Research Article

IRE1α–XBP1 signaling in leukocytes controls prostaglandin biosynthesis and pain

See allHide authors and affiliations

Science  19 Jul 2019:
Vol. 365, Issue 6450, eaau6499
DOI: 10.1126/science.aau6499

A “sUPR” target for pain management?

The unfolded protein response (UPR) is initiated when unfolded or misfolded proteins accumulate in the endoplasmic reticulum. One highly conserved arm of the UPR, the IRE1α–XBP1 signaling pathway, also plays a role in various other UPR-independent processes, including hypoxia, angiogenesis, and inflammation. Chopra et al. report that this pathway additionally regulates the production of two molecules, cyclooxygenase 2 and microsomal prostaglandin E synthase 1, that help mediate inflammation-induced pain (see the Perspective by Avril and Chevet). When elements of the IRE1α–XBP1 signaling pathway were knocked out, pain behaviors were reduced in two different mouse models of pain. Targeting this pathway may result in improved pain management therapies.

Science, this issue p. eaau6499; see also p. 224

Structured Abstract


Tissue injury triggers rapid local responses coordinated by immune cells, which dictate the maintenance and resolution of inflammation and therefore the recovery from functional impairment and pain. This inflammatory process requires high levels of protein synthesis, folding, modification, and trafficking, which are events regulated by the endoplasmic reticulum (ER). Excessive protein synthesis and handling can lead to the accumulation of misfolded proteins in this organelle, provoking a cellular state of “ER stress” and subsequent activation of the unfolded protein response (UPR). The IRE1α–XBP1 signaling pathway is an evolutionarily conserved branch of the UPR that maintains ER homeostasis while simultaneously governing various immunometabolic processes. Yet, the physiological consequences of IRE1α–XBP1 signaling in leukocytes during tissue injury and inflammation remain largely unexplored.


IRE1α–XBP1 signaling mediates the rapid induction of pro-inflammatory cytokines in myeloid cells. This pathway has also been implicated in the regulation of lipid metabolic processes that are central for programming immune cell functions in health and disease. Nonetheless, whether IRE1α–XBP1 activation in leukocytes modulates the pain that can be driven by inflammatory processes has not been studied. The scarcity of pharmaceutical products that effectively manage postoperative pain has promoted the use of opioids, in turn contributing to the opioid crisis in the United States. Identifying the key molecular pathways that endow immune cells with potent pro-algesic attributes may lead to the development of more effective and safer strategies for pain treatment. We examined whether leukocyte-intrinsic IRE1α–XBP1 signaling controls transcriptional and metabolic programs that could be implicated in inflammation and pain development.


Transcriptomic analyses of mouse bone marrow–derived dendritic cells stimulated by pattern recognition receptors revealed that IRE1α was necessary for the optimal expression of gene networks involved in eicosanoid metabolism. IRE1α deficiency blunted the normal induction of prostaglandin-endoperoxide synthase 2 (Ptgs2/Cox-2) and prostaglandin E synthase (Ptges/mPGES-1) in stimulated myeloid cells. This in turn reduced the capacity of myeloid cells to produce multiple prostaglandins, including the pro-algesic lipid mediator PGE2. We determined that upon activation by IRE1α, the functional form of transcription factor XBP1 bound to the human PTGS2 and PTGES genes to directly induce their expression and enable robust PGE2 generation. Selective loss of IRE1α or XBP1 in leukocytes decreased PGE2 biosynthesis in vivo upon challenge with pro-inflammatory stimuli and reduced pain-related behaviors in PGE2-dependent models of visceral and postsurgical pain. Blocking IRE1α activation by using small-molecule inhibitors evoked similar antinociceptive effects in both models of pain evaluated.


Our study demonstrates that the IRE1α–XBP1 arm of the UPR operates as a crucial mediator of eicosanoid metabolism and prostaglandin synthesis in myeloid immune cells by promoting the expression of both Cox-2 and mPGES-1. We determined that abrogating this pathway genetically or pharmacologically diminishes pain-related behaviors in mice. Modulating IRE1α–XBP1 signaling may be helpful to induce better analgesia with the goal of improved pain management and reduced opioid use.

Activation of the IRE1α–XBP1 pathway in leukocytes promotes PGE2 generation and pain.

ER stress, inflammatory conditions, or engagement of pattern recognition receptors, such as Toll-like receptors (TLRs), trigger IRE1α activation and generation of functional XBP1 in myeloid leukocytes. This multitasking transcription factor induces the expression of both Cox-2 and mPGES-1, which are enzymes that catalyze the synthesis of PGE2 from arachidonic acid. PGE2 is known to function as a potent lipid mediator that promotes pain by activating and sensitizing nociceptors. Disabling IRE1α–XBP1 signaling reduces behavioral pain responses in PGE2-dependent mouse models of postsurgical and inflammatory visceral pain.


Inositol-requiring enzyme 1[α] (IRE1[α])–X-box binding protein spliced (XBP1) signaling maintains endoplasmic reticulum (ER) homeostasis while controlling immunometabolic processes. Yet, the physiological consequences of IRE1α–XBP1 activation in leukocytes remain unexplored. We found that induction of prostaglandin-endoperoxide synthase 2 (Ptgs2/Cox-2) and prostaglandin E synthase (Ptges/mPGES-1) was compromised in IRE1α-deficient myeloid cells undergoing ER stress or stimulated through pattern recognition receptors. Inducible biosynthesis of prostaglandins, including the pro-algesic mediator prostaglandin E2 (PGE2), was decreased in myeloid cells that lack IRE1α or XBP1 but not other ER stress sensors. Functional XBP1 transactivated the human PTGS2 and PTGES genes to enable optimal PGE2 production. Mice that lack IRE1α–XBP1 in leukocytes, or that were treated with IRE1α inhibitors, demonstrated reduced pain behaviors in PGE2-dependent models of pain. Thus, IRE1α–XBP1 is a mediator of prostaglandin biosynthesis and a potential target to control pain.

The endoplasmic reticulum (ER) ensures the proper folding and posttranslational modification of secreted and transmembrane proteins. Diverse physiological and pathological conditions can provoke the accumulation of misfolded proteins within this organelle. These, in turn, can induce ER stress and activate the unfolded protein response (UPR). The inositol-requiring enzyme 1[α] (IRE1[α])–X-box binding protein spliced (XBP1) pathway is the most evolutionarily conserved arm of the UPR (1). When ER homeostasis is altered, the dual-enzyme IRE1α undergoes oligomerization and autophosphorylation, activating its endoribonuclease domain to excise a 26-nucleotide fragment from unspliced Xbp1 mRNA (2). This unconventional splicing event gives rise to the functional form of transcription factor XBP1, which promotes the expression of multiple genes that enhance the protein-folding capacity of the ER (2, 3).

Emerging evidence indicates that IRE1α–XBP1 signaling can also control UPR-independent cellular pathways, influencing processes such as hepatic lipogenesis (4), hypoxia responses (5), angiogenesis (6), atherosclerosis (7, 8), arthritis (9), and antitumor immunity (1012). Myeloid cells stimulated through plasma membrane–bound Toll-like receptors (TLRs) rapidly and selectively activate IRE1α–XBP1. This event is required for their optimal production of some pro-inflammatory cytokines (13). Nevertheless, the precise transcriptional and metabolic programs coordinated by IRE1α–XBP1 signaling in leukocytes under inflammatory conditions, and their physiological consequences, remain largely unexplored.

IRE1α controls transcriptional programs in myeloid cells stimulated through pattern recognition receptors

To understand how IRE1α–XBP1 activation influences global gene expression in myeloid cells, we performed unbiased transcriptomic analyses of wild-type (WT) and IRE1α-deficient bone marrow–derived dendritic cells (BMDCs) stimulated with bacterial lipopolysaccharide (LPS) (TLR4 agonist) or fungal zymosan (Dectin-1 and TLR2 agonist). Consistent with previous reports (13), WT BMDCs exposed to these microbial products exhibited IRE1α-dependent Xbp1 splicing (fig. S1, A and B) but did not show robust induction of canonical XBP1 target genes in the ER stress response (fig. S1C) or activation of other UPR branches (fig. S1D). We did not observe signs of regulated IRE1α-dependent decay (RIDD) (1416) upon LPS or zymosan stimulation because the expression levels of several genes reported to be potentially regulated by this process were not increased in BMDCs that lack IRE1α (fig. S1, E and F). IRE1α deficiency did not compromise normal BMDC generation or survival in response to granulocyte macrophage colony-stimulating factor (GM-CSF) (fig. S2, A and B). However, we identified 1792 and 2863 genes whose expression was significantly altered in IRE1α-deficient BMDCs stimulated with either zymosan or LPS, respectively, compared with their WT counterparts (Fig. 1A). There was a significant overlap of 1167 differentially regulated genes between the two stimuli (Fig. 1A), indicating a common effect of IRE1α deficiency independent of the agonist used. Ingenuity pathway analysis (IPA) for these commonly regulated genes revealed the enrichment of nine biological categories (Fig. 1B). As expected, IRE1α deficiency influenced transcriptional processes involved in posttranslational protein modification as well as cellular maintenance and survival (Fig. 1B) (17). Surprisingly, biosynthesis and metabolism of eicosanoids emerged as a major cellular function potentially regulated by IRE1α in BMDCs stimulated with LPS or zymosan (Fig. 1B). We then identified 27 regulators that not only changed expression at the mRNA level but also had a significant number of known targets enriched in the gene list (Fig. 1C, top 10 regulators are shown). Confirming previous reports (13), expression of Il6 and its associated target genes was significantly decreased in TLR-stimulated BMDCs that lack IRE1α, compared with their WT counterparts (Fig. 1C). Additionally, and corresponding with IPA denoting altered eicosanoid metabolism, prostaglandin-endoperoxide synthase 2 (Ptgs2/Cox-2) and prostaglandin E synthase (Ptges/mPGES-1) emerged as potential regulators that were markedly decreased in IRE1α-deficient BMDCs exposed to LPS or zymosan (Fig. 1C). We confirmed the down-regulation of these two enzymes at the mRNA and protein levels in stimulated BMDCs that lack IRE1α using quantitative reverse transcription polymerase chain reaction (RT-PCR) and immunoblot assays, respectively (Fig. 1, D and E). IRE1α deficiency did not affect the constitutive expression of Ptgs1/Cox-1 or Ptges2/mPGES-2 (fig. S2C), suggesting that this ER stress sensor primarily mediates the rapid induction of Ptgs2/Cox-2 and Ptges/mPGES-1 in response to inflammatory stimuli. Consequently, we hypothesized that IRE1α may be required for normal eicosanoid production by myeloid cells.

Fig. 1

IRE1α regulates the expression of Ptgs2 and Ptges. Ern1WT or Ern1KO BMDCs were left untreated or stimulated with LPS (50 ng/ml) or zymosan (25 μg/ml) for 6 hours, and global transcriptional profiles were determined with RNA-seq. (A) Number of differentially regulated genes identified in IRE1α-deficient BMDCs treated with LPS or zymosan. (B) IPA of RNA-seq data highlighting eicosanoid metabolism as one of the central biological functions altered in LPS- or zymosan-stimulated BMDCs lacking IRE1α. (C) Top 10 key regulators identified with RNA-seq analysis. Individual replicate heatmaps show log ratio of expression values versus the reference, which is the mean across samples for the “Ern1 WT” panel, and Ern1WT for the “Ern1KO/Ern1WT” panel. U, untreated; Zym, zymosan. (D) Quantitative RT-PCR for Ptgs2 and Ptges upon LPS or zymosan stimulation. (E) Representative immunoblot analyses for Cox-2 and mPGES-1 expression in Ern1WT and Ern1KO BMDCs stimulated with LPS (+, 10 ng/ml, or ++, 100 ng/ml) or zymosan (25 μg/ml). The density of each band was normalized to its own actin value, and numbers shown represent relative expression compared with Ern1WT BMDCs treated with 10 ng/ml LPS because there was undetectable expression of these proteins in unstimulated BMDCs. In (D) and (E), data are shown as mean ± SEM and are representative of at least three independent experiments. Two-tailed Student’s t test was used for statistical analysis. *P < 0.05, **P < 0.005, ***P < 0.0005.

IRE1α–XBP1 signaling is necessary for optimal prostaglandin biosynthesis

Prostaglandins are a major class of eicosanoids whose inducible biosynthesis depends on the rapid metabolism of arachidonic acid by Cox-2 (Fig. 2A) (18). These bioactive lipids participate in the regulation of diverse physiological processes such as allergy, fever, vascular permeability, and pain (19). Lipidomic analyses revealed that IRE1α deficiency did not influence basal prostaglandin levels in untreated BMDCs (Fig. 2B). However, and consistent with the observed impairment in Cox-2 induction, we identified a profound decrease in the intracellular levels of several prostaglandins, including prostaglandin E1 (PGE1), PGF, PGD2, PGE2, PGF, 15-keto PGF, D12-PGJ2, 13,14dh-15k PGE2, and PGD3 in LPS-stimulated BMDCs that lack IRE1α when compared with their WT counterparts (Fig. 2B and table S1).

Fig. 2 IRE1α promotes prostaglandin biosynthesis.

(A) Pathway depicting the main events implicated in the metabolism of arachidonic acid and prostaglandin production. Cox-1 is constitutively expressed, whereas Cox-2 is induced by pro-inflammatory stimuli. (B) Ern1WT (n = 4 independent samples) or Ern1KO BMDCs (n = 3 independent samples) were left untreated (bottom) or stimulated with LPS (50 ng/ml) for 6 hours (top), and lipidomic analyses were performed. Data are represented as volcano plots with red lines indicating a significance level of P = 0.05. Two-tailed Student’s t test with FDR correction was used for statistical analysis.

Cox-2 converts arachidonic acid to prostaglandin endoperoxide H2 (PGH2), which is subsequently metabolized by mPGES-1 to generate the potent lipid mediator PGE2 (Fig. 3A) (18). Corresponding with the decreased induction of both Cox-2 and mPGES-1 in IRE1α-deficient BMDCs stimulated with LPS (Fig. 1), there was a marked reduction in PGE2 production by these cells compared with their WT counterparts (Fig. 3, B and C). Additional IRE1α-deficient myeloid cell subsets, including primary neutrophils and macrophages, also demonstrated defective PGE2 synthesis upon LPS stimulation (fig. S3, A and B). To further confirm these findings in vivo, we administered LPS intraperitoneally to transgenic mice selectively lacking IRE1α in leukocytes (Ern1f/f Vav1cre) (20) and then quantified PGE2 production in situ. LPS exposure triggered Xbp1 splicing and the concomitant IRE1α-dependent induction of both Ptgs2 and Ptges in peritoneal leukocytes (fig. S4, A to C). Accordingly, mice lacking IRE1α in leukocytes demonstrated reduced production of peritoneal PGE2 upon LPS administration compared with that of their WT counterparts (fig. S4D). Confirming our transcriptional profiling by using an independent agonist (Fig. 1), PGE2 synthesis was also diminished in zymosan-exposed BMDCs that lack IRE1α (Fig. 3D). Similar results were observed in vivo after the intraperitoneal administration of zymosan into mice lacking IRE1α in leukocytes (fig. S4, E to I). In this setting, lipidomic analyses further confirmed that the production of Cox-2–dependent prostaglandins (PGE2, PGD2, and PGF2α) and TBX2 was reduced in cell-free peritoneal lavage from Ern1f/f Vav1cre compared with Ern1f/f mice (fig. S4, E to H). By contrast, lipoxygenase-dependent 15-HETE was unaltered (fig. S4I). XBP1 deletion phenocopied the same defects observed in IRE1α-deficient myeloid cells (Fig. 3E and fig. S3C), whereas the ablation of other ER stress sensors such as PERK (protein kinase RNA-like endoplasmic reticulum kinase) (encoded by Eif2ak3) and ATF6α (activating transcription factor 6α) did not compromise PGE2 generation in response to this treatment (Fig. 3, F and G). Thus, the IRE1α–XBP1 arm of the ER stress response is selectively required for optimal PGE2 production by LPS- or zymosan-stimulated myeloid cells.

Fig. 3 IRE1α–XBP1 is necessary for optimal PGE2 production by myeloid cells.

(A) Pathway showing the main steps implicated in PGE2 generation. Both Cox-1 and mPGES-2 are constitutively expressed, whereas Cox-2 and mPGES-1 are induced by pro-inflammatory stimuli. (B and C) ELISA-based confirmation for reduced PGE2 in supernatants from Ern1KO BMDCs upon stimulation with the indicated concentrations of LPS (B), or at different time points after stimulation with LPS at 50 ng/ml (C). (D to G) Mouse BMDCs of the indicated genotypes were stimulated with zymosan (25 μg/ml) for 6 hours, and PGE2 was quantified in culture supernatants by means of ELISA. Dots represent BMDCs generated from independent mice. (H to K) CRISPR/Cas9–based gene editing was used to ablate XBP1 in human monocyte–derived DCs, and cells were then stimulated for 6 hours with zymosan (25 μg/ml). [(H) to (J)] Quantitative RT-PCR was used to assess the indicated transcript levels, and (K) PGE2 levels were determined in the corresponding supernatants by means of ELISA. sg, single-guide; Scr, scrambled RNA. Dots represent human DC samples generated from independent donors. Data are shown as mean ± SEM and are representative of at least two independent experiments. Two-tailed Student’s t test was used for (B) and (D) to (K). Two-way ANOVA (Tukey’s test) was used for (C); *P < 0.05, **P < 0.005, ***P < 0.0005.

IRE1α-dependent induction of PGE2 was also observed in BMDCs treated with agonists for other plasma membrane–bound TLRs, whereas stimulation from endosomal TLR3, TLR8, or TLR9 had no effect (fig. S5A). These results are consistent with previous reports that demonstrated predominant IRE1α–XBP1 activation by agonists engaging plasma membrane-bound but not endosomal TLRs (13). PGE2 induction was also reduced in IRE1α-deficient BMDCs activated with phorbol myristate acetate (PMA) (fig. S5A), thus ruling out the possibility that IRE1α ablation was compromising proximal TLR signaling. Moreover, we also found diminished PGE2 production, accompanied by reduced expression of both Cox-2 and mPGES-1, in IRE1α-deficient BMDCs treated with the pharmacological ER stressor thapsigargin (fig. S5, B and C). Thus, optimal PGE2 synthesis by mouse myeloid cells undergoing ER stress or stimulated through plasma membrane–bound TLRs requires IRE1α–XBP1 activation, which promotes the expression of Cox-2 and mPGES-1.

To define whether IRE1α–XBP1 signaling also controlled inducible PGE2 production in human myeloid cells, we generated monocyte-derived DCs from the peripheral blood of healthy volunteers and then abrogated this pathway using gene-editing techniques. Transient transfection of primary human DCs with single-guide RNA (sgRNA)–Cas9 complexes targeting XBP1 effectively edited this gene and prevented the generation of its spliced (active) form upon zymosan treatment (Fig. 3H). The induction of PTGS2 and PTGES, as well as PGE2 production, were significantly diminished in zymosan-exposed human DCs that lack XBP1, compared with their WT counterparts transfected with scrambled sgRNA-Cas9 complexes (Fig. 3, I to K). Similar effects were observed when human DCs were transfected with complexes that target ERN1 (fig. S5, D and E). Thus, IRE1α–XBP1 signaling acts as a conserved mediator of inducible PGE2 production in human DCs.

XBP1s transactivates the human PTGS2 and PTGES promoters

We sought to determine the molecular mechanism by which IRE1α-activated XBP1 (XBP1s) mediates inducible PGE2 production in human myeloid cells. We analyzed the promoter regions of PTGS2 and PTGES for potential XBP1s binding sites, as previously described (21, 22), and found putative X-box–binding and Unfolded Protein Responses Element A (UPRE-A) sequences on the PTGS2 promoter (Fig. 4A). Additionally, we identified an X-box–binding region and two ETS domain–binding sites in the PTGES promoter (Fig. 4B). We hypothesized that XBP1s could operate as a driver of PTGS2 and PTGES transcription.

Fig. 4 XBP1s transactivates the PTGS2 and PTGES promoters.

(A and B) Predicted XBP1s-binding sites in promoter regions of human (A) PTGS2 and (B) PTGES. (C to F) Human primary monocyte–derived DCs were left untreated or stimulated with zymosan in the presence or absence of the ER stressor 2-DG. ChIP assays were performed by using anti-XBP1s or isotype control antibodies (Ab). The IRE1α inhibitor MKC8866 was used as indicated, and quantitative PCR was used to determine XBP1s occupancy at these promoter regions under the conditions tested. ChIP-PCR assays were performed by using three to six independent human donors. (G and H) HEK293 cells were cotransfected with XBP1s- or CHOP-expressing plasmid vectors, and luciferase reporter constructs harboring the (G) PTGS2 or (H) PTGES promoters, along with a Renilla luciferase construct for internal control. Firefly luciferase (Luc) activity was normalized to Renilla activity in each case. Data are representative of at least two independent experiments with similar results, using four technical replicates. Data are shown as mean ± SEM and expressed as fold increase in Firefly luciferase activity compared with empty vector (pcDNA3.1). Two-tailed Student’s t test was used for statistical analysis. *P < 0.05, **P < 0.005, ***P < 0.0005.

Chromatin immunoprecipitation (ChIP)–PCR was used to evaluate direct XBP1s binding to the promoter regions identified. We stimulated human monocyte-derived DCs with zymosan alone or in combination with 2-deoxy-d-glucose (2-DG), which inhibits N-linked protein glycosylation, producing ER stress and robust IRE1α–XBP1 activation (23). Zymosan exposure increased XBP1s binding to the predicted PTGS2 and PTGES promoter regions, and concomitant treatment with the ER stressor 2-DG markedly enhanced these effects (Fig. 4, C and D). Disabling the IRE1α ribonuclease (RNase) domain by using the selective pharmacological inhibitor MKC8866 (2427) abrogated XBP1s binding to these promoters in zymosan-stimulated human DC undergoing ER stress (Fig. 4, C and D). XBP1s also bound the GFPT1 promoter, as previously reported (23), whereas promoter regions of pri-miR-21 devoid of XBP1s-binding sites were not enriched in these assays (Fig. 4, E and F). Furthermore, luciferase reporter assays using human embryonic kidney (HEK) 293 cells demonstrated that XBP1s was sufficient to transactivate the human PTGS2 and PTGES promoters in a dose-dependent manner (Fig. 4, G and H). By contrast, the PERK-controlled ER stress transcription factor C/EBP homologous protein (CHOP) showed no effect in this reporter system (Fig. 4, G and H). Thus, IRE1α-activated XBP1s promotes inducible PGE2 biosynthesis by directly driving the transcriptional induction of both PTGS2 and PTGES.

IRE1α–XBP1 signaling in leukocytes promotes pain behaviors

PGE2 generated by induction of Cox-2 and mPGES-1 engages EP1-4 receptors on peripheral sensory neurons and the central nervous system to promote pain responses (2830). We postulated that mice lacking IRE1α in leukocytes would demonstrate reduced pain behaviors because of their impaired capacity to induce PGE2 production in response to inflammatory stimuli (fig. S4). Two classical PGE2-dependent models of pain were used to test this hypothesis: an acetic acid–based model for inflammatory visceral pain (3134) and a paw incision model of postsurgical pain (35). We injected 0.9% v/v of acetic acid intraperitoneally into either Ern1f/f or Ern1f/f Vav1cre mice, and writhing behaviors were monitored over time by a blinded observer. The number of writhing events recorded within the first 30 min (Fig. 5A), as well as PGE2 levels in cell-free peritoneal lavage samples (fig. S6A), were significantly reduced in Ern1f/f Vav1cre mice compared with their IRE1α-sufficient counterparts. We also performed automated unbiased and blinded tests to evaluate how the inflammatory visceral pain caused by acetic acid administration affected the normal ambulatory capacity of the host. IRE1α-sufficient (Ern1f/f) mice demonstrated lower displacement ability than that of their Ern1f/f Vav1cre littermates, as evidenced by a significant decrease in their total ambulatory counts and times after acetic acid injection (Fig. 5, B and C). Reduced writhing behaviors were also observed in mice selectively lacking XBP1 in leukocytes (Xbp1f/f Vav1cre) (Fig. 5D), thus confirming that canonical IRE1α–XBP1 signaling mediates this response. Peritoneal leukocytes demonstrated constitutive IRE1α-dependent Xbp1 splicing that was maintained upon acetic acid administration (fig. S6B), and loss of IRE1α in these cells decreased their normal expression of Ptges by ~50% (fig. S6C). No alterations in interleukin-1β (IL-1β), IL-6, or tumor necrosis factor–α (TNFα) expression at the mRNA or protein levels were found in this milieu at the same time point analyzed (fig. S6, D to G). Thus, leukocyte-intrinsic IRE1α enabled the rapid production of PGE2 without modulating these pro-inflammatory cytokines upon intraperitoneal acetic acid administration. Similar IRE1α-driven writhing responses were observed in male and female mice, indicating that this phenotype was not markedly influenced by sex (fig. S7). To determine whether disabling IRE1α–XBP1 signaling pharmacologically could reduce inflammatory visceral pain, we used two inhibitors of IRE1α: the kinase domain–specific inhibitor KIRA6 (36, 37) and the RNase domain–specific inhibitor MKC8866 (2427). These compounds were independently administered intraperitoneally 6 hours and 30 min before acetic acid injection. Treatment with either KIRA6 or MKC8866 reduced Xbp1s and Ptges expression in peritoneal leukocytes (Fig. 5, E and F) and significantly diminished the number of writhing behaviors after acetic acid injection (Fig. 5, G and H). The administration of a similar dose of celecoxib, a nonsteroidal anti-inflammatory drug that inhibits Cox-2 and limits prostanoid production, also decreased writhing behaviors (fig. S8), confirming the role of PGE2 in this behavioral response. Thus, IRE1α–XBP1 activation promotes visceral pain in the acetic acid–based model.

Fig. 5 IRE1α–XBP1 signaling in leukocytes promotes inflammatory visceral pain.

(A to C) 0.9% v/v acetic acid (5 ml/kg) was administered intraperitoneally to Ern1f/f (n = 13) or Ern1f/f Vav1cre (n = 17) mice. Each group contained both male and female mice. (A) Writhing behaviors after acetic acid injection were recorded every 5 min for 30 min. Total ambulatory time (B) and total ambulatory counts (C) for mice of the indicated genotypes were monitored for 15 min after acetic acid injection. (D) 0.9% v/v acetic acid (5 ml/kg) was injected intraperitoneally into Xbp1f/f (n = 10) or Xbp1f/f Vav1cre (n = 11) mice (mix of males and females) and writhing was recorded every 5 min for 30 min. (E to H) Pharmacological inhibition of IRE1α reduces visceral pain. WT C57BL/6J mice were given KIRA6 (25 mg/kg) or MKC8866 (20 mg/kg) intraperitoneally 6 hours and 30 min before challenge with 0.9% v/v acetic acid (5 ml/kg). In (E) and (F), the indicated mRNA transcript levels were determined by means of quantitative RT-PCR in leukocytes recovered from peritoneal lavage samples after treatment with IRE1α inhibitors. Writhing behaviors after acetic acid administration were recorded in mice receiving (G) KIRA6 or (H) MKC8866 and compared with their corresponding vehicle controls. All data are presented as mean ± SEM. Two-way ANOVA (Sidak’s test) was used for (A), (D), (G), and (H). Two-tailed Student’s t test was used for (B) and (C). One-way ANOVA (Dunnett’s test) was used for (E) and (F). *P <0.05, **P < 0.005, ***P < 0.0005. Data corresponding to behavioral responses were generated in two independent experiments.

We next evaluated whether IRE1α deficiency in leukocytes could also influence postoperative pain, which is a PGE2-mediated process commonly treated with Cox-2 inhibitors (3840). A surgical incision was made in the right hind paw of either Ern1f/f or Ern1f/f Vav1cre mice, and nonreflexive pain-related behaviors were monitored over time and analyzed in comparison with baseline measurements before surgery. We observed IRE1α-dependent Xbp1 splicing in CD45+ leukocytes sorted from the injury site 24 hours after surgery (fig. S9A). Although the proportions of neutrophils, macrophages, and DCs infiltrating the lesions at this time point were not altered (fig. S9B), we identified a significant reduction in the number of Cox-2–expressing leukocytes infiltrating the injured tissues (surgical site) in Ern1f/f Vav1cre mice, compared with their littermate controls (Fig. 6, A to D). Weight-bearing distribution tests indicated that Ern1f/f Vav1cre mice showed a greater capacity to use their injured paws as compared with their IRE1α-sufficient counterparts. They displayed a more balanced hind-paw distribution at early stages after incision and recovered significantly faster (Fig. 6E). These effects were not associated with differential body weight in the two genotypes (fig. S10A). Additional parameters indicative of spontaneous nonevoked pain behaviors, such as facial expression by grimace score (Fig. 6F) and guarding behavior by using the guarding score (Fig. 6G), were also decreased in mice lacking IRE1α in leukocytes. Mechanical threshold, determined by using von Frey–induced withdrawal reflex analysis, was comparable in Ern1f/f versus Ern1f/f Vav1cre mice after surgery (fig. S10B), which is consistent with the negligible role of peripheral Cox-2 in evoked punctate mechanical hypersensitivity previously reported in rodent models of acute pain (4144). The numbers of flinches (fig. S10C) and paw perimeter (fig. S10D) also remained unchanged in Ern1f/f versus Ern1f/f Vav1cre mice after paw incision. The role of leukocyte-intrinsic IRE1α in these postsurgical responses was comparable in male (fig. S11, A to G) and female (fig. S12, A to G) mice. However, we found that male Ern1f/f Vav1cre mice showed reduced impairment and faster recovery of rearing activity in comparison with their Ern1f/f counterparts (fig. S11H), whereas female mice did not (fig. S12H).

Fig. 6 IRE1α activation mediates spontaneous pain behaviors postsurgery.

(A to D) Cox-2 expression in leukocytes infiltrating the paw after surgery. (A) Representative confocal microscopy images of ipsilateral paw tissue from Ern1f/f or Ern1f/f Vav1cre mice stained with fluorescently labeled antibodies specific for CD45 (red) or Cox-2 (green), and DAPI (blue). White arrows denote cells coexpressing CD45 and Cox-2. Scale bars, 100 μm. (B) Magnified images of the indicated insets are shown. Scale bars, 50 μm. (C) Quantification of total CD45+ cells, and of Cox-2-expressing CD45+ leukocytes (D) in the paw 48 hours after surgery. Data are shown as mean ± SEM. *P < 0.05. (E to G) A surgical incision was made in the right hind paw of Ern1f/f (n = 11 to 24) or Ern1f/f Vav1cre (n = 11 to 29) mice. Each group comprised both male and female mice. Animals of the indicated genotypes were monitored for (E) spontaneous hind paw weight bearing distribution, (F) grimace score, and (G) guarding score. Grimace and guarding scores were determined 8 hours postsurgery. (H to J) Pharmacological inhibition of IRE1α reduces postoperative pain behaviors. C57BL/6J male mice (n = 8/group) were administered vehicle or MKC8866 (20 mg/kg) intraperitoneally 6 hours and 30 min before a surgical incision was made in the right hind paw. Animals were monitored then for (H) spontaneous weight-bearing distribution, (I) grimace score, and (G) guarding score. Grimace and guarding scores were determined 8 hours after surgery. For (F), (G), (I) and (J), data are presented as median ± 25 to 75% confidence interval (boxes) with smallest and largest values (whiskers). Data in all other panels are shown as mean ± SEM. Two-tailed Student’s t test was used for (C) and (D). Two-way ANOVA (Sidak’s test) was used for (E) and (H). The Mann-Whitney U test was used for (F), (G), (I), and (J). *P < 0.05, **P < 0.005, ***P < 0.0005.

To determine whether pharmacological targeting of IRE1α could also modulate postsurgical pain, we administered MKC8866 before paw incision surgery and monitored pain responses thereafter. IRE1α inhibition improved nociceptive functional behaviors, as demonstrated by a more balanced weight distribution when compared with that of vehicle-treated mice (Fig. 6H). Grimace and guarding scores after surgery were also significantly reduced in mice receiving MKC8866 (Fig. 6, I and J). In contrast to our observations when using Ern1f/f Vav1cre mice, we found reduced flinching activity after paw incision in MKC8866-administered groups (fig. S13A), suggesting a pro-algesic role for IRE1α in additional nonleukocyte cells in this setting. Rearing activity was unchanged upon IRE1α targeting (fig. S13B), indicating that complete inhibition of IRE1α may be required for altering this specific behavior in male mice after paw incision. Consistent with our results using conditional IRE1α-deficient mice, mechanical hypersensitivity remained unaltered upon administration of MKC8866 (fig. S13C). Similar behavioral responses were observed in mice treated with KIRA6 (fig. S14). As a positive control, we administered a comparable dose of celecoxib following the same scheme and route described above. Similar to IRE1α inhibition, we observed a more balanced weight-bearing distribution as well as diminished grimace and guarding scores after paw incision in mice that received celecoxib, compared with vehicle-treated mice (fig. S15, A to C). Flinches, rearing activity, and mechanical threshold after paw incision remained unaffected upon celecoxib treatment (fig. S15, D to F). This is consistent with previous reports demonstrating that Cox-2 inhibition at the comparable dose chosen does not influence mechanical hypersensitivity in models of acute pain (41, 42). Thus, mice lacking IRE1α–XBP1 in leukocytes exhibit reduced behavioral pain responses in two distinct PGE2-dependent models of pain. Furthermore, targeting IRE1α pharmacologically can modulate these pain behaviors in vivo.


Here, we present molecular and functional evidence that reveals an unexpected function for the ER stress sensor IRE1α as a central mediator of prostaglandin biosynthesis and behavioral pain responses in mice. Our findings suggest a previously unappreciated mechanism in which IRE1α activates transcription factor XBP1 to sustain expression of two rate-limiting enzymes that are necessary for optimal prostaglandin production—namely, Cox-2 and mPGES-1. More effective pain management strategies are needed in the clinic, especially in light of the devastating opioid crisis that the United States currently faces (45). The pharmacological modulation of IRE1α–XBP1 signaling may represent an alternative approach for pain control with the potential of producing better analgesia, diminished opioid requirements, and reduced opioid side effects. Future studies will be needed to determine whether IRE1α–XBP1 signaling also regulates additional physiological and pathological processes driven by prostaglandins such as pregnancy, fever, allergy, and immunosuppression in cancer.

Materials and methods

RNA isolation, quantitative RT-PCR, and Xbp1 splicing assays

Total RNA was isolated using RNeasy Mini kit or QIAzol lysis reagent (Qiagen) according to the manufacturer’s instructions. RNA (0.1 to 1 μg) was reverse-transcribed to generate cDNA using the qScript cDNA synthesis kit (Quantabio). Quantitative RT-PCR was performed using PerfeCTa SYBR green fastmix (Quantabio) and TaqMan Universal PCR master mix (Life Technologies) on a QuantStudio 6 Flex real-time PCR system (Applied Biosystems). Normalized gene expression was calculated by comparative threshold cycle method using ACTB or Actb as a control. Xbp1 splicing assays were performed as previously described (46). PCR products were separated by electrophoresis through a 2.5% agarose gel and visualized by ethidium bromide staining. Primers used in this study are described in table S2.

Transgenic mice

Atf6f/f, Eif2ak3f/f, Vav1cre, and CD11ccre mice were obtained from The Jackson Laboratory. Xbp1f/f and Ern1f/f mice have been previously described by our groups (4, 47). We generated conditional knockout mice lacking ATF6, IRE1α, or XBP1 in leukocytes by crossing Atf6f/f, Ern1f/f, or Xbp1f/f animals, respectively, with the Vav1cre strain that allows selective gene deletion in hematopoietic cells (20). Crossing Eif2ak3f/f mice with CD11ccre animals generated mice devoid of PERK in DCs. All mouse strains were on a full C57BL/6J background. Mice were housed in specific pathogen-free animal facilities at Weill Cornell Medical College, Memorial Sloan Kettering Cancer Center, and Wake Forest University. Mice were handled in compliance with Weill Cornell Institutional Animal Care and Use Committees procedures. Mice used for behavioral pain tests were housed at Wake Forest School of Medicine, in accordance with the Wake Forest University Guidelines on the ethical use of animals. The Institutional Animal Care and Use Committee of Wake Forest University approved all pain-related experiments. Animals were housed under a 12-hour light–dark cycle, with food and water ad libitum.

Primary cell isolation and generation

Murine BMDCs were generated by incubation of flushed, single suspended, bone marrow cells isolated from mice of the indicated genotypes in complete RPMI media [RPMI supplemented with l-glutamine, 10% fetal bovine serum (FBS), HEPES, sodium pyruvate, non-essential amino acids, β-mercaptoethanol, and penicillin-streptomycin] containing 20 ng/ml of recombinant GM-CSF (Gemini or Peprotech). Media was replenished on day 6, and cells were harvested on day 7 and used directly for subsequent in vitro functional assays.

Human monocyte-derived DCs were generated by positively isolating CD14+ cells (Miltenyi, catalog number 130-050-201) from blood/buffy coats using Ficoll-gradient centrifugation and plated in complete RPMI media containing 10% FBS and human recombinant GM-CSF (Peprotech) at 1000 IU/ml and IL-4 (Peprotech) at 500 IU/ml for 7 days. Cells were then harvested and used for subsequent in vitro assays (48).

Mouse primary macrophages were generated by incubating flushed, single suspended, bone marrow cells from mice of the indicated genotypes in DMEM F12 50/50 media supplemented with L-glutamine, 10% FBS and penicillin–streptomycin, and containing 20 ng/ml recombinant M-CSF (Peprotech) and 1 ng/ml recombinant IL-3 (Peprotech) for 3 days in bacteriological plates. On day 4, non-adherent cells were washed and plated in tissue culture-treated dishes at 1×105 cells/ml in media containing 20 ng/ml recombinant M-CSF. On day 6, media was replaced and cells were harvested and used for stimulation on day 7.

Primary neutrophils were isolated directly from the bone marrow of Ern1f/f or Ern1f/f Vav1cre mice using negative selection (Miltenyi, catalog #130-097-658) according to the manufacturer protocol. In all cases, isolation purity was greater than 80%. All stimulations were performed in 96-well plates at a volume of 200 μl of media and supernatants were collected after the indicated time points.

Flow cytometry-based analysis

Murine bone marrow-derived BMDCs were washed with phosphate-buffered saline (PBS), FcγR-blocked using TruStain fcXTM (anti-mouse CD16/32, Biolegend, clone 93, 5 μg/ml) and then stained with antibodies specific for CD11c (Biolegend, clone N418, PE-Cy7, 1 μg/ml) and MHC-II (Tonbo, clone M5/114.15.2, FITC, 1.25 μg/ml), along with live/dead staining using 4′,6-diamidino-2-phenylindole dihydrochloridehydrate (DAPI) (Thermo Fischer Scientific, 0.5 μg/ml). Data were acquired on an LSR II instrument (BD Biosciences).

Single-cell suspensions from ipsilateral paws (described below) were washed, FcγR-blocked using TruStain fcXTM, as described above, and stained with antibodies specific for CD45 (BD Biosciences, clone 30-F11, PE-CF594, 1 μg/ml), CD11c (Biolegend, clone N418, APC, 1 μg/ml), MHC-II (Tonbo, clone M5/114.15.2, FITC, 1.25 μg/ml), Ly-6G (Tonbo, clone 1A8, APC-Cy7, 1 μg/ml), CD11b (Tonbo, clone M1/70, PerCP-Cy5.5, 1 μg/ml), F4/80 (Biolegend, clone BM8, PE, 1 μg/ml) along with live/dead staining using DAPI (Thermo Fischer Scientific, 0.5μg/ml). Live CD45+ cells were sorted using BD Aria II SORP cell sorter at the Flow Cytometry Core facility of Weill Cornell Medicine. All flow cytometry data were analyzed with FlowJo software V10 (TreeStar).

Lipidomic analyses

5×106 Ern1WT or Ern1KO BMDCs were stimulated with 50 ng/ml LPS in six-well plates. Cells were collected after 6 hours, and washed with ice-cold PBS. Cell pellets were then frozen at −80°C until further analysis. Cell pellets were suspended in 850 μl of ice-cold PBS and homogenized using a probe sonicator on ice (three cycles of 10 s each, power and frequency). The homogenate was diluted with 150 μl of methanol containing 10 ng each of prostaglandin E1-d4, resolvin D1-d5, leukotriene B4-d4, 15-HETE-d8, and arachidonic acid-d8, and 100 ng each of cholesteryl heptadecanoate and triheptadecanoyl glycerol (all served as internal standards for the LC-MS analysis). The samples were applied to C18 solid phase extraction cartridge (StrataX C18, Phenomenex) and the lipids were extracted following published procedures (49, 50) with following modifications: The SPE cartridges were eluted with isooctane–ethyl acetate (9:1) first for non-polar lipids (sterol esters, neutral sphingolipids, and triglycerides) before eluting the fatty acyl lipidome with methanol containing 0.1% formic acid. The lipidomic analysis was performed by the Lipidomics Core Facility at Wayne State University by LC-MS using standard protocols. The procedures followed were essentially as described earlier for eicosanomic analysis (5153) and by other published procedures for fatty acids, sterol esters, triacyl glycerols, and sphingolipids (5456). Volcano plots were generated in R studio using the bioconductor limma package (57).

Immunoblot assays

BMDCs were washed twice in cold PBS and cell pellets were lysed using RIPA lysis buffer (150 mM sodium chloride, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, and 50 mM Tris pH 8.0) supplemented with protease and phosphatase inhibitors (Roche). Homogenates were centrifuged at 22,000×g rpm for 30 min at 4°C, and the supernatants were collected. Protein concentrations were determined using BCA protein assay kit (Thermo Fisher Scientific). Equivalent amounts of protein were separated via SDS–PAGE and transferred to PVDF membranes (Immobilon, Millipore) following standard protocols. Membranes were blotted with primary antibodies anti-Cox-2 (Cell Signaling, catalog #12282, 1:1,000), anti-mPGES-1 (Cayman Chemicals, catalog #160140, 1:200), and anti-β actin (Cell Signaling, catalog #4967, 1:1,000), followed by incubation with an HRP-conjugated anti-rabbit secondary antibody (Thermo Fischer Scientific, catalog # G-21234, 1:5,000). SuperSignal West Pico and Femto chemiluminescent substrates (Thermo Fisher Scientific) were used to image blots in a FlourChemE instrument (ProteinSimple).


2.5 × 105 cells were stimulated with the indicated compounds and time points, and PGE2 in the supernatants were measured using the PGE2 enzyme-linked immunosorbent assay (ELISA) kit (Enzo Lifesciences, Cat# ADI-900-001). If different number of cells were plated, PGE2 levels were normalized to 2.5 × 105 cells/well. Cell viability counts were comparable in all cases. Plates were read at 405 nm using a Varioskan Instrument (Thermo Fischer Scientific). LPS (cat# L2630) and zymosan (cat# Z4250) were purchased from Sigma-Aldrich (St. Louis, MO). Pam3CSK4 (100 ng/ml), HKLM (5x107 cells), Poly I:C HMW (2 μg/ml), Poly I:C LMW (2 μg/ml), FLA-ST (500 ng/ml), FSL-1 (50 ng/ml), ssRNA (40 μg/ml), ODN-GPG (5 μM) were all purchased from Invivogen (cat# tlrl-kit1mw). PMA (50 μg/ml) was purchased from EMD Millipore (cat# 524400).

In vivo measurement of prostaglandins using mass spectrometry


HPLC-grade methanol, acetonitrile, and water used for sample purification and LC–MS/MS analysis were JT Baker-brand (ThermoFisher Scientific). The deuterated internal standards PGE2-d4, PGD2-d4, PGF-d4, 6-keto-PGF-d4, thromboxane B2-d4 and 15-HETE-d8 were purchased from Cayman Chemicals. Authentic standards of these analytes were also purchased from Cayman. Formic and acetic acid were purchased from Sigma-Aldrich. Acquity chromatography columns were purchased from Waters Corp. (Milford, MA). Solid phase extraction (SPE) cartridges (Strata C18-E, 200 mg/6 ml) were purchased from Phenomenex (Torrance, CA). LC–MS/MS analysis was performed on a Shimadzu Nexera system in-line with a SCIEX 6500 QTrap. The QTrap was equipped with a TurboV Ionspray source and operated in negative ion mode. SCIEX Analyst software (ver 1.6.2) was used to control the instruments and acquire and process the data.

Sample collection and purification

Peritoneal lavages were obtained by flushing the abdominal cavity with 3 ml of PBS (pH 7.4). The wash was centrifuged at 400×g for 5 min and supernatants were stored at −80°C. One milliliter of cell-free supernatant was removed from −80°C storage, placed on ice, and allowed to thaw. A 10-μl aliquot of internal standard solution was then added to each sample and the sample was vortexed. The samples were then transferred to a 13 × 100-mm test tube. The original sample vials were rinsed with 0.2 ml of 1:1 water:methanol and vortexed. This rinse was also transferred to the 13 × 100-mm test tube. The transferred samples were then diluted with 4.0 ml of 2% acetic acid (aqueous). The diluted samples were applied to pre-conditioned Strata C18 SPE cartridges. The loaded cartridges were washed with 3 ml of 2% acetic acid (aq) followed by 2 ml of 2% acetic acid (aq) with 10% methanol. Finally, the analytes were eluted from cartridges into 12 × 75-mm test tubes with 2 ml of acetonitrile. The samples were dried under nitrogen, capped and stored at −20°C until LC–MS analysis.

Prostaglandin Analysis

Immediately prior to LC–MS analysis, the samples were reconstituted in 50 μl of MeOH and 60 μl of H2O, vortexed and transferred onto a 96-well plate. The samples were analyzed on the above-referenced LC–MS system. The analytes were chromatographed on an Acquity UPLC BEH C18 reversed-phase column (5.0- × 0.21-cm; 1.7 μm) which was held at 40°C. A gradient elution profile was applied to each sample; after an initial hold for 0.4 min, %B was increased from 8% (initial conditions) to 70% over 4.0 min, and held at 70% for 1.1 min. Then the column was returned to initial conditions for 1.5 min prior to the next injection. The flow rate was 330 μl/min and component A was water with 0.05% formic acid, whereas component B was acetonitrile with 0.05% formic acid.

All analytes were detected by the SCIEX 6500 QTrap via selected reaction monitoring (SRM) in negative ion mode. The SRM transition (Da) and collision energy (volts) are given for each analyte: PGE2 (351.2 to 233.1, –15); PGE2-d4 (355.2 to 237.1, –15); PGD2 (351.2 to 189.1, –28); PGD2-d4 (355.2 to 193.1, –28); PGF (353.2 to 193.1, –34); PGF-d4 (357.2 to 197.1, –34); 6-keto-PGF (369.2 to 163.1, –34); 6-keto-PGF-d4 (373.2 to 167.1, –34); Thromboxane B2 (369.2 to 169.1, –20); Thromboxane B2-d4 (373.2 to 173.1, –20); 15-HETE (319.2 to 219.1, –25); 15-HETE-d8 (327.2 to 226.1, –25). Analytes were quantitated by stable isotope dilution against their deuterated internal standard. Data were normalized to total number of peritoneal lavage cells recovered.

ChIP assays

Human monocyte-derived DC were incubated in complete RPMI medium (11.1 mM glucose and 4 mM l-glutamine) in the presence or absence of 10 mM 2-DG and treated with 1 mg/ml zymosan, as previously described (23). Cells were then washed and fixed in 1% formaldehyde for ChIP assays. Cross-linking was terminated using 0.125 M glycine. Nuclear extracts were collected and resuspended in a lysis buffer containing a high salt concentration. Chromatin sonication was carried out using a Bioruptor device from Diagenode (Liege, Belgium). The chromatin solution was precleared by adding Protein A/G PLUS-Agarose for 30 min at 4°C under continuous rotation. After bead removal, isotype control or anti-human XBP1s antibodies (Clone Poly6195, Biolegend) was added for overnight incubation at 4°C, and then incubation with Protein A/G PLUS-Agarose was carried out for 2 hours at 4°C. Beads were pelleted by centrifugation at 16,000×g and sequentially washed with lysis buffer high salt, wash buffer, and elution buffer. Cross-links were reversed by heating at 67°C in a water bath, and the DNA bound to the beads isolated by extraction with phenol/chloroform/isoamylalcohol. Irrelevant antibody (Ab) and sequences of the Pri-miR-21 promoter were used as control of binding specificity. The IRE1α-specific inhibitor utilized in these assays was MKC8866 (2427). Results are expressed as percentage of input. Primers used for ChIP-PCR are in table S2. Primer sequences are numbered in bp from the transcription initiation site, but in the case of Pri-miR-21 it is numbered from the mRNA sequence, which is encoded in chromosome 17, GRCh38.p7. This was selected because of its lack of putative XBP1s-binding sequences.

RNA-sequencing and bioinformatic analyses

RNA was isolated using RNeasy MinElute kit (Qiagen) from LPS- or zymosan-stimulated murine bone marrow-derived DCs. All samples passed RNA quality control examined by Agilent Bioanalyzer 2100, and mRNA libraries were generated and sequenced at the Weill Cornell Epigenomics Core Facility. RNA-sequencing (RNA-seq) data was aligned using bowtie2 (58) against the mm10 mouse genome and RSEM v1.2.12 software (59) was used to estimate gene-level read counts using Ensemble transcriptome information. DESeq2 (60) was used to estimate significance of differential expression difference between any two experimental groups and gene expression changes of at least 1.2 fold were considered significant if passed false discovery rate (FDR) < 5% thresholds. Gene set enrichment analysis was performed using QIAGEN’s Ingenuity® Pathway Analysis software (IPA®, QIAGEN Redwood City, using “Canonical Pathways,” “Diseases & Functions,” and “Upstream Regulators” options. Enrichment results with at least ten deregulated genes were considered and pathways that passed FDR<5%, functions with P-value < 10−7 and regulators with P-value<0.001 were considered significant. Only functions and regulators with significant predicted activation states (|Z-score|>2) were reported. Functions were additionally filtered to remove entries specific to cancer cell lines and immune cell types. Significance of overlap was calculated with hypergeometric test. RNA-seq data was deposited to Gene Expression Omnibus (GEO) ( under accession no. GSE131404.

Gene editing in human monocyte-derived DCs

Human CD14+ monocytes were isolated from peripheral blood and plated at a density of 5 × 106 cells in 3 ml of RPMI supplemented with human recombinant GM-CSF at 1000 IU/ml and IL-4 at 500 IU/ml as described above. On day 6, DCs were prepared for transfection by washing with serum-free PBS and resuspension in RPMI medium supplemented with human recombinant GM-CSF and IL-4, at the same concentrations described above. DCs were then reverse-transfected on a 96-well plate by adding 2.5 × 105 cells in suspension onto 150 nM sgRNA-CAS9 ribonucleoprotein complexes containing lipofectamine CRISPRMAX transfection reagent (Invitrogen). All materials for sgRNA-Cas9 complex generation were purchased from Integrated DNA Technologies, and prepared as instructed (61). The final sgRNA-Cas9 and CRISPRMAX complex concentrations per well were 50 nM and 1% (vol/vol), respectively. Forty-eight hours post-transfection, genetic ablation of target genes was assessed via quantitative RT-PCR.

The 20-nucleotide CRISPR-RNA (crRNA) targeting human XBP1 (Homo sapiens chromosome 22, GRCh38.p12, NC_000022.11) is directed at the genomic sequence 5′-TGCACGTAGTCTGAGTGCTGCGG-3′ (the 3 additional nucleotides highlighted in bold represent the protospacer adjacent motif, or PAM). This target sequence corresponds to exon 4 of the human XBP1 transcript and was manually chosen by identifying a 20-base pair fragment immediately upstream of the highlighted PAM (62). The PAM was selected such that Cas9-mediated target DNA cleavage and resulting nucleotide would perturb XBP1u recognition and splicing by activated IRE1α (2, 63). The most likely on- and off-target effects of the manually selected CRISPR sequence were then analyzed using the Broad Institute’s Genetic Perturbation Platform (64). To validate the genomic editing capacity of the crRNA, quantitative RT-PCR was performed on total RNA isolated from cells transfected with sgRNA-Cas9 complexes containing the XBP1 crRNA described above. The reverse primer for XBP1s quantification via quantitative RT-PCR anneals to the same nucleotides as the XBP1 crRNA target site. Therefore, the primers can only efficiently amplify intact, unperturbed XBP1s cDNAs. The primers for evaluating deletion efficacy are listed in Table S2.

The genomic target sequence for the crRNA directed at human ERN1 (Homo sapiens chromosome 17, GRCh38.p12, NC_000017.11) is 5′-ATGTAGAGGATTCCATCTGACCC-3′. This sequence was generated and chosen using the Zhang Lab’s crRNA design tool (65). To validate the genomic editing capacity of this crRNA, quantitative RT-PCR was performed on total RNA isolated from cells transfected with sgRNA-Cas9 complexes containing ERN1 crRNA. XBP1s levels were used to assess the genetic perturbation of IRE1α (2, 63), using the primer pair specified in table S2. The scrambled crRNA contains a 20-nucleotide sequence that was computationally designed to be non-targeting within the human genome (61). The RNA sequence for this non-targeting control is 5′-CGUUAAUCGCGUAUAAUACG-3′.

Plasmid constructs and luciferase reporter assays

Expression constructs used for luciferase-based assays are pcDNA3.1 XBP1s (NM_001079539.1), pcDNA3.1 CHOP (NM_001195053.1) while reporter constructs used are pGL3-PTGS2 promoter (−1.2 kb/+137) and pGL3-PTGES promoter (−1.3 kb/+35). All plasmids were generated at VectorBuilder.

For dual luciferase assays, 2 × 104 HEK293FT cells were plated overnight in a 96-well plate and were transfected with the indicated plasmids using Lipofectamine 3000 (Thermo Fischer Scientific). Eighteen ng of reporter and 2 ng of Renilla plasmid were cotransfected with various ratios (w:w) of expression plasmids (reporter: expression plasmid = 1:1, 1:3 or 1:5) and pcDNA3.1, which was added to reach a total of 200 ng of DNA/well. After 36 to 48 hours, cells were washed with PBS and were lysed in Passive Lysis Buffer according to the manufacturer's protocol (Dual luciferase reporter assay system, Promega, catalog #E1960) (3). Firefly and Renilla luciferase activities were measured in 96-well plates using an automated luminometer (Luminoskan Ascent, Thermo Fischer). Firefly luciferase activity was normalized to its own Renilla activity.

Administration of pharmacological inhibitors

C57BL/6J mice were twice injected intraperitoneally with small-molecule inhibitors (at 6 hours and then 30 min) before acetic acid challenge or paw incision. KIRA6 (Medchem Express, cat# HY-19708), MKC8866 (Medchem Express, cat# HY-104040), or celecoxib (Sigma-Aldrich, cat# PZ0008) were used as indicated. The vehicle used for KIRA6, MKC8866, and celecoxib administration was 7% Tween-80, 3% ethanol, and 90% saline, as previously described (36). Briefly, anhydrous compounds were dissolved at 50 mg/ml concentration in dimethyl sulfoxide (DMSO), and then further dissolved in the vehicle at the desired concentrations so that a final volume of 200 μl was administered to each mouse via intraperitoneal injection.

Single-cell suspensions from mouse paws

Mice were perfused transcardially with 20 ml of 0.1 M phosphate buffer one day after paw incision. Both anterior and posterior parts of the injured or non-injured paw were dissected in a petri dish containing 2 ml of RPMI 164 medium (Gibco). Tissue was dissected into small pieces using surgical scissors, then transferred to a tube containing 2 ml of 0.5 mg/ml of Type II collagenase (Worthigton) in RPMI 1640 (Gibco) and incubated for 2 hours at 37°C shaking at 700 rpm. The enzymatic reaction was stopped by adding 4 ml of 2% FBS (Sigma) in 0.1 M phosphate buffer. Digested tissue was passed through a 40-μm nylon mesh (BD Biosciences) using a syringe plunger. Cell suspension was centrifuged at 450×g for 5 min at 4°C and resuspended in 1 ml of 2% FBS in 0.1 M phosphate buffer. Total cell number and cellular viability were determined using trypan blue staining and a hemocytometer. Cells were stored at −80°C in FBS containing 10% DMSO until subsequent flow cytometry analyses were performed.

Paw incision surgery in mice

Plantar incision surgery was performed as previously described (35). Briefly, mice were anesthetized with isoflurane in oxygen (4% induction, 1.5 to 2% for maintenance) and the right hind paw was aseptically cleaned with 10% povidone–iodine solution. Then, a 5-mm incision was made in the glabrous hind-paw skin from the heel to the base of the toes using a No. 11 scalpel and sterile technique. The underlying muscle and ligaments were elevated with a curved forceps and stretched for 6 to 8 s, without incising them. The incision was closed using 5.0 nylon mattress sutures.

Paw inflammation

Paw perimeter was measured in both left and right hind paws before the surgery and after every behavioral evaluation. The procedure was performed in a consistent manner using a 4.0 silk thread, which was placed around the center of the surgery in the right paw and at the same level in the paw contralateral to surgery. An increase in the paw perimeter was considered to reflect augmented inflammation of the affected paw.

Behavioral tests

All behavioral measurements were performed by a blinded observer before and after surgery (postoperative days 1 to 21) or acetic acid intraperitoneal injection (0 to 30 min). Animals were acclimated to the testing devices and/or places for 3 days, and baseline measurements were taken for at least 4 consecutive days before surgery or acetic acid injection.

Writhing spontaneous pain behaviors were evaluated after intraperitoneal injection of 0.9% acetic acid (v/v, 5 ml/kg). The number of writhing responses was quantified immediately after acetic acid injection for 30 min in 5-min intervals by an observer blinded to genotype. Writhings induced by acetic acid are overt stretching behaviors indicative of abdominal pain, a phenomenon that is dependent upon mPGES-1 and PGE2 (31, 32). Spontaneous exploratory activity (ambulatory counts and ambulatory time) was measured for 15 min immediately after injection of acetic acid in independent groups. We used a commercially available equipment and software, which was set according to the manufacturer's instructions (Med Associates, St. Albans, VT). Briefly, mice were placed in the center of an acrylic box (42.5 × 42.5 × 37.5 cm) with banks of 16 beams transmitters spaced 2.5 cm apart in both the x and y directions and 7 cm above the floor of the box (z plane).

Various spontaneous (non-evoked) pain-related behaviors were also evaluated postsurgery: weight bearing, grimace score, guarding score, rearing, and flinching activity. Mice were placed in individual acrylic chambers on an elevated mesh floor for 30 to 45 min before testing. Following the acclimatization period, the number of total vertical rearings and paw flinches were quantified during a 2-min period. Vertical rearings were defined as the number of times that the animal stood supporting its weight on both hind limbs. Vertical rearings are a normal behavior in rodents. Thus, a reduction in this behavior was indicative of a protective response to pain produced by movement, which mimics pain induced by surgeries in humans. Spontaneous flinching of the affected paw was quantified every time that the animal shaked the affected paw without any stimulation. Flinches of the injured paw is a pain-related behavior that is indicative of breakthrough pain, similar to intense spontaneous spike of pain in humans with postoperative pain. When all animals had been evaluated for flinching behaviors, they were observed again in sequential order for guarding behaviors using the guarding score method as follows (66). Mice were observed for 1 min every 10 min for a total of three times and scores are averaged for a final value. A score was given depending on the position in which the injured paw is found during the majority of the 1-min scoring period, and the following score system was used: 0 = full weight bearing of the paw is present if the wound is blanched or distorted by the mesh; 1 = the area of the wound touched the mesh without blanching or distorting; and 2 = the paw is completely off the mesh. After completion of guarding assessment, the animals were observed again to evaluate facial expressions using the Mouse Grimace Scale (67). Five facial expression parameters were evaluated: orbital tightening, nose bulge, cheek bulge, ear position (ear pulled backward), and whisker position (backward, forward, clumped together). Each parameter was scored using the following scale: 0 = not present; 1 = moderately present, and 2 = severely present. This evaluation was performed two times 5 min apart, and the values were averaged to have a single score per time point after surgery. These values were used to subtract the baseline value of the grimace scale (before surgery), and therefore the presented value for each animal is the mouse grimace scale difference score.

Mechanical hypersensitivity, an evoked pain-related behavior, was evaluated after all previous assessments. Mechanical withdrawal thresholds were calculated using the up-down method and applying force with calibrated von Frey filaments (0.07 G, 0.17 G, 0.40 G, 0.60 G, 1.04 G, 1.37 G, and 2.0 G, Stoeling, Wood Dale, IL, United States) to the plantar aspect of the paw for 5 s. Withdrawal of paws or flinching in response to a given applied force was noted as a positive response.

Hind paw weight bearing distribution was determined using an incapacitance tester apparatus (Stoelting, IL, version 5.64). This is a test for non-reflexive behaviors that represents a spontaneous pain-related behavior that mimics postoperative pain behaviors in humans (protection of the surgery site from normal activities). Before surgery, animals were habituated for at least 3 days to the apparatus, in which animals stand with each hind paw resting on individual weight plates inside an acrylic chamber. The apparatus measures the body weight distributed between the two hind paws over a 3-s period, and provide the average measurement. We used the average value of each hind paw to determine the weight distribution ratio (ipsilateral/contralateral side). A ratio below one indicates a greater weight bearing on the contralateral paw and was therefore considered as a pain-related behavior.


Mice were anesthetized with isoflurane (3 to 4% in oxygen) and perfused transcardially with 20 ml of filtered 0.1 M phosphate-buffered saline (PBS) followed by 20 ml of 4% formaldehyde. Tissue around the injured paw was collected by making a rectangular incision around the injury about 1.5 mm apart from the center of the surgery. Skin and muscle associated with the incision were collected and postfixed for 3 hours in 4% formaldehyde at 4°C. Tissue was stored at 4°C in 30% sucrose solution for 72 hours before sectioning. Slices of tissue were cut at 18 μm using optimal cutting temperature compound (Sakura Finetek) in a Leica cryostat and placed on coated slides (VWR International). Slides were then washed three times for 5 min with 0.1 M PBS and blocked using a solution of 3% normal donkey serum (NDS) + 0.3% Triton X-100 in 0.1 M PBS for 1 hour at room temperature. Primary antibodies used were rabbit anti-Cox-2 (Cell Signaling, catalog #12282, 0.203 μg/ml) and rat anti-CD45 (BioRad, catalog #MCA1388, 5 μg/ml). Tissues with primary antibodies were incubated overnight at 4°C. Tissues were then washed three times for 5 min with 0.1 M PBS and incubated 2 hours at room temperature with corresponding secondary antibodies: Cyanine 2-conjugated donkey anti-rabbit immunoglobulin G (IgG) (Jackson Immuno Research Labs, catalog #711-225-152, 3.75 μg/ml) and Cyanine 3-conjugated donkey anti-rat IgG (Jackson Immuno Research Labs, catalog #712-165-150, 3.75 μg/ml). Finally, slides were rinsed three times and mounted using anti-fade medium containing DAPI (Invitrogen) to allow visualization of cell nuclei.

At least three images per slide were taken at 20X from areas adjacent to the incision, using a Nikon Eclipse Ni fluorescent microscope system (Nikon, Japan) equipped with a Nikon DS-Qi1Mc digital camera (Nikon). All images were acquired using the NIS elements software Version 4.2 (Nikon, Japan). In each micrograph, CD45+ or Cox-2+ cells were quantified by a blinded observer in three random squares of 100 μm2 each using Adobe Photoshop CS6 (Adobe Systems). The percentage of Cox-2+ cells was then calculated in relationship to the total CD45+ cells by a blinded observer. For colocalization studies, images were acquired with an Olympus FV1200 confocal microscope using a 40X objective lens (UPLFLN40XO/1.3) and 405 nm, 488 nm, and 559 nm laser lines. Confocal images were prepared with Olympus Fluoview Version 4.2b software and Adobe Photoshop software. All images were taken from adjacent areas of the surgical wound ipsilateral to paw incision.

Statistical analyses

All statistical analyses were performed using the GraphPad Prism 7.0 software. Comparisons between two groups were assessed using unpaired two-tailed Student’s t-test, unless otherwise stated. All grouped data (time course) were analyzed using two-way analysis of variance (ANOVA) and Sidak’s multiple comparisons test. Grimace and guarding scores postsurgery were analyzed using the Mann–Whitney U test. Data are presented as mean ± standard error of the mean (SEM), unless otherwise stated. P-values lower than 0.05 were considered to be statistically significant.

Supplementary Materials

  • * Present Address: Vertex Ventures HC, 345 California Avenue, Palo Alto, CA 94306, USA.

References and Notes

Acknowledgments: We thank J. McCormick for expert assistance with flow cytometry and all members of the Weill Cornell Epigenomics Core Facility for help with RNA-seq. We thank K. Maddipati and the lipidomics facility at Wayne State University for their support with lipidomics experiments and analyses. We also thank D. Simmons, S. Bettigole, and all members of the Cubillos-Ruiz, Glimcher, and Romero-Sandoval laboratories for helpful suggestions and critical reading of this manuscript. We are grateful to D. K. Morales for creating the model shown in the summary page. Funding: Our research was supported by the Early-Career Investigator Award W81XWH-16-1-0438 of the Department of Defense (J.R.C.-R.), The Pershing Square Sohn Cancer Research Alliance (J.R.C.-R.), Weill Cornell Medicine Funds (J.R.C.-R. and L.H.G.), Department of Anesthesiology-Wake Forest School of Medicine Funds (E.A.R.-S.), NIH grant R01CA112663 (L.H.G.), Plan Nacional de Salud y Farmacia Grant SAF2017-83079-R (M.S.C.), and NIH grant 1S10OD017997-01A1 (L.J.M.). Author contributions: S.C. conceived, designed and conducted in vitro and in vivo experiments, analyzed data, and wrote the manuscript. P.G. designed and performed CRISPR-related experiments along with other in vitro and in vivo analyses. P.A.A.-V., M.M.F., and E.A.R.-S. performed pain-related in vivo and in vitro experiments, conducted behavioral tests in mice, and analyzed data. M.S. performed in vitro and in vivo experiments. L.J. performed analysis of lipidomic data. S.A. and M.S.C. designed and performed experiments related to ChIP-PCR. S.G. and M.M.F. performed pain-related and immunofluorescence experiments. C-S.C., T.A.S., and C.T. performed diverse in vitro experiments. T.I. provided Ern1-floxed transgenic mice. P.J.K. and L.J.M. performed and analyzed mass spectrometry assays for in vivo samples. A.V.K. carried out advanced computational analyses of RNA-seq data. A.J.D. and K.S. contributed with ideas and experimental systems, analyzed data, and reviewed the manuscript. L.H.G. provided Xbp1-floxed mice and critical resources, analyzed and interpreted data, and reviewed the manuscript. E.A.R.-S. and J.R.C.-R. conceived and designed the research, performed experiments, analyzed and interpreted data, wrote the manuscript, and administered the project. Competing interests: L.H.G. and J.R.C.-R are cofounders of and scientific advisors for Quentis Therapeutics. L.H.G. is a former director of Bristol-Myers Squibb Pharmaceuticals, currently serves on the Board of Directors of and holds equity in GlaxoSmithKline Pharmaceuticals and Waters Company, and is on the scientific advisory boards of Abpro, Kaleido, and Repare Therapeutics. Data and materials availability: All data are available in the main text or the supplementary materials. RNA-seq data was deposited to GEO ( under accession no GSE131404. Ern1-floxed or Xbp1-floxed mice may be obtained under a materials transfer agreement from T.I. or L.H.G., respectively.
View Abstract

Stay Connected to Science

Navigate This Article