Experimental Anti-Inflammatory Drug Semapimod Inhibits TLR Signaling by Targeting the TLR Chaperone gp96

Semapimod (CNI-1493, N, N’-bis [3, 5-diacetylphenyl] decanediamide tetrakis [amidinohydrazone]) was initially designed as a bulky arginine mimetic to limit arginine transport and NO production during inflammation (1). In addition to the expected inhibition of inflammatory cytokine-induced arginine transport in macrophages, Semapimod attenuated inflammation and protected against lethal endotoxemia (1). Inhibition of arginine uptake and NO production was not the only mechanism responsible for the anti-inflammatory effect of Semapimod: the drug inhibited LPS-induced inflammatory cytokine release by macrophages at concentrations at least 10 times lower than that required for the inhibition of arginine uptake (2). Semapimod inhibits inflammatory responses not only in macrophages/monocytes, but also in other cell types, including endothelial cells (3), dendritic cells (4), and enterocytes (5). Since its discovery, Semapimod has been reported to have a beneficial effect in a broad range of experimental and clinical inflammatory conditions, such as acute endotoxemia (6, 7), bacterial infection (8), malaria (9), arthritis (10, 11), autoimmune encephalomyelitis (12), Alzheimer disease (13), pancreatitis (14), allograft rejection (15), cancer (16, 17), postoperative ileus (18, 19), and Crohn disease (20, 21).

The mechanism of action of Semapimod is not well understood. It inhibits activating phosphorylation of MAPKs of p38, JNK, and ERK families in response to inflammatory stimuli (21–23), but it does not directly inhibit these kinases. Although Semapimod directly inhibits c-Raf (21), a MAPK kinase kinase upstream of ERK, this does not explain blockade of activation of p38 and JNK, which are independent of c-Raf. Semapimod has been found to directly inhibit deoxyhypusine synthase, an enzyme that catalyzes posttranslational modification of the translation initiation factor 5A, which might explain the antiviral and antimalaria effects (9, 24), but not the blockade of inflammatory cytokine production. Upon intracranial injection, Semapimod potently activates the cholinergic anti-inflammatory pathway involving the vagus nerve (25, 26); however, this does not explain the drug’s anti-inflammatory effects in vitro.

The intestinal epithelium becomes largely refractory to the TLR ligands following bacterial colonization, which has been extensively demonstrated in the adult, microbiota-associated intestine. However, the naive epithelium of the neonates possesses TLR responsiveness similar to that of the professional innate immune cells (5, 27–31). TLR signaling in the epithelium plays critical role in the pathogenesis of necrotizing enterocolitis, a disease coincident with the onset of bacterial colonization of the gut (29, 32–34). We are interested in Semapimod because it improves outcomes of experimental necrotizing enterocolitis (35). Because Semapimod is not absorbed in the intestine (19), it is an attractive drug for organ-targeted therapy of intestinal inflammatory disorders.

In this study, we demonstrate that in enterocytes, Semapimod inhibits TLRs by targeting their common molecular chaperone gp96.

IEC-6, HEK293, and SW480 cell lines were grown as recommended by the supplier (American Type Culture Collection, Manassas, VA). IEC-6 cells were used at passages 17–28. For all experiments, cells were plated at 4 to 5 × 10⁴/cm² and grown overnight to 70–90% confluence. Cell viability was determined by Trypan blue staining. Reagents were purchased from the following suppliers: Semapimod, Medkoo Biosciences (Chapel Hill, NC); recombinant canine gp96, catalog number ADI-SPP-766, Enzo Life Sciences (Farmingdale, NY); recombinant human HSP90, catalog number SPR-101A, StressMarq Biosciences (Victoria, BC, Canada); LPS from Escherichia coli 0127:B8, MG132, geldanamycin, radicicol, and N-ethyl carboxamidoadenosine (NECA), Sigma-Aldrich (St. Louis, MO); tripalmytoyl cysteine-serine-(lysine)₄ (Pam₃CSK₄), Tocris Bioscience (Bristol, U.K.); ultrapure flagellin from Salmonella typhimurium, Invivogen (San Diego, CA); recombinant rat IL-1β, PeproTech (Rocky Hill, NJ); peroxynitrite, Cayman Chemical (Ann Arbor, MI); and the ATP-desthiobiotin kit, Thermo Scientific (Rockford, IL). Abs were from the following sources: gp96 (H-212) and TLR4 (H80), Santa Cruz Biotechnology (Santa Cruz, CA); TLR9 (SAB2104136) and FLAG M2, Sigma-Aldrich; phospho-p38, p38, phospho-MAPK kinase 3/6, and IκBα, Cell Signaling Technology (Danvers, MA); cyclooxygenase-2 (COX-2), Cayman Chemical; inducible NO synthase (iNOS), BD Biosciences (San Jose, CA); MyD88, Abcam (Cambridge, MA); and HSP90, StressMarq Biosciences. Synthetic oligonucleotide 5′-TCGTCGTTTCGTCCGGCGCGCCGG-3′ was used as CpG DNA.

The mTLR4-Flag plasmid (catalog number 13087; constructed by R. Medzhitov) was obtained from Addgene (Cambridge, MA). HEK293 human embryonic kidney cells were transiently transfected using the calcium phosphate precipitate method. Control cells were transfected with the empty pFLAG-CMV2 vector (Sigma-Aldrich).

Standard procedures were used as described previously (5). For quantitative protein measurements, band densities on underexposed Western blots were determined using GelDoc imaging system and Quantity One software (Bio-Rad, Hercules, CA). Immunofluorescence images were acquired on LSM 700 confocal system (Carl Zeiss Microimaging, Thornwood, NY). For comparisons, sections were mounted and processed on the same slide. Identical acquisition settings and image adjustments were used. Surface and 1-μm subsurface signal intensity was measured using the ImageJ software (National Institutes of Health) by scanning randomly chosen cells along a horizontal line across the center of the nucleus.

The standard procedure was used (Santa Cruz Biotechnology), except that cells were lysed in Nonidet P-40 (NP-40) buffer (1% NP-40, 100 mmol NaCl, 20 mmol Tris [pH 8], and 0.5 mmol PMSF).

IEC-6 cells were lysed with NP-40 buffer containing 2.5 mmol MgCl₂ for 10 min at 4°C, and the lysate was cleared by high-speed centrifugation. Incubation with ATP-desthiobiotin, adsorption to streptavidin-agarose beads, washing, and elution were performed as recommended by the manufacturer. Eluted proteins were separated on 20-cm 10% polyacrylamide gel, and protein bands were visualized by silver staining. Protein bands were identified at CHLA Proteomics Core. Briefly, bands excised from gel were digested with trypsin, and resulting peptides were identified using liquid chromatography-mass spectroscopy. Protein identity was established by database search for matching peptides.

The 20-μl reactions (20 mmol HEPES [pH 7.2], 50 mmol NaCl, 2.5 mmol MgCl₂, 0.1 mmol DTT, 200 μmol [100 μCi/ml] ATP-γ-[³⁵S], and 100 μg/ml gp96, with or without Semapimod) were incubated 30 min at 37°C and loaded onto drained and precooled 1 ml Sephadex G25 spin columns (GE Healthcare Life Sciences). The columns were spun for 2 min at 2000 rpm and 4°C. Radioactivity of flow-through was determined by scintillation counting. In control reactions, gp96 was substituted for equivalent amount of autoclaved porcine collagen. Following background subtraction, data were expressed as percent of ATP-γ[³⁵S] binding in the absence of the inhibitor.

The 20-μl reactions (20 mmol HEPES [pH 7.2], 50 mmol NaCl, 2.5 mmol MgCl₂, 0.1 mmol DTT, 20 μmol [100 μCi/ml] γ-[³²P] ATP, and 100 μg/ml gp96 or HSP90, with or without Semapimod) were incubated 3 h at 37°C. 0.5 ml 5% w/v suspension of activated charcoal (Norit A) equilibrated with 20 mmol HEPES (pH 7.2), 50 mmol NaCl, 5 mmol EDTA, and 200 μmol ATP was mixed with samples, and after 10-min agitation at room temperature, activated charcoal was removed by centrifugation. Radioactivity of the clear supernatant was measured by Čerenkov counting. In control reactions, gp96 was substituted for equivalent amount of autoclaved porcine collagen. Following background subtraction, data were expressed as percent of charcoal-absorbed radioactivity in the absence of inhibitor.

Quantitative data were expressed as means ± SD. Data were compared using unpaired t test.

Because Semapimod has been reported to inhibit a variety of inflammatory responses, it was not clear what aspects of inflammation are affected. To define targets of Semapimod in the inflammatory cascade, we examined effects of this drug on inflammatory responses in intestinal epithelial cells elicited by a variety of stimuli. The IEC-6 cell line was chosen because these are primary untransformed epithelioid cells with presumably intact innate immune machinery. Pretreatment with Semapimod blocked LPS-induced, but not IL-1β–induced, activating phosphorylation of p38 MAPK and its upstream activators MAPK kinases 3/6 (Fig. 1A). These protein kinases are critical mediators of transcriptional induction of COX-2, the rate-limiting enzyme in the biosynthesis of inflammatory prostanoids. MAPK kinases 3/6 mediate inflammatory cytokine- and stress-induced, but not LPS-induced, expression of COX-2 in enterocytes (5). As expected, Semapimod also blocked LPS-induced but not IL-1β–induced expression of COX-2 (Fig. 1B). Semapimod failed to block IL-1β–induced expression of another key inflammatory factor, iNOS. iNOS was not appreciably induced by LPS either in the presence or absence of Semapimod (Fig. 1B). Induction of COX-2 by osmotic shock, oxidative stresses, or proteasome blockade was unaffected (Fig. 1C). Semapimod blocked activating phosphorylation of p38 MAPK induced by Pam₃CSK₄ and CpG DNA, the agonists of TLR2 and -9, respectively (Fig. 1D), but not by the TLR5 agonist flagellin (Fig. 1F). Semapimod did not significantly affect cell viability (Fig. 1E), which rules out effects of alarmons released from dead cells. Thus, Semapimod appears to specifically block responses mediated by a subset of TLRs including TLR2, -4, and -9.

To elucidate whether Semapimod blocks responses mediated by proinflammatory signaling pathways other than p38, we examined effects of this drug on activation of the NF-κB pathway, as judged by the levels of the inhibitory subunit IκBα. IκBα is rapidly degraded in response to inflammatory stimuli, which relieves inhibition of the NF-κB and leads to transcriptional activation of inflammatory genes. Semapimod blocked LPS-induced (Fig. 2A), but not IL-1β–induced (Fig. 2B), degradation of IκBα. Semapimod also blocked degradation of IκBα induced by CpG DNA and Pam₃CSK₄ (Fig. 2D), but not by flagellin (Fig. 2E). Thus, Semapimod blocks TLR ligand–induced activation of both p38 MAPK and NF-κB signaling, suggesting that the blockade occurs early in the signaling cascade, before branching into the p38 and NF-κB pathways.

According to Fig. 2A, Semapimod inhibits degradation of IκBα induced by LPS concentrations up to 1 μg/ml, but is ineffective at 5 μg/ml, suggesting that inhibition can be overcome by high concentration of LPS. To examine this effect in more detail, we studied inhibition at various concentrations of LPS and Semapimod. In the absence of Semapimod, LPS caused a dose-dependent increase in p38 phosphorylation at concentrations between 1 and 100 ng/ml (Fig. 3A). Semapimod concentrations from 0.02 to 10 μmol progressively shifted the response curve to the right (Fig. 3A). At any concentration up to 10 μmol, Semapimod failed to significantly inhibit response to ≥5 μg/ml LPS. The IC₅₀ of Semapimod for response to 20–1000 ng/ml LPS is ∼0.3 μmol (Fig. 3B). Similar results were obtained for inhibition of responses to CpG DNA, the agonist of TLR9 (Supplemental Fig. 1). Thus, Semapimod appears to desensitize responses to TLR ligands in a dose-dependent fashion.

We next examined the time course of inhibition. Semapimod inhibited LPS-induced IκBα degradation when applied before or simultaneously with, but not after application of LPS (Fig. 3C). Therefore, Semapimod acts almost instantaneously, but is ineffective once the response has been initiated.

Recruitment of the cytosolic adapter protein MyD88 to the plasma membrane-localized TLRs is one of the earliest events in inflammatory signaling (36). To test whether Semapimod affects this event, we examined changes in intracellular localization of MyD88 following stimulation with LPS, with or without pretreatment with Semapimod. In untreated cells, MyD88 localized to myddosomes, granule-like structures dispersed throughout the cell (37). LPS caused pronounced localization of MyD88 to the cell surface in the absence, but not in the presence of Semapimod (Fig. 4). Therefore, Semapimod blocks LPS-induced cell surface localization of MyD88, consistent with blockade of MyD88 recruitment to the TLR4 complexes on the cell surface.

Because TLRs associate with multiple ATP-binding proteins including protein kinases TNFR-associated factor 6, TAK1, RIP2 (38, 39), and heat shock proteins (40, 41), we performed a pulldown assay for ATP-binding proteins in the presence or absence of Semapimod. Cell lysates prepared from IEC-6 cells treated with or without LPS in the presence or absence of Semapimod were incubated with ATP-desthiobiotin, a reagent that covalently binds to ATP-binding sites in proteins. ATP-desthiobiotin–modified proteins were then collected on streptavidin-agarose beads, and bound proteins were analyzed by gel electrophoresis and silver staining. LPS treatment did not cause any detectable change in the spectra of bound proteins. However, Semapimod treatment resulted in disappearance of a prominent band of ∼100 kDa, regardless of LPS treatment (Fig. 5A). This band was excised, and the corresponding protein was identified as gp96 using trypsin digestion, liquid chromatography, and database-linked mass spectroscopy. The identity of the ∼100-kDa species as gp96 was further confirmed by Western blot. gp96-immunoreactive species was not recovered by ATP-desthiobiotin pulldown if cells were treated with Semapimod (Fig. 5B). Interestingly, Semapimod did not interfere with the ATP-desthiobiotin pulldown of HSP90, the cytosolic paralog of gp96 (Fig. 5B). These results demonstrate that Semapimod specifically abrogates modification of gp96 with ATP-desthiobiotin.

gp96 is a glycoprotein chaperone of the HSP90 family that facilitates folding, assembly, and trafficking of a limited number of client proteins, most notably TLR signaling complexes (42–46). gp96 resides in the endoplasmic reticulum (ER) and possesses intrinsic ATPase activity, which is required for its chaperone function (47–50). gp96 deficiency obliterates TLR signaling due to failure of trafficking of functional TLR complexes to the cell surface (44). It was therefore plausible that Semapimod inhibits multiple TLRs via its effect on gp96.

Protection from modification by ATP-desthiobiotin by Semapimod indirectly indicates that the latter inhibits ATP-binding and/or ATPase activity of gp96. To test such inhibition directly, we examined effects of Semapimod on ATP binding and ATPase activities of purified gp96 in vitro. ATP binding was measured by incubating the gp96 protein with ³⁵S-labeled ATP-γS, a nonhydrolysable analog of ATP, followed by spin column size-exclusion chromatography on Sephadex G25 (GE Healthcare Life Sciences), and determining radioactivity of the high m.w. (flow-through) fraction. Semapimod inhibited ATP binding in a dose-dependent fashion (IC₅₀ ≈0.4 μmol, Fig. 6A). ATPase activity was measured by incubating gp96 with γ-[³²P] ATP, followed by adsorption of released inorganic phosphate on activated charcoal and determining the radioactivity of unhydrolysed ATP. Semapimod inhibited gp96 ATPase in a dose-dependent fashion (IC₅₀ ≈0.3 μmol, Fig. 6B), which is close to the IC₅₀ for ATP binding and innate immune response inhibition in IEC-6 cells (Fig. 3, Supplemental Fig. 1). At the concentration that completely inhibits gp96 ATPase, Semapimod did not appreciably inhibit the ATPase activity of HSP90, the cytosolic paralog of gp96 (Fig. 6B, inset). Therefore, Semapimod directly and specifically inhibits ATP-binding and ATPase activities of gp96.

To gain an additional insight into the mechanism of action of Semapimod, we sought to elucidate whether known gp96 inhibitors block LPS signaling like Semapimod and whether Semapimod inhibits trafficking of TLR signaling complexes similar to other gp96 inhibitors. To answer the first question, we examined LPS-induced IκBα degradation and p38 MAPK phosphorylation in IEC-6 cells pretreated with geldanamycin, radicicol, or NECA. These drugs have been shown to inhibit gp96 by associating with its nucleotide-binding pocket (51–55). Geldanamycin and radicicol indeed inhibited responses to LPS, however, only upon prolonged exposure of ≥3 h (Fig. 7A). By contrast, Semapimod blocked LPS signaling almost instantaneously (Fig. 3C). NECA failed to block LPS signaling at any concentration tested up to 20 μmol (Fig. 7A). Although geldanamycin and radicicol inhibit LPS signaling, their effect is slow, which implies a different mechanism of action and is consistent with the blockade of TLR trafficking to the cell surface. To answer the second question, we examined effects of Semapimod on subcellular localization of TLR4 and TLR9 in SW480 enterocytes. This cell line of human origin allowed the use of proven anti-human TLR4 and TLR9 Abs. Responses of SW480 cells to LPS and CpG DNA are similar to those of IEC-6 cells. The bulk of TLRs localized to granules dispersed throughout the cell body, as previously reported (56, 57). Exposure to Semapimod caused dramatic accumulation of both TLRs in the perinuclear space (Fig. 7B), which is consistent with retention in the ER. Thus, Semapimod affects subcellular localization of TLRs in a manner expected of an inhibitor of gp96 chaperone function. In contrast, Semapimod is different from geldanamycin and radicicol in its rapid effect on TLR4 signaling, which is consistent with inhibition of the pre-existing TLR4 signaling complexes. Semapimod is similar to the other two gp96 inhibitors in its ability to block intracellular trafficking of the TLRs (Fig. 8).

Geldanamycin has been reported to disrupt gp96-TLR complexes, as judged by coimmunoprecipitation (42). To test whether Semapimod acts by dissociating gp96-TLR complexes, we transiently transfected HEK293 cells with FLAG-tagged TLR4 and examined effect of Semapimod on coimmunoprecipitation of gp96 and TLR4. Semapimod failed to block coimmunoprecipitation of these two proteins (Fig. 7C), indicating that its effects on TLR4 does not involve physical dissociation of gp96-TLR4 complexes.

In this report, we demonstrate that in primary enterocyte cell culture, Semapimod inhibits signaling by agonists of TLR2, -4, and -9, but not by TLR5 agonist, the inflammatory cytokine IL-1β, or cellular stresses. Effects of Semapimod on responses to LPS are dose-dependent with regard to concentrations of both Semapimod and LPS. In cells treated with Semapimod, higher concentrations of LPS are required to elicit the same response than in untreated cells, and Semapimod fails to block responses to LPS applied at concentrations of ≥5 μg/ml. gp96, an ER-associated chaperone of the HSP90 family, which is critically involved in assembly and trafficking of the TLR signaling complexes, was identified as a target of Semapimod using a pulldown assay for the ATP-binding proteins. Semapimod inhibits ATP binding and ATPase activity of purified gp96 in a dose-dependent fashion in vitro. Like the known gp96 inhibitors geldanamycin and radicicol, Semapimod impedes trafficking of TLR4. However, Semapimod inhibits TLR4 signaling much faster than geldanamycin or radicicol, which is consistent with direct inhibition of TLR signaling, but not with indirect effect via impaired receptor trafficking.

The fact that Semapimod does not alter the dose-response curves for TLR ligands, but rather shifts them to higher ligand concentrations, indicates that this drug acts by reducing the affinity of TLRs to their ligands rather than by inhibiting receptor signaling. Even at highest Semapimod concentrations tested, the responses, including p38 phosphorylation or IκBα degradation, were not inhibited as long as high concentrations of TLR ligands were used. This mode of action is consistent with a model by which the high, but not the low affinity of TLR receptor complexes to their cognate ligands depends on gp96 (Fig. 8).

Several lines of evidence argue that Semapimod blocks innate immune responses at the level of TLRs. First, because this drug inhibits neither IL-1β–induced activation of NF-κB or expression of iNOS, nor IL-1β– or stress-induced activation of p38 MAPK or expression of COX-2, it does not appear to directly target the intracellular NF-κB or p38 MAPK signaling cascades. The blockade of TLR ligand–induced IκBα degradation and p38 MAPK phosphorylation indicates that Semapimod targets a common upstream activator shared by NF-κB and p38 MAPK pathways, which is consistent with blockade at the receptor level. Second, Semapimod apparently decreases the affinity of the TLR4 signaling complex to its ligand, LPS. At concentrations up to 10 μmol, Semapimod has no effect on responses to LPS applied at high concentrations. Such behavior is most easily explained by targeting the receptor complex, but it is inconsistent with blockade of a downstream signaling mediator. Third, Semapimod blocks recruitment of the MyD88 adapter, the earliest detectable event in TLR signaling. Because TLRs and IL-1R share the key elements of their downstream signaling cascades (58), the failure of Semapimod to inhibit signaling from the IL-1R also argues against a downstream signaling mediator as target.

Using a pulldown assay for ATP-binding proteins, we have identified gp96, the ER paralog of the HSP90 chaperone, as a direct target of Semapimod. Judging by abrogation of gp96 pulldown by Semapimod in the ATP-desthiobiotin-streptavidin assay, this drug interferes with modification of gp96 by ATP-desthiobiotin, an ATP derivative that covalently attaches to the ATP-binding pockets in proteins. Blockade of ATP-desthiobiotin modification indirectly indicated that Semapimod inhibits the ATP-binding activity of gp96. To corroborate inhibition of ATP binding, we examined effects of Semapimod on ATP-binding and ATPase activities of purified gp96 and found that both were inhibited in a dose-dependent fashion and with similar IC₅₀. It remains unknown whether Semapimod inhibits ATP binding/ATPase activities of gp96 by interaction with the ATP-binding pocket or an allosteric site. X-ray crystallography is needed provide an answer to this question.

The facts that Semapimod inhibits TLR signaling at the receptor level and that it directly targets gp96, the essential chaperone for TLRs (45, 46), strongly argue that Semapimod inhibits TLR signaling via its effect on gp96. Close IC₅₀ values for inhibition of TLR signaling and for inhibition of gp96 ATP binding and ATPase activities provide further support for this idea. Insensitivity of IL-1β responses to Semapimod is in agreement with known independence of IL-1R biogenesis from gp96 (45). Although our data strongly implicate gp96, they cannot formally rule out other targets of Semapimod in TLR signaling. However, auxiliary proteins shared by IL-1R and TLRs, as well as receptor-specific auxiliary proteins such as MD-2 or CD14, can be excluded as targets. If main receptor subunits (TLR2, -4, and -9 proteins) are targets, one would expect different IC₅₀ for each receptor, which was not the case.

Semapimod is similar to geldanamycin and radicicol, the two known gp96 inhibitors, in its ability to block intracellular trafficking of the TLRs and thus to act as one would expect of a true gp96 inhibitor. However, unlike the other two inhibitors for which the effect on TLR4 signaling is slow, Semapimod inhibits TLR4 signaling fast. Unlike geldanamycin or radicicol, Semapimod fails to inhibit the ATPase activity of the closely related HSP90 and is thus not a generic HSP90 family inhibitor. A plausible explanation for Semapimod’s fast effect on responses to LPS is inhibition by this drug of gp96 associated with cell-surface high-affinity TLR signaling complexes. This explanation is consistent with increased sensitivity to LPS upon surface expression of gp96 protein (43) or treatment with low concentrations of extrinsic gp96 (59), as well as competition between gp96 and LPS for cell-surface binding (60).

Our results provide mechanistic explanation as to why Semapimod is effective in experimental NEC. The neonates are generally supersensitive to the TLR ligands (31). The naive intestinal epithelium of the neonates responds to TLR ligands (27, 28, 30), and these responses may play critical role in the pathogenesis of NEC (32–34, 61). Thus, Semapimod may prevent NEC by blocking TLR ligand signaling in the neonatal intestine.

Our study provides an insight into the roles of gp96 in TLR signaling complexes. It is generally believed that gp96 participates in TLR signaling by facilitating correct folding, plasma membrane insertion, and trafficking of receptor signaling complexes to the cell surface (40). If this is true, one could expect gp96 inhibitors to block TLR signaling slowly, as pre-existing plasma membrane TLR complexes should not be affected. Slow (hours) inhibition of responses to LPS is what we indeed observed using the classic gp96 inhibitors geldanamycin and radicicol. However, unlike these two inhibitors, Semapimod desensitizes LPS responses almost instantaneously, which indicates inhibition of signaling by pre-existing TLR complexes. Therefore, if Semapimod inhibits TLR signaling via gp96, the latter somehow participates in high-affinity TLR ligand sensing. The idea of gp96 participating in cell-surface TLR signaling is not entirely novel. Soluble gp96 and LPS have been shown to compete with each other for binding polymorphonuclear neutrophils (60). Externally added gp96, although not an effective TLR agonist per se, enhances responses to TLR2 and TLR4 ligands in dendritic cells (59). Forced surface expression of gp96 causes TLR4 hyperresponsiveness (43). LPS responses and LPS binding to cells are inhibited by an externally added peptide inhibitor of gp96 (62, 63). All of these data indicate that surface gp96 enhances TLR signaling. Although Semapimod acts like a classic gp96 chaperone inhibitor in its ability to block TLR trafficking, it possesses a novel property of rapidly inhibiting TLR signaling, presumably by targeting gp96 on the cell surface. Accordingly, Semapimod may find use as pharmacologic tool for dissecting the complex role of gp96 in TLR function.

Surprisingly, Semapimod does not affect responses to flagellin, the ligand of TLR5, despite the known dependence of TLR5 trafficking on gp96 (46). One might speculate that although TLR5 depends on gp96 for its biogenesis, gp96 may not be required for efficient flagellin sensing by the TLR5 receptor complex.

Identification of Semapimod as inhibitor of TLR signaling may shed light on the mechanisms of action of this drug in a variety of inflammatory disorders.

We thank Mary Beth Amrine, Alexandria Lee, and Rudolph Davis for help with experiments and Christopher Gayer and G. Esteban Fernandez for critical reading of the manuscript.

The authors have no financial conflicts of interest.