β Boswellic Acid Blocks Articular Innate Immune Responses: An In Silico and In Vitro Approach to Traditional Medicine

Osteoarthritis (OA) is hallmarked as a silent progressive rheumatic disease of the whole joint. The accumulation of inflammatory and catabolic factors such as IL6, TNFα, and COX2 drives the OA pathophysiology into cartilage degradation, synovia inflammation, and bone destruction. There is no clinical available OA treatment. Although traditional ayurvedic medicine has been using Boswellia serrata extracts (BSE) as an antirheumatic treatment for a millennium, none of the BSE components have been clinically approved. Recently, β boswellic acid (BBA) has been shown to reduce in vivo OA-cartilage loss through an unknown mechanism. We used computational pharmacology, proteomics, transcriptomics, and metabolomics to present solid evidence of BBA therapeutic properties in mouse and primary human OA joint cells. Specifically, BBA binds to the innate immune receptor Toll-like Receptor 4 (TLR4) complex and inhibits both TLR4 and Interleukin 1 Receptor (IL1R) signaling in OA chondrocytes, osteoblasts, and synoviocytes. Moreover, BBA inhibition of TLR4/IL1R downregulated reactive oxygen species (ROS) synthesis and MAPK p38/NFκB, NLRP3, IFNαβ, TNF, and ECM-related pathways. Altogether, we present a solid bulk of evidence that BBA blocks OA innate immune responses and could be transferred into the clinic as an alimentary supplement or as a therapeutic tool after clinical trial evaluations.


Introduction
The worldwide number of patients suffering from rheumatic diseases such as osteoarthritis (OA) equals the whole United States of America (USA) population. That is more than 300 million people that will suffer their whole life from daily chronic pain and disability with no option for an efficient treatment [1][2][3]. Currently, OA is defined as a complex heterogenic disease [4][5][6] with no clinical [7][8][9] or molecular [10][11][12] consensus on the disease pathophysiology. It is agreed to be a silent progressive degradative disease that hinders the early diagnosis of articular cartilage degradation, synovial inflammation [13,14], and/or bone alteration [15,16]. Then, biomechanical, metabolic, and aging-related alterations accumulate to the point that the whole joint space narrows due to cartilage destruction [17][18][19]. Chronic articular destruction activates inflammatory, catabolic, and pain pathways that lead to acute pain that might bring permanent disability [17,18,20]. Great efforts have been committed to different drug discovery approaches; pharmacomodulation [21], drug repurposing [22,23], and the re-discovery of natural products from traditional medicine, namely, the first Chinese medicine book Hunagti Neiching [24] and the Indian Ayurvedic medicine [25].  [67] that resolves the adenosine diphosphate (ADP) binding site and has successfully been used for virtual screening simulations [68,69]. All protein database (PDB) crystallographic structures were refined with the protein preparation wizard module implemented in the Schrödinger suite [70]. A grid was generated at the ligand sites to attempt a molecular docking using the extra-precision (XP) scoring functions implemented in Glide [71,72]. Binding energies were computed at the molecular mechanics (MM)/generalized born surface area (GBSA) [73,74] following the described theory [75]. Additional molecular dynamics (MD) calculations were performed, aiming to confirm the predicted stability of the best-ranked drug-target system. Specifically, MD trajectories were generated with the graphic processing unit (GPU)-accelerated version of Desmond that successfully allows them to mimic interactions in a drug design [76]. The MD simulation is completed by producing trajectories of 100 ns. All calculations were conducted with Schrödinger (Schrödinger Release 2021-3).
Only early cell passes (P) comprised of P 0 -P 3 (primary cells) and P 5 -P 35 (non-primary cells) were used to keep the cellular phenotype unaltered. Cells were seeded in cell plates (P12 or P100) and after 4 h, the medium was renewed with FBS deprived medium [22,23,77,78]. All cell culture reagents were purchased from Sigma-Aldrich (Sant Louis, MO, USA) except otherwise mentioned.

Proteome Profiling
Protein levels were analyzed by enzyme-linked immunosorbent assay (ELISA) and matrix-assisted laser desorption/ionization (MALDI)-time of flight (TOF). Briefly, proteins were isolated from cell cultures using RIPA 1X buffer (Merck, Darmstadt, Germany) with HALT protease and phosphatase inhibitors (Thermo Fisher Scientific, Waltham, MA, USA) and quantified using the bovine serum albumin (BSA) Pierce Bradford assay (Ther-moFisher, USA).
The cellular and secreted (supernatant) levels of IL6 and IL1β proteins were determined in joint cells by mouse and human ELISA antibodies (Abbkine, Wuhan, China) and quantified (450 nm and 570 nm subtraction) in a Multiskan SkyHigh plate reader (Thermo Fisher Scientific, Waltham, MA, USA). Similarly, cellular proteome and secretome were determined in mouse and human chondrocytes by qualitative shotgun data-dependent acquisition (DDA) MALDI-TOF method [86][87][88][89]. The DDA was performed in a micro-liquid chromatography (mLC) mass spectrometry (MS)/MS attached to a hybrid quadrupole-TOF/triple-TOF 6600 and output data was filtered by 5% false discovery rate (FDR) and p ≤ 0.05 [89]. The ortholog migration algorithm from REACTOME conversed mouse and human proteomes [90]. The Venn diagram and interactome from FunRich software analyzed protein profiles [91]. The molecular function and pathway enrichment from REACTOME identified different medical and biological associations [90].

Statistical Analysis and Sample Size
The sample size (n) of the whole study was established in 6 independent biological replicas from different patients or cell passes and 3 dependent technical replicas. The availability of non-primary cells of early pass was unlimited. On the contrary, the availability of primary cells isolated from OA patients or healthy corpses was very limited. Patients were excluded from the study (CAEIG-2016/258) if they were diagnosed with microcrystalline arthritis or any other infectious process. Moreover, even if they were not diagnosed, a further analysis of the fresh sample was also needed. Similarly, the presence of any other pathology apart from OA directly excluded those samples. All included samples presented sufficient cartilage and bone to perform the cell isolation. Nevertheless, only 3 healthy samples and 3 OA samples presented sufficient synovium to be removed and processed to isolate synoviocytes. Biochemical assays (Griess and MTT) required 10 4 cells (P96 plate), transcriptomics (RT-qPCRs) required 10 5 cells (P12), metabolic assays required 3 × 10 5 cells (P6) and proteomics (ELISA and MALDI-TOF) required 3.5 × 10 6 cells (P100) per data point. Overall, the use of primary cells was optimized to acquire the largest amount of data points. MALDI-TOF proteomics analysis was the most cell-demanding technique and, considering that thousands of technical replicas runs are performed, we optimized its sample size to 3 independent biological replicas.
All datasets were expressed as the mean ± standard error mean (SEM) and normalized by the "Control" treatment data. Gene expression (mRNA) data was obtained by the RT-qPCR threshold cycle (∆∆Ct), also known as the ∆∆ quantification cycle (∆∆Cq). Gene expression was normalized by reference gene HPRT1 and analyzed by comparative transcriptomics. Fold-change (log 2 ) units were used in the proteome heatmaps or column charts.
The whole manuscript used The New England Journal of Medicine (NEJM)'s statistical significance method of representation. Specifically, a p-value of: <0.001 (***), <0.002 (**), <0.033 (*), and <0.12 (blank space or ns). Data distribution (Shapiro-Wilk and Kolmogorov-Smirnov normality tests) determined the use of parametric or non-parametric tests. Data from primary human cells were analyzed by non-parametric Kruskal-Wallis or Mann-Whitney test if needed. Otherwise, data were analyzed by parametric One Way-ANOVA and Tukey post-test or unpaired t-test Welch's correction if needed.

Determining the Proteome Profile of TLR4/IL1R-Stimulated Primary Human OA Chondrocytes
Anti-OA properties have been associated with BSE mix through NFκB pathway inhibition; nonetheless, the results were not consistent enough. Consequently, no international organizations such as the Food and Drug Administration (FDA) or the European Medicines Agency (EMA) have approved BSE mix or any individual component as a drug for clinical use. As such, our study focused on three pillars: the BSE mix single element BBA, the use of a primary joint cell model, and the implementation of novel techniques such as computational chemistry.
Potential BBA targets described in the literature are downstream of the TLR4/IL1R signaling pathway, including NFκB. In this sense, we analyzed the secretome and proteome of TLR4/IL1R-activated human OA chondrocytes (hOC) isolated from OA patients to validate novel and already described inhibitory targets ( Figure 1).
The use of animal or cellular mouse models is essential in translational medicine research as an initial step due to the scarcity of primary human OA experimentation. In this sense, we replicated the cellular proteome analysis workflow in TLR4/IL1R-activated mouse ATDC5 chondrocytes (Figure S1A-D). Interestingly, the mouse proteome profile identified a similar pattern of induced and inhibited proteins to the human proteome profile ( Figure S1A,B). Moreover, both TLR4 and IL1R-activated mouse ATDC5 chondrocytes put at the centermost of the interactome ( Figure S1C) the pleiotropic Polyubiquitin C (Ubc), a protein known to promote NFκB pathway activation [92]. In line with this, categories associated with MAPK/p38, NFκB, and ROS were found enriched among others, whereas ECM pathways were found depleted ( Figure S1D,E).

Predicting the Binding Affinity of β Boswellic Acid to Potential Proteome Profile Targets
Secretome and proteome pointed to NFκB, MAPK/p38 (MAPK14), TLR4, IL1R, and NLRP3 as the most promising targets for BBA inhibitory properties. The first three develop their activity in the cytosol; therefore, BBA needs to cross the cell membrane to perform inhibitory effects, whereas the others are innate immune receptors bound to a membrane. In this sense, through computational chemistry we explored the binding affinity of β boswellic acid (BBA) towards these biological targets ( Figure 2).    was ranked (F) according to the predicted binding energy and displayed the best result (pose) in each analysis. The more negative energy, the stronger the interaction ligand-target is. The selected computational order ranks relative to the affinity; (G) BBA cell permeability (red dots) using a quantitative structure-activity relationships (QSAR) model trained with cell permeability data of 190 compounds (blue dots) and obtaining a log P e = 5.38; (H) molecular dynamics (MD) trajectories (100 ns) and root means-square deviation (RMSD) were used to measure the structural dynamic stability of a ligand docked in a target binding cite.
The development of NFκB signaling inhibitors such as the Cmpd1 inhibitor uses the IKK as a prime target [66]. This inhibitor binds to the IKKβ at the hinge loop connecting the N and C lobes, a region that recognizes the adenine within the adenosine triphosphate (ATP) [66]. Our simulations (Figure 2A,F) revealed that Cmpd1 is located at that region with significant large energy (−59.91 kcal/mol). On the contrary, 11-keto-β-boswellic acid (KBA, −18.11 kcal/mol), acetyl-11-keto-β-boswellic acid (AKBA, −18.96 kcal/mol), or BBA (−13.88 kcal/mol) are anchored with less negative binding energy; that is, with a less intense interaction (Figure 2A,F). Consequently, these data demonstrated that the observed activity for BBA and derivatives was not associated with the IKK-related NFκB inhibition mechanism. The unsuccessful interaction results motivated us to assess the interaction of BBA against NFκB using the Müller et al. NFκB p50 (1SVC) [93] homodimer bound to DNA. We explored the whole protein surface, and the best-ranked poses were retained for MM/GBSA refinement. Interestingly, the simulations suggested that BBA binds NFκB with a strong interaction energy of −31.92 kcal/mol ( Figure 2B,F).
This successful result led us to study targets upstream of NFκB such as the human TLR4 complex. This complex is formed by two subunits of TLR4 protein (Chain A and B) and two subunits of accessory molecule MD2 [94] (Figure 2C,F). TLR4 is a promiscuous innate immune response (IIR) receptor capable of binding to a large diversity of host (DAMPs) and non-host (PAMPs) derived structures. Our molecular model for TLR4 was designed using the crystal deposited in the PDB with code 3VQ1 [95] ( Figure 2C,F) Interestingly, the binding affinity ( Figure 1A) of BBA towards the human TLR4 complex (−45.15 Kcal/mol for Chain A) was found energetically on par with the specific pharmacological TLR4 inhibitor CLI-095/TAK242 (−47.29 Kcal/mol). Furthermore, similar results were obtained for curcumin (−58.49 Kcal/mol) and mygalin (−65.98 Kcal/mol), two other candidates as TLR4 inhibitors for clinical practice (Figure 2A,F). Moreover, a similar trend was observed for TLR4 complex Chain B (Figure 2A,F).
Considering that computational analysis ranked BBA-TLR4 as the most promising target for BBA, we decided to further assess two TLR4-related receptors namely the IL1R and TLR3 complexes ( Figure 2D-F). The binding affinity of IL1R-BBA (−4.12 Kcal/mol) was found lower than IL1R-Methotrexate (−34.54 Kcal/mol), the main IL1R inhibitor ( Figure 2D,F). The TLR3 shares with the TLR4 part of the signaling cascade but it is activated by dsDNA and inactivated by inhibitors such as T5626448 (−12.19 Kcal/mol); nevertheless, the binding affinity TLR3-BBA (+2.18 Kcal/mol) was found inexistent ( Figure 2E,F). Finally, we addressed the potential affinity NLRP3-BBA, a signaling pathway downstream to TLR4 and IL1R. We compared the binding energy of the best docking pose predicted for BBA against the NLRP3 agonist ADP and antagonist MCC950 [96]. According to the positive binding energy shown in our simulations, BBA (+7.35 kcal/mol) does not fit into that position, while both MCC950 (−69.80 kcal/mol) and ADP (−31.31 kcal/mol) efficiently dock to the binding site ( Figure 2F).
These accumulated results evidenced a predilection of BBA to bind the receptor TLR4; nonetheless, they do not exclude the possibility of directly interacting with NFκB after crossing the cellular membrane. We used a 190-compound dataset to train [97] (training blue points) the Quantitative Structure Activity Relationships (QSAR) [98] software to predict the effective drug cell permeability coefficient (log P e ). Indeed, trained QSAR predicted (testing red points) an effective permeability coefficient of BBA (log P e = 5.38) that matches the naloxone permeability, a drug known to cross the cell membrane but also capable of binding and inhibiting the TLR4 signaling [23] (Figure 2G).
This potential dual dynamic is interesting and not yet documented by the literature; onwards, we focused on the BBA binding effect on the TLR4 complex. In fact, we performed molecular dynamics (MD) simulations and root means-square deviation (RMSD) monitoring to test BBA-TLR4 docking strength against the CLI-095-TLR4 complex ( Figure 2H). Specifically, RMSD values reveal that BBA (red line) is stabilized with an RMSD variation of 3 Å after 20 ns, whereas the reference inhibitor TAK242 required a larger RMSD window of 5 Å after 40 ns ( Figure 2H). The smaller RMSD variation associated with BBA confirms the stability of the complex under biological conditions. Consequently, MD simulations ensured dynamic binding stability (100 ns trajectory) and RMSD monitoring ensured struc-tural stability through a key TLR4 residue (Arg90) comparable for both BBA-TLR4 and CLI-095-TLR4 ( Figure 2H).
Altogether, computational pharmacology found the TLR4 complex as the most probable target with large binding interaction energy comparable to the value predicted for one of the golden standard inhibitors, namely, CLI-095 or TAK242.

BBA Effects on Cell Viability and ROS Secretion in Chondrocytes
Computational docking predicted a higher binding affinity of BBA towards the human TLR4 complex. This BBA feature was compatible with TLR4 antagonism. Moreover, BBA was predicted not to bind the IL1R complex. Previous studies reported that TLR4 blockade inhibited both TLR4 and IL1R signaling [22]. Thus, we validated the in silico results of BBA actions in an in vitro model to determine its potential modulation of the TLR4/IL1R signaling cascade.
Among the different pathways associated with TLR4/IL1R signaling, ROS generation was found highly enriched in the human and mice proteome profiles (Figures 2I,J and S1D,E). Specifically, NO generation was a common denominator between TLR4 and IL1R activation (Figures 2I,J and S1D,E). In fact, NO has been associated with the development of OA, the promotion of innate immune responses, the increase of articular catabolism, and the generation of more ROS ( Figure 3A) [99][100][101]. Before proceeding to determine the effects of potential therapeutic BBA (0.10-5 µM) doses on ROS generation, we compared the basal innate immune response tone (mRNA) between primary human chondrocytes isolated from OA patients or healthy corpses ( Figure 3B). Primary human OA chondrocytes (hOC) presented higher mRNA levels for all the studied IIR factors ( Figure 3B). Furthermore, OA joint cells were found more sensitive to TLR4/IL1R activation than healthy ones ( Figure 3B). Next, we evaluated the balance between cell viability, the accumulation of NO subproduct NO 2-, and NOS2 gene expression (mRNA) to determine the effects of potential therapeutic BBA (0.10-5 µM) doses on ROS generation in primary hOC ( Figure 3C,D) as well as in mice ATDC5 chondrocytes ( Figure S1F-I).
Considering these results, we evaluated whether BBA effects were bound to the modulation of the expression of TLR4 or IL1R mRNA levels. Gene expression of TLR4 and IL1R showed an increase in mRNA levels after LPS/IL1β stimulation but no modulation after BBA co-treatment ( Figures 3E,F and S1K,L).

BBA Effects on TLR4/IL1R-Mediated IIRs in Primary Human OA Chondrocytes
Once the inhibitory effects of BBA in the NOS2 axis were demonstrated, we further explored other TLR4/IL1R-modulated pathways found enriched in the proteome analysis including MAPK p38/NFκB, IFNαβ, and TNF (Figure 1; Supplementary data S2). The MyD88-dependent MAPK p38/NFκB downstream pathway is central to TLR4/IL1R signaling. We evaluated the mRNA expression of common specific elements in the MAPK p38 and NFκB pathway, namely, HSP90AA1, MAPK14 or p38, and IL6 in primary human OA chondrocytes (hOC). We found increased mRNA levels of HSP90AA1, p38, and IL6 after TLR4/IL1R activation ( Figure 3G-I) and decreased ones after cotreatment with BBA (0.10-1 µM), CLI-095 (1 µM), and siTLR4 (1 µM) in a dose-response dynamic ( Figure 3G-I). To validate these data, we further tested the intracellular and secreted levels of IL6 in these cells. The TLR4/IL1R activation increased secreted ( Figure 3J) and cellular ( Figure 3K) protein concentration of IL6 (up to 584 pg/mL) while BBA and CLI-095 cotreatment inhibited it. Consistent with a decreased activation of MAPK p38/NFκB pathway by BBA, we observed in TLR4/IL1R-activated hOC a reduction in the total phosphoproteome after BBA treatment ( Figure 3L). Interestingly, cotreatment with BBA or CLI-095 also reduced the protein concentration of IL6 and phosphoproteome levels in non-stimulated cells ( Figure 3J-L).
Next, we evaluated the TNF pathway element TNFα, the MyD88-independent IFNαβ pathway element IFNβ1, and highly studied innate immune response factors associated with the TLR4/IL1R pathway including MAPK p38/NFκB downstream signaling factors in the OA context. Remarkably, most of these factors were directly identified in the secretome and proteome profiles (

BBA Effects on TLR4/IL1R-Mediated IIRs in Primary Human OA Osteoblasts
The OA disease progression begins degrading the cartilage and eventually alters the bone structure ( Figure 4A). As such, we evaluated the BBA inhibitory effects on the TLR4/IL1R downstream pathways in primary human OA osteoblasts (hOB) isolated from OA patients. Remarkably, primary hOB presented higher mRNA levels for all the studied IIR factors. Similarly, hOB was more sensitive to TLR4/IL1R activation than healthy ones ( Figure 4B). In this sense, we used BBA (0.10-1 µM), CLI-095 (1 µM), and siTLR4 (1 µM) doses that did not alter primary hOB cell viability ( Figure 4B-A').

BBA Effects on TLR4/IL1R-Mediated IIRs in Human OA Synoviocytes
The pathophysiology from early-to-late OA regularly causes inflammation of the synovium; in turn, inflamed synovia promotes the alteration of other joint tissues ( Figure 5A). Consequently, we evaluated the BBA inhibitory effects in human synoviocytes. In this sense, primary human OA synoviocytes (hOS) presented higher mRNA levels for all TLR4/IL1R-associated innate immune factors ( Figure 5B). In fact, hOS were more sensitive to TLR4/IL1R activation than healthy ones ( Figure 5B). Nevertheless, the scarcity of primary human OA synoviocytes forced us to switch to human SW982 synoviocytes (hSW). Notably, the evaluation of BBA inhibitory effects on the TLR4/IL1R downstream pathways did not use doses of BBA (0.10-1 µM), CLI-095 (1 µM), and siTLR4 (1 µM) that altered hSW cell viability ( Figure 5C-Z).
The use of BBA, CLI-095, and siTLR4 downregulated (mRNA, protein, and phosphoproteome status) all factors except IL1R ( Figure 5D-Z). Likewise, neither BBA nor CLI-095 inhibited TLR4 mRNA levels, only the siTLR4 was capable ( Figure 5D-Z). Overall, the intensity of TLR4/IL1R-activation on hSW synoviocytes mRNA and protein levels was higher compared to hOC chondrocytes and hOB osteoblasts. Nonetheless, BBA, CLI-095, or siTLR4 co-treatment effects were very similar in the three human joint cells: chondrocytes (

BBA Effects in Joint Cells on the NLRP3 Inflammasome Pathway
The proteome enrichment analysis of TLR4/IL1R-activated primary human OA chondrocytes identified specific pathways downstream of these receptors but upstream of relevant key factors such as MAPK p38 or NFκB. The transcriptomic and protein analysis further supported that these pathways could contribute to the mechanism of action of BBA inhibitory effects. Remarkably, the proteome profile identified the NLRP3 inflammasome signaling as a pathway enriched after the activation of TLR4 and IL1R (Figures 1 and S1). Furthermore, the activation of the NLRP3 pathway triggers CASP1 that catalyzes the transformation of pro-IL1β into mature IL1β, the canonical agonist of IL1R1.
As such, we decided to further explore this potential mechanism of action in primary human OA joint cells. In TLR4/IL1R-activated hOC, gene expression (mRNA) of NLRP3, CASP1, and IL1β was highly increased (Figure 6A-C). In contrast, treatment with BBA, CLI-095, or siTLR4 reduced the induced mRNA levels ( Figure 6A-C). Interestingly, NLRP3 silencing with siRNA (siNLRP3) recapitulated the blocking effects of BBA, CLI-095, and siTLR4 in TLR4/IL1R-activated hOC ( Figure 6A-C). Furthermore, CASP1 activity assay and IL1β ELISA protein assay were used to validate these results ( Figure 6D-F). Stimulation with LPS or IL1β increased CASP1 activity and cellular and secreted IL1β protein ( Figure 6D-F). In line with previous results, treatment with BBA and CLI-095 reduced CASP1 activity and cellular and secreted IL1β levels to basal levels ( Figure 6D-F).
All in all, the mechanism of action of BBA inhibitory effects on TLR4/IL1R-mediated IIR has proven to be directly associated with the NLRP3 pathway as well as MAPK p38 and NFκB, IFN, and TNF pathways (Figure 7). Furthermore, BBA inhibitory effects might also come from the partial interruption of the positive feedback between IIR factors, ROS, DAMPs/PAMPs, and the TLR4/IL1R crosstalk (Figure 7).

Discussion
We present evidence that β boswellic acid or BBA selectively binds to the TLR4 complex and inhibits in primary human joint cells the TLR4/lL1R-mediated innate immune responses associated with the activation of ROS synthesis, MAPK p38/NFκB, IFNαβ, TNF, and NLRP3 signaling pathways. Furthermore, we determined in stimulated and non-stimulated articular cells that TLR4 inhibition by BBA or other known inhibitors is associated with NLRP3 downregulation. Moreover, we revealed that this downregulation was enough to recapitulate the effects of BBA or TLR4 inhibition on TLR4/lL1R-mediated IIR. Altogether, we provided relevant data supporting BBA as a strong TLR4 inhibitor that may help to control OA-associated IIR.
The BBA is extracted from Boswellia serrata and it is present in the BSE mix of triterpenes, essential oils, and polysaccharides that it is commonly taken as an infusion or food complement in traditional ayurvedic medicine to treat inflammatory diseases such as asthma, inflammatory bowel disease, RA, and OA. Interestingly, the evidence failed to demonstrate that BSE anti-inflammatory effects were a consequence of 5-lipoxygenase [27] inhibition. On the contrary, the BSE element BBA has recently been described to inhibit NFκB, a downstream signaling pathway of TLR4/IL1R [103,104]. Remarkably, TLR4 and IL1R are two innate immune receptors associated with the development of multiple rheumatic diseases [17,[38][39][40][41][50][51][52]. The IL1R has been used as a therapeutic target for diseases such as OA [17], RA [50], or gout [105]. However, there is no clinically available TLR4 inhibitor to treat rheumatic inflammatory diseases [22,106]. Considering the traditional use of BSE as antirheumatic and the new potential properties of BBA, we decided to determine whether BBA was able to block TLR4 and inhibit OA-associated IIR in joint cells.
Considering that a functional TLR4 receptor is required for full IL1R signaling, the first approach was to analyze the primary human OA chondrocytes secretome and proteome ( Figure 1) changes linked to TLR4 and IL1R activation [22]. We obtained key data on the TLR4/IL1R downstream signaling pathways. Specifically, MAPK p38/NFκB, TNF, IFNαβ, ROS, and NLRP3 downstream pathways were found enriched whereas ECM anabolic pathways were found depleted in the proteome of stimulated chondrocytes. These proteomic changes were consistent with increased inflammatory and catabolic OA-like phenotype [107][108][109]. Likewise, the enrichment in ROS-related pathways including NO synthesis was in concordance with OA worsening associated with oxidative stress [110,111]. Similar results were obtained in mice ATDC5 chondrocytes proteome ( Figure S1).
Next, through computational pharmacology, we investigated BBA potential inhibition of key factors implicated in major pathways found in the secretome and proteome of TLR4/IL1R-activated primary human OA chondrocytes ( Figure 2). We predicted a greater binding affinity of BBA to the human TLR4 complex than other elements suggested in the literature such as NFκB [93] and IKK [66,112]. Furthermore, this binding affinity of BBA to TLR4 was found equivalent to known antagonists such as CLI-095. Interestingly, the docking analysis predicted that BBA could not bind to any domain of IL1R, NLRP3, or TLR3 complex. Although docking score results found the binding of BBA to TLR4 the most energetically and structurally stable, the BBA effective permeability coefficient does not discard BBA from crossing the cellular membrane and binding to NFκB, as the literature suggested. Nonetheless, molecular dynamics simulations presented additional evidence of BBA-TLR4 binding stability, suggesting that specific TLR4 inhibition may be the main mechanism of action ( Figure 2).
The combination of anti-TLR4 predicted activity of BBA and the TLR4/IL1R targets identified from the proteome analysis supported the in vitro study of BBA effects in TLR4/IL1R-activated human chondrocytes, osteoblasts, and synoviocytes. The relevance of this predicted BBA activity was underpinned by the observation of increased levels of OA-associated factors as well as increased sensitivity to TLR4 and IL1R agonists in chondrocytes, osteoblasts, and synoviocytes isolated from OA patients in contraposition to healthy ones.
Considering that the specific TLR4 inhibitor CLI-095 and the anti-TLR4 siRNA replicated the same BBA inhibitory effects in all the cellular studies [113], and neither CLI-095 nor BBA downregulated TLR4 and IL1R expression, it was confirmed that TLR4 inhibition is the mechanism underlying BBA effect. In agreement with this, previous reports with other TLR4 inhibitors have obtained similar results about the key role of TLR4 inhibition on IL1R signaling [22,23]. Remarkably, BBA downregulated basal expression levels of several IIR factors suggesting that TLR4 constitutive activation contributes to the expression of multiple IIR factors in joint cells.
Despite BBA not binding to to IL1R, it exhibited outstanding effects in blocking IL1Rmediated IIR. This was consistent with the described anti-inflammatory effects of TLR4 inhibition on the IL1-mediated RA animal model [114]. The NLRP3 inflammasome pathway has been associated with the IIR in multiple rheumatic diseases [115,116], including OA [117]. Moreover, this pathway was substantially enriched in chondrocytes upon the activation of TLR4 and ILR1 receptors. In line with this, the NLRP3 inflammasome pathway has been identified downstream of TLR4/IL1R crosstalk [117,118]. Specifically, it participates in the synthesis and release of IL1β, which generates DAMPs that activate TLR4 [17] promoting an inflammatory positive feedback loop. Accordingly, we investigated whether NLRP3 signaling pathway modulation could be responsible for BBA effects. Interestingly, in all the articular cells studied, BBA was able to downregulate the TLR/IL1R-induced expression of the NLRP3 axis (NLRP3, CASP1, and IL1β) as well as the enzymatic activity of CASP1, and the concentration of cellular and secreted IL1β protein. Similar results were obtained in non-activated OA joint cells. Consistent with this, NLRP3 and TLR4 silencing recapitulated the inhibitory effects of BBA and CLI-095 in OA joint cells. These data demonstrated that NLRPL3 signaling is tightly related to TLR4 and confirmed that BBA effects on IL1R signaling are TLR4-and NLRP3-dependent. Overall, the local use of BBA on the articular tissues might be open to consideration.

Conclusions
We demonstrate that BBA inhibits OA-associated inflammatory and catabolic processes in chondrocytes, osteoblasts, and synoviocytes without altering the joint cell viability. We prove that the specific binding of BBA to the TLR4 complex inhibits TLR4 and IL1R signaling and downregulates MAPK p38/NFκB, NLRP3 inflammasome, and other OA-associated pathways. Moreover, we identified the TLR4-NLRP3 axis as an underlying mechanism of the BBA effect on IL1R signaling. Altogether, we present a solid bulk of evidence that BBA blocks OA innate immune responses and could be transferred into the clinic as an alimentary supplement or as a therapeutic tool after clinical trial evaluations. Funding: This research was funded by Sociedad Española de Reumatologia (SER)-Fundación Española de Reumatologia (FER) via 2019 edition of "Beca para proyectos de investigación NO financiados por Agencias Públicas" code name "Búsqueda de nuevos fármacos bloqueantes de la inflamación asociada a TLR4 en condrocitos humanos Artrósicos (PNFSER18ALI)" section "Artrosis"; the Instituto de Salud Carlos III (ISCIII) & Fondo Europeo de Desarrollo Regional (FEDER) via "Fondo de Investigación Sanitaria" [grant numbers PT20/00009, FI20/00210, CM20/00186, RD21/0002/0025, PI19/01446, PI16/01870, CPII20/00026, CP15 Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Written informed consent has been obtained from the patient(s) to publish this paper.

Data Availability Statement:
The data presented in this study are available in the article and Supplementary Materials.