Background

Phosphodiesterases (PDEs) are the only known enzymes capable of breaking down cyclic nucleotides such as cyclic adenosine monophosphate (cAMP) or cyclic guanosine monophosphate (cGMP), and are divided into 11 families (PDE1–11) [1, 2]. Because of their critical role in various cellular functions, they are emerging as a target of new therapeutics for inflammation [3]. Among cAMP-hydrolyzing PDEs, PDE4 accounts for most cAMP-dependent PDE activity in the central nervous system (CNS) [4]. Thus, PDE4 inhibition has been an important strategy for resolving neuroinflammation, conferring neuroprotection, and promoting cognitive function in various CNS disease models by increasing cAMP levels [5, 6]. The PDE4 family consists of four homolog genes (PDE4A–D), among which PDE4B and PDE4D are highly expressed in the brain and may be good candidate targets for drugs to treat CNS disorders [1, 5, 7].

Roflumilast, a selective PDE4 inhibitor that inhibits all PDE4 subtypes to similar extents (with a half-maximal inhibitory concentration [IC50] of 0.2–0.9 nM against PDE4As, PDE4Bs and PDE4Ds and 3–4.3 nM against PDE4Cs) [8]. It has anti-inflammatory effects and is approved by the Food and Drug Administration (FDA) for the treatment of severe chronic pulmonary disease [8, 9]. Moreover, in one study, roflumilast enhanced memory in an object-location task and spatial Y-maze in rodents [10]. In a study of an APP/PS1 mouse model of Alzheimer’s disease, roflumilast (0.1, 0.2, and 0.4 mg/kg) administered intragastrically at a volume of 0.1 mL/10 g for 3 weeks restored performance in the Morris water maze, novel object recognition, and passive avoidance tasks and attenuated neuronal cell injuries and apoptosis via PDE4B/4D-mediated cAMP/CREB/BDNF signaling in the hippocampus and cerebral cortex [11]. Subcutaneous administration of roflumilast (1 mL saline containing 3% DMSO, 3 mg/kg) decreased tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β), and IL-6 levels and the number of apoptotic neurons in the brain following subarachnoid hemorrhage [12].

Zatolmilast (BPN14770), a selective allosteric PDE4D inhibitor, which binds to a primate-specific, N-terminal region with IC50 values of 7.8 nM and 7.4 nM against PDE4D7 and PDE4D3, respectively [13]. It is in a phase 3 clinical study as a therapeutic agent for fragile X syndrome, a rare genetic disorder [14, 15]. It increases brain cAMP levels and the phosphorylation of cAMP response element-binding protein (CREB) and brain-derived neurotrophic factor, markers of memory consolidation. It can also enhance memory and cognitive function and protect neurons, and is a potential clinical treatment for Alzheimer’s disease [16, 17].

Microglia and macrophages are widely distributed in the brain and have vital roles in innate immune responses and neural homeostasis [5, 18]. Under normal brain conditions, microglia remove damaged neurons and cell debris; in their resting phenotypes, they perform various physical functions to support neurogenesis, release neurotrophic factors, and regulate brain development [19,20,21]. However, abnormal activation of microglia alters their cell morphology and functionality, causing chronic neuroinflammation that in turn sustains glial cell activation via the production of various inflammatory mediators and cytokines, including nitric oxide (NO), TNF-α, IL-1β, and IL-6 [19, 22]. Such neuroinflammatory states are highly correlated with various CNS diseases, including Parkinson’s disease, Alzheimer’s disease, and Huntington’s disease. Therefore, inflammatory mediator and cytokine reduction is a potential therapeutic strategy for neuroinflammatory diseases [23, 24].

In this study, we examined the pharmacological effects of roflumilast and zatolmilast on lipopolysaccharide (LPS)-induced neuroinflammation in vitro and in vivo, and evaluated their pharmacokinetics in the brain.

Results

Effects of various concentrations of roflumilast and zatolmilast on NO and inflammatory cytokine production in LPS-exposed BV-2 cells

The immortalized murine microglia cell line BV-2 was used as a substitute for primary microglia in the LPS-induced neuroinflammation model [25]. To exclude potential cytotoxicity at the experimental concentrations, we assessed BV-2 cell viability after roflumilast and zatolmilast exposure using the WST-8 assay (Fig. 1A). Nontoxic concentrations (≤ 10 µM) of roflumilast and zatolmilast were used in subsequent experiments.

Fig. 1
figure 1

Effects of roflumilast and zatolmilast on nitric oxide (NO) and tumor necrosis factor-α (TNF-α) production in lipopolysaccharide (LPS)-induced BV-2 cells. (A) Roflumilast and zatolmilast cytotoxicity in BV-2 cells was assessed at concentrations up to 100 µM using the WST-8 assay. (B, C) NO and TNF-α production after 1 h pretreatment with roflumilast or zatolmilast with or without LPS (1 µg/mL) exposure for an additional 24 h. Culture supernatant was collected and analyzed for NO and TNF-α levels (n = 6 per group). NO levels were measured using the Griess test, and TNF-α levels were detected via enzyme-linked immunosorbent assay. The NO and TNF-α levels are presented as a percentage of untreated control cell levels. The data are expressed as the mean ± standard deviation. ###P < 0.001 versus the control group; **P < 0.01, ***P < 0.001 versus the LPS group. (D) Western blotting results of phosphorylated (p-)NF-κB and NF-κB protein expression. Full-length blots are shown in Supplementary Figure S1

Full size image

Then we investigated whether these inhibited LPS-induced production of NO, an inflammatory mediator, and TNF-α, an inflammatory cytokine, in BV-2 cells. Compared to the untreated control, NO and TNF-α production were significantly increased after LPS treatment. Pretreatment with roflumilast showed a dose-dependent decrease in NO and TNF-α levels (Fig. 1B); however, zatolmilast did not have this effect (Fig. 1C). The calculated IC50 values of roflumilast against NO and TNF-α were 311.5 nM and 12.2 nM, respectively.

Effects of roflumilast and zatolmilast on the NF-κB signaling pathway in LPS-exposed BV-2 cells

Crosstalk between PDE4 activity and the NF-κB pathway may play a major role in modulating inflammatory cytokine levels [1, 26]. To investigate the molecular mechanisms underlying the anti-inflammatory effects of the PDE4 inhibitors roflumilast and zatolmilast, we assessed their ability to inhibit NF-κB and p-NF-κB expression via Western blotting analysis (Fig. 1D). LPS treatment raised p-NF-κB levels compared to the control group. Roflumilast prevented the LPS-induced increase in p-NF-κB expression in a concentration-dependent manner, whereas zatolmilast did not.

Effects of roflumilast and zatolmilast on inflammatory cytokine production in LPS-treated mice

Next, we assessed the effects of roflumilast and zatolmilast on LPS-induced neuroinflammation in vivo to validate the in vitro experiments (Fig. 2). Roflumilast and zatolmilast were administered to mice orally once daily for 4 consecutive days before intraperitoneal injection of LPS (10 mg/kg). Whereas LPS administration alone increased the levels of TNF-α, IL-1β, and IL-6 in both plasma and brain compared to the control, the levels in the roflumilast group were reduced. Zatolmilast did not reduce inflammatory cytokine production in vivo.

Fig. 2
figure 2

Anti-neuroinflammatory effects of roflumilast and zatolmilast in lipopolysaccharide (LPS)-treated mice. (A) Tumor necrosis factor-α (TNF-α), (B) interleukin-1β (IL-1β), and (C) IL-6 levels in plasma and brain homogenates of mice administered roflumilast and zatolmilast. C57BL/6J mice were administered roflumilast or zatolmilast (10 mg/kg) orally once daily for 4 consecutive days. Then, LPS (10 mg/kg) was injected intraperitoneally 2 h after the final drug administration. Mice were sacrificed and plasma and brain samples collected 8 h after the LPS injection. TNF-α, IL-1β, and IL-6 levels were determined by enzyme-linked immunosorbent assay. Data are expressed as the mean ± standard deviation (n = 7–9 per group). ###P < 0.001 versus the control group; **P < 0.01, ***P < 0.001 versus the LPS group

Full size image

Bidirectional MDR1-MDCK monolayer permeability of roflumilast and zatolmilast

Drug interactions with P-glycoprotein (P-gp), a major efflux transporter encoded by the MDR1 gene, may be the primary mechanism hindering drug passage to the brain [27]. Therefore, we performed MDR1-MDCK monolayer permeability assays of roflumilast and zatolmilast (3, 10, and 30 µM) in both the apical-to-basolateral and basolateral-to-apical directions for 2 h. The Papp values of roflumilast and zatolmilast in the apical-to-basolateral direction were > 25.0 and 3.7 × 10–6 cm/s, respectively, and increased in a concentration-dependent manner (Table 1). The efflux ratios (ERs) of roflumilast and zatolmilast were < 0.95 and 1.92, respectively. According to FDA guidelines, Papp < 1 × 10–6 cm/s indicates poor permeability, Papp of 1–10 × 10–6 cm/s indicates moderate permeability, and Papp > 10 × 10–6 cm/s indicates high permeability. In addition, ER > 2 indicates the occurrence of drug efflux. Considering the FDA guidelines, roflumilast and zatolmilast exhibited high and moderately high permeability, respectively. These findings suggest that they are not P-gp substrates.

Table 1 Apparent permeability coefficients (Papp) and efflux ratios for roflumilast and zatolmilast (3, 10, and 30 µM) using the MDR1-MDCK monolayer assay
Full size table

Pharmacokinetics in mice

To clarify the pharmacokinetics of roflumilast and zatolmilast, we constructed mean plasma concentration–time curves and assessed non-compartmental pharmacokinetic parameters after intravenous and oral administration in mice. The results are presented in Fig. 3; Tables 2 and 3. After intravenous dosing (1 mg/kg), the AUCs of roflumilast and zatolmilast were 4.89 and 81.9 µg h/mL, respectively. These compounds had low systemic clearance and volume of distribution of 0.01–0.2 L/h/kg and 0.07–0.59 L/kg, respectively. At a dose of 10 mg/kg, roflumilast had a time-to-peak drug concentration (Tmax) of 2.33 h and peak concentration (Cmax) of 0.15 µg/mL. Zatolmilast had a Tmax of 2 h and Cmax of 2.53 µg/mL following oral administration. The oral bioavailabilities of roflumilast and zatolmilast were 2.3 and 3.7, respectively.

Fig. 3
figure 3

Plasma concentration–time curves of roflumilast and zatolmilast after intravenous and oral administration in mice. Data are expressed as the mean ± standard deviation (n = 3). (A) Intravenous administration (1 mg/kg); (B) Oral administration (10 mg/kg)

Full size image
Table 2 Non-compartmental pharmacokinetic parameters of roflumilast and zatolmilast following intravenous administration (1 mg/kg) in mice
Full size table
Table 3 Non-compartmental pharmacokinetic parameters of roflumilast and zatolmilast following oral administration (10 mg/kg) in mice
Full size table

Brain disposition in mice

Finally, we evaluated the brain disposition of roflumilast and zatolmilast in mice. Plasma and brain tissue samples were obtained at 4 h after oral administration (10 mg/kg), and the drug concentrations were measured via LC-MS/MS. The results are presented in Table 4. The unbound plasma and brain concentrations were 0.39 ng/mL and 0.07 ng/g for roflumilast and 9.57 ng/mL and 1.7 ng/g for zatolmilast, respectively. The calculated Kp,uu values of roflumilast and zatolmilast were 0.17 and 0.18, respectively.

Table 4 Concentrations of roflumilast and zatolmilast in mouse plasma and brain tissues and the unbound brain-to-plasma partition coefficient (Kp,uu) values following oral administration (10 mg/kg) in mice
Full size table

Discussion

We evaluated the in vitro and in vivo bioactivities of two selective PDE4 inhibitors, roflumilast and zatolmilast, against LPS-induced neuroinflammation, as well as their ability to cross the blood–brain barrier. Roflumilast, a PDE4 inhibitor, attenuated the production of NO and the inflammatory cytokines TNF-α, IL-1β, and IL-6 in both BV-2 microglial cells and C57BL/6J mice. However, zatolmilast, a PDE4D inhibitor, showed no anti-inflammatory effects. We further confirmed that the anti-inflammatory effects of roflumilast occurred via attenuation of p-NF-κB levels.

Barber et al. [26] reported that PDE4A4 was upregulated in lung macrophages of subjects with COPD; this is thought to play a major role in the pathogenesis of COPD. This suggests that PDE4A4 might be a PDE4 isoform-specific therapeutic target for COPD. On the other hand, analysis of mRNA levels reveals that PDE4B and PDE4D are the most abundant subfamilies throughout the human, monkey, and rat brains [28]. In addition, TNF-α increases PDE4B expression and nuclear translocation in microglia and PDE4B is highly expressed in activated microglia after traumatic brain injury and spinal cord injury [29]. Thus, it is reasonable to believe that roflumilast may have anti-neuroinflammatory activity primarily by inhibiting PDE4B in BV-2 cells and mice, unlike the anti-inflammatory effect of PDE4A inhibition in COPD.

A crucial mechanism of LPS-induced neuroinflammation is the toll-like receptor 4 (TLR4) signaling pathway. In the response to LPS, TLR4 is an important mediator of several inflammatory pathways and binds to MyD88 to activate NF-κB [30]. It is mainly expressed in microglia of the CNS, and abnormal expression causes CNS injury, including the degeneration of neurons, synaptic loss, and neuronal cell death [23, 31]. This process is mediated by the production of inflammatory cytokines such as TNF-α, IL-1β, and IL-6, increased activity of iNOS, COX-2, β-secretase, and γ-secretase, Aβ accumulation, and oxidative stress [32, 33]. Therefore, LPS can induce neuroinflammation and amyloidogenesis in neurons [34, 35].

The PDE4–cAMP–protein kinase A (PKA)–CREB–NF-κB signaling pathway is closely related to inflammatory responses. PDE4 inhibition increases cAMP levels, in turn activating PKA. PKA activity modulates CREB phosphorylation and NF-κB transcription [5]. LPS injections increase neuroinflammation and damages the BBB [36], where PDE4B is the main PDE4 subfamily engaged in the LPS response [37, 38]. By contrast, TLR ligation has no effect on PDE4D gene expression [7]. Thus, PDE4B inhibition, but not PDE4D, may be necessary for the treatment of TLR-dependent neuroinflammation.

To assess the PDE4 inhibitors roflumilast and zatolmilast for use for the treatment of CNS disorders, we assessed MDR1-MDCK monolayer permeability and mouse brain disposition. To treat brain diseases effectively, a drug must penetrate the BBB [39]. Therefore, identifying the disposition of drugs in the brain is an important step in CNS drug development. In the pharmaceutical industry, Kp,uu,brain is often used in pharmacokinetic–pharmacodynamic assessments of biological effects on the brain and to predict an effective dose [40]. In the present study, both roflumilast and zatolmilast had ERs < 2.0 in the MDR1-MDCK monolayer, indicating that they are not substrates of P-gp. In vivo, both compounds had long Tmax values of 2–4 h and long terminal half-lives > 7 h after oral dosing (10 mg/kg). Plasma concentrations of roflumilast at all times after oral dosing were above the IC50 value against TNF-α (12.2 nM; 4.9 ng/mL), correlating with reduced TNF- α levels in roflumilast-treated mice.

Compounds with Kp,uu,brain > 0.3–0.5 can penetrate the BBB [41]. Their Kp,uu,brain values < 0.18 suggest poor CNS penetration by both roflumilast and zatolmilast. Although the unbound brain concentrations of roflumilast and zatolmilast at Tmax were 0.07 ng/g (0.17 nM) and 1.7 ng/g (4.2 nM), respectively, the value of roflumilast was similar to the PDE4B inhibitory concentration (0.2 nM). The treatment of inflammatory diseases by PDE4 inhibitors such as rolipram and roflumilast has been limited due to side effects such as emesis and vomiting [42, 43]. The calculated oral emetic dosages of roflumilast in mice and rats are 9–30 mg/kg [9]. Robichaud et al. reported that PDE4D, but not PDE4B, is responsible for emesis and other side effects [43]. Selective PDE4B inhibitors have anti-inflammatory effects without side effects [43]. This evidence makes it reasonable to believe that inhibitors with higher selectivity for PDE4B vs. PDE4D should have fewer side effects.

Conclusions

This study demonstrates differential pharmacological effects of two PDE4 inhibitors in both in vitro and in vivo LPS-induced neuroinflammation models. The findings suggest selective PDE4B inhibition as a potential therapeutic target for neuroinflammatory diseases.

Methods

Chemicals and reagents

Roflumilast and zatolmilast were purchased from MedChemExpress (Monmouth Junction, NJ, USA). LPS from Escherichia coli (0111:B4), Griess reagent, and disopyramide were purchased from Sigma Aldrich (St. Louis, MO, USA). Mouse TNF-α, IL-1β, and IL-6 enzyme-linked immunosorbent assay (ELISA) kits were purchased from R&D Systems (Minnesota, MN, USA). High-performance liquid chromatography (HPLC)-grade solvents were used for HPLC tandem mass spectrometry (MS/MS) analysis. The other chemicals and reagents were of the highest grade available.

Cell culture

BV-2 cells, a murine microglial cell line, were purchased from AcceGen (Fairfield, NJ, USA) and cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) (Corning, Riverfront, NY, USA) supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 µg/mL streptomycin. MDR1-MDCK cells, a Madin–Darby canine kidney cell line, were obtained from Dr. Piet Borst (Netherlands Cancer Institute, Amsterdam, the Netherlands) and cultured in DMEM containing 10% FBS, 100 U/mL penicillin, and 100 µg/mL streptomycin. Cells were incubated with 5% CO2 and an atmosphere of 95% humidity at 37℃.

Cell viability assay

Cell viability was assessed using the WST-8 assay with Cyto X reagent (LPS Solution, Daejeon, Korea). BV-2 microglial cells were seeded at a density of 5 × 105 cells/mL in 96-well plates. After incubation for 24 h, the cells were treated with 0–100 µM of either roflumilast or zatolmilast and exposed for 24 h. Finally, Cyto X solution was loaded in each well and incubated for 20 min. Absorbance was measured at 450 nm using a spectrophotometer (BioTek, Winooski, VT, USA).

In vitro LPS-induced neuroinflammation model

BV-2 cells were seeded at a density of 5 × 105 cells/mL in 96-well plates. After 24 h, the cell culture medium was replaced with DMEM containing 1% FBS, and cells were treated with various concentrations of roflumilast and zatolmilast for 1 h. Finally, 1 µg/mL LPS was added to the cell culture medium for 24 h to induce inflammation.

In vivo LPS-induced neuroinflammation mouse model

Eight-week-old male C57BL/6J mice (body weight 22–26 g) were purchased from Orient Bio Inc. (Seongnam, Korea). The animals were acclimatized to a temperature-controlled room at 23℃ with a 12 h light/12 h dark cycle for 7 days before the experiment; the mice received water and food ad libitum. All procedures complied with the Guidelines for the Care and Use of Laboratory Animals of the Animal Ethics Committee of the Korea Research Institute of Chemical Technology.

To assess anti-neuroinflammatory effects of roflumilast and zatolmilast in vivo, C57BL/6J mice were divided into four groups (n = 7–9 in each group): control group, LPS group, LPS + roflumilast 10 mg/kg group, and LPS + zatolmilast 10 mg/kg group. The tested compounds were solved in the vehicle (10% N-methyl-2-pyrrolidone, 60% polyethylene glycol 400, 30% distilled water) immediately before the experiment. The dosing volume was 5 mL/kg. In the treatment groups, roflumilast and zatolmilast were administered orally once daily for 4 consecutive days. On day 4, all groups (except the control) received an intraperitoneal injection of LPS (10 mg/kg) 2 h after drug administration. Finally, 8 h after LPS injection, the mice were sacrificed via CO2 inhalation. Plasma and brain samples were collected and stored at − 80℃ until analysis.

Measurement of NO

Frozen brain tissues in cold 0.1 M phosphate-buffered saline were homogenized using the FastPrep-24™ Classic homogenizer (MP Biomedicals, Irvine, California, USA). The homogenates were centrifuged at 15,000 rpm for 10 min, and the supernatants were collected. To assess NO levels, the cell culture supernatant and brain tissue pellet samples were reacted with an equal volume of Griess reagent at room temperature for 10 min. Absorbance was measured at 540 nm using a spectrophotometer (BioTek).

TNF-α, IL-1β, and IL-6 ELISA

To measure TNF-α, IL-1β, and IL-6 levels in each sample, ELISA was performed using mouse TNF-α, IL-1β, and IL-6 ELISA kits (R&D Systems). The absorbance was measured at 450 nm using a spectrophotometer (BioTek).

Western blotting analysis

BV-2 cells were treated with roflumilast (1 and 10 µM), zatolmilast (1 and 10 µM), or 0.5% dimethyl sulfoxide for 1 h. Then the cells were incubated with or without LPS (1 µg/mL) for 4 h. Cells were harvested using trypsin and washed twice in cold Dulbecco’s phosphate-buffered saline and lysed with RIPA lysis buffer (Elpis, Daejeon, Korea) containing protease and phosphatase inhibitor mini tablets (Thermo Fisher, Waltham, MA, USA). Protein lysate concentrations were determined using the Pierce BCA Protein Assay kit (Thermo Fisher). Protein samples (5 µg) were separated via 5–20% sodium dodecyl sulfate polyacrylamide gel electrophoresis and then transferred to polyvinylidene fluoride membranes (ATTO, Tokyo, Japan). After transfer, membranes were blocked using Tris-buffered saline containing 0.5% Tween 20 and 5% bovine serum albumin at room temperature for 1 h. The membranes were incubated overnight at 4℃ with primary antibodies at a dilution ratio of 1:1000. NF-κB (cat no. #8242, Cell Signaling Technology, MA, USA), phosphorylated (p-)NF-κB (cat no. #3033, Cell Signaling Technology), and GAPDH (cat no. #5174, Cell Signaling Technology) were used as primary antibodies. Subsequently, blots were detected using horseradish peroxidase-conjugated secondary antibodies (1:10,000 dilution) and the ECL substrate (Millipore Corporation, Billerica, MA, USA) at room temperature for 1 h. Protein expression levels were measured using WSE-6100 LuminoGraph (ATTO). The bands were analyzed using the ImageJ software (NIH, Bethesda, MD, USA).

LC-MS/MS analysis

LC-MS/MS analyses were performed on an Agilent 1200 Series HPLC System (Agilent Technology, Palo Alto, CA, USA) coupled to an API 4000 Qtrap mass spectrometer equipped with an electrospray-ionization source operated in positive-ionization mode (AB SCIEX, Foster City, CA, USA). Multiple reaction monitoring (MRM) mode was used for targeted analysis. The optimized instrument conditions were as follows: curtain gas, 10 psi; ion spray voltage, 5500 V; temperature, 550℃; nebulizing gas (GS1), 50 psi; heating gas (GS2), 50 psi. MRM was m/z 403.1 → 167.1 for roflumilast, 406.1 → 290.1 for zatolmilast, and 340.1 → 239.2 for disopyramide (internal standard). All data were measured using Analyst Software version 1.4.2 (AB SCIEX).

The LC chromatograph was equipped with a Unison UK-C18 column (100 × 2.0 mm i.d., 3 μm; Imtakt, Portland, Oregon, USA). The mobile phase consisted of 0.1% formic acid in water (A) and acetonitrile (B), and was filtered and degassed before use. Roflumilast and zatolmilast were analyzed in isocratic elution mode or gradient elution mode. The isocratic elution mode conditions were 0–2.5 min, 80% B; the flow rate was 0.3 mL/min. The gradient elution mode conditions were 0–0.2 min, 5% B; 0.2–2.3 min, linear increase to 95% B; 2.3–2.4 min, 95% B; 2.4–2.5 min, linear decrease to 5% B; 2.5–5 min, 5% B; the flow rate was 0.35 mL/min.

MDR1-MDCK monolayer permeability analysis

MDR1-MDCK cells were seeded at a density of 2 × 105 cells/well on the apical side of 24-transwell plates with a pore size of 0.4 μm (Corning Costar, Cambridge, MA, USA). Each basolateral chamber was filled with 0.6 mL medium, and each apical side with 0.2 mL medium containing cells. The plates seeded with cells were incubated for 3 days. The culture medium was replaced on both the apical and basolateral sides 2 days after seeding. Before the experiment, the transepithelial electrical resistance was measured using an epithelial voltohmmeter (Millicell ERS-2; Millipore), where 300 ohms or greater was considered to indicate the formation of a cell monolayer with tight junctions. The cells were rinsed twice with pre-warmed Hank’s balanced salt solution (HBSS) containing 10 mM of HEPES and were incubated in a CO2 incubator for 30 min. Before the transport assay, roflumilast, zatolmilast, and reference compounds (atenolol, propranolol, and loperamide) were diluted with HBSS containing 10 mM HEPES to prepare the desired concentrations. The final concentrations of roflumilast and zatolmilast were 3, 10, and 30 µM. After equilibration, the solutions containing the test or reference compounds were added to either the apical side (0.2 mL) or basolateral side (0.6 mL) of the wells, and the plates were incubated in a shaking incubator (37℃, 200 rpm). Samples were collected every 30 min. To test for apical-to-basolateral flow, 110 µL sample was removed from each basolateral chamber; to test for basolateral-to-apical flow, 110 µL sample was taken from each apical chamber; then, each well was replenished with 110 µL fresh HBSS. To each sample, 50 µL internal standard (5 ng/mL disopyramide) in water was added. Samples were prepared for analysis using solid-phase extraction, and the sample concentrations were analyzed via LC-MS/MS. The permeability of the tested compounds was evaluated by calculating the apparent permeability coefficient (Papp, cm/s) and the efflux ratio (ER). The Papp was calculated as follows:

$${\text{P}}_{app}=\frac{\text{d}\text{Q}}{\text{d}\text{t}}\times \frac{1}{\text{A}}\times \frac{1}{{\text{C}}_{0}}$$

where dQ/dt is the initial permeability rate of the compound across the cells (ng/s), A is the surface area of the cell monolayer (cm2), and C0 is the initial concentration of the donor side (ng/mL).

From the apical-to-basolateral and basolateral-to-apical Papp values, the ER was calculated as follows:

$${\rm{Efflux}}\,{\rm{Ratio}} = \frac{{{{\rm{P}}_{app\,\,\left( {Basolateral\, \to \,Apical} \right)}}}}{{{{\rm{P}}_{app\,\,\left( {Apical\, \to \,Basolateral} \right)}}}}$$

In vitro protein-binding assay using equilibrium dialysis

The roflumilast and zatolmilast solutions were prepared to a final concentration of 5 µM by spiking 6 µL 500 µM stock solution in 594 µL pooled mouse plasma and mouse blank brain homogenate. A rapid equilibrium dialysis device with inserts (cat no. #90,112, Thermo Fisher) was used for the protein-binding assay. First, 100 µL roflumilast or zatolmilast solution was loaded into the red chamber and 350 µL prepared BupH™ phosphate-buffered saline (cat no. #28,372, Thermo Fisher) was loaded into the white chamber. The samples were incubated for 4 h in a shaking incubator (37℃, 300 rpm). After incubation, 50 µL buffer from the white chamber was added to 50 µL fresh mouse plasma or brain homogenate, and 50 µL dosing solution from the red chamber was added to 50 µL fresh buffer to adjust for matrix effects. To each sample, 200 µL ice-cold acetonitrile with internal standard was added. The samples were centrifuged at 4,000 rpm for 20 min and the supernatants were collected for LC-MS/MS analysis. The unbound fraction of drug (Fu) in plasma and brain tissue was calculated as follows based on previous studies [45, 46]:

$${{\text{F}}_{{\text{u,plasma}}\,{\text{or}}\,{\text{dh}}}}\,\left( \% \right) = \frac{{{C_{protein-free\,compartment}}}}{{{C_{protein-containing\,compartment}}}} \times 100$$
$${{\text{F}}_{{\text{u,}}\,{\text{brain}}}}\,\left( \% \right) = \frac{{1/D}}{{\left[ {\left( {\frac{1}{{Fu,\,dh}}} \right) - 1} \right] + \left( {\frac{1}{D}} \right)}}$$

where Cprotein-free compartment and Cprotein-containing compartment, D, and dh are the concentrations of drug in the buffer chamber and drug solution chamber, dilution factor, and diluted brain homogenate, respectively.

The values were calculated using the measured diluted free fractions and the respective dilution factor.

Pharmacokinetics and brain disposition in mice

Male C57BL/6J mice (aged 8 weeks, 22–26 g) were purchased from Orient Bio and used to assess the pharmacokinetics of roflumilast and zatolmilast. All mice were acclimated to a temperature-controlled room at 23℃ for 7 days and fasted for 12 h before the experiment.

For the pharmacokinetic study, mice were administrated roflumilast or zatolmilast intravenously and orally. Blood samples were collected at 0.083, 0.5, 1, 2, 4, 8, and 24 h after drug administration. Plasma samples were obtained via centrifugation at 15,000 rpm for 3 min and stored at − 20℃ until LC-MS/MS analysis.

For the brain disposition evaluation, 4 h after oral administration (10 mg/kg), the mice were sacrificed via CO2 inhalation. Residual blood in the brain was removed from circulation via cardiac perfusion with saline solution. Blood and brain samples were collected and stored at − 20℃ until LC-MS/MS analysis.

Plasma and brain standard calibration curves were prepared to final concentrations of 1.95–8000 ng/mL. For both plasma and brain samples, 30 µL each sample was vortexed with 270 µL acetonitrile containing disopyramide as the internal standard for protein precipitation. The samples were centrifuged at 15,000 rpm for 10 min and the supernatants were analyzed by LC-MS/MS. The brain-to-plasma ratio (Kp) and unbound Kp (Kp,uu) were calculated as follows:

$${\text{K}}_{\text{P}}= {\text{C}}_{\text{b}\text{r}\text{a}\text{i}\text{n}}/{\text{C}}_{\text{p}\text{l}\text{a}\text{s}\text{m}\text{a}}$$
$${\text{K}}_{\text{p},\text{u}\text{u}}={\text{K}}_{\text{p}} \times \frac{{\text{f}\text{u}}_{\text{b}\text{r}\text{a}\text{i}\text{n}}}{{\text{f}\text{u}}_{\text{p}\text{l}\text{a}\text{s}\text{m}\text{a}}}$$

Pharmacokinetic analysis

Pharmacokinetic parameters were calculated via non-compartmental analysis using Phoenix WinNonlin, ver. 8.3 (Pharsight, Mountain View, CA, USA). The area under the plasma concentration–time curve from 0 to 24 h (AUClast) and infinity (AUCinf) were determined following the linear trapezoidal rule.

Statistical analysis

All experimental data were collected in triplicate and are expressed as the mean ± standard deviation. Statistical analyses were evaluated using one-way analysis of variance followed by Tukey’s post hoc test with GraphPad Prism 9.4 (GraphPad Software, San Diego, CA, USA). Significant differences were defined as p values < 0.05, < 0.01, or < 0.001.