A novel mode of action for COX‐2 inhibition: Targeting ATPase domain of HSP90 induces ubiquitin degradation of new client protein COX‐2

Dear Editor, Cyclooxygenase 2 (COX-2) is the main target of nonsteroidal anti-inflammatory drugs (NSAIDs),1 and it is rapidly expressed in response to extracellular factor stimulation like lipopolysaccharide (LPS) in mouse monocyte macrophage 264.7 (RAW264.7) cells.2 Herein,we reveal the existence of a novel mechanism that intervenes the formation of the heat shock protein 90 (HSP90)/COX-2 complex and induces the ubiquitin-proteasomal degradation of COX-2. Baicalein, the aglycone of baicalin, is a key antipyretic component of the classic traditional Chinese medicinal herb Scutellaria baicalensis Georgi (Huangqin),3 showed better antipyretic effects and nitric oxide inhibition than baicalin (Figure 1A-C). We synthesized and evaluated several molecular probes based on baicalein (Figure 1D, Figures S1-S7), and the potential targets captured by baicalein probe modified magnetic microspheres were identified by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, protein profiling analysis, and cell co-localization imaging (Figure 1E-G). The results of these analyses showed that HSP90 was the most likely target of baicalein. HSP90, the most common chaperone protein, could help its client proteins to fold correctly to exert their biological activity.4 Then, two known client proteins of HSP90—JNK and AKT—were dephosphorylated by baicalein (Figure 1H). The binding affinity between baicalein and HSP90, as detected by surface plasmon resonance (SPR), was approximately 26.07 μM (Figure 2A, Figure S8), while baicalin could not bind to HSP90 (Figure S9). The adenosine triphosphatase (ATPase) domain is considered the most important in HSP90. When HSP90 binds to adenosinetriphosphate (ATP), transient dimerization occurs in the N-terminal region, leading to HSP90 inactivation.5 The binding affinity between ATP and HSP90 was 0.6 mM, as detected by SPR; this affinity increased approximately six-fold to 3.5 mM when baicalein competed with ATP

to bind to HSP90 ( Figure 2B). Geldanamycin (Gel), an HSP90 ATPase domain inhibitor, weakened the fluorescence resonance energy transfer of the baicalein probecoumarin/HSP90 ( Figure 2C, Figure S10). Moreover, baicalein markedly inhibited the ATPase activity of HSP90 ( Figure 2D). 6 Molecular docking studies illustrated that baicalein interacted with the amino acid residues-SER-52, ASP-93, PHE-138, and THR-184-in the ATPase domain of HSP90 with a binding energy of −8.19 kcal/mol ( Figure 2E). These key amino acid sites are highly conserved in human, rat, mouse and horse species ( Figure 2F). [7][8][9] The on-off image demonstrated that baicalein had a faster binding rate and slower dissociation rate than the mutants in the SPR assay ( Figure 2G, Figures S11-S14). In addition, baicalein caused a 0.8 • C decrease in the protein melting temperature (ΔT) ( Figure 2H). Circular dichroism spectroscopy revealed that the optical activity of HSP90 was partially changed after incubation with baicalein ( Figure 2I). These results suggest that baicalein targets the ATP-binding domain of HSP90.
A co-immunoprecipitation (Co-IP) test was designed to explore any unknown potential client proteins of HSP90 responsible for the antipyretic effects exerted by baicalein ( Figure S15). The results of this test demonstrated that COX-2, instead of COX-1, could be captured by HSP90 in LPS-stimulated RAW264.7 cells ( Figure 3A). By integrating data on Co-IP protein profiling (Table S1), client proteins of HSP90 and proteins related to inflammation and heat-clearing from GeneCards, we constructed a Venn diagram, which suggested COX-2 as a new client protein of HSP90 ( Figure 3B). The binding affinity between COX-2 and HSP90 was 1.17 μM, as detected by microscale thermophoresis. After pretreatment with baicalein, the specific binding aforementioned almost disappeared ( Figure 3C, Figure S16).
Next, molecular docking analysis and molecular dynamics simulation were performed, which estimated the ∆G bind values of the HSP90/COX-2 and HSP90/baicalein+COX-2 complexes as −55.9 kcal/mol and −47.5 kcal/mol, respectively. The potential binding sites contributed to ∆E ele that weakened by baicalein were suggested as GLU-62, LEU-209 in HSP90 and THR-70, ARG-95 in COX-2 ( Figure 3D,E and Figures S17-S19; Table  S2). Transmission electron microscopy and Co-IP results further demonstrated that baicalein promoted the dissociation of HSP90 and COX-2 ( Figure 3F,G). Since COX-2 is a substrate of the ubiquitin-proteasome system, 10 the evidence of colocalization of HSP90 and COX-2 revealed that baicalein instead of baicalin induced COX-2 degradation, which could be reversed by the ubiquitin-proteasome inhibitor MG-132 ( Figure 3H). Moreover, COX-2 levels were decreased by baicalein administration ( Figure S20).
Almost no normal COX-2 protein was expressed in RAW264.7 cells, which were stimulated by LPS under baicalein treatment for 120 min, and the proteins sized approximately 35 kDa were predicted to be ubiquitindegraded COX-2 proteins ( Figure 4A). Consistent with this, several ubiquitin-conjugates were formed following MG-132 treatment ( Figure 4B). As expected, the expression of COX-1 and HSP90 at the protein level changed minimally ( Figure 4C,D). The degradation of COX-2 induced by baicalein almost disappeared when the cells were pretreated with HSP 90α/β siRNA ( Figure S21). Hence, baicalein could decrease the expression of downstream effectors of COX-2, including prostaglandin E2 (PGE2), interleukin-6 (IL-6) and tumor necrosis factor-α (TNF-α), in LPS-stimulated RAW264.7 cells by inhibiting HSP90 in a manner similar to that of Gel ( Figure 4E-G).
Moreover, the effects of some known HSP90 inhibitors, such as Gel, AT13387, cisplatin (Cis), cucurbitacin D and five representative NSAIDs (aspirin [ASP], paracetamol, ibuprofen, indometacin and mefenamic acid), on COX-2 degradation were compared. Only HSP90 ATPase domain inhibitors markedly decreased the expression of COX-2 at the protein level ( Figure 4H). In a rat model of pyrexia, the HSP90 ATPase domain inhibitors, Gel (1 mg/kg) and baicalein (1 mg/kg), exhibited significant antipyretic activity compared with Cis (1 mg/kg) and ASP (20 mg/kg) within 30 min. Interestingly, this effect lasted for 240 min ( Figure 4I, Figure S22). Meanwhile, MG-132 reduced the antipyretic effect of baicalein ( Figure S23). HSP90 ATPase domain inhibitors inhibited the increase in COX-2 levels induced by LPS more stably and durably than ASP (Figure 4J). However, there was no effect on COX-1 levels under the same conditions ( Figure 4K). Moreover, PGE2, IL-6 and TNF-α levels in the serum demonstrated the same effects at the cellular level ( Figure 4L-N).
In conclusion, COX-2 was identified as a new client protein of HSP90. HSP90 ATPase domain inhibitors trigger the dissociation of the HSP90/COX-2 complex and induce the degradation of COX-2, which plays an antipyretic role (Figure 4O). Our findings demonstrate another mechanism for inhibiting COX-2 and present a new strategy for the development of antipyretic and analgesic drugs.

C O N F L I C T O F I N T E R E S T
The authors have declared that no competing interest exists.

F I G U R E 1
The antipyretic activity evaluation of baicalein, the target fishing, and co-localization of baicalein with its target protein