects of morphine (such as hyperalgesia, tolerance, and withdrawal responses) (11, 14). This study was undertaken to investigate the interactions between PD-1 and MOR in the PNS and CNS. Our findings demonstrated that PD-1 is required for morphine antinociception and MOR signaling in DRG and spinal cord neurons in mice. Nivolumab, a U.S. Food and Drug Administration–approved anti–PD-1 monoclonal antibody, also inhibited morphine-induced antinociception in nonhuman primates (NHPs) after spinal administration.


Loss of PD-1 results in a reduction in morphine antinociception in mice

We first examined whether morphine antinociception would be altered in Pd1 knockout (Pd1−/−) mice after subcutaneous injection [1, 3, and 10 mg/kg, subcutaneously (sc)]. Tail-flick testing in wild-type (WT) mice revealed a rapid (<0.5 hours) and dose-dependent increase in tail-flick latency that lasted more than 3 hours after morphine treatment (P < 0.0001; Fig. 1A and fig. S1, A to C). The duration and potency of the antinociceptive effect was reduced in Pd1−/− animals (P < 0.0001; Fig. 1A and fig. S1, A to C). Hot-plate testing also showed an impairment of morphine-induced antinociception in Pd1−/− mice (P < 0.0001; Fig. 1A and fig. S1, D to F). The dose-response curve showed a right shift in morphine antinociception in Pd1−/− mice compared to WT animals (P < 0.05, tail-flick and hot-plate tests; Fig. 1B). Intrathecal morphine administration in the spinal cord [2 nmol, intrathecally (it)] also elicited marked antinociception in tail-flick and hot-plate tests in WT mice; but this action was compromised in Pd1−/− mice (P < 0.0001, tail-flick and hot-plate tests; Fig. 1C). Collectively, these results suggest that morphine antinociception requires PD-1 via both peripheral and central actions.

Fig. 1 Morphine antinociception is diminished in mice lacking PD-1.

(A) Subcutaneous morphine antinociception (10 mg/kg, sc) in WT mice and Pd1−/− mice in tail-flick (left) and hot-plate (right) tests. Saline injection in WT mice and Pd1−/− mice was included as vehicle. (B) Dose responses and ED50 values of subcutaneous morphine antinociception in tail-flick (left) and hot-plate (right) tests in WT and Pd1−/− mice. *P < 0.05 and ****P < 0.0001, WT versus Pd1−/−, unpaired two-tailed t test. (C) Intrathecal morphine antinociception (2 nmol, it) in WT and Pd1−/− mice in tail-flick (left) and hot-plate (right) tests. Saline injection in WT and Pd1−/− mice was included as vehicle. (D) Intrathecal antinociception of DAMGO (MOR agonist), DPDPE (DOR agonist), or U69593 (KOR agonist) in tail-flick test. (A, C, and D) *P < 0.05, ***P < 0.001, and ****P < 0.0001, WT versus Pd1−/−, two-way ANOVA, followed by Bonferroni’s post hoc test. (E to G) Intrathecal morphine antinociception (2 nmol) in LLC-induced bone cancer pain (E), CFA-induced inflammatory pain (F), and SNL-induced neuropathic pain (G) in WT and Pd1−/− mice. n.s., no significance. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, versus preinjection baselines. ##P < 0.01, ###P < 0.001, and ####P < 0.0001, WT versus Pd1−/−, two-way ANOVA, followed by Bonferroni’s post hoc test. Data are expressed as means ± SEM. Sample sizes are indicated in brackets.

Opioids induce antinociception through the μ, δ, and κ opioid receptors (MOR, DOR, and KOR, respectively) (11). Intrathecal injection of specific agonists DAMGO (MOR), DEDPE (MOR), and U69593 (KOR) each evoked antinociception in WT mice (Fig. 1D and fig. S2, A to C). Pd1−/− mice only exhibited reduction in DAMGO but not DPDPE- or U69593-induced antinociception (P = 0.0030 and 0.0487 for DAMGO in tail-flick and hot-plate tests, respectively; Fig. 1D and fig. S2, A to C). Thus, PD-1 primarily affects the MOR-mediated antinociception.

Opioids are a mainstay treatment for cancer pain, which often manifests after bone metastasis (15). We assessed whether morphine would attenuate cancer pain in WT and Pd1−/− mice using a bone cancer model (16). Femoral inoculation of Lewis lung cancer cells resulted in severe cancer pain, as indicated by reductions in mechanical threshold (Fig. 1E). Spinal morphine administration completely reversed tumor-induced mechanical allodynia in WT mice (P < 0.0001; Fig. 1E and fig. S2D). However, the antiallodynic effect of morphine was largely compromised in Pd1−/− mice (P < 0.0001; Fig. 1E). The antihyperalgesic effect of morphine in Hargreaves test was also compromised in Pd1−/− mice (P < 0.0001; fig. S2, E and F). Furthermore, intrathecal morphine evoked antiallodynic effects in inflammatory pain and neuropathic pain in WT mice, induced by complete Freund’s adjuvant (CFA) and spinal nerve ligation (SNL), respectively; also in these cases, morphine’s antinociception was compromised in Pd1−/− mice (P < 0.0001 in both CFA and SNL models; Fig. 1, F and G).

Anti–PD-1 treatment with nivolumab diminishes morphine antinociception in WT mice

Next, we tested whether impaired morphine antinociception in Pd1−/− mice could be recapitulated by treatment of nivolumab, a human anti–PD-1 monoclonal antibody previously shown to evoke allodynia in WT naïve mice (7). Systemic pretreatment with nivolumab [10 mg/kg, intravenously (iv)], given 1 hour before the morphine injection, reduced morphine antinociception in the tail-flick and hot-plate tests (P < 0.0001, tail-flick and hot-plate tests; Fig. 2, A and B, and fig. S3, A and B). Furthermore, spinal pretreatment with nivolumab (1 μg, it; given 30 min before morphine) also decreased morphine antinociception (2 nmol, it; P < 0.0001 for both tail-flick and hot-plate tests; Fig. 2C). Dose-response analysis revealed increases in ED50 (median effective dose) values of morphine antinociception in Pd1−/− mice and nivolumab-treated mice, as compared to WT mice (fig. S3C). ED50 values were 2.8 mg/kg (tail-flick test) and 4.1 mg/kg (hot-plate test) in WT mice, 5.5 mg/kg (tail-flick test) and 9.3 mg/kg (hot-plate test) in Pd1−/− mice, and 8.2 mg/kg (tail-flick test) and 8.2 mg/kg (hot-plate test) in nivolumab-treated mice (fig. S3C). To assess for the possibility of immunomodulation in this process, we further tested the effects of spinal pretreatment with nivolumab on morphine antinociception in CB-17 mice with deficiency in T cells and B cells. Nivolumab also decreased morphine antinociception (2 nmol, it) in adaptive CB-17 immune cell–deficient mice (P = 0.0117 and 0.0069 for tail-flick and hot-plate tests, respectively; fig. S3D).

Fig. 2 Morphine antinociception is decreased in WT mice after pretreatment with anti–PD-1 antibody nivolumab.

(A) Subcutaneous morphine antinociception (10 mg/kg, sc) after intravenous pretreatment with nivolumab (10 mg/kg, iv) in tail-flick (top) and hot-plate (bottom) tests. Nivolumab (Nivo) or control human IgG4 was intravenously injected 60 min before subcutaneous morphine injection. (B) Dose-response curve of morphine-induced antinociceptive effects (MPE % at the peak time point) in nivolumab (10 mg/kg, iv)–pretreated mice. (C) Spinal antinociception of morphine after intrathecal injection (2 nmol) in mice pretreated with IgG4 or nivolumab (1 μg, it) in tail-flick (top) and hot-plate (bottom) tests. Nivolumab or IgG4 was intrathecally injected 30 min before intrathecal morphine injection. (A and C) *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, morphine IgG4 versus morphine nivolumab, two-way ANOVA, followed by Bonferroni’s post hoc test. Data are expressed as means ± SEM. Sample sizes are shown in brackets.

In addition to male mice, morphine antinociception was also decreased in female Pd1−/− mice (P = 0.0155 for tail-flick test and P = 0.0070 for hot-plate test) and blunted by nivolumab in female WT mice (P = 0.0025 for tail-flick test and P = 0.0102 for hot-plate test; fig. S4, A to D). Together, our data demonstrate that Pd1 deletion or PD-1 blockade causes reduction in morphine antinociception in naïve animals and also in animals with inflammatory, neuropathic, and cancer pain. These results also suggest a functional interaction between PD-1 receptor and MOR.

PD-1 is coexpressed with MOR and regulates MOR function in DRG neurons

To determine the mechanisms by which PD-1 regulates opioid antinociception, we examined the interaction of PD-1 and MOR at the cellular level. In the DRG, MOR is mainly expressed by small-diameter nociceptive neurons (13). Using a sensitive RNAscope assay, we observed a high percentage of colocalization of Pd1 and MOR (Oprm1) mRNAs in DRG neurons (Fig. 3, A and B). To determine the mRNA expression of Pd1 and Oprm1 in individual DRG neurons, we quantified the number of fluorescence-labeled puncta in positive neurons, showing that about 45% neurons expressed Pd1 mRNA, 40% neurons expressed Oprm1 mRNA, and 30% DRG neurons expressed both mRNAs (Fig. 3C). Among Pd1+ cells, ~60% coexpressed Oprm1 (Fig. 3C).

Fig. 3 PD-1 and MOR are coexpressed in mouse DRG neurons and have functional interaction.

(A and B) In situ hybridization (RNAscope) images of Oprm1 mRNA (A) and Pd1 mRNA (B) in mouse DRG neurons of WT mice. Sixteen pairs of numbers show 16 neurons coexpressing both mRNAs. White stars (A) and yellow stars (B) show single-labeled neurons. Scale bar, 20 μm. (C) Percentage of positive neurons for Pd1 and Oprm mRNA expression and coexpression in DRG. Neurons (1244) from three mice were analyzed. (D) Double staining showing colocalization of PD-1 and MOR in DRG. Yellow arrows indicate double-labeled neurons. Bottom: Enlarged boxes show coexpression of both receptors on the cell surface indicated by yellow arrows. Scale bars, 10 μm. Right: PD-1 immunostaining with a PD-1 blocking peptide. Blue 4′,6-diamidino-2-phenylindole (DAPI) staining labels cell nuclei. Scale bar, 10 μm. (E) Proximity ligation assay (PLA) shows positive signals of PD-1/MOR interaction in cytoplasm (arrows) and axons (arrowheads) of cultured DRG neurons (nine images from two repeats). (F) Co-IP showing PD-1/MOR interaction in DRG. DRG lysate was immunoprecipitated (IP) with control IgG or MOR antibody and immunoblotted with PD-1 or MOR antibody. This experiment was repeated in triplicate. (G and H) Morphine (10 μM) effects on calcium currents in small-diameter DRG neurons (<25 μm) of WT and Pd1−/− mice. (G) Traces of calcium currents. (H) Time course of relative calcium currents. *P < 0.05, **P < 0.01, and ****P < 0.0001, Pd1−/− versus WT group, two-way ANOVA, followed by Bonferroni’s post hoc test. (I) Double staining for MOR and PD-1 in human dorsal root. PD-1 was labeled by nivolumab (1 mg/ml). Arrows indicate the double-labeled nerve fibers. Scale bars, 250 μm (left) and 50 μm (right). Data are expressed as means ± SEM. Sample sizes are indicated in brackets.

Immunohistochemistry also showed high degree of colocalization of PD-1 and MOR immunoreactivity (IR) in DRG neurons (Fig. 3D and fig. S5A). Colocalization was especially observed on the cell surface of small-diameter DRG neurons (Fig. 3D). Specificity of PD-1–IR was confirmed by a lack of staining after blocking peptide treatment (Fig. 3D) and was also tested previously in Pd1−/− mice (7). The MOR expression did not change in Pd1−/− mice (P = 0.3637 versus WT; fig. S5, B and C). Furthermore, we observed colocalization of PD-1 and MOR in axons of the mouse spinal nerve (fig. S5D), indicating axonal transport of PD-1 and MOR from cell bodies to peripheral nerve axons.

We further examined possible PD-1/MOR interaction using proximity ligation assay (PLA) in dissociated DRG neurons (17) and coimmunoprecipitation (co-IP) in DRG tissue. PLA analysis revealed positive fluorescence signals on cell bodies and axons in 44% of cultured mouse DRG neurons (Fig. 3E and fig. S5E), and staining was absent in negative control with omission of primary antibody or after treatment with blocking peptide (fig. S5E). Co-IP experiment demonstrated that anti-MOR antibody pulled down PD-1 in mouse DRG lysates [8.7-fold of control immunoglobulin G (IgG); Fig. 3F]. These results indicate close proximity of PD-1 and MOR and possible interaction between these two receptors in native DRG neurons.

Opioids inhibit pain by suppressing the function of calcium channels in primary afferent neurons (18). Accordingly, we recorded calcium currents in dissociated small-diameter DRG neurons (<25 μm) in WT and Pd1−/− mice using whole-cell patch clamp (18). Morphine (10 μM) produced a 35% reduction in calcium currents in WT DRG neurons (Fig. 3, G and H). However, this reduction was blunted in DRG neurons from Pd1−/− mice (P < 0.0001; Fig. 3, G and H). Thus, PD-1 is a required signaling element for morphine to suppress the calcium channel function in mouse DRGs.

We previously showed that PD-1–IR is expressed on the surface of human DRG neurons (7). We further examined PD-1 and MOR expression in human dorsal roots. Coexpression of PD-1 and MOR was observed on the nerve axons of the human dorsal roots (Fig. 3I). The co-IP assay also suggested a possible interaction of MOR and PD-1 in human dorsal root lysates (fig. S5F).

PD-1 disruption impairs suppression of nociceptive transmission by morphine in spinal cord neurons

Inhibition of calcium channels in the central terminals of DRG neurons by opioid treatment also inhibits neurotransmitter release in the spinal cord dorsal horn (SDH) via presynaptic regulation (11). Double staining in SDH revealed PD-1 colocalization with MOR in axons and axonal terminals (fig. S6A) and with calcitonin gene-related peptide (CGRP), a neuropeptide that is present in primary afferent central terminals (fig. S6B). The results suggest that spinal PD-1 may originate from DRG neurons.

To test the function of PD-1 in SDH synaptic transmission, we prepared spinal cord slices and recorded spontaneous excitatory postsynaptic currents (EPSCs) (sEPSCs) in outer lamina II neurons (Fig. 4, A to E). These interneurons are predominantly excitatory and form a nociceptive circuit with primary C-afferents and lamina I projection neurons (19). Bath perfusion of WT spinal cord slices with morphine (10 μM) led to a marked inhibition of sEPSC frequency (44.8% inhibition, P < 0.0001; Fig. 4, A and C). However, in Pd1−/− SDH neurons, the degree of morphine-induced inhibition was diminished (22.7%, P = 0.0068; Fig. 4, B, D, and E). Morphine (10 μM) caused a mild inhibition of sEPSC amplitude (10.7%) in WT mice (P = 0.0013) but had no such inhibition in Pd1−/− mice (P = 0.09561; Fig. 4, A to E). No significant difference was observed in the sEPSC amplitude in Pd1−/− cells compared to WT after morphine treatment (P = 0.2692; Fig. 4E). Morphine suppressed sEPSC frequency with an IC50 (median inhibitory concentration) of 1.07 and 15.24 μM in WT and Pd1−/− samples, respectively (fig. S7, A to C). DAMGO’s (0.5 μM) inhibition on sEPSCs was also diminished in Pd1−/− neurons compared to WT cells (P = 0.0154; fig. S7, D and E), suggesting a specific involvement of MOR. Thus, Pd1 deletion diminishes MOR-mediated suppression of spinal nociceptive synaptic transmission.

Fig. 4 Morphine’s inhibition on spinal cord nociceptive neurons in WT and Pd1−/−mice and the effects of nivolumab.

(A to E) Morphine’s effects on spinal excitatory synaptic transmission in lamina II neurons of spinal cord slices in WT and Pd1−/− mice. (A and B) Recording traces of spontaneous excitatory postsynaptic currents (sEPSCs) and cumulative histograms of the interevent interval and amplitude of sEPSCs in neurons from WT (A) and Pd1−/− (B) mice, before (Control) and after morphine treatment (10 μM). The histograms were examined for 1 min before morphine treatment (338 events in WT mice and 415 events in Pd1−/− mice) and 1 min after morphine washout (85 events in WT mice and 235 events in Pd1−/− mice). D indicates distance between two curves. (C and D) sEPSC frequency and amplitude before and after morphine (10 μM) treatment in WT and Pd1−/− mice. (E) Percentage changes in sEPSC frequency (top) and amplitude (bottom) in WT and Pd1−/− mice. Unpaired two-tailed t test. (F and G) Morphine-induced outward currents in lamina II neurons of spinal cord slices of WT and Pd1−/− mice. (F) Traces of outward currents in WT and Pd1−/− mice. (G) Left: Chart distribution of neurons with and without outward current. Right: Amplitude of morphine-induced outward currents in WT and Pd1−/− mice. (H) Effects of control IgG (left) and nivolumab (right) on morphine’s inhibition of sEPSCs in lamina II neurons of spinal cord slices of WT mice. Top: Recording traces of sEPSCs. Bottom: sEPSC frequency and amplitude. (I) Percentage changes in sEPSC frequency (top) and amplitude (bottom) showing the effects of IgG or nivolumab on morphine’s inhibition of sEPSCs. (A and B) Kolmogorov-Smirnov test; (C, D, and H) paired two-tailed t test; (E, G, and I) unpaired two-tailed t test. Data are expressed as means ± SEM. Sample sizes are indicated in brackets.

Opioids also activate potassium channels to generate outward currents and hyperpolarization in neurons (20). Morphine (10 μM) evoked outward currents in 36.7% (5 of 14) SDH lamina II neurons of WT mice with an amplitude of 15.8 ± 1.5 pA (Fig. 4, F and G). Morphine-evoked outward currents were significantly smaller in Pd1−/− mice (7.8 ± 1.8 pA; P = 0.0104), despite a similar response rate (30.7% of 13 neurons; Fig. 4, F and G). In addition, potassium currents were blocked by Tertiapin-Q, a high-affinity blocker for inward-rectifier K+ channels, suggesting an involvement of G protein–gated inwardly rectifying potassium (K+) (GIRK) channels (fig. S7F).

We next investigated whether nivolumab would alter the effects of morphine on SDH neurons of WT mice. Perfusion of spinal cord slices with nivolumab at a very low concentration (100 ng/ml ≈ 0.7 nM) significantly reduced the morphine’s inhibition of sEPSC frequency (P = 0.0004, compared with human IgG4 control; Fig. 4, H and I). Nivolumab at 0.7 nM without morphine did not affect sEPSC frequency but did increase sEPSC frequency at higher concentrations (P = 0.0229 and P = 0.0351 for 2.1 and 10.5 nM, respectively; fig. S8, A to F). Treatment with control IgG4 did not affect the morphine-induced inhibition of sEPSC (Fig. 4H and fig. S8). Nivolumab (0.7 nM) also suppressed morphine’s inhibition on sEPSC amplitude (P = 0.0257; Fig. 4I).

PD-L1 and morphine act together to inhibit synaptic transmission and nociception

As a major ligand of PD-1, PD-L1 produces antinociception in mouse models of pathological pain (7). A dose-response comparison showed that PD-L1 and morphine inhibit sEPSC frequency at IC50 of 0.60 nM and 1.07 μM, respectively (figs. S7C and S9, A and B). PD-L1 failed to inhibit sEPSCs in SDH neurons lacking Pd1 (fig. S9C). In naïve animals, Pd1 deficiency led to an increase in sEPSC frequency (P < 0.0001) but not amplitude (P = 0.7208) (fig. S9D), indicating that PD-1 is an endogenous negative regulator of synaptic transmission in mouse SDH neurons.

We further assessed whether PD-L1 and morphine would act together to modulate nociception. At very low concentrations, PD-L1 (0.075 nM) and morphine (0.01 μM) each produced mild inhibition of sEPSC frequency (5 and 8%, respectively), but coapplication of PD-L1 and morphine at these same low concentrations produced a much greater inhibition (40%) of sEPSCs, suggesting a collaborative enhancement of these two compounds in inhibition of sEPSC at low concentrations (P < 0.0001; Fig. 5, A to D). This enhancement was also observed in behavioral testing. Intrathecal PD-L1 (0.075 nmol) failed to produce any antinociception, and intrathecal morphine (0.75 nmol) only produced very mild and transient antinociception. However, intrathecal coapplication of PD-L1 and morphine at these low doses produced a much greater degree of antinociception (P < 0.0001 for both tail-flick and hot-plate tests; Fig. 5E).

Fig. 5 PD-L1 and morphine produce collaborative inhibition of spinal synaptic transmission and nociception in WT mice.

(A to C) Top: sEPSC traces showing the effects of PD-L1 (A), morphine (B), and PD-L plus morphine (C) in lamina II neurons of spinal cord slices. The upper and lower traces in each panel were obtained from the same neurons. Bottom: sEPSC frequency and amplitude. Data were analyzed by paired two-tailed t test. (D) Percentage changes in sEPSC frequency (left) and amplitude (right) under the treatment of PD-L1, morphine, and PD-L1 plus morphine. (E) Antinociception of PD-L1 plus morphine after intrathecal administration in tail-flick test (left) and hot-plate test (right). PD-L1 or saline was injected 30 min before morphine injection. (A to C), paired two-tailed t test; (D) one-way ANOVA; (E) two-way ANOVA, followed by Bonferroni’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.0001, and ****P < 0.0001, saline + morphine versus PD-L1 + morphine. Data are expressed as means ± SEM. Sample sizes are indicated in brackets.

Next, we asked whether PD-L1 would inhibit nociception via opioid receptors. Perfusion of spinal cord slices with the opioid receptor antagonist naloxone (1 μM) had no effects on basal sEPSC frequency and amplitude (fig. S10, A and B) but completely blocked the inhibitory actions of morphine (10 μM; fig. S10, C and D). Naloxone also blocked the PD-L1’s inhibition of sEPSC frequency and amplitude (fig. S10, E and F). Furthermore, intrathecal injection of PD-L1 (10 μg) reduced mechanical hypersensitivity in experimental bone cancer in mice, but this antinociceptive effect was abrogated by intrathecal naloxone (5 nmol; fig. S11, A to C). To determine whether PD-L1 could directly activate MOR, we used a cell line that lacks PD-1 expression but has high expression of MOR and promiscuous Gq proteins coupled to calcium signaling. At all the concentrations (30 to 3000 ng/ml), PD-L1 failed to activate MOR to trigger Ca2+ responses (fig. S11, D and E). As a positive control, DAMGO (0.5 μM) evoked a robust Ca2+ increase, which was blocked by naloxone (fig. S11, D and E). Thus, PD-L1 may indirectly interact with MOR to modulate synaptic transmission and nociception.

Loss of PD-1 enhances opioid-induced hyperalgesia and spinal long-term potentiation

Opioid-induced hyperalgesia/allodynia not only occurs after chronic exposure but also may happen after acute treatment (21). Intrathecal DAMGO (0.5 nmol) elicited much faster mechanical allodynia in Pd1−/− mice, which was evident at 3 hours in von Frey test using both threshold (P < 0.0001; Fig. 6A) and frequency measurements (P = 0.0031, 0.4 g; Fig. 6B). Opioid withdrawal has been shown to induce long-term potentiation (LTP), a critical form of synaptic plasticity underlying the pathogenesis of pain (21, 22). We recorded for LTP in lamina II neurons in spinal cord slices from WT and Pd1−/− mice after electrical stimulation of the attached dorsal root. Perfusion of DAMGO (0.5 μM, 3 min) in WT neurons initially induced a depression of evoked EPSCs (eEPSCs) during the perfusion, and this depression was diminished in Pd1-deficient neurons (P = 0.0488; Fig. 6, C to E). After DAMGO withdrawal, the depression phase was followed by a recovery phase in WT and Pd1−/− mice, but the duration of recovery was shortened in Pd1-deficient neurons (P = 0.0102 versus WT; Fig. 6, C, D, and F). After the recovery phase, WT neurons developed a moderate LTP (20% increase in eEPSC amplitude; Fig. 6, C, D, and G). In sharp contrast, LTP was much greater in Pd1−/− mice (P = 0.0463; Fig. 6G). In addition, LTP occurred more rapidly in Pd1−/− mice (Fig. 6D). Together, these findings suggest that loss of PD-1 might cause faster and more robust development of opioid-induced hyperalgesia/allodynia as a result of enhanced LTP in SDH neurons.

Fig. 6 DAMGO-induced hyperalgesia and spinal LTP are enhanced in Pd1−/− mice.

(A and B) Paw withdrawal threshold (A) and paw withdrawal frequency to 0.16-g filament (B) change after DAMGO (0.5 nmol, it) in WT and Pd1−/− mice. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, WT versus Pd1−/−, two-way ANOVA, followed by Bonferroni’s post hoc test. (C to G) LTP of EPSC amplitude in lamina II neurons of spinal cord slices after DAMGO withdrawal in WT and Pd1−/− mice. (C) Recording traces of eEPSC before, during, and after DAMGO (0.5 μM, 3 min) in SDH neurons of WT and Pd1−/− mice. (D) Time course of eEPSC amplitude of WT and Pd1−/− neurons. Two-way ANOVA. (E) Effects of DAMGO on eEPSC amplitude, compared to the baseline, in WT and Pd1−/− mice during the drug perfusion. (F) Latency (min) of recovery of DAMGO-induced depression of eEPSC amplitude, compared to the baseline (100%) in WT and Pd1−/− mice. (G) Maximum potentiation (% of baseline) of eEPSC amplitude at 30 to 36 min after DAMGO (0.5 μM) perfusion in WT and Pd1−/− mice. (E to G) Unpaired two-tailed t test. Data are expressed as means ± SEM. Sample sizes are indicated in brackets.

Loss of PD-1 enhances chronic opioid-induced tolerance and microgliosis in SDH

Chronic morphine exposure (six daily injections; 10 mg/kg, sc) induced antinociceptive tolerance in WT and Pd1−/− mice in tail-flick testing (Fig. 7A, left) and hot-plate testing (Fig. 7B, left). The maximum possible effect (MPE) analysis revealed that morphine evoked tolerance in WT mice on day 4 in tail-flick testing and on day 3 in hot-plate testing [P < 0.0001; Fig. 7, A and B (right)]. In Pd1−/− mice, the induction was faster: Morphine induced tolerance on day 2 in both tail-flick and hot-plate tests [P = 0.0163 and 0.0256, respectively; Fig. 7, A and B (right)].

Fig. 7 Chronic morphine-induced tolerance and microgliosis in SDH are potentiated in Pd1−/− mice.

(A and B) Tail-flick (A) and hot-plate (B) tests showing the latency (left) and maximum possible effect (% MPE; right) of morphine antinociception using a 6-day treatment paradigm (10 mg/kg, sc, once daily) in WT and Pd1−/− mice. *P < 0.05 and ****P < 0.0001. versus day 1 baseline antinociception, two-way ANOVA. Morphine antinociception was tested 30 min after each injection. (C and D) Microglial activation (microgliosis) in SDH of WT and Pd1−/− mice after 4-day morphine exposure (10 mg/kg, sc, once daily). (C) Low-magnification (top) and high-magnification (bottom) images of IBA1 immunostaining. Scale bars, 500 μm. (D) Quantification of IBA1 IR in SDH. **P < 0.01, unpaired two-tailed t test, WT versus Pd1−/− mice. Data are expressed as means ± SEM. Sample sizes are indicated in brackets.

Because microglial activation in SDH after chronic opioid exposure regulates opioid tolerance (23, 24) and PD-1 was implicated in microglia signaling (25), we investigated microglia activation (microgliosis) in SDH of WT and Pd1−/− mice after 4-day morphine exposure (10 mg/kg, sc, daily). This course of morphine treatment did not trigger microgliosis in WT mice, as revealed by IBA1 immunostaining, but induced a marked microgliosis in Pd1−/− mice (P = 0.0040; Fig. 7, C and D). Thus, enhanced microglial activation in the SDH plays an active role in the facilitation of opioid-induced tolerance in PD-1 deficiency.

We examined additional adverse effects of opioids, including constipation and withdrawal response in WT and Pd1−/− mice. Gastrointestinal transit (GIT) assay revealed increased baseline gastrointestinal movement in Pd1−/− compared to WT mice (P = 0.0190; fig. S12A). Subcutaneous morphine (10 mg/kg) induced a substantial reduction of gastrointestinal movement in WT mice, and this deficit was improved in Pd1−/− mice (P = 0.0265; fig. S12A). We also tested morphine-induced withdrawal response after chronic morphine exposure (26) by examining naloxone-induced [2 mg/kg, intraperitoneally (ip)] jumping response for 30 min in WT and Pd1−/− mice. Naloxone-induced jumping was potentiated in Pd1−/− mice within the first 10 min (P = 0.0001 and P = 0.0039; fig. S12B), suggesting a moderate modulation of opioid withdrawal response by PD-1.

Spinal infusion of nivolumab blocks morphine antinociception in NHPs

NHPs have been tested for new pain therapeutics including opioids because of their phylogenetic proximity to humans (27). We examined the actions of nivolumab on nociception and morphine antinociception in NHPs. To minimize immunomodulatory response of this human antibody and target the spinal cord pain circuit, we delivered drug via a catheter implanted into the intrathecal cerebrospinal space of NHP (Fig. 8A) (28). Tail-withdrawal testing in a 46°C water bath revealed a dose-dependent hyperalgesia after intrathecal infusion of nivolumab (10, 30, and 100 μg) in NHPs (Fig. 8, A and B). The induction of hyperalgesia, corresponding to a reduction in tail-withdrawal latency, was very rapid, peaking within 1 hour, suggesting a possible neuronal hyperactivity (Fig. 8B). This hyperalgesia was maintained at 6 hours but fully recovered at 24 hours (P < 0.0001, n = 5 NHPs; Fig. 8B). By contrast, the isotype control serum (IgG4) had no effect on thermal threshold even at the highest dose (100 μg, it; Fig. 8B). Thus, nivolumab is sufficient to produce hyperalgesia in NHPs (85% reduction of thermal latency at the dose of 100 μg; Fig. 8, A and B).

Fig. 8 Morphine antinociception is blocked after pretreatment with nivolumab in NHPs.

(A) Paradigm of intrathecal injection, drug treatment, and tail-withdrawal test in 46°C water bath for testing thermal hyperalgesia. (B) Thermal threshold changes after intrathecal nivolumab (10, 30, and 100 μg). ****P < 0.0001, compared with control human IgG4 treatment, two-way ANOVA, followed by Bonferroni’s post hoc test. (C) Paradigm of intrathecal injection, drug treatment, and tail-withdrawal test in 50°C water bath for testing morphine antinociception. (D to F) Effects of intrathecal pretreatment with nivolumab (30 and 100 μg) on spinal antinociception of morphine (30 μg). (D) Tail-withdrawal latency before and after nivolumab administration. (E) % MPE, ****P < 0.0001, IgG4 + morphine versus nivolumab + morphine, two-way ANOVA, followed by Bonferroni’s post hoc test. (F) Area under the curve (AUC) of % MPE; P = 0.0002 and P < 0.0001, one-way ANOVA, followed by Bonferroni’s post hoc test. n = 5 NHPs per group. Nivolumab or control antibody (human IgG4) was intrathecally injected 60 min before intrathecal morphine injection. Data are expressed as means ± SEM.

Last, we tested whether nivolumab would impair morphine antinociception in NHPs as in rodents. Nivolumab was given 1 hour before morphine injection, and tail-withdrawal testing was conducted in a 50°C water bath to test morphine antinociception (Fig. 8C) (29). Intrathecal infusion of morphine (30 μg) evoked full antinociception for >3 hours, as shown by tail-withdrawal latency (Fig. 8D), MPE (Fig. 8E), and area under the curve (AUC) (Fig. 8F). Intrathecal pretreatment with nivolumab (30 and 100 μg) produced a dose-dependent inhibition of morphine antinociception in NHPs (P < 0.001, n = 5 animals), whereas IgG4 (100 μg, it) had no effect (Fig. 8, D to F). Nivolumab (100 μg) largely blocked morphine antinociception in NHPs (>85% reduction, P < 0.001; n = 5; Fig. 8F). These data strongly indicate that anti–PD-1 treatment could block morphine antinociception in NHPs.


We have demonstrated a previously unrecognized role of PD-1 in regulating opioid antinociception in rodents and NHPs, with several lines of cellular, functional, and behavioral evidence demonstrating a cross-talk between PD-1 and MOR. First, PD-1 and MOR are colocalized in DRG neurons, spinal nerve axons, and spinal cord axonal terminals. The proximity and potential interaction between these two molecules were further evidenced by co-IP and PLA experiments. Second, morphine’s antinociception in tail-flick and hot-plate testing in naïve mice as well as its antiallodynic actions in inflammatory, neuropathic, and bone cancer pain models are all diminished in Pd1−/− mice. Third, morphine’s antinociceptive cellular actions, including suppression of calcium currents in DRG neurons, inhibition of sEPSCs in SDH neurons, and induction of outward currents in SDH neurons, were all compromised in Pd1−/− mice. Morphine’s signaling deficits in Pd1−/− mice can be recapitulated by nivolumab treatment in adult WT mice, arguing against developmental regulation in Pd1−/− mice. Future studies are warranted to determine how PD-1 interacts with MOR at the cellular and molecular levels.

We also tested for adverse effects from opioid treatment, including hyperalgesia/allodynia, tolerance, and withdrawal response in WT and Pd1−/− mice. These excitatory adverse effects of opioids (morphine and DAMGO) are all potentiated in Pd1−/− mice. In contrast, constipation, an inhibitory adverse effect of opioid, was improved in Pd1−/− mice. Pd1 deficiency or blockade in WT mice also increased basal synaptic transmission (EPSC) and LTP in SDH neurons, whereas exogenous PD-L1 suppressed EPSC in these neurons. Together, these observations indicate an overwhelming inhibitory action of PD-1 on neurons. Our previous study showed that activation of PD-1 by PD-L1 activates the Src homology region 2 domain-containing phosphatase-1 (SHP-1) in nociceptors to suppress neuronal excitability via inhibition of sodium channels and activation of potassium channels (7). Thus, PD-1 may serve as a neuronal suppressant in the PNS and CNS. It remains to be investigated whether PD-1 acts as a pan-neuronal inhibitor in different brain regions.

There are two caveats to this study. First, although we revealed neuronal modulation by PD-1 in opioid-induced antinociception and hyperalgesia, we did not rule out the contribution of PD-1–mediated immunomodulation to these opioid-induced processes. PD-1 is highly expressed by T cells and antigen-presenting cells such as dendritic cells and macrophages to inhibit immunity (30). However, PD-1 in microglia acts to inhibit the function of these CNS immune cells (25). Our data showed that loss of PD-1 potentiated chronic morphine-induced microgliosis in the SDH, contributing to enhanced opioid tolerance in Pd1−/− mice. In addition, PD-1 expressed on other immune cell types such as T cells and macrophages may also play a role in opioid tolerance. However, the acute effect of PD-1 on opioid antinociception is very likely mediated by neuronal PD-1, given its rapid actions within a few minutes to tens of minutes. To determine the specific contribution of each cell type in morphine antinociception and tolerance, conditional knockout mice with Pd1 deletion in different neuronal and immune cell types are needed. Second, the downstream signaling of PD-1 remains to be determined. The phosphatase SHP-1 is required for PD-1 to regulate the activities of sodium channels and potassium channels and pain sensitivity (10). It will be of great interest to investigate the involvement of SHP in opioid antinociception and tolerance.

Nivolumab’s approval for treating various cancers (31, 32) and the use of opioids as mainstay treatments for cancer pain (10, 33) render major clinical implications for our findings in this study. We found colocalization and interactions between MOR and PD-1 in human dorsal root tissue and demonstrated nivolumab blockade of spinal morphine antinociception in NHPs. Thus, anti–PD-1 immunotherapy may interfere with opioid analgesia in patients with cancer by disrupting the PD-1–MOR interaction in peripheral nerve and DRG, as well as in the spinal cord particularly with disruption of the blood-brain barrier in some patients with cancer. In addition, PD-L1 might be used to treat clinical pain and enhance opioid analgesia in noncancer patients. A recent case report showed that PD-1 immunotherapy elicited severe itch (pruritus) in an 88-year-old woman with a 7-month history of lung adenocarcinoma, and remarkably, treatment with naloxone resulted in substantial relief within 1 hour (34), suggesting a correlation of PD-1 and MOR in humans. This finding also suggested that naloxone may inhibit proinflammatory cytokine production in peripheral mononuclear cells in this patient (34). Future clinical trials are warranted to test morphine-induced analgesia and morphine-induced pruritus in patients receiving anti–PD-1 immunotherapy and explore the underlying immune and neuronal modulations.


Study design

This study was designed to evaluate the interactions between PD-1 and MOR in the PNS and CNS. First, we examined the role of PD-1 in modulating opioid antinociception in tail-flick and hot-plate tests and in mouse models of bone cancer pain, inflammatory pain, and neuropathic pain by using Pd1−/−mice and/or the anti–PD-1 antibody nivolumab. Second, we checked the expression patterns of PD-1 and MOR using RNAscope and immunohistochemistry, and the interaction between PD-1 and MOR was tested by PLA and co-IP. Third, we evaluated how PD-1 regulates opioid signaling using electrophysiology recordings in dissociated DRG neurons and spinal cord slices. Fourth, we assessed the side effects of morphine including hyperalgesia, tolerance, GIT, and withdrawal responses. Last, we tested whether nivolumab would impair morphine antinociception in NHPs. The sample size and power calculation was determined on the basis of our experience with the experimental models, anticipated biological variables, and previous literature (7, 35), and differences were analyzed using two-tailed t test, one-way analysis of variance (ANOVA), and two-way ANOVA. All data were included in analyses. All experiments have been replicated successfully, and all results reported have been reliably reproduced. Sample sizes and replicates are shown in the figure legends. Animals were randomly assigned into different cages and groups before experiments. In behavior tests, the investigators were blinded to drug administrations and group assignments.


Mouse PD-L1 (no. ab130039) and human IgG4 (no. ab90286) were purchased from Abcam. Nivolumab (OPDIVO), a humanized anti–PD-1 antibody, was from Bristol-Myers Squibb. Morphine sulfate was obtained from West-Ward Pharmaceuticals. Naloxone (no. 1453005), CFA (no. F5881), DAMGO (no. E7384), DPDPE (no. E3888), and U69593 (no. U103) were obtained from Sigma-Aldrich. Tertiapin-Q (no. 1316) was obtained from Tocris. Fura-2 AM (#F14185) was obtained from Invitrogen.


Pd1 (Pdcd1) knockout mice with a C57BL/6 background and NOD.CB-17-Prkdcscid mice were purchased from the Jackson laboratory (stock nos. 021157 and 001303) and maintained at the Duke animal facility (7). Young mice (aged 5 to 7 weeks of both sexes) were used for electrophysiological studies in the spinal cord and DRG neurons. Adult male mice (aged 8 to 10 weeks), including knockout mice and corresponding WT control mice, as well as some CD1 mice (Charles River Laboratories), were used for behavioral and pharmacological studies. Female mice were also used for behavioral tests in Figs. 1 (E and G), 6 (A and B), and 7 (A and B) and fig. S4. Mice were group housed on a 12-hour light/12-hour dark cycle at 22° ± 1°C with free access to food and water. Animals were randomly assigned to each group. Sample sizes were estimated on the basis of our previous studies for similar types of behavioral, biochemical, and electrophysiological analyses (7, 35). Two to five mice were housed in each cage. Animal experiments were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Duke University Institutional Animal Care and Use Committee.


Five adult male and female rhesus monkeys (Macaca mulatta), 9.3 to 13.8 kg, were kept at an indoor facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (Frederick, MD). They were individually housed in species-specific rooms with environmental controls set to maintain 21° to 25°C, 40 to 60% relative humidity, and a 12-hour light (06:30 to 18:30)/12-hour dark cycle. Their daily diet consisted of 20 to 28 biscuits (Purina Monkey Chow; Ralston Purina Co., St. Louis, MO), fresh fruits, and water ad libitum. Small amounts of primate treats and various cage enrichment devices were supplied as forms of environmental enrichment. All monkeys have been previously trained in the warm water tail-withdrawal assay. All animal care and experimental procedures were conducted in accordance with the Guide for the Care and Use of Laboratory Animals as adopted and promulgated by the U.S. National Institutes of Health (Bethesda, MD) and approved by the Institutional Animal Care and Use Committee in Wake Forest University School of Medicine (Winston-Salem, NC). This study is reported in accordance with the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines for reporting experiments involving animals (36).

Mouse pain models and behavioral testing

Inflammatory pain. Peripheral inflammation was induced by intraplantar injection of 20-μl CFA (Sigma-Aldrich) in a hind paw under brief anesthesia with isoflurane.

Neuropathic pain. To produce SNL, animals were anesthetized with isoflurane, and the L6 transverse process was removed to expose the L4 and L5 spinal nerves. The L5 spinal nerve was then isolated and tightly ligated with 6-0 silk thread.

Bone cancer pain. Lewis lung carcinoma cell line (LLC1) was obtained from the American Type Culture Collection. Before the inoculation, the cancer cells were digested with 0.25% trypsin and made into a suspension of cells (1 × 108/ml) in phosphate-buffered saline. Mice were anesthetized with 3% isoflurane, and a 0.5- to 1-cm superficial incision was made near the knee joint to expose the patellar ligament. Then, a 25-gauge needle was inserted at the site of the intercondylar notch of the left femur into the femoral cavity. The needle was connected with a 10-μl microinjection syringe containing 2-μl suspension of 2 × 105 tumor cells, as well as 2-μl absorbable gelatin sponge solution for the closure of the injection site. The contents of the syringe were slowly injected into the femoral cavity for a duration of 2 min. To further prevent leakage of tumor cells, the outside injection site was sealed with silicone adhesive (Kwik-Sil, World Precision Instruments). The wound was then closed with silk sutures and reswabbed.

Drug injection in mice

For intravenous injection, anti–PD-1 antibody (nivolumab, 10 mg/kg in 100-μl saline) or control antibody (human IgG4) was administered into the tail vein of mouse. For intrathecal injection, spinal cord puncture was made by a Hamilton microsyringe (Hamilton) with a 30-gauge needle between the L5 and L6 levels to deliver reagents (5 μl) to the cerebral spinal fluid. For subcutaneous injection, morphine (1, 3, 10, and 30 mg/kg in 250-μl saline) was injected beneath the back skin with a 30-gauge needle.

Behavioral tests of pain in mice

All animals were habituated to testing environment for at least 2 days before the baseline testing. All the tests were concluded blindly.

Tail-flick test. Mice were gently held by hand with a terry glove. The exposed distal part of the tail (3 cm) was immersed into 50°C hot water. The tail-flick latency was defined as the time required for a mouse to flick or remove its tail out of hot water. A maximum cutoff value of 15 s was set to avoid thermal injury. Tail-flick latency was assessed before and after drug injection. Data are also expressed as the MPE, calculated as MPE (%) = 100 × [(postdrug response − baseline response)/(cutoff response − baseline response)]. The MPE (%) data from each animal were also converted to AUC.

Hot-plate test. For all the experiments, hot-plate test was conducted after tail-flick test. Mice were placed on the hot plate at 53°C, and the reaction time was scored when animal began to exhibit signs of pain avoidance such as jumping or paw licking. A maximum cutoff value of 40 s was set to avoid thermal injury.

Mechanical sensitivity. For CFA-induced inflammatory pain, SNL-induced neuropathic pain, and bone cancer pain, as well as morphine-induced hyperalgesia, mechanical sensitivity was assessed by von Frey hairs. Animals were habituated to the testing environment daily for at least 2 days before baseline testing. Animals were confined in boxes placed on an elevated metal mesh floor, and the hind paws were stimulated with a series of von Frey hairs with logarithmically increasing stiffness (0.02 to 2.56 g, Stoelting), presented perpendicularly to the central plantar surface. We determined the 50% paw withdrawal threshold by the up-down method. To determine mechanical allodynia, we also tested paw withdrawal frequency using a low-threshold filament (0.4 g).

Thermal sensitivity was tested using Hargreaves radiant heat apparatus (IITC Life Science). The basal paw withdrawal latency was adjusted to 9 to 12 s, with a cutoff of 20 s to prevent tissue damage.

Opioid-induced tolerance in mice

Chronic morphine treatment was used to induce antinociceptive tolerance. Briefly, mice were given injection of morphine (10 mg/kg, sc) daily for 6 days. The tail-flick and hot-plate latency was tested before and 1 hour after each morphine injection. The MPE (%) value was calculated to compare the antinociceptive effects and tolerance development. Additional WT and Pd1−/− mice were used to examine microgliosis 4 days after morphine treatment.

Tail-withdrawal test and drug injection in NHPs

The warm water tail-withdrawal assay was used to evaluate nociceptive responses modulated by ligands administered intrathecally (28, 37). These monkeys have been implanted with the intrathecal catheter for drug infusion. They seated in primate restraint chairs, and the lower part of their shaved tails (about 15 cm) was immersed into a thermal flask containing water maintained at either 46° or 50°C.Water at 46°C was used as a non-noxious stimulus to detect allodynic/hyperalgesic responses, and 50°C water was used as an acute noxious stimulus to assess antinociceptive effects (27). Tail-withdrawal latencies were measured at each temperature using a computerized timer by experimenters who were unaware of the experimental conditions. If monkeys did not remove their tails within 20 s (cutoff), the flask was removed, and a maximum time of 20 s was recorded. Test sessions began with baseline measurements at each temperature. Subsequent tail-withdrawal latencies were measured at multiple time points after intrathecal administration of human IgG4 (100 μg) or nivolumab (0, 10, 30, or 100 μg). In addition, a single dose (0, 30, or 100 μg) of nivolumab or human IgG4 (100 μg) was administered intrathecally 1 hour before morphine (30 μg) to determine whether nivolumab interferes with morphine antinociception.

Whole-cell patch-clamp recording in mouse spinal cord slices

Mice of both sexes (5 to 7 weeks old) were anesthetized with urethane (1.5 to 2.0 g/kg, ip). The lumbosacral spinal cord was quickly removed and placed in ice-cold dissection solution (240 mM sucrose, 25 mM NaHCO3, 2.5 mM KCl, 1.25 mM NaH2PO4, 0.5 mM CaCl2, and 3.5 mM MgCl2), equilibrated with 95% O2 and 5% CO2. The slices were saturated with 95% O2 and 5% CO2. After spinal extraction under anesthesia, animals were euthanized by decapitation. Transverse spinal slices (300 to 400 μm) were prepared using a vibrating microslicer (VT1200S Leica). The slices were incubated at 32°C for at least 30 min in artificial cerebrospinal fluid (ACSF) (126 mM NaCl, 3 mM KCl, 1.3 mM MgCl2, 2.5 mM CaCl2, 26 mM NaHCO3, 1.25 mM NaH2PO4, and 11 mM glucose), equilibrated with 95% O2 and 5% CO2. The slices were placed in a recording chamber and completely submerged and perfused at a flow rate of 2 to 4 ml/min with ACSF, which was saturated with 95% O2 and 5% CO2 and maintained at room temperature. Lamina II neurons in lumbar segments were identified as a translucent band under a microscope (BX51WIF; Olympus) with light transmitted from below. Whole-cell voltage-clamp recordings were made from outer lamina II neurons by using patch pipettes fabricated from thin-walled, fiber-filled capillaries (38). Patch pipette solution used to record sEPSCs contained 135 mM K-gluconate, 5 mM KCl, 0.5 mM CaCl2, 2 mM MgCl2, 5 mM EGTA, 5 mM Hepes, and 5 mM Mg–adenosine triphosphate (pH 7.3 adjusted with KOH, 300 mOsm). The patch pipettes had a resistance of 8 to 10 M. The sEPSC recordings were made at a holding potential (VH) of −70 mV in the presence of 10 μM picrotoxin and 2 μM strychnine. Signals were acquired using an Axopatch 700B amplifier. The data were stored and analyzed with a PC using pCLAMP 10.3 software. sEPSC events were detected and analyzed using Mini Analysis Program version 6.0.3. The histograms of sEPSC were examined for the events in 1 min before morphine treatment and in 1 min after morphine washout. The D value was measured by the Kolmogorov-Smirnov test and indicates the distance between two curves (39). Morphine-induced outward currents were recorded in whole-cell patch-clamp configuration and analyzed by Clampfit 10.6. Briefly, the current value was defined by the peak amplitude value of an outward current subtracted by the amplitude of baseline current under voltage-clamp mode. In addition, some neurons were treated with Tertiapin-Q (100 nM), a blocker of GIRK channels (40). In all cases, n refers to the number of the neurons studied. All drugs were bath applied by gravity perfusion via a three-way stopcock without any change in the perfusion rate.

Spinal LTP of C fiber–eEPSCs was elicited in lamina IIo neurons of spinal cord slices by stimulating the dorsal roots attached to the slices using whole-cell voltage-clamp recording (21, 22). The dorsal root electrical stimulation was performed by an electrode with a constant current source of pulse at an interval of 30 s. The stimulus (0.1 ms, 30-s interval, 120 to 300 μA) was applied at an intensity of 1.2 times of the threshold to elicit EPSCs and prevent a conduction block of action potential in the dorsal root. DAMGO (0.5 μM) was bath-applied for 3 min to induce LTP, and amplitudes of individual C fiber–eEPSCs were normalized to predrug (control) values. LTP was defined as >20% increase in amplitude after DAMGO treatment.

Statistical analysis

All data were expressed as means ± SEM, as indicated in the figure legends. Statistical analyses were completed with GraphPad Prism 6.1. Behavioral data were analyzed using two-tailed Student’s t test (two groups), one- or two-way ANOVA (repeated measures over a time course), followed by post hoc Bonferroni test. Electrophysiological data were tested using one-way ANOVA (for multiple comparisons) or two-tailed Student’s t test (two groups). The criterion for statistical significance was P < 0.05.


Materials and Methods

Fig. S1. Dose response of subcutaneous morphine antinociception in WT and Pd1−/− mice.

Fig. S2. Antinociception of MOR, DOR, or KOR agonist in WT and Pd1−/− mice.

Fig. S3. Morphine antinociception in WT and immunodeficient CB-17 mice pretreated with nivolumab.

Fig. S4. Morphine antinociception in female Pd1−/− mice and WT mice pretreated with nivolumab.

Fig. S5. PLA showing PD-1/MOR interaction in cultured mouse DRG neurons and coexpression of PD-1/MOR on mouse dorsal roots and co-IP analysis in human dorsal roots.

Fig. S6. PD-1 is coexpressed with MOR and CGRP in axonal terminals in mouse SDH.

Fig. S7. Effects of morphine, DAMGO, and PD-L1 on sEPSCs in lamina IIo neurons of spinal cord slices in WT and Pd1−/− mice.

Fig. S8. Anti–PD-1 treatment with nivolumab increases sEPSC frequency in lamina IIo neurons of spinal cord slices.

Fig. S9. Effects of PD-L1 on sEPSCs and outward currents in lamina IIo neurons of spinal cord slice in WT and Pd1−/− mice.

Fig. S10. PD-L1–induced inhibition of sEPSC in lamina IIo neurons in spinal cord slice and the effect of naloxone.

Fig. S11. PD-L1–induced antinociception in mice and MOR-mediated calcium signaling in vitro.

Fig. S12. Morphine-induced GIT and naloxone-induced withdrawal behavior in WT and Pd1−/− mice.

Data file S1. Raw data.

Reference (41)


Funding: This work was supported by NIH R01 grant DE17794 (to R.-R.J.), DA032568 (to M.-C.K.), and DA044775 (to M.-C.K.). Author contributions: Z.W. conducted the behavioral tests and electrophysiology in DRG neurons; C.J. conducted the electrophysiology in spinal cord slices; Q. He conducted the immunohistochemistry; M.M. conducted the in situ hybridization; Q. Han and S.B. conducted the immunoprecipitation; K.W. prepared the mouse model of bone cancer. H.D. and M.-C.K. conducted the experiments in NHPs. R.-R.J. supervised the project; and R.-R.J., Z.W., C.J., and M.-C.K. wrote the paper. We also thank R.-R.J.’s laboratory members Y. Huh and C. Donnelly for reading and editing the manuscript. Competing interests: R.-R.J. is a consultant of the Boston Scientific and received research support from the company. He serves at Board of Directors of Ascletis Pharma. These activities are not related to this study. Related to this study, R.-R.J. filed a patent “Methods and kits for treating pain” (PCT 16/612,909) from the Duke University. M.-C.K. had research contracts with Biogen MA Inc. and Orexigen Therapeutics in the past 3 years. These activities are not related to this study. The other authors declare that they have no competing financial interests. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article