Endoplasmic Reticulum Stress Contributes to Nociception via Neuroinflammation in a Murine Bone Cancer Pain Model

Bone cancer pain is a severe complication of metastatic or advanced malignancy, which is characterized by allodynia and hyperalgesia.¹  Bone cancer pain can compromise the patient’s quality of life and impose a heavy burden on society.²  To date, the mechanism underlying bone cancer pain has not been completely understood. As a result, existing pharmacologic agents cannot relieve pain efficiently and satisfactorily.^3,4 

As a protective mechanism, endoplasmic reticulum stress is responsible for folding and trafficking of proteins under a series of stress conditions.⁵  Endoplasmic reticulum stress has been indicated to be involved in numerous neurodegenerative diseases.^6–8  Activation of the unfolded protein response by accumulation of unfolded proteins in endoplasmic reticulum initiates three endoplasmic reticulum stress pathways, which are mediated by RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α, inositol-requiring enzyme 1, and the activating transcription factor 6.⁹  Endoplasmic reticulum stress has been proposed to be involved in pain development. All pathways of endoplasmic reticulum stress have been found to be activated in the peripheral nervous system in neuropathic and inflammatory pain.^10,11  Activated RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α in astrocytes and activated activating transcription factor 6 in neurons were found in the central nervous system in neuropathic pain.¹²  However, the role of endoplasmic reticulum stress in the development of bone cancer pain has still not been fully elucidated.

Various models have been proposed to describe the contribution of inflammation to pain. The production of TNF-α and interleukin 1β promotes the development of allodynia in peripheral nerve injury¹³  and spinal cord injury.¹⁴  Neuroinflammation drives widespread chronic pain via central sensitization.¹⁵  Our previous studies also confirmed the upregulation of TNF-α, interleukin 1β, and interleukin 6 in the spinal cord during bone cancer pain.¹⁶  Several studies have indicated that inflammatory mediators serve as the inducers of endoplasmic reticulum stress. TNF-α stimulation of synovial fibroblasts increased the expression of endoplasmic reticulum stress markers.¹⁷  Similarly, the increased levels of inflammatory mediators TNF-α and interleukin 1β in rheumatoid arthritis resulted in an increase in endoplasmic reticulum stress in dendritic cells, fibroblast-like synoviocytes, T cells, and B cells.¹⁸  On the basis of these findings, we assumed that the production of inflammatory mediators in the spinal cord during bone cancer pain serves as a trigger for endoplasmic reticulum stress.

Endoplasmic reticulum stress affects various cell signaling processes, including apoptosis¹⁹  and inflammation. The crosslink between inflammation and endoplasmic reticulum stress has been proposed in recent studies. Three unfolded protein response signaling pathways are related to the production of inflammatory mediators via the activation of the transcription factor nuclear factor-κB, one of the key modulators in inflammatory gene transcription.²⁰  Inhibition of RNA-dependent protein kinase-like endoplasmic reticulum kinase reduced the expression of interleukin 6, chemokine ligand 2, and chemokine ligand 20 in astrocytes.²¹  Additionaly, endoplasmic reticulum stress increased the production of interleukin 1β, interleukin 6, and interleukin 23 in response to lipopolysaccharide.²²  However, previous studies have not determined whether neuroinflammation could be modulated by endoplasmic reticulum stress in bone cancer pain.

In this study, we tested the hypothesis that endoplasmic reticulum stress could be triggered by the production of inflammatory mediators in the spinal cord and would exacerbate the development of mouse bone cancer pain. Inhibition of endoplasmic reticulum stress by the inhibitors 4-PBA and GSK2606414 downregulated the neuroinflammation in the spinal cord and improved the nociceptive behaviors in bone cancer pain mice.

Male C3H/HeN mice (20-25 g, 5 weeks of age; Vital River Experimental Animal Corporation of Beijing, China) were used. Mice were housed with free access to water and food and in a 12-h light–dark cycle. All experiments were carried out between 9:00 am and 5:00 pm. All experiments were performed in strict accordance with the guidelines and approved by the Animal Care and Use Committee of the Medical School of Nanjing University (Nanjing, China).

NCTC 2472 osteolytic sarcoma cells were cultured in NCTC 135 medium (Sigma, USA) with 10% horse serum (Gibco, USA) and maintained in a 5% CO₂ atmosphere at 37°C (Thermo Forma, USA).

The establishment of the bone cancer pain model was described previously by °Schwei et al.²³  Animals were anesthetized intraperitoneally with pentobarbital sodium at a dose of 50 mg/kg. Right knee arthrotomy was performed after general anesthesia. Next, 20 μl of α–minimum essential medium (Thermo Fisher Scientific, USA) containing 2 × 10⁵ osteolytic sarcoma cells was injected into the intramedullary space of the right femur, whereas the sham group received injections of 20 μl of α–minimum essential medium alone. After sealing the drill hole with bone wax, the wound was closed with 4-0 silk sutures (Ethicon, USA). Animals recovered from anesthesia on a heated blanket.

Pregnant B6 mice were anesthetized with isoflurane, and the fetuses were removed on embryonic day 14. A microscope was used to remove the meninges and blood vessels of the embryonic spinal cord. After digestion with 0.05% trypsinase at 37°C for 20 min, spinal tissues were dissociated softly 10 times. The supernatant was sieved with a 70-μm cell strainer (Falcon, USA) and centrifuged for 5 min at 1,000 rpm. The cells were resuspended in Dulbecco’s Modified Eagle Medium (Biological Industries, USA) containing 10% fetal bovine serum (Gibco), 2 mM glutamine (Sigma), 25 mM glucose, and 1% penicillin/streptomycin (Gibco) and plated onto poly-L-lysine (Sigma, USA)–coated six-well plates at the density of 1 × 10⁶/cm². The medium was completely changed to neurobasal medium (Gibco) containing 2% B27 (Gibco), 2 mM glutamine, and 10-μl/ml penicillin/streptomycin 4 h later. Cells were cultured at 37°C in a 5% CO₂ atmosphere (Thermo Forma). One-half medium change was performed on day 2, and cells were allowed to grow for 7 to 9 days.

In in vivo experiments, recombinant murine TNF-α (Peprotech, USA), lipopolysaccharides (lipopolysaccharide, Sigma), 4-phenylbutyrate (4-PBA, Sigma), and GSK2606414 (1-[5-(4-Amino-7-methyl-7H-pyrrolo[2,3-d]pyrimidin-5-y)-2,3-dihydro-1H-indol-1-yl]-2-[3-(trifluoromethyl)phenyl]ethanone; Tocris, United Kingdom) were dissolved in normal saline before administration. TNF-α (5 ng/5 μl) and lipopolysaccharide (100 ng/5μl) were intrathecally injected in control mice. On the basis of previous studies, 4-PBA was intraperitoneally (1 mg/100 μl) and intrathecally (40 μg/5 μl, 80 μg/5 μl, 120 μg/5 μl) administered to bone cancer pain mice on day 14 after the operation. Meanwhile, GSK2606414 was intrathecally administered to bone cancer pain mice on day 14 after the operation at a dose of 50 μg/5 μl, 100 μg/5 μl, and 200 μg/5 μl on the basis of a previous study. The intrathecal injection was performed as previously described by Hylden and Wilcox.²⁴ 

In in vitro experiments, TNF-α and lipopolysaccharide were dissolved in culture media at a dosage of 100 nM. Primary spinal neurons were treated with TNF-α and lipopolysaccharide for 12 h and 24 h, respectively.

Mechanical allodynia and spontaneous pain in mice were tested before operation (day 0) as well as 4, 7, 10, 14, 21, and 28 days after operation in each group, 0, 1, 3, 6 h after administration of recombinant murine TNF-α, lipopolysaccharide, and vehicle, and 0, 1, 2, 3, 4, 5, 7, 10 h after administration of 4-PBA, GSK2606414, and vehicle. Experimenters of all behavioral tests were blinded to the group assignment data.

Paw withdrawal mechanical thresholds of the right hind paw were measured using von Frey filaments (0.16, 0.4, 0.6, 1.0, 1.4, 2.0 g; Stoelting, USA) and the up-down method as previously reported.²⁵  Mice were placed in transparent plexiglass compartments with a wire mesh bottom for a 30-min acclimatization period, after which the von Frey filaments were stuck upright to the plantar surface and the lowest filament stimulus strength that caused paw flinching or withdrawal was recorded.

Mice were placed in transparent plexiglass compartments with a wire mesh bottom and acclimatized for a 30-min acclimatization period, which was followed by observation of the number of flinching episodes of the right hind paw more than 2 min. Each mouse was tested five times.

After deeply anesthetizing the mice with pentobarbital (50 mg/kg, intraperitoneally), mice were euthanized at days 0, 4, 7, 10, 14, 21, and 28 after operation and 3 h after administration. The spinal tissues were removed and stored at -80°C for further study. Samples of spinal tissue or cells were homogenized in Radio Immunoprecipitation Assay Lysis Buffer (10 μl/mg for tissue, Beyotime Biotechnology, China) with phenylmethyl sulfonyl fluoride and rested on ice for 30 min. Next, the samples were centrifuged at 12,000 rpm, 4°C for 20 min and the supernatant was collected. The concentration of each protein sample was tested by the Bicinchonininc Acid Protein Assay Kit. Protein samples were subjected to sodium dodecylsulphate - polyacrylamide gel electrophoresis and then transferred onto a polyvinylidene fluoride membrane. The membranes were incubated with primary antibodies for binding immunoglobulin protein (BIP; 1:1,000, CST, USA), RNA-dependent protein kinase-like endoplasmic reticulum kinase (1:200, Santa Cruz, USA), p-RNA-dependent protein kinase–like endoplasmic reticulum kinase (1:200, Santa Cruz), eukaryotic initiation factor 2α (1:200, Santa Cruz), p-eukaryotic initiation factor 2α (1:200, Santa Cruz), inositol-requiring enzyme 1α (1:200, Santa Cruz), p-inositol-requiring enzyme 1α (1:1000, Abcam, USA), activating transcription factor 6α (1:200, Santa Cruz), TNF-α (1:200, Santa Cruz), interleukin 1β (1:1,000, Abcam), interleukin 6 (1:1,000, Abcam), and β-actin (1:4,000, Abcam). After incubation with the secondary antibody (1:10,000, Millipore, USA) and electro chemi luminescence solution, images were captured and analyzed using a cooled charge coupled device system (Tanon, China).

Total RNA was isolated from mouse spinal cord by Trizol reagent (Invitrogen, USA), and messenger RNA was reverse transcribed into cDNA by using the HiScript II First Strand cDNA Synthesis Kit (Vazyme, China). All reverse transcription reactions were performed with cDNA and ChamQ Universal SYBR qPCR Master Mix (Vazyme, China) and run in an ABI StepOnePlus Real-Time PCR System (Applied Biosystems, USA). The following primers were used in this study: TNF-α (F): CGAGTGACAAGCCTGTAGCCC, TNF-α (R): GTCTTTGAGATCCATGCCGTTG; interleukin 1β (F): TCAGGCAGGCAGTATCACTCA, interleukin 1β (R): GGAAGGTCCACGGGAAAGAC; interleukin 6 (F): ACA ACCACGGCCTTCCCTAC, interleukin 6 (R): TCTCATTTCCACGATTTCCCAG; GAPDH (F): GGAAAGCTGTGGCGTGAT, and GAPDH (R): AAGGTGGAAGAATGGGAGTT.

After general anesthesia described above, the animals were transcardially perfused with normal saline and 4% paraformaldehyde at day 14 after operation and 3 h after intrathecal administration. The lumbosacral enlargements were removed and fixed in 4% paraformaldehyde for 6 h, and then dehydrated in 30% sucrose for 48 to 72 h at 4°C. Cells were fixed with methanol. After freezing with the optimal cutting temperature compound, tissues were cut into 20-μm sections by using a freezing microtome. After washing with phosphate buffered saline, the sections and cells were blocked with 10% goat serum containing 0.3% Triton and incubated with primary antibodies for BIP, p-RNA-dependent protein kinase–like endoplasmic reticulum kinase, p-eukaryotic initiation factor 2α, GFAP (mouse, 1:100, Cell Signaling Tech, USA), ionized calcium binding adapter molecule 1 (mouse, 1:300, Abcam), neuronal nuclei (mouse, 1:1,000, Abcam), and GAD65 plus GAD67 (rabbit, 1:100, Abcam) separately overnight at 4°C. After washing with phosphate buffered saline, sections were incubated with Alexa 488-conjugated goat anti-rabbit (1:3,000, ThermoFisher) and Alexa 594-conjugated goat anti-mouse (1:3,000, ThermoFisher) secondary antibodies. Next, the sections were transferred on slides and incubated with DAPI (Abcam). Images were captured using a laser-scanning confocal microscope (Olympus, Japan).

The femur tissue was fixed with 10% paraformaldehyde and decalcified in EDTA decalcification solution for 1 to 2 weeks. After dehydration, the tissue was embedded in paraffin and sliced into 5-μm sections. After staining with hematoxylin and eosin reagents, images were captured using a laser-scanning confocal microscope (Olympus).

After anesthetization, mice were perfused with normal saline and 2.5% glutaraldehyde. The lumbosacral enlargements were removed and kept in 2.5% glutaraldehyde at 4°C Cells were collected 12 h after TNF-α stimulation or 24 h after lipopolysaccharide stimulation and fixed in 2.5% glutaraldehyde at 4°C. Samples were fixed in 1% osmium tetroxide and dehydrated in graded ethanol. After embedding in epoxy resin, the samples were cut into 60-mm sections and stained with uranyl acetate and lead citrate. Images were captured using Hitachi 7100 electron microscopy.

The number of animals used in each study was based on our previous experience with this design.²⁶  No a priori statistical power calculation was used to guide sample size. In response to peer review, sample sizes were increased to six in the Western blotting, quantitative polymerase chain reaction, and immunofluorescence. We randomized animals to the different treatment groups and blinded the experimenter to the drug treatment to reduce selection and observation bias. None of the variables had missing data, and no outliers appeared in our study. Statistical analysis was performed using SPSS 22.0 (IBM Corporation, USA). Data are presented as mean ± SD. Results from the behavioral study were analyzed using two-way repeated measurements ANOVA followed by post hoc tests (Bonferroni test) to assess differences at each time point between groups and one-way ANOVA followed by Bonferroni test to assess differences over time within groups. Results from western blotting and quantitative polymerase chain reaction were analyzed using one-way ANOVA followed by Bonferroni test or independent t tests for between-group comparisons. Normal distribution assumption was analyzed using Q-Q plots. All tests were two-tailed, and P < 0.05 was considered as the level of significance.

Intrafemur implantation of NCTC 2472 sarcoma cells was performed to establish mouse bone cancer pain model. Changes in nociceptive behaviors were measured by determining the paw withdrawal mechanical threshold (fig. 1A) and number of spontaneous flinches (fig. 1B) on days 0 (baseline), 4, 7, 10, 14, 21, and 28 after the operation. The baseline paw withdrawal mechanical threshold and number of spontaneous flinches did not differ significantly between sham group and tumor group (P > 0.05, n = 8, respectively). A statistically significant reduction in paw withdrawal mechanical threshold and a statistically significant increase in number of spontaneous flinches were found on day 4 after the operation in comparison with the baseline both in the sham group (baseline vs. day 4: P = 0.003 in paw withdrawal mechanical threshold; P < 0.001 in number of spontaneous flinches; n = 8) and the tumor group (baseline vs. day 4: P = 0.002 in paw withdrawal mechanical threshold; P < 0.001 in number of spontaneous flinches; n = 8). The acute hyperalgesia may have resulted from the surgery since the nociceptive behaviors recovered from day 7 in the sham group. In the tumor group, the paw withdrawal mechanical threshold decreased from day 10 (P = 0.003 on day 10; P = 0.001 on day 14; P < 0.001 on day 21; P = 0.001 on day 28; n = 8), whereas the number of spontaneous flinches persistently increased for 28 days (P < 0.001, n = 8, respectively). The paw withdrawal mechanical threshold and number of spontaneous flinches both showed statistically significant differences compared with those in the sham group on days 10, 14, 21, and 28 after the operation (P < 0.001, n = 8, respectively). Specific descriptive statistics (mean ± SD) of behavior tests were shown in Supplemental Digital Content, table 1 and table 2 (https://links.lww.com/ALN/C138). Hematoxylin and eosin staining of the femur slices showed infiltration of tumor cells in the marrow cavity, discontinuous bone trabeculae, and destruction of bone cortex in bone cancer pain mice on day 14 after the operation (fig. 1C).

The role of endoplasmic reticulum stress in inflammatory pain and neuropathic pain has been proposed. To determine whether endoplasmic reticulum stress was involved in the mouse bone cancer pain model, several endoplasmic reticulum stress-associated proteins were examined (fig. 2A). As the marker of initiation of endoplasmic reticulum stress, the expression level of BIP (GRP76) was elevated on days 10 (P < 0.001, n = 6), 14 (P < 0.001, n = 6), 21 (P <0.001, n = 6), and 28 (P = 0.029, n = 6) in the tumor group compared with baseline (day 0; fig. 2B). The phosphorylation of RNA-dependent protein kinase-like endoplasmic reticulum kinase (fig. 2D) and eukaryotic initiation factor 2α (fig. 2C) increased briefly on day 4 (P < 0.001, n = 6, respectively) and persistently from day 14 after the operation in the tumor group (P < 0.001, n = 6, respectively). The phosphorylation of inositol-requiring enzyme 1α (fig. 2F) increased only on day 28 after the operation in the tumor group (P < 0.001, n = 6). No differences were found in activating transcription factor 6α expression during the postoperative period (fig. 2E, P > 0.05, n = 6, respectively). In the sham group, the expression of endoplasmic reticulum stress markers showed no statistically significant differences during the postoperative period (fig. S1, P > 0.05, n = 6, respectively). Because the increased expression level of p-RNA-dependent protein kinase–like endoplasmic reticulum kinase and p-eukaryotic initiation factor 2α were maintained from day 14 to day 28, whereas the expression of p-inositol-requiring enzyme 1α increased just on day 28, we considered the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway to be a key modulator in mouse bone cancer pain model. Next, electron microscopy was used to evaluate ultrastructural changes in endoplasmic reticulum on day 14 after the operation (fig. 2G). The cisternae of endoplasmic reticulum swelled obviously in the tumor group, compared with the normally narrow endoplasmic reticulum cisternae in the sham group. Also, there were more dark particles in the tumor group, which might be accumulated vesicles for degradation, and further study was required. These data suggest that activation of endoplasmic reticulum stress in mouse bone cancer pain model mainly depends on the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway.

We next assessed the cellular localization of p-RNA-dependent protein kinase–like endoplasmic reticulum kinase and p-eukaryotic initiation factor 2α expression. Double immunofluorescence staining was performed. The results showed substantial p-RNA-dependent protein kinase–like endoplasmic reticulum kinase and p-eukaryotic initiation factor 2α expression in the spinal dorsal horn, which was mostly colocalized with neuronal nuclei (a neuronal marker; fig. 3, A and B). To further classify the endoplasmic reticulum stress neurons, we performed double-labeling with BIP and GAD (the marker of γ-aminobutyric acid–mediated [GABAergic] interneurons). The results showed that BIP was localized in the GABAergic interneurons (fig. 3C). We also examined the expression of p-RNA-dependent protein kinase–like endoplasmic reticulum kinase and p-eukaryotic initiation factor 2α in astrocytes and microglia. Partial colocalization with glial fibrillary acidic protein (GFAP; an astrocyte marker) in the spinal cord was found (Supplemental Digital Content, figs. S2A and S2B, https://links.lww.com/ALN/C138). No colocalization with ionized calcium binding adapter molecule 1 (a microglial marker) was found (Supplemental Digital Content, figs. S2C and S2D, https://links.lww.com/ALN/C138). These results indicate that the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway was activated in dorsal horn neurons in mouse bone cancer pain model.

The expression of inflammatory mediators was tested by quantitative polymerase chain reaction and Western blotting. The results of quantitative polymerase chain reaction showed increased expression of TNF-α, interleukin 1β, and interleukin 6 from day 14 after the operation in the tumor group (fig. 4A). The increased expression of TNF-α was 50-fold on day 14 compared with the basal value (P < 0.001, n = 6), whereas sevenfold higher expression of interleukin 1β (P < 0.001, n = 6) and sixfold higher expression of interleukin 6 (P < 0.001, n = 6) were noted on day 28 after the operation. Similarly, Western blotting showed increased expression of TNF-α (P < 0.001, n = 6; fig. 4C), interleukin 1β (P < 0.001, n = 6; fig. 4D), and interleukin 6 (P = 0.002 on day 10, P < 0.001 on day 21, P < 0.001 on day 28, n = 6; fig. 4E) since day 14 after the operation in the tumor group (fig. 4B).

Considering the upregulation of TNF-α, interleukin 1β, and interleukin 6 reported above, we next explored whether these inflammatory mediators could activate endoplasmic reticulum stress. Studies have indicated the increased expression of interleukin 1β and interleukin 6 after lipopolysaccharide stimulation in various cells.^27,28  Therefore, we treated primary spinal neurons with TNF-α (100 nM) and lipopolysaccharide (100 nM) to further verify the endoplasmic reticulum stress caused by inflammation. The results of immunostaining showed that the expression of p-RNA-dependent protein kinase–like endoplasmic reticulum kinase (P < 0.001 in the TNF-α group, P = 0.001 in the lipopolysaccharide group, n = 6; fig. 4, F and G) and p-eukaryotic initiation factor 2α (P < 0.001 in the TNF-α group, P = 0.04 in the lipopolysaccharide group, n = 6; fig. 4, H and I) was increased in primary spinal neurons treated with TNF-α and lipopolysaccharide in comparison with the vehicle-treated neurons. Additionally, we examined the endoplasmic reticulum stress in the primary cortical neurons after TNF-α and lipopolysaccharide stimulation. Western blotting and electron microscopy showed the evaluated endoplasmic reticulum stress in neurons treated with TNF-α and lipopolysaccharide (Supplemental Digital Content, fig. S3, https://links.lww.com/ALN/C138). These results suggest that endoplasmic reticulum stress might be triggered by upregulation of inflammatory mediators in mouse bone cancer pain model.

We next explored whether intrathecal administration of TNF-α (5 ng) or lipopolysaccharide (100 ng) could trigger endoplasmic reticulum stress in the spinal cord in control mice. Decreased paw withdrawal mechanical threshold (P < 0.001, n = 8, respectively) and increased number of spontaneous flinches (P < 0.001, n = 8, respectively) were found at 1 h and 3 h after injection both in TNF-α– and lipopolysaccharide-treated mice in comparison with baseline. Expression levels of the endoplasmic reticulum stress markers BIP, p-RNA-dependent protein kinase–like endoplasmic reticulum kinase, p-eukaryotic initiation factor 2α, and p-inositol-requiring enzyme 1α were increased in both TNF-α– and lipopolysaccharide-treated mice in comparison with the vehicle-treated mice at 1 h after injection (P < 0.001, n = 6, respectively). These results further demonstrated that inflammatory mediators might be potent inducers of endoplasmic reticulum stress in pain development.

Because the findings above showed that increased expression of inflammatory mediators in lipopolysaccharide-stimulated primary neurons resulted in activation of the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway of endoplasmic reticulum stress, we aimed to determine whether the pathway was involved in bone cancer pain model. For this assessment, we used the endoplasmic reticulum stress inhibitor 4-PBA (a small molecule proposed to facilitate the correct folding of nascent proteins). Tumor-bearing mice were administered 4-PBA by intraperitoneal injection (1 mg/100 μl) and intrathecal injection (40 μg/5 μl, 80 μg/5 μl, 120 μg/5 μl) on day 14 after the operation. The expression of BIP, p-RNA-dependent protein kinase–like endoplasmic reticulum kinase, p-eukaryotic initiation factor 2α, p-inositol-requiring enzyme 1α, and activating transcription factor 6α were decreased by 4-PBA, which represented inhibition of endoplasmic reticulum stress (Supplemental Digital Content, fig. S4, https://links.lww.com/ALN/C138). The findings showed a statistically significant increase in paw withdrawal mechanical threshold and decrease in number of spontaneous flinches in the bone cancer pain plus 4-PBA group (40 μg/5 μl; paw withdrawal mechanical threshold: P = 0.031 at 1 h; P = 0.001 at 2 h; P = 0.002 at 3 h; P < 0.001 at 4 h; P = 0.001 at 5 h; P = 0.041 at 6 h; number of spontaneous flinches: P < 0.001 at 1 h; P < 0.001 at 2 h; P < 0.001 at 3 h; P < 0.001 at 4 h; P = 0.003 at 5 h; n = 8) and bone cancer pain plus 4-PBA group (80 μg/5 μl; paw withdrawal mechanical threshold: P = 0.002 at 1 h; P = 0.001 at 2 h; P = 0.002 at 3 h; P = 0.003 at 4 h; P = 0.004 at 5 h; P = 0.006 at 6 h; number of spontaneous flinches: P = 0.001 at 1 h; P < 0.001 at 2 h; P < 0.001 at 3 h; P < 0.001 at 4 h; P < 0.001 at 5 h; n = 8, respectively) that appeared at 1 h after administration and persisted for 6 h, in comparison with the baseline. Meanwhile, these mice showed statistically significant differences in comparison with bone cancer pain plus vehicle mice. However, there was no change in paw withdrawal mechanical threshold and number of spontaneous flinches in the bone cancer pain plus 4-PBA (120 μg/5 μl) mice (fig. 6, A and B). Moreover, the effect of intrathecal analgesia was better than that of intraperitoneal injection (paw withdrawal mechanical threshold in bone cancer pain plus 4-PBA (intraperitoneal 1 mg) group vs. bone cancer pain plus 4-PBA (intrathecal 40 μg) group vs. bone cancer pain plus 4-PBA (intrathecal 80 μg) group: 0.90 ± 0.283 vs. 1.28 ± 0.354 vs. 1.65 ± 0.396 at 3 h; number of spontaneous flinches in bone cancer pain plus 4-PBA (intraperitoneal 1 mg) group vs. bone cancer pain plus 4-PBA (intrathecal 40 μg) group vs. bone cancer pain plus 4-PBA (intrathecal 80 μg) group: 5.00 ± 0.756 vs. 3.50 ± 0.926 vs. 0.88 ± 0.641 at 3 h; n = 8). Specific descriptive statistics (mean ± SD) of behavior tests were shown in Supplemental Digital Content, table 3 and table 4 (https://links.lww.com/ALN/C138).

Next, we administered GSK2606414 intrathecally to selectively inhibit the activation of the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway and further verify the role of the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway in bone cancer pain model. The expression of p-RNA-dependent protein kinase–like endoplasmic reticulum kinase and p-eukaryotic initiation factor 2α decreased, whereas no change was noted in p-inositol-requiring enzyme 1α and activating transcription factor 6α expression (Supplemental Digital Content, fig. S4, https://links.lww.com/ALN/C138). Similar to the findings obtained with 4-PBA treatment, time- and dose- dependent antinociception occurred after GSK2606414 administration (fig. 6, C and D). The paw withdrawal mechanical threshold and number of spontaneous flinches in the bone cancer pain plus GSK2606414 (100 μg/5 μl; paw withdrawal mechanical threshold: P = 0.004 at 1 h; P = 0.001 at 2 h; P = 0.016 at 3 h; number of spontaneous flinches: P = 0.01 at 3 h; 5.00 ± 0.19, P = 0.004 at 4 h; P = 0.048 at 5 h; n = 8) and bone cancer pain plus GSK2606414 (200 μg/5 μl; paw withdrawal mechanical threshold: P < 0.001 at 1 h; P < 0.001 at 2 h; P = 0.004 at 3 h; P = 0.002 at 4 h; P = 0.004 at 5 h; number of spontaneous flinches: P = 0.038 at 2 h; P < 0.001 at 3 h; P = 0.007 at 4 h; n = 8) group improved after administration. Specific descriptive statistics (mean ± SD) of behavior tests were shown in Supplemental Digital Content, tables 5 and 6 (https://links.lww.com/ALN/C138).

Studies have revealed the modulation of inflammatory mediators by endoplasmic reticulum stress. To determine whether endoplasmic reticulum stress could modulate neuroinflammation in mouse bone cancer pain model, we examined the expression of TNF-α, interleukin 1β, and interleukin 6 after injection. Treatment with the endoplasmic reticulum stress inhibitor 4-PBA decreased the expression of TNF-α, interleukin 1β, and interleukin 6 in the spinal cord (fig. 7, A and B) in comparison with bone cancer pain plus vehicle mice (TNF-α and interleukin 1β: P < 0.001 for 40-μg and 80-μg treatments; interleukin 6: P = 0.036 for 40-μg treatment, P = 0.012 for 80-μg treatment, P < 0.001 for 120-μg treatment; n = 6). Similar results were found in the GSK2606414-treated mice. The expression levels of TNF-α, interleukin 1β, and interleukin 6 decreased in the mice that received 100-μg and 200-μg treatment (TNF-α: P = 0.002 for 100-μg treatment, P < 0.001 for 200-μg treatment; interleukin 1β: P < 0.001 for 100-μg and 200-μg treatments; interleukin 6: P < 0.001 for 200-μg treatment; n = 6; fig. 7, E and F).

We also examined the activation of astrocytes and microglia in the spinal cord after treatment. The results showed that both 4-PBA and GSK2606414 could inhibit the activation of astrocytes (P < 0.001, n = 6, respectively; fig. 7, C, D, G, and H). There were no differences in the activation of microglia after treatment in comparison with that in the bone cancer pain plus vehicle mice (P > 0.05, n = 6, respectively; Supplemental Digital Content, fig. S5; https://links.lww.com/ALN/C138). These results indicate that the analgesic effect of the inhibition of endoplasmic reticulum stress was modulated by reduced neuroinflammation in the spinal cord.

In this study, we found activation of the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway of endoplasmic reticulum stress in spinal neurons, along with increased levels of inflammatory mediators TNF-α, interleukin 1β, and interleukin 6 in mouse bone cancer pain model. Stimulation of primary neurons with TNF-α (100 nM) and lipopolysaccharide (100 nM) resulted in the upregulation of the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway. Intrathecal administration of TNF-α (5 ng) and lipopolysaccharide (100 ng), respectively induced hyperalgesia and upregulation of endoplasmic reticulum stress markers in control mice. Inhibition of endoplasmic reticulum stress and the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway improved the nociceptive behaviors in a time- and dose-dependent manner, along with decreased expression of TNF-α, interleukin 1β, and interleukin 6 and suppressed activation of astrocytes in the spinal cord.

Endoplasmic reticulum stress is caused by various physiologic or pathologic conditions such as disturbance of Ca² homeostasis or exposure to oxidative stress and during protection of cells against various pathologic stresses. There are three endoplasmic reticulum stress sensors: RNA-dependent protein kinase-like endoplasmic reticulum kinase, inositol-requiring enzyme 1, and activating transcription factor 6, which are bound to BIP to maintain inactivated in normal conditions. When endoplasmic reticulum stress is triggered, these sensors are dissociated and three signaling pathways are activated.²⁹  Activation of endoplasmic reticulum stress markers and pathways have been reported in various diseases. Autophagy mediated by inositol-requiring enzyme 1–TRAF2-JNK pathway and RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α-C/EBP homologous protein (CHOP) pathway was suggested in the development of cancer.^30,31  Activation of inositol-requiring enzyme 1α–XBP1-JNK pathway and RNA-dependent protein kinase-like endoplasmic reticulum kinase–ATF4-TRB3 pathway was found in metabolic diseases.³²  Additionally, increased expression of BIP, p-RNA-dependent protein kinase–like endoplasmic reticulum kinase, and p-eukaryotic initiation factor 2α was found in various neurodegenerative diseases.³³ 

Studies targeting endoplasmic reticulum stress in nociception development have been conducted. In a neuropathic pain model, endoplasmic reticulum stress was promoted in the spinal dorsal horn and mediated spinal sensitization.³⁴  Moreover, endoplasmic reticulum stress in the dorsal root ganglion contributed to the development of pain hypersensitivity after nerve injury.³⁵  In inflammatory pain, increased expression of BIP and p-eukaryotic initiation factor 2α was found in trigeminal ganglion.¹⁰  Meanwhile, upregulation of BIP and activating transcription factor 6 was found in the superficial spinal dorsal horn in a formalin-induced rat pain model.³⁶  In this study, we observed activation of the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway and inositol-requiring enzyme 1 pathway in spinal neurons during the development of mouse bone cancer pain, but we observed no difference in activating transcription factor 6 expression. The activation of the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway persisted from days 14 to days 28, whereas the inositol-requiring enzyme 1 pathway was activated only on day 28. Therefore, we assumed that the RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α pathway was responsible for the progressive bone pain, whereas the inositol-requiring enzyme 1 pathway may be responsible for the terminal stage of mouse bone cancer pain.

Accumulating evidence indicates that inflammation plays a major role in nociception development. Activation of glial cells in the central nervous system leads to the release of proinflammatory mediators, which induce central sensitization and contribute to the development of chronic pain.¹⁵  Additionally, excessive inflammation in the peripheral and central nervous system contribute to the initiation and maintenance of neuropathic pain.³⁷ 

The mechanisms underlying endoplasmic reticulum stress and inflammation are interlinked.³⁸  Inflammation is the inducer of endoplasmic reticulum stress,³⁹  whereas persistent endoplasmic reticulum stress can activate and aggravate the inflammatory response. The relationship between endoplasmic reticulum stress and inflammation has been revealed in numerous diseases. Inflammation-induced endoplasmic reticulum stress was found in inflammatory bowel disease.³⁹  The modulation of the NLRP3 inflammasome by inositol-requiring enzyme 1α contributed to the development of Alzheimer disease. In the multiple sclerosis mouse model, endoplasmic reticulum stress-induced activation of the JAK1/STAT3 axis led to the expression of interleukin 6.⁴⁰  Moreover, endoplasmic reticulum stress-mediated inflammation in macrophages followed intracerebral hemorrhage.⁴¹  However, few studies have explored the interaction between endoplasmic reticulum stress and inflammation in nociception development.

In this study, primary neuronal cells were stimulated with TNF-α and lipopolysaccharide, respectively to investigate whether endoplasmic reticulum stress could be triggered by inflammatory mediators in neurons. Increased expression of p-RNA-dependent protein kinase–like endoplasmic reticulum kinase and p-eukaryotic initiation factor 2α was found in primary spinal neurons after stimulation of TNF-α and lipopolysaccharide. Studies have revealed that cortical neuron was different from spinal neurons. For example, TNF-α caused apoptosis in hippocampal neurons but not spinal cord neurons.⁴²  Meanwhile, TNF-α induced long-term potentiation in spinal cord neurons but long-term depression in hippocampal neurons.⁴³  In consideration of different mechanism of these neurons TNF-α acted on, we further investigated whether endoplasmic reticulum stress was induced in primary cortical neurons after TNF-α– and lipopolysaccharide- stimulation. Similar results were found as in primary spinal neurons.

Intrathecal injections of TNF-α and lipopolysaccharide were performed to further explore the effects of inflammatory mediators on endoplasmic reticulum stress. The dosages of TNF-α and lipopolysaccharide were used as described by Shen et al.⁴⁴  and Reeve et al.⁴⁵  Intrathecal injection of TNF-α 5 ng or lipopolysaccharide 100 ng induced hyperalgesia in control mice, along with increased expression of endoplasmic reticulum stress markers. These results indicate that endoplasmic reticulum stress was triggered by inflammatory mediators in nociception development. Moreover, decreased expression of TNF-α, interleukin 1β, and interleukin 6 and suppressed activation of astrocytes were noted after inhibition of endoplasmic reticulum stress by 4-PBA and GSK2606414, which indicated the modulation of inflammation by endoplasmic reticulum stress in bone cancer pain model. In summary, these results provide evidence for the close relationship between inflammation and endoplasmic reticulum stress in the development of mouse bone cancer pain.

4-PBA is a small molecule which can facilitate the correct folding of nascent proteins and suppresses endoplasmic reticulum stress. Studies have indicated that 4-PBA can alleviate hypertension,⁴⁶  attenuate endothelial inflammation and permeability in acute lung injury,⁴⁷  and attenuate neuropathic pain¹¹  and inflammatory pain³⁶  by inhibition of endoplasmic reticulum stress. GSK2606414 is the selective inhibitor of RNA-dependent protein kinase-like endoplasmic reticulum kinase–eukaryotic initiation factor 2α. Studies have revealed that GSK2606414 can reduce the levels of p-RNA-dependent protein kinase–like endoplasmic reticulum kinase and p-eukaryotic initiation factor 2α and restore protein synthesis rates.²⁹ 

Studies have indicated the antinociceptive effects of intraperitoneal administration of 4-PBA in neuropathic pain at a dosage of 100 mg/kg,¹¹  but few studies have focused on the direct analgesic effect on the central nervous system. In this study, we injected 4-PBA intrathecally at three dosages (40 μg/5 μl, 80 μg/5 μl, 120 μg/5 μl) to investigate the dose effect on central nervous system. We also performed intraperitoneal administration to explore the analgesic effect via different modes of administration. As shown in the results, in comparison with intraperitoneal administration of 4-PBA, intrathecal administration appeared to show a better analgesic effect and longer analgesic time. Meanwhile, intrathecal administration of 4-PBA showed a time- and dose-dependent manner effect. Nociceptive behaviors were significantly improved at the dosage of 40 μg and 80 μg. However, there was no change in nociceptive behaviors at the dosage of 120 μg, which may have been caused by the complete inhibition of endoplasmic reticulum stress. As an adaptive mechanism, endoplasmic reticulum stress is a double-edged sword.⁴⁸  On the one hand, endoplasmic reticulum stress triggered by accumulation of unfolded proteins in the endoplasmic reticulum could mediate autophagy⁴⁹  and endoplasmic reticulum associated protein degradation,⁵⁰  and then degrade cytoplasmic constituents, damaged organelles, and intracellular pathogens and maintain the intracellular homeostasis. On the other hand, chronic or persistent endoplasmic reticulum stress could induce apoptosis by activating CHOP⁵¹  and the caspase-12⁵²  pathway and ultimately lead to cell apoptosis. Thus, inhibition of endoplasmic reticulum stress appropriately to eliminate its harmful effects while maintaining the protective effects may be the key to endoplasmic reticulum stress as a therapeutic target.

Studies have indicated that estrogen and estrogen receptor played an important role in bone cancer pain model.⁵³  The relationship between endoplasmic reticulum stress and estrogen signaling has also been proposed. Endoplasmic reticulum stress could regulate estrogen signaling,⁵⁴  whereas estrogen could reduce endoplasmic reticulum stress.⁵⁵  To rule out the effect of estrogen, we used male mice in our study.

In conclusion, our results suggest that endoplasmic reticulum stress was activated in the spinal neurons by upregulation of inflammatory mediators in the spinal cord in mouse bone cancer pain model. Downregulation of neuroinflammation via the inhibition of endoplasmic reticulum stress attenuated nociception in bone cancer pain model. Our findings might partially explain the mechanism underlying the role of endoplasmic reticulum stress in pain development and indicate a potential analgesic approach targeting endoplasmic reticulum stress.

Supported by the National Natural Science Foundation of China (Beijing, China) grant Nos. 81671087, 81471129, 81771142, 81870875; National Key Research and Development Program of China (Beijing, China; SQ2018YFC200044), and a grant from Jiangsu Commission of Health (Jiangsu Provincial Key Medical Discipline, Jiangsu, China).

The authors declare no competing interests.