KUNA115 (Mishra et al., under review), KUNB106 (Mishra et al., in press), KUNG65 (compound 30 in Crowley et al., 2017), KU-32 (compound A4 in Ansar et al., 2007), and KU177 (compound 12c in Zhao et al., 2011) were synthesized by the Blagg laboratory using the cited protocols. The identity of the ligands was confirmed by high resolution mass spectrometry and nuclear magnetic resonance, while the purity of the compounds was confirmed to >95% by high performance liquid chromatography. Gedunin (#33-871-0), Celastrol (#32-031-0), 17-AAG (#AAJ66960MC), and Cycloheximide (#AC357420010) were obtained from Fisher Scientific. DAMGO (#1171) was obtained from Tocris/R&D. Morphine sulfate pentahydrate was obtained from the NIDA Drug Supply Program. All compounds except for DAMGO and morphine were prepared as DMSO stock solutions and diluted into a vehicle solution prior to injection. DAMGO was prepared in a stock solution of sterile USP water and morphine in sterile USP saline; morphine was prepared fresh prior to each experiment. Matched vehicle controls were included for each drug injection. The vehicles used were: 2% DMSO and 98% sterile USP water for KUNA115, KUNB106, KUNG65, KU177, Cycloheximide; 1% DMSO and 99% sterile USP water for KU-32 and 17-AAG; 10% DMSO, 10% Tween80, and 80% sterile USP water for gedunin and celastrol; sterile USP water for DAMGO; and sterile USP saline for morphine. Drug powders were stored at −20°C under desiccation or as recommended by the manufacturer, and stock solutions were stored at −20°C.

CRISPR gene editing constructs were obtained from Genecopoeia as all-in-one DNA vectors containing universal promoters driving expression of the gRNA and Cas9 gene, along with a neomycin resistance gene for mammalian cell selection and an mCherry gene for visualization (pCRISPR-CG vector). The constructs were pre-designed by Genecopoeia to target each mouse gene. They included a universal negative control vector that expresses all the same elements with a non-targeting gRNA (#CCPCTR01-CG01-B), and constructs to target Hsp90α (#MCP229411-CG01), p23 (#MCP232080-CG12), Cdc37 (#MCP231406-CG12), and PEBP1 (#MCP231756-CG01) as a further negative control.

Each DNA vector was amplified for use using standard bacterial transformation, and an endotoxin-free maxi-prep kit to reduce inflammation upon injection. Each vector was also validated by restriction digest. The Hsp90α vector was validated in an in vitro experiment. Mouse 66.1 breast cancer cells were cultured as described in Edwards et al. (2018). The cells were electroporated with 10 μg of DNA per cuvette, then selected with 500 μg/ml of G418 until the cells recovered and began growing again. At this point, the cells were harvested and analyzed by Western blot as described below. The in vivo delivery and validation of all vectors are also described below.

Male and female CD-1 (a.k.a. ICR) mice from 4 to 8 weeks of age were used for all experiments and were obtained from Charles River Laboratories. The mice were recovered for at least 5 days after shipping prior to use and housed no more than five per cage. All mice were housed in the University of Arizona’s AAALAC-accredited vivarium with temperature and humidity control, 12 h light/dark cycles, and standard chow and water available ad libitum. All experiments performed were approved by the University of Arizona’s IACUC, and all experiments were in accordance with the NIH Care and Use of Laboratory Animals handbook.

Mice were randomly assigned to experimental groups in age-matched cohorts, and the experimenter was blinded to treatment group identity by the delivery of coded drug vials. We utilized a post-surgical paw incision pain model, with the surgery performed as described in our earlier work (Lei et al., 2017). Drugs or CRISPR DNA constructs were delivered by the intracerebroventricular (icv) route in a 5 μl volume, also performed as described in Lei et al. (2017). For drug treatments, the paw incision surgery was performed, and while the mice were still under anesthesia, an icv injection of inhibitor drug or vehicle was performed. The mice then recovered from both the surgery and injection for 1 or 24 h prior to opioid injection and pain measurement. For CRISPR experiments, 4 μg of DNA was complexed with Turbofect in vivo transfection reagent (#FERR0541 from Fisher Scientific) according to the manufacturer’s instructions and injected icv daily from days 1 to 3. The mice then recovered, with the paw incision surgery performed on day 9 and opioid injection and pain measurement performed on day 10. Our in vivo CRISPR protocol is based on the protocol reported in Sandweiss et al. (2017).

Mechanical pain/allodynia on the incised paw was measured using Von Frey filaments with the up-down method, as performed in our earlier work and the literature (Chaplan et al., 1994; Lei et al., 2017; Edwards et al., 2018). Pre- and post-CRISPR and pre- and post-surgical baselines were measured to determine any impact of the treatment on baseline responses prior to the injection of morphine. Mechanical allodynia was measured in a 2–3 h time course after the injection of morphine.

To analyze brain signaling changes, mice were injected icv with KUNA115 as above for 24 h, followed by icv injection of DAMGO for 10 min. The mice were sacrificed by rapid cervical dislocation, and the periaqueductal gray (PAG) region was rapidly dissected on an ice-cooled metal block and snap-frozen in liquid nitrogen. G418-selected populations of 66.1 cells transfected with CRISPR constructs described above were also harvested for protein analysis. The cells were washed with ice-cold dPBS and the cells recovered by adding lysis buffer and scraping with a plastic cell spatula. The methods for protein extraction from both brain regions and cell lysates, the composition of our lysis buffer, and the protocol for performing the Western blot are all reported in our earlier work (Lei et al., 2017).

We used the following antibodies for our Western analysis: phospho-Akt (#50-191-224, Fisher Scientific); total-Akt (#50-190-279, Fisher Scientific); phospho-ERK (#50-191-932, Fisher Scientific); total-ERK (#50-191-129, Fisher Scientific); phospho-JNK (#9255, Cell Signaling); total-JNK (#9252, Cell Signaling); Hsp70 (#4872, Cell Signaling); STAT3 (#9139, Cell Signaling); GAPDH (#PIMA515738, Fisher Scientific); Hsp90α (#MA110892, Fisher Scientific); and mCherry (#NBP196752SS, Fisher Scientific). The antibodies were generally used at 1:1,000 in 5% BSA in TBST rocking overnight at 4°C. We used goat anti-rabbit or goat anti-mouse IRDye 680 or 800 secondary antibodies from LiCor Biosciences at 1:5,000–10,000 in 5% non-fat dry milk in TBST at room temperature for 1 h. The resulting data was imaged using an Odyssey Fc imager from LiCor Biosciences. The data were quantitated using Scion Image, derived from NIH ImageJ. Phospho-protein signal was normalized to total protein signal from the same sample (e.g., pERK normalized to tERK), while total protein signal was normalized to the housekeeping gene GAPDH (e.g., STAT3 normalized to GAPDH). These normalized data were further normalized to the vehicle-treated control animals within each experiment.

CRISPR-mediated knockdown in the brain of the target proteins Hsp90α, p23, and Cdc37 was validated post hoc in treated mice as above using immunohistochemistry. The mice were first perfused with 4% paraformaldehyde in saline, and the brains removed and frozen as a block in OTC medium. The brains were sectioned using a cryotome with 20 μm sections and mounted on Leica Xtra slides (#NC0215141, Fisher Scientific); the frozen sections were dried at room temperature for 15 min and stored at −20°C until use.

For Hsp90α: sections warmed to room temperature, then washed in TBS (20 mM Tris, 150 mM NaCl, pH 7.2) for 10 min. Blocked for 2 h in a humidified chamber at room temperature (10% goat serum, 0.3% Triton-X100 in TBS). Incubated with primary antibody (#380-003, Synaptic Systems) at 1:50 in 5% goat serum, 0.3% Triton-X100 in TBS overnight at 4°C. Washed 3 × 10 min in TBS, followed by anti-rabbit Alexa 488 secondary antibody (#A11034, Fisher Scientific) at 1:200 in 5% goat serum, 0.3% Triton-X100 in TBS for 1 h at room temperature. The slides were then washed 3 × 10 min in TBS, dried at room temperature for 10 min, then mounted.

For p23 and Cdc37: sections warmed to room temperature, then washed in phosphate buffered saline (PBS) for 10 min. Heat-induced antigen retrieval performed for 20 min at 95°C in sodium citrate buffer (10 mM sodium citrate, 0.05% Tween20, pH 6.0). The slides were cooled to room temperature for 20 min, then washed in PBS with 0.1% Tween20 (PBST) twice, and PBS once. The sections were then blocked for 30 min in a humidified chamber at room temperature [3% fetal bovine serum (FBS) in PBS]. After blocking, the sections were incubated with primary antibody overnight at 4°C (p23-1:1,000 of #MA3414 from Fisher Scientific; Cdc37-1:100 of #MA3029 from Fisher Scientific; both in 3% FBS in PBS). The slides were then washed 3 × 10 min in PBST, followed by secondary antibody at room temperature (p23-1:200 of anti-mouse Alexa594, #A11032 from Fisher Scientific, in 3% FBS in PBS for 1 h. Cdc37-1:200 of anti-mouse Alexa488, #A11031 from Fisher Scientific, in 3% FBS in PBST for 30 min). The slides were then washed 3 × 10 min in PBST and dried at room temperature for 10 min before mounting.

All sections were imaged using a standard fluorescent microscope using the appropriate filters for Alexa488 (blue/green) and Alexa594 (green/red). The knockdown of all targets was broadly apparent across the brain. The pontine reticular nucleus (PRN) was chosen as the site of imaging and quantitation. Images were taken from 4 to 6 adjacent sections from each animal. The fluorescence intensity divided by the area of the image was calculated for each section, and the 4–6 sections per animal averaged to produce a single mean value counted as N = 1.

All data is reported as the mean ± SEM. All statistical analysis was performed using GraphPad Prism 8.0. All behavioral data is reported as raw threshold values without normalization; Western blot and immunohistochemistry data normalized as described in those sections above. Statistical comparisons for behavioral data and ERK/JNK/Akt Western data performed using a two-way ANOVA with Fisher’s Least Significant Difference post hoc test. Comparisons of immunohistochemistry data and Hsp70/STAT3 Western data performed using an Unpaired 2-Tailed t-test. In all cases significance was set as a p-value of <0.05. For the dose/response experiment using KUNA115, the area under the curve (AUC) was calculated for each dose and treatment using Prism 8.0, and graphed by log dose and treatment group. Linear regression analysis was performed, and the parameters of those fitted lines used to calculate the potency (A₅₀) as previously described (Lei et al., 2017). The sample sizes and technical replicates for each experiment are described in the Figure Legends.

To identify the active Hsp90 isoform in regulating opioid anti-nociception in the brain, we performed a screen of isoform-selective small molecule inhibitors. KUNA115 is selective for Hsp90α (Mishra et al., under review), KUNB106 for Hsp90β (Mishra et al., in press), and KUNG65 for Grp94 (Crowley et al., 2017). All inhibitors were delivered at a screening dose of 0.1 nmol by the icv route for 24 h, a model predicated on our earlier studies with the non-selective Hsp90 inhibitor 17-AAG (Lei et al., 2017). Over a full morphine dose range of 1–10 mg/kg, we found that KUNA115 strongly blocked morphine anti-nociception in paw incision pain, suggesting the involvement of the isoform Hsp90α (Figure 1A). Dose/response analysis for this experiment revealed an A₅₀ potency value of morphine of 8.95 mg/kg for Vehicle-treated mice, in line with literature values, validating the experiment (Figure 1B). The dose/response curve for KUNA115 meanwhile was so flat it did not give a feasible value (calculated A₅₀ of 88, 139, 382 mg/kg; Figure 1B). These results show that an Hsp90α-selective inhibitor strongly blocked morphine anti-nociception in this pain model in line with our previous results using a non-selective inhibitor (Lei et al., 2017). We also found a similar result with female mice, suggesting no sex differences with this target and model (Figure 1C).

When we tested the other isoform-selective inhibitors, we found no differences for KUNB106 (Hsp90β; Figure 1D) and KUNG65 (Grp94; Figure 1E). These results do suggest that Hsp90α alone is active in the brain for opioid signaling regulation. As a further control, we tested the impact of an alternate site inhibitor KU-32, which binds to the C-terminal region of Hsp90 unlike the ATP pocket targeted 17-AAG but is similarly non-selective between isoforms (Ansar et al., 2007). The results with KU-32 are the same as for KUNA115 above and 17-AAG (Lei et al., 2017), further validating the results and suggesting a bona fide role for Hsp90 in regulating opioid anti-nociception (Figure 1F).

We also performed additional experiments to further define the impact of KUNA115/Hsp90α on opioid anti-nociception. All experiments above were carried out with a 24 h recovery, leaving the time course of KUNA115 unknown. We thus performed a paw incision experiment as above with only a 1 h KUNA115 treatment; this resulted in a strong loss of morphine anti-nociception, similar to the 24 h results above (Figure 1G). This finding suggests that KUNA115 has a relatively rapid onset that is sustained for 24 h or more. We also controlled for potential impacts of KUNA115 on mechanical thresholds without pain or opioids present. We found that a 24 h KUNA115 treatment as above had no impact on baseline mechanical thresholds, suggesting the results above are due to a specific impact on the opioid system (Figure 1H). This conclusion is further supported by our earlier work in which we found no impact of Hsp90 inhibitor treatment on motor performance in the Rotarod test (Lei et al., 2017).

We next sought to measure the impact of Hsp90α-selective inhibition on opioid signaling. We combined KUNA115 treatment with DAMGO stimulation in the brain, which is a highly selective, potent, and efficacious MOR agonist. We analyzed ERK and JNK MAPK, Akt, Hsp70, and STAT3 by Western blot in the PAG region of the brain (Figure 2A). This region was chosen based on our earlier studies using non-selective inhibitors (Lei et al., 2017); the PAG is also a key region in the pain modulatory circuitry (Heinricher et al., 2009). We found that both ERK and JNK MAPK phosphorylation was stimulated by DAMGO in Vehicle-treated mice, however, stimulation over baseline was lost with KUNA115 treatment (Figure 2B). With JNK MAPK, KUNA115 significantly raised the unstimulated baseline, as we saw for ERK MAPK with 17-AAG treatment (Lei et al., 2017). With the kinase Akt, KUNA115 treatment tended to increase both unstimulated and DAMGO-stimulated phospho-Akt levels, so that the KUNA115/DAMGO group was significantly elevated over Vehicle/Vehicle baseline (Figure 2B). Lastly, KUNA115 treatment had no impact on Hsp70 protein levels, unlike 17-AAG treatment (Lei et al., 2017), while it significantly decreased total protein levels of the signaling regulator STAT3 (Figure 2B). Together these results demonstrate that Hsp90α has a specific role in regulating opioid signal transduction in the brain suggesting potential involvement in opioid anti-nociception, similar to what we showed for 17-AAG treatment and ERK MAPK (Lei et al., 2017).

To confirm the role of Hsp90α in regulating opioid anti-nociception, we used CRISPR/Cas9 DNA constructs to knockdown Hsp90α protein expression broadly across the brain in adult mice (similar approach to Sandweiss et al., 2017). We first validated our CRISPR construct in vitro using mouse 66.1 cells. Transfection of a negative control CRISPR vector or a vector targeting the protein PEBP had no impact on Hsp90α protein levels, while our vector targeting Hsp90α reduced protein levels by ~90% (Figure 3A). We could also detect expression of the mCherry protein in all cells with a CRISPR vector, verifying successful transfection of all constructs (Figure 3A).

We next treated mice with the Hsp90α or negative control CRISPR vectors as described in the “Materials and Methods” section, and validated successful target knockdown by immunohistochemistry. We found broad knockdown across the entire brain and selected the PRN and PAG for analysis. We could detect a strong signal in both cell bodies and apparent dendritic fields that was strongly reduced by Hsp90α CRISPR treatment (Figure 3B). Quantitation of fluorescent signal revealed a significant reduction of 43.9% (Figure 3C). Thus validated, we next tested the impact of Hsp90α knockdown on morphine anti-nociception (Figure 3D). We found that Hsp90α CRISPR treatment fully blocked anti-nociception in paw incision pain (Figure 3D), very similarly to KUNA115 (Figure 1) and 17-AAG (Lei et al., 2017), confirming the role of Hsp90α in regulating opioid anti-nociception in the brain.

We first used the co-chaperone-selective inhibitors gedunin (p23, Brandt et al., 2008), celastrol (Cdc37, Zhang et al., 2008), and KU177 (Aha1, Zhao et al., 2011) in paw incision pain as for the isoform inhibitors above. Both gedunin (p23; Figure 4A) and celastrol (Cdc37; Figure 4B) strongly reduced morphine anti-nociception in paw incision pain, very similar to the Hsp90 inhibitors above. However, KU177 (Aha1) had only a slight impact on opioid anti-nociception, suggesting it may not have a significant role (Figure 4C).

We next moved forward with CRISPR/Cas9 gene editing to confirm that p23 and Cdc37 regulate opioid anti-nociception as we did for Hsp90α above. IHC analysis showed a similar broad knockdown across the brain with CRISPR treatment, particularly apparent in the PRN (Figure 4D). Quantitation revealed significant decreases of 36.3% for p23 and 46.0% for Cdc37 (Figure 4E). We next tested the impact of targeted CRISPR treatment for these proteins in paw incision pain, and found that both p23 and Cdc37 CRISPR knockdown fully blocked morphine anti-nociception, very similar to Hsp90α inhibition above or 17-AAG treatment (Lei et al., 2017; Figure 4F). These results confirm that both p23 and Cdc37 regulate opioid anti-nociception in the brain.

Lastly, we sought to identify part of the molecular biology mechanism by which Hsp90 regulates opioid anti-nociception. We combined 24 h Hsp90 inhibition as above and in our previous work (Lei et al., 2017) with treatment of the translation inhibitor cycloheximide in the brain 30 min prior to morphine treatment. We found that cycloheximide had no impact on morphine anti-nociception in vehicle-treated mice; however, cycloheximide fully restored morphine anti-nociception back to vehicle-treated levels in mice treated with the pan-Hsp90 inhibitor 17-AAG (Figure 5A). We found the same results with KUNA115, confirming that Hsp90α regulates translation during morphine anti-nociception (Figure 5B). These results strongly suggest that rapid translation within 30 min of morphine treatment is required for Hsp90 inhibition to impact opioid signaling and anti-nociception.