**3. Discussion**

In animal models, Dufour et al. [12] suggested that budesonide (BUD) complexed with 2-hydroxypropyl-β-cyclodextrin (HPβCD) might be an alternative to BUD alone in the treatment of smoking-induced COPD. The current study was designed to characterize the effect of the BUD:HPβCD complex on the response of human alveolar epithelial cells (A549) or human monocytes (A-THP1) to a mix of hydrogen peroxide and lipopolysaccharides (H2O<sup>2</sup> + LPSs) mimicking stressful effects including those from cigarette smoke [28,44,45] or from environmental toxicants. We characterized the effect of the BUD:HPβCD complex on (i) ROS generation (oxidative stress), (ii) Akt phosphorylation (PI3K/Akt signaling activation), (iii) HDAC2 phosphorylation (HDAC2 inhibition of activity), and (iv) IL-8 release (inflammatory response) in comparison with the effects induced by BUD or HPβCD.

We demonstrated the protective effect afforded by BUD:HPβCD against cytotoxicity and ROS generation induced by oxidative and inflammatory stress in comparison with BUD. The effect observed for BUD:HPBCD was comparable to that observed with HPBCD and might be limited by cholesterol. We also demonstrated (i) the involvement of the canonical molecular pathway including ROS generation, decrease in PI3K/Akt activation, decrease in HDAC2 activity and insensitivity to glucocorticoid in the effect induced by BUD:HPβCD, (ii) maintenance of IL-8 decrease with BUD:HPβCD—even BUD at a high concentration (100 µM) induced a slightly higher effect—and (iii) the absence of improvement in glucocorticoid insensitivity with BUD:HPβCD in comparison with BUD, in conditions where HDAC2 was inhibited.

Improvement of cell viability after oxidative and inflammatory stress induced by BUD:HPβCD is likely due to HPβCD and linked to a decrease in ROS generation. The literature has reported that cyclodextrins, including HPβCD, may improve the toxicological profile of drugs by complexing them [46,47]. Additionally, the antioxidant potential of HPβCD has been reported. Anraku et al. [48] showed HPβCD remove pro-oxidants such as uremic toxins from the blood in a rat model of chronic renal failure. Zimmer et al. [49] showed that HPβCD decreases aortic ROS generation in a mouse model of atherosclerosis. Other reports reviewed by López-Nicolás et al. [50] described HPβCD as a protective agent of lipophilic nutrients and antioxidants against oxidation in foods. The demonstration of HPβCD's antioxidant potential is interesting given the major role played by oxidative stress in numerous pathologies including COPD [51]. Here, the molecular mechanism leading to a decrease in ROS is still unclear but a direct effect through the interaction of HPβCD with H2O<sup>2</sup> (Figure S3) is unlikely.

An indirect effect through changes in biophysical membrane properties could be suggested as an alternative explanation. We initially suggested that membrane cholesterol would play a major role in the occurrence of BUD:HPβCD-related cytoprotective effects. It has been extensively demonstrated that βCDs, including HPβCD, can interact with lipid membranes, and change membrane biophysical properties [17,35] closely related to signal transduction. This agrees with our previous experiments on giant unilamellar vesicles (GUVs), since we demonstrated BUD:HPβCD and HPβCD disrupted the liquid-disordered/liquid-ordered (Ld/Lo) phase separation observed in the presence of cholesterol for the benefit of the Ld phase, a process hindered in the presence of cholesterol [27]. Here, we observed an increase in BUD:HPβCD-related antioxidant effects in cholesterol-depleted cells suggesting that cholesterol might hinder ROS generation. The BUD:HPβCD-related antioxidant effect was preserved and even increased in cells partially depleted in cholesterol (50% cholesterol depletion after 30 min exposition to MβCD at 5 mM; no or very small cholesterol depletion induced by HPβCD for 2 h at the highest concentrations used in this work; Figure S4). Extracellular mechanisms are unlikely since we observed (i) no cellular uptake of HPβCD over the entire incubation period (Figure S5), and (ii) no neutralization of extracellular signals potentially responsible for oxidative stress, namely H2O<sup>2</sup> and free radicals (Figure S3). BUD:HPβCD and HPβCD-related cytoprotective effects could be seen as a membrane-mediated mechanism involving membrane lipid disorganization with limited lipid extraction after 2 h (cholesterol extraction induced by BUD:HPβCD or HPβCD reached 18% and 12%, respectively, while no cholesterol extraction in cholesterol-depleted cells was observed (Figure S4)), agreeing with the work of Lopez et al. [52].

One remaining question is the cross-talk between the antioxidant effect and inhibiting effect on oxidant-induced PI3K/Akt signaling. NAC and Vit C concentrations that inhibit more than 75% of ROS generation after 2 h of incubation (Figure S6) showed a protective effect against H2O<sup>2</sup> + LPS-induced increase in PI3K/Akt signaling of about 36% (NAC) and ~65% (Vit C). The effect was not related to the effect of LPS which might induce an increase in PI3K/Akt signaling [53,54] since we observed that H2O<sup>2</sup> + LPS-induced increase in PI3K/Akt signaling was almost exclusively associated with the presence of H2O<sup>2</sup> (Figure S7). The activity of endogenous antioxidant enzymes GSH peroxidase against H2O<sup>2</sup> [55], difference in the location within the bilayer between the effect induced by BUD:HPBCD and the location of enzymes involved in ROS generation or PI3K/Akt activation could be also involved.

Focusing on the final attempt for BUD:HPBCD, meaning its ability to decrease the release of inflammatory cytokines after oxidant and inflammatory stress, we could have expected a higher anti-inflammatory effect of the BUD:HPβCD complex compared to the BUD alone. Dufour et al. [12] in a murine asthma model showed that similar anti-inflammatory effects could be obtained with a 2.5-fold lower BUD concentration when given as a complex with HPβCD. Zimmer et al. [49] reported anti-inflammatory effects of HPβCD in vivo in a mice model of atherosclerosis. At the cellular level, George et al. [56] assumed an anti-inflammatory property of HPβCD after showing that its presence along with plasticized poly(vinyl chloride) (PVC) reduced LPS-induced TNF-α expression in human monocyte-like U937 cells while PVC alone had no effect. Matassoli et al. [57] showed that HPβCD can inhibit LPS-induced TNF-α secretion in primary human monocytes. The higher effect on reduction of IL-8 release induced by BUD at a high concentration (100 µM) as compared to the effect of BUD:HPβCD could be linked to an inflammatory effect observed at high doses of HPβCD [58]. Cell-type cellular components of inflammation [59] and changes in the release of BUD from HPβCD hydrophobic cavity, depending upon the concentrations, could also play a role.

Lastly, in a potential translational perspective, the design of studies and concentrations have to be questioned. First, cells were exposed with BUD and H2O<sup>2</sup> + LPS at the same time, meaning that BUD had time to prevent the inflammatory response before the decrease in glucocorticoid sensitivity induced by oxidative stress could take place. We reproduced experiments by changing the time course and by preincubating cells for 30 min with the oxidant and inflammatory stress before the incubation of cells with BUD:HPβCD for 2 h. No differences were observed. Another critical parameter would be the equilibrium between the free and bound forms of BUD or HPBCD [60]. In the presence of a lipophilic membrane, drug partitioning from the complex into the membranes can occur, promoting drug release from the CD hydrophobic cavity. The latter point agrees with our work. Indeed, we showed that in pure phospholipid monolayers there is an increase in membrane surface pressure with the BUD:HPβCD complex but not with HPβCD. This increase is usually associated with the insertion of a molecule within the monolayer. Since the only difference between HPβCD and the BUD:HPβCD complex is the presence of BUD, we could assume that BUD was inserted within the membrane. The critical importance of the equilibrium between free and complexed budesonide was also evidenced when we determined the effect of a mix of BUD and HPβCD on IL-8 release for cells treated with increasing concentrations of trichostatin, a pharmacological inhibitor of HDAC2. The protective effect against IL-8 release of the mixture was higher than that afforded by the complex (Figure S8).

Second, BUD concentrations and/or amount of oxidant and inflammatory stressors used are relevant for patho-physio-logical conditions. BUD dry powder for inhalation (Pulmicort®), was recommended for COPD patient administration—up to 1000 µg/day on average. If we assume that about 30% of the nominal dose inhaled with a dry powder inhaler might reach the lungs [61,62], therefore 300 µg of BUD dry powder in Pulmicort® administered in patients might reach the lungs. If the 300 µg of BUD will disperse in the lung lining fluid (20–40 mL in a human of 70 kg) [63], then the pulmonary BUD concentration could be approximately 17–35 µM, which is in the range of concentrations used in our study (1–100 µM). However, we must remain cautious since the amount of

BUD deposited in the lung is difficult to predict. The question of the relevance of the quantity of H2O<sup>2</sup> + LPS is also raised. Here again, it appears difficult to properly assess the exposition of alveolar cells to H2O<sup>2</sup> and LPS during smoking—e.g., many factors should be considered, such as the frequency of smoking, the number and type of cigarettes smoked per day, the duration of smoking, the distribution of smoke in the lungs, the half-life of each molecular species generated in cigarette smoke, their own biodisponibility, and so on. Nakayama et al. [28] and Hasday et al. [45], respectively, reported that extract amounts of H2O<sup>2</sup> ranging from 500 nmol to 4 µmol of H2O<sup>2</sup> per cigarette and 6 to 9 µg of active LPS per gram of a cigarette can be extracted. The amount of H2O<sup>2</sup> used in this work appears less important (up to 200 nmol in 200 µL), whereas the amount of LPS appears in the same range (up to 10 µg in 100 µL).

In conclusion, we demonstrated the anticytotoxic, antioxidant, anti-inflammatory properties of the BUD:HPβCD complex with protective activity against PI3K/Akt signaling activation and HDAC2 inhibition induced by oxidative stress. The antioxidant and anticytotoxic properties appeared essentially due to HPβCD while the anti-inflammatory properties appeared mainly to be due to BUD. Further investigations are clearly needed for a more complete view of the potential of the BUD:HPβCD complex in other relevant models of oxidative stress-induced glucocorticoid insensitivity in vitro or in vivo.
