KS-40070 is synthesized in the Korea Research Institute of Chemical Technology (Park et al., 2021). The purity of KS-40070 was determined to be 99.9%. Analytical evaluation was conducted using a Waters ACQUITY H-Class PLUS UPLC system (Waters Corporation, Milford, MA, United States) coupled with an SQD2 mass spectrometry detector. Chromatographic separation was performed on an ACQUITY UPLC BEH C18 column (1.7 μm, 2.1 × 50 mm). The mobile phase consisted of buffer A (water containing 0.1% formic acid) and buffer B (chromatographic-grade acetonitrile containing 0.1% formic acid).

3T3-L1 cells (ATCC #CL-173, mouse preadipocyte) were purchased from the American Type Culture Collection (Manassas, VA, United States). 3T3-L1 cells were seeded 5 × 10⁵ in 6-well plates coated with collagen (Corning, Corning, NY, United States) and maintained in Dulbecco’s modified Eagle’s medium (DMEM, Gibco/Invitrogen, Carlsbad, CA, United States) supplemented with 10% fetal bovine serum (FBS), 100 μg/mL penicillin, and 100 μg/mL streptomycin. Cells were grown at 37°C in a 5% CO₂ incubator with 95% humidified air. After cells reached complete confluence, cells were treated with differentiation-inducing cocktail, DMI medium containing dexamethasone (DEX, final concentration 2.5 μΜ), isobutyl-1-methylxanthine (IBMX, final concentration 0.5 mM), insulin (final concentration 2 μg/mL) and 3–4 days that was then replaced with insulin (2 μg/mL) for another 4 days. Medium was replaced every 2 days.

Human adipose-derived mesenchymal stem cells (ADMSC; CB-ADMSC, # CEFO-ADMSC) were used to generate 3D human adipose tissue-mimicking model, as previously described by Park et al. (Park et al., 2020). Brefly, ADMSCs were purchased from CEFO Bio (Seoul, Korea) and maintained using a Human Adipose-derived MSC Kit (CEFO Bio) supplemented with 100 μg/mL penicillin and 100 μg/mL streptomycin. Cells were grown at 37°C in a 5% CO₂ incubator with 95% humidified air. After cells reached complete confluence, they were treated with DMIT cocktail containing DEX (final concentration 10 μΜ), IBMX (final concentration 1 mM), insulin (final concentration 10 μg/mL), and indomethacin (final concentration 40 μΜ) for 12 days. Medium was replaced every 2 days. Alginate beads for 3D cell culture were produced by mixing ADMSC, alginate (2%, w/v), gelatin (0.5%), collagen (40 μL/mL), and serum free medium that was subsequently incubated for 30 min at 37°C. Cells suspended in alginate solution were transferred by syringe with an 18G needle into 50 mL tubes containing 1% calcium chloride. Excess calcium chloride was removed by incubating alginate beads in serum free medium for about 5 min. Medium was replaced and transferred to a 25 T flask and then incubated overnight for stabilization. The next day, alginate beads were transferred to 6-well plates, and differentiation was initiated. All cell lines were used passage under 6 (At higher 6 passage, the lipid differentiation rate is decreased.)

Cells were washed twice with Dulbecco’s Phosphate Buffered Saline (DPBS) and RIPA lysis buffer (Thermo Fisher Scientific, Waltham, MA, United States) containing a protease inhibitor cocktail tablet (Sigma Aldrich, St. Louis, MO, United States) and phosphatase inhibitor cocktail tablet (Roche, Basel, Switzerland) were added for lysis. Whole cell lysates were centrifuged at 13,000 rpm, after which supernatants were collected. Protein concentration in samples was determined using a BCA protein assay kit (Thermo Fisher Scientific). Then, 20 µg of total protein was loaded into a Bolt 4%–12% Bis-Tris Gel (Invitrogen) and transferred to a PVDF membrane (Merck Millipore, Burlington, MA, United States). Blocked membranes were incubated with antibodies. Protein bands were visualized with the Super Signal West Femto Maximum Sensitivity Substrate kit (Thermo Fisher Scientific) and analyzed by chemiluminescence (Biorad, Hercules, CA, United States). DYRK1B, DYRK1A, PPARγ, C/EBPα, perilipin, FABP4, adiponectin, phospho-Akt, Akt, phospho-FOXO1A, phospho-GSK α/β, GSK3α/β and β-actin were purchased from Cell Signaling Technology (Danvers, MA, United States). FOXO1A, Glucose-6-phosphatase (G6Pase) and tumor necrosis factor-α (TNF-α) were purchased from Abcam (Cambridge, UK). Western blot density results were analyzed with ImageJ software provided by the National Institutes of Health (Bethesda, MD, United States).

DYRK1B siRNA and negative control siRNA were purchased from Thermo Fisher Scientific. 3T3-L1 cells were transfected using the Lipofectamine RNAiMAX reagent (Thermo Fisher Scientific) according to the manufacturer’s protocol. Final concentration of siRNA was 250 pmol. After transfection, cells were treated with DMI containing insulin (final concentration 2 μg/mL), IBMX (final concentration 0.5 mM), and DEX (final concentration 2.5 μΜ) for 2 days that was then replaced with insulin (2 μg/mL) for another 4 days. Medium was replaced every 2 days.

A LanthaScreen Eu Kinase Binding Assay (Invitrogen) was performed according to the manufacturer’s protocol. Briefly, compounds were diluted into kinase buffer A. Tracer solution in kinase and kinase/antibody solutions were then prepared, mixed with the compound, and incubated for 1 h at room temperature. Acceptor/tracer emission (665 nm) and antibody/donor emission (615 nm) were measured and then used to calculate the emission ratio. Human DYRK1B recombinant protein (#PV4649), human DYRK1A recombinant protein (#PV3785), human DYRK2 recombinant protein (#PV6331), 5× kinase buffer A (#PV3189), kinase tracer 236 (#PV5592), and LanthaScreen™ Eu-anti-GST antibody (#PV5594) were purchased from Invitrogen.

Lipid droplets were stained using boron-dipyrromethene (BODIPY) 493/503 (Invitrogen). Nuclei were stained using 4′,6-diamidino-2-phenylindole (DAPI, Cell Signaling Technology). For fixative, cells were washed with DPBS and then incubated in a formalin solution for 10 min. After removing the formalin solution with DPBS, cells were permeabilized using 0.25% Triton X-100 for 10 min. After permeabilization, solution was removed with DPBS. The cells were stained with BODIPY solution for 10 min. Stained cellular lipid droplets were washed with DPBS, and cells were then stained with DAPI solution for 1 min. Afterwards, cells were washed with DPBS and visualized using a fluorescence microscope.

Furthermore, lipid accumulation was estimated in living cells using an AdipoRed™ Adipogenesis Assay Reagent (PT-7009, Lonza, Basal, Switzerland). Cells were washed, AdipoRed solution (30 μL/mL in DPBS) was added, and cells were incubated for 20 min at room temperature. The 3 dimensional (3D) cells were transferred to micro tubes, and cells in the alginate beads were isolated using 16 mM EDTA. After shaking for 5 min, lysates were centrifuged at 13,000 rpm for 1 min 30 s. After removing the supernatant, the red pellet was suspended with DPBS, then it was transferred to a plate and measured using a microplate reader (Molecular Devices, San Jose, CA, United States) at an excitation wavelength of 485 nm and an emission wavelength of 538 nm. DNA was quantified using a Blue Fluorometric dsDNA Quantitation Kit (Invitrogen). Hoechst 33,342 with TNE buffer (1.25 μL/mL) was added to the plate, incubated for 5 min at room temperature, and measured using a microplate reader at an excitation wavelength of 355 nm and an emission wavelength of 460 nm.

Four-week-old male (13–15 g) C57BL/6J mice were purchased from Orient Bio (Seongnam, Korea). All animal experiments were conducted according to the guidelines for animal experimentation and approved by the institutional animal care and use committee of KRICT (Approval number: 2019-7A-01–02). All animals were housed in plastic cages and allowed free access to water and food. They were maintained in a room illuminated daily from 07:00 to 19:00 (12/12 light/dark cycle) under controlled temperature (23°C ± 1°C), ventilation (10–12 air changes per hour), and humidity (55% ± 5%). At the end of the experiments, mice were euthanized using carbon dioxide inhalation in accordance with institutional guidelines for animal care and use.

Obesity/diabetes was induced in C57BL/6J mice by feeding a high fat diet (Diet Research Inc., New Brunswick, NJ, United States) with 60% of its energy from fat for 5 weeks starting at 5 weeks old. Rosiglitazone (Combi-Blocks, San Diego, CA, United States) and Xenical (Sigma-Aldrich) were used as reference compounds. Mice were divided into six groups (n = 7) and fed normal diet (ND, Lean), DIO (Diet-induced obesity mouse), DIO with 10 mg/kg rosiglitazone, DIO with 20 mg/kg Xenical, DIO with 50 mg/kg of KS-40070, and DIO with 100 mg/kg of KS-40070 for 9 days. All compounds were dissolved in 0.5% carboxymethyl cellulose and orally administered twice a day (10:00 and 17:00).

Blood samples of mice (50 μL) were collected after overnight fasting from the retro-orbital plexus and stored in heparin-coated capillary tubes. Blood samples were immediately centrifuged for 15 min at 13,000 rpm and then stored at −20°C prior to assay. Concentrations of total cholesterol, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), triglyceride (TG), ALT, and AST were measured using a Response 920 automated biochemical analyzer (DiaSys, Holzheim, Germany).

After mice were sacrificed, white adipose tissue (related tissues of metabolic disease, subcutaneous and visceral fat) was collected, fixed in 10% formalin, embedded in paraffin, and sectioned to a thickness of 5 μm. For hematoxylin and eosin (H&E) staining, sections were deparaffinized with xylene three times, rehydrated by pass -through in stepped concentrations of alcohol baths (100%, 90%, 80%, and 70%), and then washed briefly under running water. Sections were stained in hematoxylin solution (Sigma-Aldrich) for 15 min and washed under running water for 3 min, then incubated in bluing solution for 5 s. Sections were subsequently stained with eosin (Daejung Chemical, Siheung, Korea) for 3 s, washed under running water for 3 min, and then dehydrated via pass-through in stepped concentrations of alcohol baths (70%, 80%, 95%, and 100%), followed by clearing with xylene three times. Stained sections were mounted with mounting media (Fisher Chemical, Pittsburgh, PA, United States) and visualized using light microscopy (DS-Ri2, Nikon, Tokyo, Japan).

For the OGTT, mice were orally administered 0.2 g/mL glucose (Sigma-Aldrich) after overnight fasting. Blood samples (50 μL) were obtained at 15, 30, and 60 min afterwards from the retro-orbital plexus and transferred to heparin-coated capillary tubes. For the ITT, non-fasted blood samples were drawn prior to insulin injection (0.55 U/kg, i.p.; Sigma-Aldrich, USA) and at 15, 30, and 60 min after insulin administration. The blood samples were immediately centrifuged for 15 min at 13,000 rpm and then stored at −20°C prior to assay. Glucose in plasma was measured using a Response 920 automated biochemical analyzer, (DiaSys, Holzheim, Germany). Plasma insulin levels were measured using an ultrasensitive mouse/rat insulin ELISA kit (FUJIFILM Wako Shibayagi corporation, Shibukawa, Japan). Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) values to indicate insulin sensitivity were calculated based on the following formula: (fasting insulin [μU/mL]xfasting glucose [mg/dL])/405) (Ko et al., 2014).

To build a three-dimensional (3D) structure of human DYRK1B was generated using SWISS-MODEL server (https://swissmodel.expasy.org/). The sequence of hDYRK1B was obtained from the UniProtKB database (https://www.uniprot.org; accession number Q9Y463). For a 3D template of hDYRK1B, sequence similarity search was performed using BLAST (Basic Local Alignment Search Tool) (Camacho et al., 2009) available on the website. The template X-ray crystal structure, hDYRK1A (PDB code 6EIF) was selected as the top score sequence with high resolution. The sequence identity and similarity between hDYRK1B and hDYRK1A are 83.3% and 92.8% respectively.

To predict the binding mode of compound, KS-40070, we performed a docking study using Schrodinger Suite 2020–4 (Schrödinger, LLC, New York, NY, United States, 2020). The DYRK1B 3D structure was obtained from homology modeling, and the protein preparation was revised using Protein Preparation Wizard in Maestro. The receptor grid box for docking was generated 25 Å × 25 Å × 25 Å size centered on complexed ligand, which was copied from the superimposition of hDYRK1A in complex with inhibitor. Compounds were minimized using an OPLS3e force field with a dielectric constant of 80.0 in MacroModel v13.0. Molecular docking of compounds was performed using the Glide v8.9 with SP method. The proposed binding models of compounds for DYRK1B were represented using Discovery Studio 2020 (Dassault Systèmes BIOVIA, San Diego, CA, United States, 2020).

Graphical results are presented as mean ± standard error of the mean (S.E.M). Statistical significance was analyzed using GraphPad Prism software (GraphPad Software Inc., La Jolla, CA, United States) and determined by student’s t-test or a one-way analysis of variance followed by Tukey’s multiple-comparison test. A P-value of less than 0.05 was considered statistically significant.

We assessed the upregulation of DYRK1B proteins during adipogenesis by monitoring expression change throughout the differentiation process in 3T3-L1 cells via Western blot analysis. As depicted in Figures 1A,B, DYRK1B expression exhibited a progressive increase during adipogenesis, reaching saturation on 8 days of differentiation. Subsequently, to elucidate the effect of DYRK1B on adipocyte differentiation, we conducted a gene silencing assay. This assay revealed that the knockdown of DYRK1B not only led to a reduction in DYRK1B expression but also resulted in decreased levels of Perilipin, a protein associated with lipid droplet formation (Figures 1C–E) (Itabe et al., 2017). BODIPY staining further demonstrated a depletion in lipid droplet (Figures 4D,E). These results suggest that DYRK1B affects lipid formation and is related to obesity.

We investigated a new class of DYRK1B inhibitors and developed KS-40070 (Figure 2A) (Park et al., 2021). KS-40070 exhibited strong inhibitory efficacy against DYRK1B, inhibiting the activity of DYRK1B by more than 95% at 5 μM, with an IC₅₀ value of 18 nM (Figure 2B). To verify the selectivity of KS-40070 for DYRK, LanthaScreen Kinase Binding Assay was performed. The IC₅₀ value of human DYRK2 was greater than 10 μM (data not shown). In particular, KS-40070 showed strong specific inhibition against DYRK1B and relatively high selectivity against DYRK2.

The dose-dependent effects of KS-40070 on lipid droplet accumulation in 3T3-L1 cells were investigated by staining lipid droplets with BODIPY and nuclei with DAPI. The differentiation of preadipocytes into adipocytes was confirmed, and lipid droplet accumulation was effectively inhibited at 10 μM of KS-40070 (Figures 3A,B). In addition, we analyzed the effects of KS-40070 on lipid accumulation in 3D human ADMSCs by staining lipids with AdipoRed and dsDNA with Hoechst 33,342 (Figure 3C).

Western blot was performed to determine whether KS-40070 affects the expression of DYRK1B and other adipogenesis-related markers. KS-40070 reduced the expression of DYRK1B and other adipogenesis transcription factors, such as PPARγ and C/EBPα, in a dose-dependent manner and adipocyte-specific markers were also effectively decreased at 5 μM (Figure 3D).

Our experimental data suggest that KS-40070 regulates FOXO1A via the inactivation of Akt (Figure 3E). Treatment with KS-40070 reduced the expression of phospho-Akt. Akt regulates adipogenesis through the inactivation and phosphorylation of several downstream targets. KS-40070 also reduced the expression of mTOR, phospho-GSK-3β, and phospho-FOXO1A, which are downstream substrates of Akt, in 3T3-L1 adipocytes. FOXO1A acts as an inhibitor in the adipogenic differentiation process. Akt-dependent FOXO1A phosphorylation promotes the translocation of FOXO1A from the nucleus to the cytoplasm, thereby suppressing its adipogenesis-inhibiting effect (Gross et al., 2008; Chen et al., 2018). KS-40070 inhibits the nuclear-cytoplasmic translocation of FOXO1A through the inactivation of Akt. Consequently, FOXO1A remains localized to the nucleus, where it suppresses adipogenic transcription factors such as PPARγ and C/EBPα.

To verify whether the anti-obesity effect of KS-40070 is the result of specifically inhibiting DYRK1B, a siRNA-based knock-down assay was performed to specifically inhibit DYKR1B expression (Figure 4). DYRK1B siRNA selectively inhibited DYRK1B protein expression, and Perilipin expression level, as an adipogenic differentiation marker, was also extremely reduced in DYRK1B targeted siRNA-transfected cells (Figures 4A–C). The effect of DYRK1B inhibition on anti-adipogenesis was also demonstrated by lipid vesicle staining using BODIPY (Figures 4D,E). In the case of negative siRNA cells, lipid vesicles increased by induction of adipogenic differentiation, but in the case of DYRK1B specific siRNA-transfected cells, lipid vesicles were hardly produced even when adipogenic induction occurred. This anti-adipogenic effect was confirmed in the same way in cells treated with KS-40070, and these results show that the decrease in DYRK1B expression is related to anti-adipogenesis, and that the anti-adipogenic effect of KS-40070 is also a phenomenon that appears by targeting DYRK1B.

Pharmacokinetic studies were performed to inform the design of in vivo DIO studies (data not shown). KS-40070 was absorbed at a T_max (time for C_max) of 0.5 ± 0 h and C_max (maximum plasma concentration) of 0.069 ± 0.026 h. T_1/2 (terminal half-life) was approximately 7 h. The AUC ∞ value (area under the plasma concentration time curve from time to zero) was 0.826 ± 0.082 μg h/mL in the intravenous administration group, and 0.0123 ± 0.05 μg h/mL in the oral administration group; the resulting F_t (oral bioavailability) was 12.4%. As such, DIO mice received twice-daily oral administrations of the experimental or control compounds in the DIO study.

In DIO mice study, the drugs were administered for 9 days twice daily and body weight and food intake were measured twice weekly. KS-40070 groups were 50 mg/kg and 100 mg/kg and 10 mg/kg rosiglitazone and 20 mg/kg Xenical were treated as positive controls. Body weights in the rosiglitazone and Xenical group showed no change. However, body weight gain of the KS-40070 treated groups were reduced in a dose-dependent manner (Figures 5A,B).

Cholesterol, triglycerides, HDL and LDL cholesterol level were measured to determine whether KS-40070 could improve plasma lipid profiles. In the vehicle group of DIO mice, all of these values were significantly increased. The rosiglitazone treated group showed slightly decreased plasma lipid profiles in DIO mice; the Xenical-treated group showed no changes. The KS-40070 treated groups, however, demonstrated significantly improved plasma lipid profiles such as cholesterol, triglycerides and LDL-cholesterol (Figures 5C–F).

To investigate whether KS-40070 decreased fat weight in DIO mice, subcutaneous and visceral fat weights were measured after mice were sacrificed. In 100 mg/kg KS-40070 treated group, subcutaneous fat weights were significantly decreased compared to vehicle group (Figures 6A,B). These results coupled with body weight change results suggest that reduction of fat weight is associated with body weight loss. At this time, subcutaneous fat weight was decreased more than visceral fat weight in KS-40070 treated group. Therefore, we performed a histological examination of the subcutaneous fat to evaluate the change in the size of adipocytes. Reference compound, Xenical-treated group had no effect, but the KS-40070 treated group reduced the size of subcutaneous fat cells (Figure 6C).

We assessed whether KS-40070 suppression of DYRK1B restore glucose tolerance and insulin resistance in DIO mice after oral treatment with 50 or 100 mg/kg, twice daily for 9 days. In the oral glucose tolerance test, plasma glucose levels as determined by the AUC of the glucose concentration curve were significantly reduced after administration of 10 mg/kg rosiglitazone and 100 mg/kg KS-40070 compared to vehicle (Figures 6D,E). Using the same model, we also measured KS-40070 effects on insulin sensitivity with an insulin tolerance test (Figures 6F,G). Plasma fasting insulin concentrations and HOMA-IR index of KS-40070 treated mice were significantly lower than those of the DIO-vehicle mice (Figures 6H,I).

To analyze the potential hepatotoxicity of KS-40070, the weight of the mouse liver and H&E staining were performed after autopsy. There were no differences in organ appearance and AST/ALT ratios among the groups (data not shown). H&E staining results showed that lipid accumulation was increased in the DIO-vehicle group (arrows), and hepatic lipid accumulation was reduced by KS-40070 administration (Figure 7A). Consistently, Western blot analysis was employed to investigate the impact of KS-40070 on the expression of fatty liver-related markers. The results demonstrated a dose-dependent decrease in the expression of fatty liver-related markers, including FAS, TNFα, TIMP1, and TGFβ, following treatment with KS-40070 (Figure 7B). These markers are known to be involved in various aspects of fatty liver pathogenesis, including lipid metabolism, inflammation, and fibrosis (Guo et al., 2022; Kotlyarov, 2023; Vachliotis and Polyzos, 2023). Thus, these results suggest that KS-40070 shows promise as a potential therapeutic agent for the treatment of non-alcoholic fatty liver disease (NAFLD) and related metabolic disorders without inducing hepatotoxic effects.

In order to predict the binding modes of KS-40070 for DYRK1B, we performed molecular docking studies with DYRK1B homology model (Figure 8). The pyrazolopyrimidine ring binds to the hinge region, which forms two hydrogen bonds with Glu191 backbone CO and Leu193 backbone NH, and hydrophobic interactions with Ala138, Val174 and Leu246. Phenylacetamide group is oriented towards solvent exposed area by forming hydrophobic interaction with Ile117. 4-methoxyphenyl ring forms hydrophobic interactions with Val258 and Phe122.