Preliminary Assessment of the Anti-inflammatory Activity of New Structural Honokiol Analogs with a 4′-O-(2-Fluoroethyl) Moiety and the Potential of Their 18F-Labeled Derivatives for Neuroinflammation Imaging

Neolignans honokiol and 4′-O-methylhonokiol (MH) and their derivatives have pronounced anti-inflammatory activity, as evidenced by numerous pharmacological studies. Literature data suggested that cyclooxygenase type 2 (COX-2) may be a target for these compounds in vitro and in vivo. Recent studies of [11C]MPbP (4′-[11C]methoxy-5-propyl-1,1′-biphenyl-2-ol) biodistribution in LPS (lipopolysaccharide)-treated rats have confirmed the high potential of MH derivatives for imaging neuroinflammation. Here, we report the synthesis of four structural analogs of honokiol, of which 4′-(2-fluoroethoxy)-2-hydroxy-5-propyl-1, 1′-biphenyl (F-IV) was selected for labeling with fluorine-18 (T1/2 = 109.8 min) due to its high anti-inflammatory activity confirmed by enzyme immunoassays (EIA) and neuromorphological studies. The high inhibitory potency of F-IV to COX-2 and its moderate lipophilicity and chemical stability are favorable factors for the preliminary evaluation of the radioligand [18F]F-IV in a rodent model of neuroinflammation. [18F]F-IV was prepared with good radiochemical yield and high molar activity and radiochemical purity by 18F-fluoroethylation of the precursor with Boc-protecting group (15) with [18F]2-fluoro-1-bromoethane ([18F]FEB). Ex vivo biodistribution studies revealed a small to moderate increase in radioligand uptake in the brain and peripheral organs of LPS-induced rats compared to control animals. Pretreatment with celecoxib resulted in significant blocking of radioactivity uptake in the brain (pons and medulla), heart, lungs, and kidneys, indicating that [18F]F-IV is likely to specifically bind to COX-2 in a rat model of neuroinflammation. However, in comparison with [11C]MPbP, the new radioligand showed decreased brain uptake in LPS rats and high retention in the blood pool, which apparently could be explained by its high plasma protein binding. We believe that the structure of [18F]F-IV can be optimized by replacing the substituents in the biphenyl core to eliminate these disadvantages and develop new radioligands for imaging activated microglia.


Introduction
Neuroinflammation is a complex inflammatory process within the CNS and occurs as a response to infection, protein aggregation, trauma, and ischemia [1]. Positron emission tomography (PET) is an in vivo molecular imaging technique that provides unique capabilities for detecting biochemical changes in activated CNS glia (microglia and astrocytes) to quantify neuroinflammation [2]. Cyclooxygenase (COX) enzymes catalyze prostaglandin synthesis and are involved in the activation of inflammatory pathways leading to the release of cytokines, chemokines, nitric oxide, and reactive oxygen species [3]. Both known isoforms, COX-1 and COX-2, are expressed in the brain. COX-1 is constitutively expressed in the CNS (as a "house-keeping" enzyme) and is not upregulated by inflammation, although some studies have shown that COX-1 may play a prominent role in neuroinflammation [4]. The inducible type of cyclooxygenase COX-2 is rapidly and dramatically overexpressed under inflammatory conditions and, therefore, is one of the most attractive molecular targets for PET imaging of neuroinflammation. When developing successful COX-2 radioligands, in addition to their high affinity and selectivity to COX-2, many other factors should be considered, such as penetration of the blood-brain barrier (BBB), chemical stability in plasma, suitable kinetic parameters, metabolic profile excluding the entry of metabolites into the brain, low nonspecific binding, and lipophilicity (LogD 7.4 ) ranging from 2 to 4, etc. [5,6]. Selective COX-2 inhibitors (COXIBs) are the main class of anti-inflammatory drugs currently used clinically, a prominent representative of which is celecoxib. Several derivatives and structural analogues of celecoxib labeled with carbon-11 (T 1/2 = 20.4 min) and fluorine-18 (T 1/2 = 109.8 min) have been developed for COX-2 imaging by PET over the past decade [7]. Of these, [ 11 C]TMI is the most promising radioligand for in vivo targeting of COX-2 in neuroinflammation, since it has demonstrated uptake even to constitutive expression of COX-2 in baboon brains [8]. Exploring a 2-(4methylsulfonylphenyl) pyrimidine scaffold, radioligands [ 11 C]MC1 [9] and [ 18 F]FMTP [10] have recently demonstrated in vivo binding to COX-2 in the brain of LPS-injected primates and mice, respectively. [ 11 C]MC1 also successfully visualized the upregulation of COX-2 in patients with rheumatoid arthritis in first-in-human studies [9].
An alternative to clinically used COXIBs could be structural analogs of natural phenolic compounds, which are attracting research attention for their anti-inflammatory capacity [11,12]. In addition, these compounds do not have the serious adverse effects commonly associated with celecoxib and its analogs. Several flavonoid and stilbene derivatives have been labeled with PET radionuclides to target amyloid plaques, tau protein, and alphasynuclein aggregates in Alzheimer's and Parkinson's diseases, which are associated with chronic neuroinflammation [13,14].
We recently proposed to explore the scaffold of 4 -O-methylhonokiol (MH in Scheme 1A), a natural biphenolic compound known for strong anti-inflammatory and neuroprotective activities [15][16][17], to design potential COX-2 radioligands. Radioligand [ 11 C]MPbP (4 -[ 11 C]methoxy-5-propyl-1,1 -biphenyl-2-ol) showed a high brain uptake in LPS (lipopolysaccharide)-treated rats compared to control animals [18]. These results inspired us to synthesize several new structural analogs of honokiol by replacing the methoxy group in MPbP with a fluoroethoxy group for convenient labeling with longer-lived fluorine-18 (F-I-F-IV, Scheme 1B). It is known that methoxy-fluoroethoxy substitution in aromatic rings gives a more stable compound against metabolic degradation than substitution with a fluoromethoxy moiety [19]. The 18 F-labeled ligand has additional advantages in preclinical trials due to its 110 min half-life; in particular, it allows the study of biodistribution over a longer time to ensure accurate measurements and reduce the number of animals in experimental groups.
In this study, we aimed to produce an 18 F-labeled analog of honokiol and to assess its potential for targeting COX-2 under inflammatory conditions. Based on the results of screening the anti-inflammatory activity of new structural analogs of honokiol with a 4 -O-(2-fluoroethyl) moiety (F-I-F-IV), we selected a candidate for radiolabeling. The initial evaluation of the radioligand [ 18 F]F-IV was performed in a rat model of LPS-induced neuroinflammation.  4 , Na 2 CO 3 , dioxane-water, 95 • C, 8 h; (v)-BrCH 2 CH 2 F, K 2 CO 3 , acetone, reflux, 24 h; (vi)-HCl conc., acetone, 20 • C, 16 h.

Anti-Inflammatory Screening
It is known that peripheral inflammatory stimuli such as LPS, administered intraperitoneally, induce a profound immunological response in the brain, triggering microglia activation leading to neuroinflammation [20]. In our studies, LPS-induced neuroinflammation in mice was characterized by an increased expression of the pro-inflammatory cytokines IL-6 (interleukin-6) and TNF-α (tumor necrosis factor), a high level of lipid peroxide in the brain, as well as a significant increase in the number of pyknotic and damaged neurons in the hippocampus and cerebral cortex. These effects are shown in Figures 1-3. The anti-inflammatory activity of the tested compounds (F-I-F-IV), manifested in a de-crease in high levels of cytokines and lipid peroxides as well as in a decrease in neuronal pyknosis, was assessed using the reference drug celecoxib, a well-known NSAID and a selective COX-2 inhibitor.
Statistical analysis showed that LPS significantly increased serum levels of IL-6 and TNF-α, while treatment with compounds F-I-F-IV or celecoxib reduced them to levels seen in intact animals ( Figure 1A,C). Moreover, F-IV exhibited an ability to inhibit serum cytokines comparable or even superior to that of celecoxib. In brain supernatants, the TNF-α level was significantly increased in all groups of LPS-induced mice compared to the minor level of the intact group. Although a moderate inhibition of TNF-α and IL-6 concentrations by F-I-F-IV compounds in the murine brain was found, it was not statistically significant ( Figure 1B,D). To assess the neuroprotective effect of the tested compounds, histological sections of four regions of the murine brain (CA1, CA3, DG of hippocampus and posterior parietal cortex) were analyzed. In morphometric studies, the total number of neurons (N total ) and the number of pyknotic (degeneratively altered) neurons (N pykn ) in the visual field were counted, as well as the ratio N pykn /N total . The results are shown in Figure 2A-C. The parietal cortex and hippocampus are among the most susceptible brain regions to induced neuroinflammation [21]. A decrease in the linear density of neurons (N total ) by an average of 15-20% was observed in all selected regions of the brain in animals with LPS-induced neuroinflammation (LPS group, Figure 2B). This effect was accompanied by degenerative neuronal damage characterized by a decrease in cell size with changes in tinctorial properties of the cytoplasm and hypercondensation of nuclear chromatin. The number of pyknotic neurons (N pykn ) in LPS-induced animals significantly exceeded the value in the intact mice in all studied brain regions: almost 5-fold in the C1 region, 1.5-fold in CA3, more than 2-fold in DG, and 3-fold in the parietal cortex ( Figure 2B,C). The restoration of linear density and a decrease in the number of degenerative altered cells were observed in the brain of animals treated with the tested compounds. The greatest effects were noticed in the CA1 region for mice treated with F-II, F-III, and F-IV, and in CA3 for mice receiving F-I. In addition, the treatment with F-II resulted in a statistically significant decrease in the number of pyknotic neurons in the cortex. These data are likely to indicate the neuroprotective effect of the investigated compounds. Moreover, the observed therapeutic effect was slightly higher than that of the reference drug celecoxib.  It has been recognized that the involvement of lipid peroxidation as well as general oxidative stress is critical to the progression and regulation of inflammation, and lipid peroxides are key mediators of many neurodegenerative diseases [22,23]. The TBARS (Thiobarbituric Acid Reactive Species) assay has been used to measure lipid peroxidation products and to assess of the overall impact of oxidative stress agents such as LPS in animal models of neuroinflammation [24]. We hypothesized that the studied biphenyls, in addition to blocking neuroinflammation, can also suppress lipid peroxidation processes. As shown in Figure 3, the TBARS level in the brain of untreated LPS mice was 116% higher (p = 0.0205) compared to the control group.
The treatment of animals with the tested biphenyls reduced TBARS levels by 107% (F-I), 126% (F-II), 116% (F-III), and 111% (F-IV). Thus, F-I-F-IV biphenyls prevented the accumulation of lipid peroxidation products in the brain of LPS-injected mice. In contrast, treatment with celecoxib had little or no effect on lipid peroxide levels.
The obtained data indicate that our fluoroethoxy biphenyls not only have antiinflammatory activity comparable or even superior to that of celecoxib, but also exhibit pronounced antioxidant properties. In our opinion, these biphenyls merit further investigation as potential anti-inflammatory agents in other animal models of neuroinflammation and peripheral inflammation.
Compound F-IV was selected from four candidates for radiolabeling because its structure is closest to the structure of MPbP, a previously studied MH derivative, the labeled analog of which showed high uptake in the brain of LPS-induced rats [18].

In Vitro COX-2 Inhibition
The anti-inflammatory activity exhibited by MH may result from the inhibition of molecular targets in activated microglia, such as COX-2 and cannabinoid receptor type 2 (CB-2) [25,26]. The F-IV inhibitory potency to human recombinant COX-1 and COX-2 was measured by enzymatic immunoassay (EIA) kit and the half-maximal inhibitor concentration (IC 50 ) values were determined. F-IV, like MH, demonstrated good inhibitory potency against COX-2 at the submicromolar level: 0.09 and 0.06 µM, respectively (Table 1). At the same time, F-IV exhibited much lower selectivity for COX-2 (SI) than celecoxib (48 vs. >200), although slightly higher than MH and MPbP. Therefore, the question remains whether the activity of the tested biphenyl derivatives in suppressing neuroinflammation is directed only to COX-2 or to other molecular targets involved. This study c a In vitro COX-2 selectivity index SI = (COX-1 IC 50 /COX-2 IC 50 )-; b values were determined using an ovine COX-1/COX-2 assay kit; c values were determined using human COX-1/COX-2 assay kit.

Computational Analysis
We employed the alchemical free energy perturbation (FEP) method to predict the binding affinities of new compounds to COX-2 and CB-2 by calculating the relative free energy of the binding (ddG) of the ligands with these targets, and the results are given in Figure 4A,B. The complexes of CB-2 and COX-2 with F-I-F-IV are shown in Figure 4C, D. Calculated ddG between MPbP and F-IV for COX-2 was in excellent agreement with the experimental value (0.25 vs. 0.22 kcal/mol), confirming the correctness of the binding mode predicted. Both CB-2 and COX-2 have predominantly hydrophobic interactions with the tested ligands. In the case of CB-2, the ligands also have a hydrogen bond with S285 and p-p stacking interaction with F91. Analogously, COX-2 forms a bridge between S530 and OH-group of the ligands. Computed data support the preferential binding of MPbP and new fluoroethoxy biphenyls to COX-2 rather than to CB-2 in activated microglia except for the ligand F-III, which had an affinity for CB-2 very close to that of MH.

Radiochemistry
The radioligand [ 18 F]F-IV (4 -(2-[ 18 F]fluoroethoxy)-2-hydroxy-5-propyl-1,1 -biphenyl) was prepared like [ 11 C]MPbP (4 -[ 11 C]methoxy-5-propyl-1,1 -biphenyl-2-ol) from the labeling precursor 15 by a radioalkylation reaction (Scheme 2). To synthesize [ 11 Figure 5A). Using NaOH or Bu 4 NOH, a radioactive by-product was obtained in which the [ 18 F]fluoroethoxy group was in position 2 of the biphenyl structure and not in position 4 , as in [ 18 F]F-IV ( Figure 5B). The product and impurity are difficult to separate by preparative HPLC as they have close retention times. We assumed that the basic conditions of the radioalkylation reaction could facilitate the removal of the BOC protecting group from the precursor 15 [28] with the subsequent formation of biphenol having two equal positions (OH) for the attack of [ 18 F]FEB. To confirm this hypothesis, a model reaction of [ 18 F]FEB with 5-propyl-1,1 -biphenyl-2,4 -diol was carried out. The results of HPLC analysis indicated that the major reaction products were [ 18 F]F-IV and by-product in a 45/55 ratio ( Figure 5C). The cleavage of the BOC group from 15 by the action of NaOH or Bu 4 NOH is apparently catalyzed by a small amount of water [29] that is required for their dissolution in DMSO, while KOMe was dissolved directly in DMSO without the addition of water. Semi-preparative HPLC provided [ 18 F]F-IV with a radiochemical yield of 35% based on [ 18 F]FEB radioactivity, and the molar activity was [35][36][37][38][39][40] GBq/µmol at the time of administration in rats.
The radiochemical purity of the formulated radioligand exceeded 97% and amounts of chemical impurities were negligible (<1µg/mL) ( Figure 5D).
[ 18 F]F-IV was radiochemically stable in human plasma for up to 60 min (92%), and no traces of [ 18 F]defluorination were found, as measured by radio-HPLC. It is possible that the radioligand is metabolized through a different metabolic pathway.
The measured lipophilicity value (logD 7.4 ) 3.30 ± 0.10 (n = 6) for [ 18 F]F-IV was within the logD range (2-4) observed for most of the known PET radiotracers for brain imaging [6], and suggests that the radioligand is lipophilic enough to penetrate the BBB.

Ex Vivo Biodistribution Studies of [ 18 F]F-IV
Ex vivo biodistribution was intended as a preliminary assessment of the BBB permeability and overall distribution of [ 18 F]F-IV in LPS-treated and intact Wistar rats. [ 18 F]F-IV uptake by tissue was expressed as %ID/g (dose administered per gram of tissue). The obtained data in healthy animals showed a rapid distribution in most of the organs at 10 min. The radioligand seems to readily penetrate the BBB, as evidenced by the high initial rate of brain uptake with a peak value of 2.21 ± 0.64 and 2.09 ± 0.65 (%ID/g) in the pons and medulla, respectively, at 10 min post-injection, ( Figure 6A). The ID data showed that the level of radioactivity remained high in both the brain and organs for more than 30 min post-injection, then gradually decreased and stabilized. The excretion pathway was through the kidneys, and 60 min after injection there was no obvious difference in [ 18 F]F-IV uptake values between tissues. The lack of significant specific accumulation of radioactivity in the brain, heart, lungs, and kidneys most likely reflects the low baseline level of COX-2 expression in these organs in healthy animals [30]. It should be noted that relatively high initial levels of radioactivity in the blood pool persisted even 120 min after injection of [ 18 F]F-IV (0.99 ± 0.17 vs. 1.35 ± 0.06%ID/g). This was the main difference in the distribution of [ 18 F]F-IV and the previously reported [ 11 C]MPbP in organs and tissues of intact rats. To describe [ 18 F]F-IV accumulation in LPS-treated rats without or with celecoxib pretreatment (LPS block experiments), %ID/g values were corrected for radioactivity in the blood as %ID organ /%ID blood for a more accurate assessment and exclusion of possible errors related to intravenously (i.v.) administration of the radioligand [31,32]. These results are depicted in Figure 6B. [ 18 F]F-IV showed a small to moderate increase in brain uptake (5% in pons and 16% in medulla) in LPS-treated rats relative to saline-injected rats (control animals). A moderate increase in the radioactivity was observed in the peripheral organs of the LPS animals vs. controls: 12%, 16%, 20%, 20%, and 23% for lung, kidney, spleen, thymus, and liver, respectively. [ 11 C]MPbP, in contrast to [ 18 F]F-IV, showed a high uptake (up to 400%) in all examined organs and brain regions of LPS rats vs. healthy rats [18]. Blocking experiments with celecoxib were performed to clarify whether specific binding to COX-2 is responsible for the accumulation of radioligand in LPS rats. Pretreatment with celecoxib 30 min before the radioligand administration resulted in a reduced uptake of [ 18 F]F-IV in almost all the extracted organs and tissues of LPS rats relative to background (control) levels. The most significant blocking effect with a 20-32% reduction in radioactivity was observed in the brain (pons and medulla), heart, lung, and kidney, indicating that [ 18 F]F-IV probably specifically binds to COX-2 in a rat model of LPS-induced neuroinflammation. Plasma protein binding (PPB) assays were performed to elucidate the reasons for the difference in pharmacokinetic profiles of the two radioligands.

Plasma Protein Binding (PPB) for [ 18 F]F-IV and [ 11 C]MPbP
After an injection is distributed in circulating blood, radioligands bind to plasma proteins and only rest unbound (free) fractions can join to the target of interest [33]. PPB in human and rat plasma samples was assessed by ultrafiltration, which is most commonly used in PET because of its relatively short analysis time and simplicity. The data in Table  2 revealed that both radioligands had a high PPB level, but [ 18 F]F-IV is associated with proteins to a greater extent than [ 11 C]MPbP: >98 vs. 92% for rats and >99 vs. 98% for humans, respectively. The measured high PPB of [ 18 F]F-IV appears to be responsible for the high level of radioactivity in the blood, its retention in rat tissues, and for lower brain uptake (compared to MPbP) in LPS-treated rats, [34]. In this study, we did not evaluate the contribution of non-specific binding or off-target interactions of [ 18 F]F-IV with proteins and phospholipids in the brain. However, they can also occur in the biodistribution of the radioligand [ 18 F]F-IV due to its rather high lipophilicity and moderate selectivity for COX-2.

General Chemistry
All commercially available chemicals were used without any further purification. Melting points were determined by Kofler hot-stage (VEB Wägetechnik Rapido, PHMK 81/2969). Purifications by column chromatography on Merck silica gel 60 (0.035-0.070 mm) were performed. Analytical thin layer chromatography (TLC) on Merck 60 F254 silica gel plates with UV visualization (254 nm) was performed. NMR spectra were recorded on a Bruker Avance 400 spectrometer in CDCl 3 ( 1 H, 13 C and 19 F at 400.17, 100.62 and 376.54 MHz, respectively). HRMS (ESI) analysis was performed on a Bruker micrOTOF mass spectrometer. All manipulations of oxygen-and moisture-sensitive materials were conducted with a standard Schlenk technique.

Animal Studies
All experiments adhered to regulations of DIRECTIVE 2010/63/EU on the protection of laboratory animals. The procedures with animals were reviewed and approved by the Ethics Committee of Smorodintsev Research Institute of Influenza. Animal studies were carried out on male C57BL/6 mice (8-10 weeks, 20 ± 1 g) and Wistar male rats (6-8 weeks, 200 ± 10 g) from the Nursery of Laboratory Animals (Rappolovo, Leningrad region). The animals were housed in a climate-controlled room with a 12/12-h light cycle. The subjects had ad libitum access to food and water during housing. Neuroinflammation was modeled in animals by intraperitoneal (i.p.) single injection of LPS from Escherichia coli saline O111:B4 (Sigma-Aldrich) at a dose of 2 mg/kg.

Anti-Inflammatory Screening
In experiments evaluating the anti-inflammatory abilities of fluoroethoxy biphenyls, mice were injected with LPS an hour after the first administration of compounds F-I-F-IV, celecoxib, or a placebo. In total, 7 groups of 5 individuals in each were studied, including a group of intact animals. The tested compounds and celecoxib were injected in a dose of 10 mg/kg in saline containing 5% DMSO for 3 days (0.4 mL solution per animal per day). Mice were decapitated 72 h after LPS administration under xylazine-zoletil anesthesia (50 mg/kg) and the brain was immediately removed. The blood collected after decapitation was left in tubes with a serum separator for 30 min, then centrifuged at 611 g (3000 rpm) and 4 • C for 10 min. Blood and brain samples were frozen and stored at −20 • C. The brain was weighed and 2 mL of ice-cold endotoxin-free PBS containing 0.1% non-ionic detergent Triton X-100 was added to the tissue samples left under ice for 10 min until homogenized with a tissue disruptor. After centrifugation for 5 min, supernatants were collected and stored at −20 • C followed by measurements of cytokines and TBARS levels by ELISA in the serum and brain.

Cytokine Assay
Mouse cytokines (TNF-α and IL-6) were assayed using ELISA MAX TM Standard Sets adapting the procedures recommended by the manufacturer BioLegend (San Diego, CA, USA). Briefly, captured antibodies were coated with PBS pH 9.5 as recommended by the manufacturer. Avidin-horseradish peroxidase conjugate with H 2 O 2 -3,3 ,5,5 -tetramethylbenzidine (Sigma-Aldrich, St. Louis, Missouri USA) substrate was used. Plates were read by microplate spectrophotometer EPOCH2TC (Bio Tek Instruments Inc., Winooski, VT, USA). The sensitivity to all cytokines ranged from 0.01 to 500 pg/mL [38].

Histology, Cresyl Violet Staining, and Neuronal Morphology
The morphometric assessment of the total linear density of neurons (N all ) and the number of pyknomorphic neurons (N pykn ) was performed on histological sections from the brain regions of the hippocampus (CA1, CA3, DG) and posterior parietal cortex. Pyknotic cells were counted using ImageJ software (NIH). There were seven experimental groups of 3 mice each. The brain was removed and fixed entirely in 10% buffered formalin solution, and coronal sections were serially cut through the hippocampus and cortex, cassetted, marked, and embedded in paraffin blocks. Brain sections 8 µm thick were stained with Nissl cresyl violet, clarified, and covered with a coverslip. They were then studied at 400× magnification on a Leica DM1000 light-optical microscope and micrographs were taken using the ADF Pro camera software.

TBARS Assay
The supernatant diluted with buffer 1:10 was used for analysis. A total of 100 µL of tissue homogenate sample (S) or control buffer/standard solution (STD) was added to 100 µL of SDS solution according to TBARS assay Elisa kit (Cayman Chemical) and mixed. Aliquots of thiobarbituric acid solution (4 mL) were added to all of tubes, which were then boiled at 100 • C for one hour, and centrifuged at 1600 g for 10 min at 4 • C. Each sample (150 µL) was loaded in duplicate into a 96-well plate, and the absorbance was measured at 535 nm on an EPOCH2TC.

Statistical Analysis
Data were analyzed for Gaussian or normal distribution using the Shapiro-Wilk test. To assess significant differences between more than two groups of normally distributed data, we performed one-way ANOVA, followed by Holm-Sidak's post hoc analyses. Kruskal-Wallis test was applied to non-normally distributed data, followed by Dunn's post hoc analyses. All statistical analyses were performed using GraphPad Prism, Version 8.2.1.

In Vitro COX-1/2 Inhibition Assay
The inhibitory potency of F-IV for COX-1 and COX-2 was established, as reported in [18], using the COX (human) Inhibitor Screening Assay kit (Item.No. 701230, Cayman Chemical Company, Ann Arbor, MI, USA) according to the manufacturer's instructions.

Relative Binding Free Energies of Ligands F-I-F-IV with CB-2 and COX-2
To compare binding affinities of our ligands to CB-2 and COX-2, using structures 6PT0 and 5KIR from Protein Data Bank (PDB), respectively, we employed the calculations of relative binding free energies of structurally similar ligands by the alchemical free energy perturbation (FEP) method [39,40]. The FEP approach is extensively applied in drug discovery projects today and it is more accurate than the MM/GBSA (molecular mechanics combined with generalized Born surface solvation) method we used previously [18]. All the molecular modeling and simulations were performed using Schrodinger 2020-2 software. We took prepared and equilibrated receptor and ligand structures for the FEP+ calculations [41,42].
The absolute free energies for MH and F-IV were calculated from experimental data available from this study and the literature [26]. In all cases, both the FEP and cycle closure correction errors did not exceed 0. 23 Figure 5D). The identity of [ 18 F]F-IV was confirmed by co-injection with non-radioactive F-IV reference.

Plasma-Stability Study and logD 7.4 Determination for [ 18 F]F-IV
The stability of [ 18 F]F-IV was investigated in human plasma at 37 • C in the range from 0 to 60 min according to a modified protocol [44]. Phosphate-buffered saline (PBS, 0.15 M, pH 7.4) was used as a control. [ 18 F]F-IV formulated solution (10 µL,~1 MBq) was added to tubes with PBS (300 µL) and human plasma (300 µL). The mixture was vortexed and incubated at 37 • C. An aliquot of 100 µL from the sample was taken at different time points (0, 10, 20, 40, and 60 min) and mixed with 100 µL ice-cooled acetonitrile and centrifuged (3 min, 4500× g). The supernatants were analyzed by radio-HPLC using XBridge C18 analytical column.
The partition coefficient of [ 18 F]F-IV (log D 7.4 ) was determined by the shake-flask method [45] for measuring the distribution of radioactivity between 1-octanol and phosphate buffer pH 7.4. Into a tube containing 10 µL of [ 18 F]F-IV (~1 MBq), 1.5 mL of presaturated 1-octanol and 1.5 mL of phosphate buffer were added and vortexed for 15 min. Then, the tube was centrifuged at 3000 g for 5 min to ensure complete separation of the phases. Then, aliquots of 500 µL were taken from each phase into separate tubes and the radioactivity was measured in a γ-counter. The measurements were performed in triplicate. The log D 7.4 value was calculated as follows: log (counts per min in 1-octanol/ counts per min in PBS) for five independent experiments.

Ex Vivo Biodistribution
Ex vivo radioligand biodistribution was performed by direct radiometry of organs and tissue samples of Wistar rats. Each rat was injected with (12 ± 3) MBq (volume ≤ 0.2 mL) of [ 18 F]F-IV through the tail vein under anesthetic conditions. Control rats were sacrificed at 10 (n = 3), 30 (n = 3), 60 (n = 3), and 120 (n = 3) min after injection. Twenty-four hours prior to ex vivo biodistribution studies, 6 rats received an intraperitoneal (i.p.) injection of LPS (2 mg/kg) from Escherichia coli (O111: B4) or saline (1 mL/kg). For blocking studies with celecoxib, it was i.v. administered at a dose of 10 mg/kg 30 min prior to [ 18 F]F-IV injection in three animals.
Blood samples and organs of interest, namely, the brain (cerebral cortex, cerebellum, medulla, and pons), thymus, heart, lung, spleen, kidney, liver, bladder, and muscle, were removed and weighed. The radioactivity in each tissue sample was measured by a γcounter to calculate the percentage of injected dose per gram of tissue (%ID/g). Data were provided as mean ± SEM values.

Plasma Protein Binding (PPB) Determination
The ultrafiltration method [33] was used to estimate the protein-free fraction of [ 18 F]F-IV and [ 11 C]MPbP in human and rat plasma. Equal volumes (500 µL) of plasma or PBS (control) were mixed with the formulated solution of radioligand (5 MBq,50µL) and incubated at 37 • C for 15 min. After incubation, aliquots of mixture (200 µL) were transferred to the tubes for ultrafiltration (Amicon 3K; Merck) and centrifuged at 7000 g for 15 min. Equal aliquots (20 µL) of ultrafiltrate (C free ) and plasma (C total ) were measured with γ-counter. The obtained data were corrected for the radioligand binding to the filter membrane. Filtrations were performed in duplicate. The unbound (free) fraction of the radioligand was calculated as f u (%) = C free /C total × 100%.

Conclusions
Four newly synthesized 4 -O-(2-fluoroethyl) honokiol-like derivatives (F-I-F-IV) exhibited significant anti-inflammatory activities comparable or even superior to that of the selective COX-2 inhibitor celecoxib. The selected compound F-IV has good inhibitory ac-tivity to COX-2 confirmed by in vitro assays and by calculations of the relative free energy (ddG) binding to the enzyme. The specific binding to COX-2 of its labeled analogue [ 18 F]F-IV was investigated in a rat model of LPS-induced neuroinflammation. The brain uptake of [ 18 F]F-IV in LPS-treated rats was not as high as that of the previously reported [ 11 C]MPbP, but only slightly higher than in intact animals. However, blocking with celecoxib significantly decreased the uptake of [ 18 F]F-IV, indicating that the radioligand appears to bind specifically to COX-2. Plasma protein binding, in our opinion, is one of the main factors responsible for the differences in the pharmacokinetics of [ 18 F]F-IV and [ 11 C]MPbP. Additional research is currently underway to provide more information on the pharmacokinetics and tissue distribution of F-IV and to study the main pathways for modifying the biphenyl structure to provide a basis for the development of new anti-inflammatory agents and potential COX-2 radioligands.

Data Availability Statement:
The data that support the findings of this study are available from the corresponding author upon reasonable request.