Ligand‐specific recycling profiles determine distinct potential for chronic analgesic tolerance of delta‐opioid receptor (DOPr) agonists

Abstract δ‐opioid receptor (DOPr) agonists have analgesic efficacy in chronic pain models but development of tolerance limits their use for long‐term pain management. Although agonist potential for inducing acute analgesic tolerance has been associated with distinct patterns of DOPr internalization, the association between trafficking and chronic tolerance remains ill‐defined. In a rat model of streptozotocin (STZ)‐induced diabetic neuropathy, deltorphin II and TIPP produced sustained analgesia following daily (intrathecal) i.t. injections over six days, whereas similar treatment with SNC‐80 or SB235863 led to progressive tolerance and loss of the analgesic response. Trafficking assays in murine neuron cultures showed no association between the magnitude of ligand‐induced sequestration and development of chronic tolerance. Instead, ligands that supported DOPr recycling were also the ones producing sustained analgesia over 6‐day treatment. Moreover, endosomal endothelin‐converting enzyme 2 (ECE2) blocker 663444 prevented DOPr recycling by deltorphin II and TIPP and precipitated tolerance by these ligands. In conclusion, agonists, which support DOPr recycling, avoid development of analgesic tolerance over repeated administration.

mechanisms underlying the decline of analgesic responses to DOPr ligands, especially how ligand-induced trafficking contributes to development of tolerance. 14,18 In particular, single exposure to 'highly internalizing' agonists like SNC-80 but not 'low-internalizing' agonists like ARM390 was shown to abolish analgesic response to a single subsequent injection of the corresponding agonist. 18 These observations led to the proposal that ligand-specific sequestration patterns could be predictive of ligand potential to induce acute analgesic tolerance.
However, to interpret functional consequences of receptor internalization we must also consider the post-endocytic fate of internalized receptors. For example, a simple association between acute tolerance and sequestration efficacy does not hold true for deltorphin II 17 or DPDPE, both of which produce DOPr internalization comparable to SNC-80, 16,19 but do not produce acute tolerance.
We have previously shown that the potential of DPDPE to induce acute tolerance is minimized by its capacity to support receptor recycling to the membrane, which does not occur with SNC-80. 16,19 Internalization patterns per se have also failed to account for development of chronic tolerance as low and high internalizing ligands similarly result in loss of analgesic response over repeated administration. 14,15 On the other hand, the extent to which receptor recycling influences development of chronic analgesic tolerance to DOPr agonists remains to be determined.
In the present study, we sought to answer this question and found that DOPr agonists that supported receptor recycling to the membrane (deltorphin II; TIPP) induce sustained analgesia, independent of their efficacy or patterns of low (TIPP) or high (deltorphin II) internalization capacity. In contrast, ligands like SB235863 and SNC-80, which did not support membrane recovery of internalized receptors, displayed progressive loss of analgesic actions irrespective of the degree of internalization they induced or of the analgesic efficacy they displayed. The association between sustained analgesia and recycling was further demonstrated by the fact that inhibition of DOPr recycling precipitated tolerance in TIPP-and deltorphin IItreated rats.

| Materials
Drugs were purchased from different companies: deltorphin II was

| Animals
Adult male Sprague Dawley (SD) rats, weighing 235-250 g, were purchased from Charles River laboratories and housed in a controlled environment on a 12-hour light/dark cycle with free access to food and water. All experimental methods and animal care procedures were approved by the Animal Care Committee of University of Montreal (CDEA protocol , in accordance with the guiding principles as enunciated by the Canadian Council on Animal Care.

| Assessment of mechanical allodynia
Mechanical allodynia is the consequence of maladaptive neuroplasticity following nerve damage. 24 In a model of diabetic neuropathy as the one used in this study, hyperglycaemia contributes to generate reactive oxygen species which induce microglia activation and inflammatory mediators. 25,26 These mediators attenuate the activity of Gly/GABA interneurons in superficial layers 24,[27][28][29] and disrupt the inhibitory control that prevents innocuous touch stimuli from producing pain which leads to allodynia. 30

| Primary neuronal cultures
Primary neuronal cultures were prepared from rat post-natal prefrontal cortex (P0-P2) as previously described. 19 Briefly, three cultures were independently prepared from 8 to 12 pups each. Pups were cryo-anaesthe-

| Labelling and quantification of DOPr trafficking in primary neuron cultures
Immunolabelling of surface receptors for quantification of trafficking was performed as previously described. 19  Invitrogen, A21202). Alternatively, cultured neurons were allowed to recover for 60 minutes in the absence of ligand before a second round of calcium-free PBS wash was completed before fixation, permeabilization and incubation with secondary antibody. The latter procedure allowed to remove antibody bound to Flag-DOPrs that translocated to the surface during recovery ensuring the exclusive labelling of receptors that were retained intracellularly. 26 Another set of neurons was similarly treated with agonist or vehicle and then allowed to recover for 60 minutes following which they were fixed and incubated with secondary antibody without permeabilization, so as to exclusively reveal Flag-DOPrs that reappeared at the surface during recovery from internalization; measures of surface labelling were normalized to intracellular labelling produced during internalization. Recycling was thus established by taking two independent measures: (a) cytoplasmic labelling density lost during recovery from treatment and (b) gain in surface labelling density when neurons were allowed to recover from agonist-induced internalization. In experiments in which 6634449 was introduced in the medium during recovery, the concentration used was 20 μmol/L.
Cytoplasmic and surface labelling densities were quantified with ImageJ using a previously described method, 27 with small modifications. 16 In particular, total cytoplasmic labelling density was obtained by measuring fluorescence intensity within the region confined between the external and nuclear perimeters, and dividing this value by the corresponding area. Total surface labelling density was defined by calculating the ratio of fluorescence measured within internal and external perimeters of surface-labelled neurons, and the corresponding area. Nuclear labelling density (fluorescence within nuclear perimeter/nuclear area) was considered background and subtracted from total density values just described. Contours defining each of the regions of interest were first drawn on brightened images, and once the trace was completed, brightness was reset to acquisition conditions. Images F I G U R E 1 Experimental design. One week after STZ administration, repeated treatments with DOPr agonists were started. Before starting chronic treatment with ED 80 of deltorphin II, TIPP, SB235863 or SNC-80, animals were tested to establish the baseline allodynia. Immediately after, they were injected with deltorphin II at a dose corresponding to ED 50 , which was repeated at the end of treatment. Allodynia measures were taken every 15 min after each injection until analgesic effects disappeared. Chronic tolerance was evaluated by comparing analgesic response to successive i.t. injections of each agonist between days 2 and 6. Difference in analgesic response to deltorphin II ED 50 in days 1 and 8 allowed to establish tolerance induced by the different agonists to this common probe F I G U R E 2 Agonist-induced internalization in DOPrs expressed in cortical neurons. A, Primary cortical neuron cultures were transfected with Flag-DOPrs and incubated with primary antibody prior to exposing neurons to vehicle (neurobasal) or to agonists (10 µmol/L, 60 min) as indicated. By the end of treatment, drugs were washed out, antibody bound to receptors remaining at the cell surface was stripped, and cells were immediately processed to reveal receptors that translocated from the cell membrane to the intracellular compartment following constitutive (vehicle) or ligand-induced sequestration. B, Histograms show intracellular labelling density ± SEM (arbitrary units). Data were generated in three independent experiments, and the total number of neurons quantified per condition is shown in the figures. Statistical significance was established by one-way ANOVA followed by Sidak's post hoc test. The analysis revealed all ligands induced a significant increase in intracellular labelling as compared to vehicle (P < .001). Results of statistical analysis for comparisons among ligands are shown in B were acquired with a FluoView 1000 confocal laser-scanning microscope (Olympus) using a 60× objective. Gain was set for each independent experiment, using calibration slides. These consisted of vehicle-or SNC-80-treated cultures processed for intracellular labelling. Calibration was performed by adjusting gain to minimize saturation in the internalization slide while still being able to visualize intracellular labelling in the vehicle slide. Once the parameters were set, they were kept constant across all conditions in the same experiment to ensure that differences in labelling density represented differences in receptor density.

| Data analysis
Data analysis was performed using GraphPad 7 (GraphPad Software).
Statistical comparisons are detailed in the text or in figure legends.

| Acute analgesic responses of 'low-internalizing' agonists last longer than those of 'high internalizing' ones
We have previously established that internalization by DOPr agonists could predict the rate of decay of peak signalling in in vitro assays. 36 In particular, when tested in HEK cells agonists like SB235863 and TIPP induced minimal internalization and slow decay of cAMP inhibition. In contrast, SNC-80 and deltorphin II produced marked internalization and fast decay in cAMP inhibition. 36 Here, we wanted to determine whether internalization profiles had any predictive value over the duration of acute analgesic responses. As internalization capacity of different agonists differs in HEK cells and neurons, 37  Treatment with the different agonists caused DOPrs labelled at the neuron surface to translocate and accumulate in the intracellular compartment. As shown in Figure 2, sequestration induced by maximal effective concentrations of SNC-80 and deltorphin II was significantly higher than that induced by SB235863 and TIPP, corroborating that the relative internalization profiles described in HEK cells could also be observed in neurons.
To determine the duration of the analgesic response induced by agonists with different internalization capacity, we measured their ability to reduce mechanical allodynia in a model of diabetic neuropathy that had been previously shown to respond to DOPr agonists. 23 Figure 3E and Table 1), despite considerable differences in potency and maximal effect ( Figure 3F) Figure 4A). Interestingly, this was not the case for the other highly internalizing agonist deltorphin II ( Figure 4C), which maintained its analgesic response over six consecutive days of treatment. Development of analgesic tolerance over repeated administration also differed between the two agonists with low internalization profiles. In particular, the analgesic response to SB235863 was practically abolished over repeated administration ( Figure 4B), whereas analgesia by TIPP was maintained throughout treatment ( Figure 4D). It is also worth noting that chronic tolerance developed at similar rate for SNC-80 (t 1/2 : CI = 1.05-2.19 days) and SB235863 (t 1/2 : CI = 1.69-2.23 days; Figure 4E), independent of their distinct internalization profiles.
Ligand signalling efficacy affects the development of chronic tolerance simply because full agonists require lower occupancy than partial ones to induce analgesia, 38,39 allowing for 'spare receptors'.
Hence, we sought to corroborate development of chronic tolerance without introducing efficacy as confounding factor. For this purpose, we assessed how repeated administration of ED 80 doses of each of the four agonists influenced the analgesic response to a common probe. Deltorphin II, which was used as the common test ligand, was injected at a submaximal dose (ED 50 ) one day prior to the beginning of each treatment and a day after the last injection of each agonist (see Figure 1 for experimental design). By comparing the effect of this fixed dose of deltorphin II before and after each treatment, it was possible to corroborate that SNC-80 ( Figure 5A) and SB235863 ( Figure 5B) induced cross-tolerance to the common probe. On the other hand, chronic treatment with deltorphin II ( Figure 5C) and TIPP ( Figure 5D) was without effect, confirming that internalization profiles are not predictive of chronic tolerance We had previously shown that recycling is essential for maintaining analgesic responses following two consecutive acute injections of DOPr agonists. 19 Hence, we wanted to determine if preservation of analgesic response over repeated treatment was also associated with ligand ability to support recycling. To address this question, we first used a previously validated approach to compare DOPr recycling by different agonists. 19 In particular, primary cortical neuron cultures were transfected with Flag-DOPrs, and on the day of the experiment, receptors were labelled at the cell surface. Cultures were then exposed to different agonists (10 µmol/L, 60 minutes) to induce internalization. indicating that internalized receptors remained trapped within the intracellular compartment even one hour after removal of these agonists from the incubation medium. On the other hand, in cultures exposed to deltorphin II ( Figure 6C) or TIPP ( Figure 6D) mean intracellular labelling intensity was significantly reduced following recovery in the absence of ligand, indicating that receptors internalized by these agonists left the intracellular compartment upon removal of the drug. To corroborate that receptors leaving the cytoplasm were relocated to the membrane, immunoreactivity following recovery was also assessed in non-permeabilized neurons ( Figure 6E-H). Consistent with intracellular labelling patterns, surface labelling after recovery was less for SNC-80 and SB235863 than for deltorphin II and TIPP ( Figure 6I). Thus, taken together, these results show that persistence of analgesic responses upon repeated drug administration persists for recycling ligands independent of the extent of their internalization capacity.

| Recycling is essential for persistence of analgesic response over repeated exposure to DOPr agonists
A major distinction between agonists that induced tolerance over repeated administration versus those that did not is their chemical structure. Indeed, deltorphin II is a naturally occurring opioid peptide 32 and TIPP is an opioid peptide analogue. 33 On the other hand, SNC-80 34 and SB235863 35 are synthetic, non-peptide ligands. Several of the naturally occurring opioid peptides are internalized with the receptor and can be hydrolysed by the endothelin-converting enzyme 2 (ECE2), an endosomal protease that functions at acidic pH. 36 Deltorphin II is one of the ECE2 substrates, and inhibition of the enzyme interferes with recycling of deltorphin II-activated DOPrs. 35 We took advantage of this knowledge to determine whether interfering with recycling had any impact on the development of chronic analgesic tolerance by the different agonists. In a first series of experiments, we assessed whether the ECE2 inhibitor 6634449 had any effect on recycling by the different agonists. We observed that 6634449 was without effect on the redistribution of internalized receptors during recovery from SNC-80 ( Figure 6A,E) or SB235863 ( Figure 6B,F) treatments, but caused those internalized by deltorphin II ( Figure 6C,G) and TIPP ( Figure 6D,H) to remain trapped in the cytosol, blocking their recovery at the cell surface (Figures6G -6H).
Having established that the ECE2 inhibitor blocked recycling supported by deltorphin II and TIPP, we reasoned that if chronic analgesic tolerance was prevented by recycling, then the administration of 6634449 together with these agonists should precipitate tolerance.
Conversely, as the ECE2 inhibitor did not change intracellular accu- or SB235863 was similar to that observed in controls pre-injected with CSF ( Figure 7A,B), and curves representing decay of analgesic response in presence or absence of the ECE2 inhibitor were superimposed ( Figure 7A,B, insets). In particular, analgesic t 1/2 for SNC-80treated rats receiving vehicle was within a 95% CI of 1.3 to 4.9 days,  Figure 7C,D). Thus, whereas analgesic responses by either agonist F I G U R E 6 DOPr agonists have different recycling profiles that are distinctively influenced by ECE2 activity. Primary cortical neuron cultures were transfected with Flag-DOPrs and incubated with primary antibody prior to exposure to either vehicle or indicated agonists (10 µmol/L, 60 min). A-D, At the end of treatment, cells were either immediately processed for intracellular labelling (panels below) or allowed to recover from treatment (60 min; lower panels) in the presence or absence of ECE2 blocker (6634449; 20 μmol/L) as indicated. Histograms below the images show intracellular labelling density ± SEM (arbitrary units) for the total number of neurons quantified per condition immediately after treatment or following recovery in the presence or absence of 6634449 as indicated. Data were generated in three independent experiments. Mean intracellular labelling density following internalization, recovery and recovery in presence of 6634449 were compared for each agonist using one-way ANOVA. Post hoc comparisons using Sidak's test revealed no effect of recovery or of the ECE2 inhibitor for SNC-80 and SB23586. Deltorphin II and TIPP showed both an effect of recovery and of the ECE2 inhibitor, as indicated. E-H, Another set of neurons was treated as above and stripped of all surface labelling before allowing them to recover in the presence or absence of 6634449 as indicated. At the end of recovery, cells were processed for surface labelling. I, Histograms show surface labelling density ± SEM (arbitrary units) in cells that were allowed to recover in the presence or absence of 6634449 following treatment with different agonists. Data were generated in three independent experiments, and the total number of neurons quantified per condition is shown in the figures. Data were analysed with two-way ANOVA to compare membrane labelling following recovery from exposure to different agonists in the presence or absence of ECE2 inhibitor. Analysis showed an effect of agonist (P < .001), an effect of ECE2 inhibitor (P < .001) and an interaction between both factors (P ˂ .001). Post hoc comparisons using Sidak's test revealed an effect of ECE2 inhibitor for deltorphin II and TIPP, P = .001 displayed no measurable decay t 1/2 in animals pre-treated with vehicle, analgesic effects of TIPP and deltorphin II were, respectively, reduced by day 2 (P = .001) and day 4 (P < .02) in animals previously exposed to 6634449 ( Figure 7C,D). In 6634449-treated animals, analgesic t 1/2 for the partial agonist TIPP (95% CI = 0.6-1.4 days) was shorter than that of SNC-80 CI = 1.4-3.1 days; P ˂ .0001 or SB235863 (CI = 1.8-3.8 days; P < .0001), while that of deltorphin II remained ill-defined despite the observed loss of its analgesic effects (Figure 7). ECE2 activity was also essential for preventing TIPP from inducing heterologous tolerance.
Indeed, whereas heterologous tolerance by SNC-80 ( Figure 8A) and SB235863 ( Figure 8B) remained unchanged by 6634449 administration, in rats that received TIPP plus the ECE2 inhibitor, the analgesic response produced by deltorphin II (ED 50 ) was significantly reduced as compared to rats that received TIPP plus vehicle ( Figure 8D). As expected, co-administration of 6634449 with deltorphin II over a six-day period reduced analgesic effect of the drug's ED 50 ( Figure 8C).

| D ISCUSS I ON
In the present study, we used a model of diabetic neuropathy to determine whether ligand-specific trafficking profiles were predictive of DOPr agonist potential to induce analgesic tolerance. We found that ligands that supported receptor recycling to the membrane had . Pressure withdrawal thresholds for mechanical allodynia were measured at 15-min intervals immediately after each administration until return to baseline, and the area under the curve was consigned each day. Results are expressed as mean area ± SEM and correspond to five rats/ group. Statistical comparisons using two-way ANOVA revealed an effect of time for SNC-80 (P ˂ .0001) and SB235863 (P ˂ .0001) but no effect of the ECE2 inhibitor nor interaction. Comparisons for deltorphin II revealed no effect of time, an effect of the ECE2 inhibitor (P ˂ .0001) and no interaction. Post hoc Sidak's comparisons revealed differences indicated in the figure Because of their constitutive interaction with GASP-1, a sorting protein that excludes receptors from the recycling path and directs them to lysosomes, 41,42 DOPrs have been classically considered as committed for degradation. 43 If direct sorting to lysosomes was the only itinerary followed by these receptors, then internalizing ligands would systematically promote degradation of the receptor and induce analgesic tolerance. The highly internalizing agonist SNC-80, whose acute 16  Agonists that do not produce tolerance over repeated administration are highly desirable for chronic pain management. However, DOPr agonists that rely on recycling for sustained analgesic actions are all peptide ligands, 16,17,19,35 and poor biodisponibility and restricted brain penetration represent a clear obstacle for clinical application. Non-peptide DOPr agonists like JNJ-20788560, 12 morphine-6-O-sulphate (M6S) 5 and PN6047 44 induce sustained analgesic response for a similar time period as the recycling peptide agonists assessed in this study, but the mechanism underlying this prolonged analgesia remains to be elucidated. JNJ-20788560 and PN6047 efficacy to induce G protein activation is comparable to that of the full agonist SNC-80. On the other hand, their internalization capacity is markedly less than that of this standard ligand. [44][45][46] As DOPr-βarr interaction is a major determinant of DOPr internalization 1 and given that DOPr recycling requires unstable DOPr-βarr association, a plausible mechanism for these non-peptidic ligands to support recycling is weak interaction between the two proteins which interferes with maximal internalization but simultaneously promotes recycling.
In summary, the study provides evidence that DOPr agonists that support receptor recycling to the membrane can produce sustained analgesic responses over repeated administrations.

This work was supported by Natural Sciences and Engineering
Research Council of Canada Grants (No. 311997 to G.P) and the Canadian Institutes of Health Research (MOP 79432 and MOP 324876 to G.P).

CO N FLI C T O F I NTE R E S T
None.

AUTH O R S ' CO NTR I B UTI O N S
HBT and YH have tested mechanical allodynia in rats and HBT interpreted the results. HBT and IC performed the primary neuronal cultures. IC did the quantification of DOPr trafficking in primary neuron cultures. GP was responsible for the supervision of project and wrote the manuscript with HBT and RC.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.