The Cell Proliferation Kit II (XTT Chemicon) was used to evaluate cell viability, as previously described (Deiana et al., 2019, 6). Six replicates in three independent experiments were tested.

Human mesenchymal stem cells (hMSC, PromoCell, Heidelberg, Germany) were used by culturing at a density of 5 × 10⁴ cells with the mesenchymal stem cell growth medium (PromoCell) or osteogenic differentiation medium (PromoCell, Heidelberg, Germany) and incubated at 37°C in a humidified atmosphere with 5% CO₂. Differentiating cells were then used for further analyses. Human dermal fibroblast cultures were established from explanted skin biopsies taken from patient P1 [mutation: c.897T>G->p (Tyr299*), male, 8 years old], P2 [c.1019del-> p (Ser340*), female, 10 years old], and a healthy age-matched control with appropriate consent, as previously reported (Venturi et al., 2012). The cells were cultured with high-glucose DMEM (ECB7501L, EuroClone, Milano, Italy) supplemented with 10% FBS (10270-106, Gibco, Life Technologies Limited, Paisley, United Kingdom), 2 mM l-glutamine (5-10K00-H, BioConcept AG, Paradiesrain, Allschwil, Switzerland), 100 U/ml penicillin, and 100 μg/ml streptomycin (penicillin–streptomycin; ECB3001D, EuroClone, Milano, Italy), as previously described (Venturi et al., 2012).

The zebrafish experiments were performed at the CIRSAL (Interdepartmental Centre of Experimental Research Service) of the University of Verona, Italy, under ethical authorization n. 662/2019-PR of 16/09/2019.

The embryos were obtained with nacre adults (ZFIN database ID: ZDB-ALT-990423-22), according to standard procedures (Kimmel et al., 1995; Whitlock and Westerfield, 2000). Preliminary tests at 2.5, 5, and 10 µM had shown no effect on the modulation of osteogenic gene expression. Thus, the embryos were grown at 33°C in water containing 15 µM fisetin from 2 days post-fertilization (dpf). The zebrafish embryos were supplemented with fisetin up to 1 week (experimental endpoint, 9 dpf). At the end of the treatment, the zebrafish embryos were euthanized and collected for performing molecular analyses were performed, as described later. To evaluate bone formation, we performed calcein staining, as previously reported (Du et al., 2001). Then, imaging was performed by using a Leica M205FA fluorescence microscope (Leica Microsystems, Wetzlar, Germany). The stained areas were quantified with ImageJ software, as previously reported (Du et al., 2001).

Adult zebrafish (15–20 months) were grown in water in the presence of 15 µM fisetin for 14 days (for 1 week of supplementation, no effect on the modulation of osteogenic gene expression was observed). At the experimental endpoint, the adult zebrafish were euthanized and collected for the staining procedures and molecular analyses, as reported later.

Pellets from cells or zebrafish samples were collected and stored at −80°C until use. The samples were processed for RNA extraction by using an “RNeasy^® protect mini kit” (Qiagen, Hilden, Germany), as previously reported (Cecconi et al., 2019). The RNA samples were quantified using a Qubit™ RNA HS assay kit” (Invitrogen, Carlsbad, United States) and a Qubit 3 Fluorometer (Invitrogen by Thermo Fisher Scientific, REF Q3321). RNA (1 mg) was reverse-transcribed using the first-strand cDNA synthesis kit (GE Healthcare, Little Chalfont), as per the manufacturer’s instruction, and a SimpliAmp Thermal Cycler (Applied Biosystems by Thermo Fisher Scientific, REF A24812). RNA and cDNA samples were stored at −80°C until use.

To investigate gene expression modulation, we performed real-time PCR analyses as reported previously (Dalle Carbonare et al., 2019).

Briefly, predesigned, gene-specific primers and probe sets for each gene [RUNX2, hs00231692_m1; OSTERIX (SP7), hs00541729_m1; collagen, type i, alpha 2 (COL1A2), hs01028956_m1; osteonectin (SPARC), hs00234160_m1; OSTEOPONTIN (SPP1), hs00167093_m1, B2M, hs999999_m1 (housekeeping); and GAPDH, 0802021 (housekeeping)] were obtained from assay-on-demand gene expression products (Thermo Fisher Corporation, Waltham, MA, United States). In addition, the following custom primer sets (Invitrogen, Carlsbad, CA, United States) were also used: runx2a (fw GAC​GGT​GGT​GAC​GGT​AAT​GG, rv TGC​GGT​GGG​TTC​GTG​AAT​A), runx2b (fw CGG​CTC​CTA​CCA​GTT​CTC​CA, rv CCA​TCT​CCC​TCC​ACT​CCT​CC), rank (fw GCA​CGG​TTA​TTG​TTG​TTA, rv TAT​TCA​GAG​GTG​GTG​TTA​T), and housekeeping gene actb1 (fw CCC​AAA​GCC​AAC​AGA​GAG​AA, rv ACC​AGA​AGC​GTA​CAG​AGA​GA). Ct values for each reaction were determined using TaqMan SDS analysis software (Applied Biosystems; Foster City, California, United States) as reported previously. To calculate relative gene expression levels between different samples, we performed the analyses by using the 2−ΔΔCT method as previously reported (Dalle Carbonare et al., 2019).

We measured ROS levels by staining cells with the DCFDA cellular ROS detection assay kit (Abcam), as previously reported (Deiana et al., 2019). After the staining, the cells were analyzed by measuring the produced fluorescence (ex/em = 485/535 nm) in the endpoint at the VictorX4 instrument (Perkin Elmer, Milan, Italy) according to the manufacturer’s protocol.

RUNX2 protein levels were separated by using SDS-PAGE and investigated by using Western blot analyses as previously reported (Brugnara and De Franceschi, 1993). Briefly, proteins were extracted by using a RIPA buffer (Thermo Fisher Scientific, Waltham, MA, United States), and the proteins were quantified by BCA assay (Thermo Fisher Scientific, Waltham, MA, United States). The proteins were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and then were transferred onto polyvinylidene difluoride (PVDF) membranes (Thermo Fisher Scientific, Waltham, MA, United States). The PVDF membranes were then probed with the primary antibody for RUNX2 (Cell Signaling, 8486) and β-actin (BA3R) (Thermo Scientific), and secondary antibodies anti-mouse (Cell Signaling, 7076) and anti-rabbit (Cell Signaling 7074). Signals were detected using a chemiluminescence reagent (ECL, Millipore, Burlington, MA, United States), as previously reported (de Franceschi et al., 2004). Images were acquired by an LAS4000 Digital Image Scanning System (GE Healthcare, Little Chalfont, United Kingdom). The densitometric analysis was performed by ImageQuant software (GE Healthcare, Little Chalfont, United Kingdom). Protein optical density was normalized to β-actin.

Calcium deposition in differentiating osteogenic cells was evaluated by Alizarin Red staining, as previously described (Dalle Carbonare et al., 2019). Briefly, after 21 days of culture in the differentiating medium, the cells were fixed with 70% ethanol and washed with water. The cells were then stained with 40 mM Alizarin Red S for 5 min at pH 4.1 and rinsed for 15 min. The stained area was quantified with ImageJ software (NIH, Bethesda, MD, United States), as previously reported (Dalle Carbonare et al., 2019). Six independent experiments were performed.

Calcein staining in zebrafish larvae was performed, as previously reported (Du, Frenkel et al., 2001). The imaging was performed using a Leica M205FA fluorescence microscope (Leica Microsystems, Wetzlar, Germany). The stained area was quantified by using ImageJ software, as previously reported (Du et al., 2001).

Bone and cartilage staining in adult zebrafish was performed as described in Sakata-Haga et al. (2018). Briefly, the euthanized fish were immersed into the fixative solution (5% formalin, 5% Triton X-100, 1% potassium hydroxide (KOH) and gently rocked for 48 h  at room temperature (RT). Then, we proceeded either to the cartilage staining step for double staining or directly to the bone staining step for bone-only staining. For cartilage staining, the specimens were immersed into C-Staining Solution (70% ethanol, 20% acetate, 0.015–0.02% Alcian Blue)” overnight at 20°C, then washed in C-Staining medium (containing 1% Triton X-100) overnight at 20°C and washed with 50–70% ethanol. Then the protocol proceeded to the bone staining step. The specimens were immersed in B-Staining medium (20% ethylene glycol and 1% KOH) and then in “B-Staining solution (0.05% Alizarin Red S, 20% ethylene glycol, 1% KOH) overnight at 20°C. The specimens were then washed with clearing solution (20% Tween 20, 1% KOH) while gentle rocking for 12 h, and the stocking was performed in glycerol 100%. The stained area was quantified by using ImageJ software, as previously reported (Du et al., 2001).

Tests were carried out in order to evaluate the stability of fisetin in a complete culture medium (DMEM with 10% FBS and 1% PEN/STREP) at different experimental times (time = 0 vs. 8 h). Starting from a 10 mM stock solution of fisetin in DMSO, we obtained a 100 uM solution, which was analyzed by HPLC at time T₀ and after 6 h. In particular, the measurements in HPLC were performed at time T₀ and after 30 min, 1, 2, 4, and 6 h. The solution was kept at 37°C and away from direct light to simulate experimental conditions.

PLGA [poly(DL-lactide-co-glycolide), 50:50 lactide-to-glycolide ratio, CAS 26780-50-7], PVA [poly(vinyl alcohol), CAS 9002-89-5], and acetone (≥99% purity, 1.00013) were purchased from Merck. Fluorescein isothiocyanate (FITC, CAS 27072), cellulose membrane dialysis tubing (CASD9777-100), dimethyl sulfoxide (DMSO, ≥ 99% purity D-5879) were purchased from Sigma-Aldrich. Fisetin was purchased from Santa Cruz Biotechnology (SC-276440).

The protocol used for the production of PLGA nanoparticles loading fisetin [PLGA (Fis)] is based on a single emulsion evaporation method, under sterile conditions at 20°C (Makadia and Siegel, 2011; Gaglio et al., 2019). A measure of 20 mg of the polymer and 4 mM (1.14 mg) of fisetin were co-dissolved in 1 ml of DMSO 100%; the obtained organic phase was added dropwise under stirring (2,000 RPM) to 10 ml of 0.5% polyvinyl alcohol (PVA) aqueous solution and left overnight to evaporate the organic phase. Afterward, the preparation was pelleted at 4°C 11,000 rpm for 20 min (Eppendorf Centrifuge 5804R), and the nanoparticles were collected and washed twice with 10 ml of Milli-Q water. Finally, the purified nanoparticles were resuspended in 1 ml of phosphate-buffered saline (PBS) solution pH 7.4 (or NaCl 0.9%) for the subsequent analysis and storage at 4°C, otherwise freeze-dried. Empty nanoparticles were prepared with the same protocol avoiding the addition of fisetin to the reaction. Furthermore, to perform an internalization study and visualize nanoparticles inside cells, PLGA NPs co-delivering fisetin and fluorescein isothiocyanate (FITC) [PLGA (Fis and Fitc)] were prepared by dissolving 20 mg of PLGA in a mixture of 640 ul of acetone and 360 ul of fisetin/FITC DMSO solution; the molar ratio between fisetin and FITC is 1:1. The other steps were the same as described previously.

To assess the capacity of the nanoparticles to retain entrapped fisetin over time, a first release study was performed in a total volume of 1 ml, at different temperatures (4 and 37°C) and different media (water, PBS NaCl 0.9%, and citric acid pH 5). Dialysis was used to carry out the in vitro drug release studies increasing the final volume. PLGA (Fis) NPs were introduced into the dialysis bag (14,000 da molecular weight cutoff, Sigma Aldrich, D9777-100FT) and placed in 100 ml of PBS pH 7.4 containing 0.1%v/v Tween 80 as the release media and stirred at 100 rpm. Samples were collected at different time intervals and replaced with an equal volume of media to maintain the sink condition. The released fisetin was quantified using a calibration curve obtained by UV–visible spectroscopy at 360 nm (Kadari et al., 2017).

A compartmental fluidic device (commercialized as MIVO React4life S.r.l., IT) was used to perform in vitro drug efficacy tests. The 3D fluidic model was performed as follows: 1) 24-well size inserts containing human intestinal tissue (EpiIntestinal by Mattek) were placed and cultured within the device, forming two fluidically independent chambers: the donor and the receiver; 2) both chambers were filled with the culture medium; and 3) the receiver chamber was connected to the peristaltic pump to form a closed-loop fluidic circuit containing 3.8 ml medium circulating at a rate of 0.3 cm/s, to simulate the capillary flow rate. Therefore, a medium containing PLGA (fis) was added to the donor chamber.

Results were expressed as mean ± SD. The statistical analysis was assessed by Student’s paired t-test comparing each treatment to the control. Differences were considered positive when p < 0.05. For in vitro data, analyses were applied to experiments carried out at least three times. We used SPSS for Windows, version 22.0 (SPSS Inc., Chicago, IL, United States), to analyze the data.

In order to evaluate the toxic effect of fisetin, we assayed the viability of mesenchymal stem cells treated with fisetin for 24 h at concentrations ranging from 0 to 5 µM. As shown in Figure 1, fisetin did not affect cell viability at any tested concentration.

As previously reported (Molagoda et al., 2021), we observed that fisetin promotes osteogenic differentiation (Supplementary Figure S1); the upregulation of osteogenic genes (BGLAP, SPP1, SPARC, and COL1A1) was much more evident when fisetin was supplemented at a concentration of 2.5 µM in MSCs (Figure 2A). RUNX2 expression in treated cells exerted the same modulation generally observed during osteogenic differentiation. In fact, RUNX2 upregulation was observed until after 7 days (Figure 2B) (Ardeshirylajimi et al., 2014). In order to assay fisetin’s effects on oxidative stress, which affects osteogenic modulation (Atashi et al., 2015), we measured ROS levels during differentiation. As shown in Figure 2, we observed reduced ROS levels in fisetin-treated cells during the late phase of differentiation (C).

We also investigated the effects of fisetin in the late phase of osteogenic differentiation, and we cultured MSCs in an osteogenic medium with or without fisetin for 21 days. The effects of 2.5 µM fisetin supplementation were also observed during the late phase of osteogenic maturation of MSCs. As shown in Figure 3, fisetin supplementation increased the expression levels of RUNX2 downstream genes related to the late osteogenic phase (A) and to mineral deposition (B).

Fisetin effects were then compared with those of other supplemental molecules known to improve bone tissue quality. In particular, we compared the effects of fisetin, ascorbic acid, and pigment epithelium-derived factor (PEDF) supplementation, respectively, on MSC osteogenic differentiation. As shown in Figure 4, fisetin induced the upregulation of all tested osteogenic genes after 7 (A) and 14 (B) days of differentiation. Gene modulation induced by fisetin treatment was similar to that observed by treating cells with PEDF and AsA. COL1A1 and SPARC gene expressions induced by fisetin after 14 days of differentiation were more pronounced than PEDF and AsA treatment.

To evaluate the effects of fisetin in vivo, we used the Danio rerio (zebrafish) model. In particular, zebrafish larvae were grown in water at 33°C after 2 days post-fertilization (dpf) in the presence/absence of fisetin. Preliminary tests at 2.5, 5 and 10 µM showed no effect on the expression of osteogenic genes compared to untreated zebrafish. However, we observed increased osteogenic gene expression in zebrafish treated with 15 µM fisetin. Thereafter, we used 15 µM concentration in all experiments.

After 7 (9 dpf) days of treatment, zebrafish were euthanized and collected to perform gene expression analyses or staining. As shown in Figure 5A, the upregulation of osteogenic genes in fisetin-supplemented zebrafish larvae could be observed.

In order to record the direct effects of fisetin on bone mineralization, we quantified the vertebrae area evaluated by calcein staining. We observed calcein fluorescence density under microscopic inspection, and we analyzed fluorescence levels by using digital methods. Calcein staining showed an increased vertebrae area in larvae after 7 days of fisetin treatment, compared to untreated larvae (Figure 5B).

Interestingly, the upregulation of osteogenic genes was also observed in adult zebrafish (15–20 months) supplemented with fisetin for 14 days (Figure 5C). In addition, we also observed increased bone mineralization, evaluated by Alizarin Red staining (ARS) in adult zebrafish treated with fisetin for 14 days. In fact, in adult zebrafish, more intense ARS and a higher percentage of positive ARS areas were observed in fisetin-treated zebrafish than in untreated zebrafish (Figure 5D).

Considering the positive effects of fisetin on osteogenic differentiation, we speculated about its possible beneficial effects in the context of skeletal pathologies. Therefore, we investigated the effects of fisetin in ex vivo models of cleidocranial dysplasia (CCD, OMIM#119600), a dominantly inherited skeletal disease. In particular, we obtained fibroblast-like cells from two unrelated pediatric patients affected by CCD carrying mutations in exon 7 of one RUNX2 allele, as we previously reported (Dalle Carbonare et al., 2021). In our previous study, we reported that RUNX2 gene expression levels in both patients were similar to those in the normal control, whereas SPARC gene expression was lower in both patients than in the control (Dalle Carbonare et al., 2021). However, when the cells were cultured in the presence of fisetin, we observed SPARC upregulation in patient’s cells (compared to untreated patients’ cells), whereas RUNX2 levels remained unchanged, suggesting fisetin’s ability to restore osteogenic maturation in this ex vivo skeletal disorder model (Figure 6).

With the aim of providing an efficient fisetin supplementation for the improvement of bone quality, we tested the characteristics of this molecule. In order to verify fisetin’s stability, we analyzed its concentration in a culture medium at different time points. As shown in Figure 7, we observed fisetin degradation upon the observation time lapse. In particular, fisetin was almost completely degraded after 6 h. Consequently, we generated fisetin containing PLGA particles [PLGA (Fis)] to prevent its degradation in order to make the supplementation available for human use. We compared our nanoformulation with data of fisetin encapsulation into PLGA previously published by Sechi et al., Kadari et al., and Liu et al. (Table 1) (Sechi et al., 2016; Kadari et al., 2017; Liu et al., 2021).

As reported in Table 1, the nanoformulation that we prepared has a more negative ζ-potential than that reported by Kadari et al., conferring it good colloidal stability. Moreover, we achieved a higher drug loading (DL) value, corresponding to 23.51%, that is, 6 to 8 times greater than the result reported in Sechi et al. PLGA (FIS) nanoparticles show a lower particle size than all the other herein reported, with an exception of those prepared by Kadari et al. but in every case showing a better monodispersivity, as it is evident from the lower PDI value.

To further investigate the size distribution of nanoparticles, nanotracking and AFM analysis were also performed. The nanoparticles shape is quite spherical (Supplementary Figure S2), and as it is evident from Table 2, data from the three different methods are in agreement and span in the same order of magnitude, even if the techniques are based on different physicochemical properties.

Furthermore, freeze-drying does not negatively affect the samples size after resuspension and the co-encapsulation of fisetin, and FITC does not affect the size of the nanoparticles noticeably.

We also tested the fisetin release rate of PLGA nanoparticles at different temperatures (4 and 37°C) and in citric acid pH 5 in a final 1 ml volume. As it is evident from Figure 8A, after an initial burst loss, the release kinetics dropped dramatically.

The temperature increase, as expected, leads to fast kinetics, while an acidic pH seems to slow down the loss. Then, an in vitro release study on the volume subsequently used for tests on intestinal epithelial tissue was carried out (Figure 8B). PLGA (Fis) NPs exerted a sustained release trend that fits with a second-degree polynomial function.

A microfluidic system containing human intestinal epithelial tissue was created to assess the ability of PLGA (Fis) to cross the intestine (Supplementary Figure S3). As expected PLGA (Fis) could cross the intestinal epithelium since the nanoparticles with negative zeta potential are taken up by Peyer’s patches and are translocated into the blood circulation (Joshi et al., 2014). Our data are shown in Table 3: after 5-h incubation, less than 5% of the initial concentration crossed the epithelium, while a value of 30% was reached within 16 h.

The fisetin amount has been calculated using a fluorescence calibration curve (Supplementary Figure S4). Moreover, PLGA (Fis) recovered after the experiment in the volume filtered by the epithelium showed a slight increase in the size, with values of approximately 190 nm. This increase in the size of PLGA (Fis) might be due to the well-known protein corona phenomenon or to some interactions occurring during the epithelial tissue crossing.

Subsequently, in order to test the ability of PLGA (Fis) to induce osteogenic differentiation, we cultured MSCs with FITC-PLGA (Fis). In particular, we cultured MSCs in an osteogenic medium with or without FITC-PLGA (Fis) for 7 days. Fresh FITC-PLGA (Fis)s were added to every medium change (every 2 or 3 days). Four hours after the addition of the complete medium, PLGA (Fis)s are visible in intercellular spaces (Figure 9A). After 6 h of treatment, fisetin (in green) was completely incorporated into the cells (Figure 9B).

After 7 days of osteogenic differentiation, cells were collected to perform gene expression analyses. As shown in Figure 9C, we observed that PLGA (Fis) increased the expression of the osteogenic transcription factors RUNX2 and SP7 compared to free fisetin-supplemented cells. Accordingly, COL1A1, a marker of osteogenic maturation, was higher in cells treated with PLGA (Fis) than in control (Figure 9D).