Effects of Nanoparticle Properties on Kartogenin Delivery
and Interactions with Mesenchymal Stem Cells

1School of Engineering, Center for Biomedical Engineering, Institute for Molecular and Nanoscale Innovation, Brown University, 184 Hope Street, Providence, RI, USA; and 2Department of Chemistry, Brown University, 324 Brook Street,
Providence, RI, USA
(Received 11 August 2019; accepted 2 December 2019)

Associate Editor Debra T. Auguste oversaw the review of this article.

Abstract—Clinical trials with mesenchymal stem cells

(MSCs) have demonstrated potential to treat osteoarthritis, a debilitating disease that affects millions. However, these therapies are often less effective due to heterogeneous MSC differentiation. Kartogenin (KGN), a synthetic small mole-
OA Osteoarthritis
ECM Extracellular matrix
MSCs Mesenchymal stem cells
hMSCs Human mesenchymal stem cells

cule that induces chondrogenesis, has recently been explored to decrease this heterogeneity. KGN has been encapsulated in nanoparticles due to its hydrophobicity. To explore the
Core-binding factor subunit b

effect of nanoparticle properties on KGN and MSC inter- actions, here we fabricated three nanoparticle formulations
RUNX Runt-related transcription factor
TGF-b1 Transforming growth factor beta 1

that vary in hydrophobicity, size, and surface charge using nanoprecipitation: KGN-loaded poly(lactic acid-co-glycolic acid) (PLGA) nanoparticles (hydrophobic surface, negative charge, ~ 167 nm), PLGA–poly(ethylene glycol) (PEG) nanoparticles (hydrophilic surface, positive charge, ~ 297 nm), and PLGA–PEG–hyaluronic acid (HA) nanopar- ticles (hydrophilic surface, negative charge, ~ 507 nm). We
Half maximal effective concentration Molecular weight
Poly(lactic-co-glycolic acid)

PLGA–poly(ethylene glycol)

observed differences in KGN loading, release, and suspen- sion stability, with the PLGA particles exhibiting ~ 50% drug loading and PLGA–PEG–HA particles releasing the most KGN. All nanoparticles were found to interact with MSCs with evidence of increased uptake in PLGA–PEG and
PEG–HA PLGA–PEG–hyaluronic acid GAG Glycosoaminoglycan
sGAG Sulfated glycosaminoglycan PEG-bis-

PLGA–PEG–HA compared with surface association of PLGA particles. Over short times (~ 7 days), MSCs incu- bated with all KGN-loaded formulations exhibited a similar increase in sulfated glycosaminoglycans, characteristic of chondrogenic differentiation, compared with non-KGN loaded formulations.
PEG-bis-amine Poly(vinyl alcohol)

N-hydroxysulfosuccinimide N,N¢-dicyclohexylcarbodiimide

EDC N-(3-dimethylaminopropyl)-N¢-ethylcar-

Keywords—Polymer nanoparticles, Poly(lactic acid-co-gly- colic acid), Poly(ethylene glycol), Hyaluronic acid, Karto- genin, Drug release, Chondrogenic differentiation.

bodiimide hydrochloride Dichloromethane Acetonitrile

MES 2-(N-morpholino)ethanesulfonic acid

Address correspondence to Anita Shukla, School of Engineering, Center for Biomedical Engineering, Institute for Molecular and Nanoscale Innovation, Brown University, 184 Hope Street, Provi-
Fluorescein isothiocyanate Hexadecyltrimethylammonium bromide Phosphate buffered saline

dence, RI, USA. Electronic mail: [email protected]
Bethany Almeida and Yingying Wang are co-first authors and contributed equally to this work.
DMSO Dimethyl sulfoxide
EDTA Ethylenediaminetetraacetic acid disodium

ti 2019 Biomedical Engineering Society


Hydrochloric acid Trifluoroacetic acid

tremely hydrophobic, requiring solubilization in cyto- toxic organic solvents.3 By delivering KGN through a nanoparticle formulation, cell death resulting from

CCK-8 Cell Counting Kit-8
DMEM Dulbecco’s modified Eagle’s medium
ITS+ Insulin/transferrin/selenium
HPLC High performance liquid chromatography
exposure to these organic solvents can be avoided and delivery can potentially be localized to cells of interest.
Polymeric nanoparticles with hydrophobic regions can encapsulate hydrophobic molecules, improving

Room temperature
their solubility and stability.34 Various KGN nanoparticle formulations have been reported recently

1H-NMR Proton nuclear magnetic resonance (Table 1).5,12,15–18,22,30,37,39 We highlighted the main

DLS Dynamic light scattering
PDI Polydispersity index
TEM Transmission electron microscopy
EE% Encapsulation efficiency
properties credited with affecting KGN release in these studies, noting that the properties affecting release are largely varied. Although promising in vivo data has been obtained using some of these formula-

DL Drug loading tions,5,15–17,30 a detailed analysis of the effect of

ANOVA Analysis of variance


Osteoarthritis (OA), a degenerative joint disorder characterized by the degradation of articular cartilage, a decrease in joint function, and severe pain,1,4 affects 10–15% of adults over the age of 60.35 Mesenchymal stem cells (MSCs) have recently garnered significant interest for OA therapy. Intra-articular injection of MSCs into the osteoarthritic joint has been shown to decrease OA-related pain,20 promote collagen regen-
nanoparticle physical and chemical properties on fac- tors such as KGN release has not yet been reported. Thorough characterization can enable the design of nanoparticles with desired KGN loading and release properties, as well as effects on MSC uptake, viability, and differentiation, potentially leading to improved therapeutic outcomes.
Here, we report three KGN-loaded nanoparticles formulations fabricated using nanoprecipitation based on poly(lactic-co-glycolic acid) (PLGA) with varied surface modifications. PLGA was selected as a nanoparticle polymer commonly used in delivery of
hydrophobic drugs, including KGN. Specifi-

eration in full-thickness defects,
and inhibit OA
cally, we developed KGN-loaded PLGA, PLGA–

progression through host tissue engraftment and re-
poly(ethylene glycol) (PLGA–PEG), and PLGA–

lease of bioactive molecules.1,6 However, there are
PEG–hyaluronic acid (PLGA–PEG–HA) nanoparti-

limitations to MSC use. For example, current MSC therapies for OA often result in the formation of fibrocartilage as opposed to hyaline cartilage.13,20,31 MSC plasticity can also result in heterogeneous pop- ulations of cells, which is not always therapeutically desirable.27
Kartogenin (KGN), a recently discovered synthetic small molecule, has the potential to improve chon- drogenic differentiation of MSCs for OA treatments. KGN has both a chondrogenic and a chondroprotec- tive effect on MSCs,14 driving differentiation to chon- drocytes and preventing dedifferentiation, which is often observed in mature chondrocytes in vitro.14 It acts by binding intracellularly to filamin A and freeing core-binding factor subunit b (CBFb), resulting in CBFb nuclear localization where it can regulate runt- related transcription factor (RUNX) associated tran- scription and induce chondrogenesis.14 KGN over- comes many of the limitations of transforming growth factor beta 1 (TGF-b1), which is one of the most common in vitro chondrogenic supplements, including a longer half-life, less osteophyte formation, and less downstream osteogenesis.21 However, KGN is ex-
cles. We sought to vary surface properties from hydrophobic (i.e., PLGA) to hydrophilic (i.e., PLGA– PEG and PLGA–PEG–HA) and surface charge from negative (i.e., PLGA–PEG–HA and PLGA) to positive (i.e., PLGA–PEG). HA was selected as a potential targeting moiety for human (hMSCs) through the
CD44 cell surface glycoprotein. We demonstrated that changing nanoparticle properties had distinct ef- fects on nanoparticle physicochemical properties, sus- pension stability, and KGN release. Interestingly, all KGN-loaded nanoparticle formulations appeared to exhibit similar overall levels of hMSC interaction, with varying uptake vs. cell-surface association. Finally, we observed that over a short incubation time of 7 days, all KGN-loaded particles began producing proteins indicative of chondrogenic differentiation.


KGN was purchased from Abcam (Cambridge, MA). HA sodium salt (molecular weight (MW)

39 kDa) was obtained from Lifecore Biomedical (Chaska, MN). PLGA (1:1 D, L-lactic to glycolic acid, MW 15.1 kDa), PEG-bis-amine (PEG-bis-NH2 MW 2 kDa), poly(vinyl alcohol) (PVA, MW 89–98 kDa), N-hydroxysulfosuccinimide (sulfo-NHS), N,N¢-dicy- clohexylcarbodiimide (DCC), N-(3-dimethylamino- propyl)-N¢-ethylcarbodiimide hydrochloride (EDC), diethyl ether, methanol, dichloromethane (DCM), acetonitrile (ACN), 2-(N-morpholino)ethanesulfonic acid (MES), fluorescein isothiocyanate (FITC), hex- adecyltrimethylammonium bromide (CTAB), phos- phate buffered saline (PBS), dimethyl sulfoxide (DMSO), dexamethasone, L-proline, ascorbate-2- phosphate, alcian blue, sodium acetate, ethylenedi- aminetetraacetic acid disodium salt (EDTA), L-cys- teine hydrochloric acid (HCl), trifluoroacetic acid (TFA), papain, disodium phosphate, and a Kaiser test kit were obtained from Sigma-Aldrich (St. Louis, MO). Deuterated chloroform was purchased from Cambridge Isotope Laboratories, Inc. (Andover, MA). UranyLess stain solution was purchased from Electron Microscopy Sciences (Hatfield, PA). Cell Counting Kit-8 (CCK-8) was purchased from Dojindo Molecu- lar Technologies (Tokyo, Japan). High and low-glu- cose Dulbecco’s modified Eagle’s medium (DMEM) and fetal bovine serum (FBS) were purchased from Gibco-BRL (Grand Island, NY). Penicillin–strepto- mycin and trypsin–EDTA were purchased by Caisson Labs (Smithfield, UT). L-Glutamine, sodium pyruvate, TGF-b1 recombinant human protein, Quant-iTTM PicoGreenTM dsDNA assay kit and insulin/transferrin/
selenium (ITS+) premix were purchased from Fisher Scientific (Hampton, NH). A Blyscan sulfated gly- cosaminoglycan (sGAG) assay was purchased from Biocolor Ltd (Carrickfergus, UK). Bone mar- row-derived hMSCs (33 year old female) were pur- chased from Lonza (Hopkinton, MA). All chemicals were of analytical reagent quality or high performance liquid chromatography (HPLC) grade. Ultrapure wa- ter (18.2 MX cm Milli-Q, Millipore Sigma, Billerica, MA) was utilized in all experiments requiring water.

Fabrication of PLGA, PLGA–PEG, and PLGA–PEG–HA Nanoparticles
The PLGA–PEG copolymer was synthesized as previously reported.38 All nanoparticles were prepared using nanoprecipitation. PLGA or PLGA–PEG (50 mg) was dissolved in ACN (10 mL) and added to 0.2% w/v PVA in water (100 mL) at 0.5 mL/min. KGN or FITC (5 mg) were included in the polymer solution prior to precipitation for KGN- or FITC- loaded formulations. This solution was stirred at 60 tiC for 4 h before collecting the nanoparticles by cen-

trifugation [15,000 rpm, 20 min, room temperature (RT, 20 tiC)] and washing three times with water.
For PLGA–PEG–HA nanoparticles, HA was de- salted by dialyzing against 0.1 M HCl. De-salted HA was solubilized in 0.01 M MES buffer (5 mg/mL, pH 5.5) and activated using EDC and sulfo-NHS (1:5:5 molar ratio) for 2 h at RT under N2. Blank or KGN- loaded PLGA–PEG nanoparticles in MES buffer were added to the activated HA solution (3:1 w/w HA:- nanoparticle), and stirred for 24 h at RT under N2. The nanoparticles were collected by centrifugation (15,000 rpm, 20 min, RT) and washed three times.

Nanoparticle Characterization Characterization of Nanoparticle Size and Charge
Hydrodynamic diameter, polydispersity index (PDI), and f potential were measured at 25 ti C at a scattering angle of 90ti using dynamic light scattering (DLS) (Zetasizer Nano ZS90, Malvern Instruments, UK). Transmission electron microscope (TEM) (JEOL 2100F, 200 kV) samples were prepared by dropping aqueous nanoparticle suspensions onto carbon-coated copper grids and then staining with UranyLess stain solution.

Quantification of HA Content on PLGA–PEG–HA Nanoparticles
A CTAB turbidity assay was used to determine the amount of HA conjugated on PLGA–PEG–HA nanoparticles.25 Briefly, 50 lL of sodium acetate buf- fer (0.2 M, pH 5.5) was added to 50 lL of HA stan- dards or nanoparticle wash samples (described in ‘‘Materials and Methods’’) and incubated at 37 ti C for 10 min. Pre-warmed CTAB solution (100 lL, 25 mg/
mL in 0.2% v/v NaOH at 37 ti C) was then added to each sample, and absorbance was read at 570 nm within 10 min. The HA content was calculated using Eq. (1).
HA content
¼ initial HA reactant mass ðmgÞ ti unreacted HA mass ðmgÞ
nanoparticle mass ðmgÞ

Quantification of KGN Loading in Nanoparticles
Encapsulation efficiency (EE%) and drug loading (DL) of KGN was determined using a 1260 Infinity II HPLC (Agilent, CA). Wash samples (described in ‘‘Materials and Methods’’) and KGN standards were analyzed with a C18 column with a mobile phase of ACN:water with 0.1% TFA (10:90–100:0 v/v in 10 min followed by 100% ACN for 10 min). Absorbance was

monitored at 274 nm at the KGN retention time of 7.232 min. EE% and DL were calculated using Eqs. (2) and (3).
¼ initial KGN mass ðmgÞ ti free KGN in supernatant ðmgÞ
ti 100%
initial KGN mass ðmgÞ

Cell viability
ti negative control ti
¼ ti negative control ti ti 100
Positive controls of cells with no treatment and negative controls with no cells or nanoparticles were included.

DL ¼
initial KGN mass ðmgÞ ti free KGN in supernatant ðmgÞ
final nanoparticle mass ðmgÞ

Nanoparticle Interactions with hMSCs
hMSCs (passage 6) were plated at 10,000 cells/cm2 into 6-well plates. Following equilibration in serum-free media for 1 h, FITC-loaded nanoparticles were added

Assessing KGN Release and Suspension Stability In Vitro
To determine KGN release in vitro, KGN-loaded nanoparticle formulations were suspended in PBS (~ 40 mg/mL) and incubated at 37 ti C with shaking. Every 3 days for 21 days, the nanoparticle suspensions were centrifuged (14,000 rpm, 10 min) to collect the supernatant, and an equal volume of fresh PBS was added. KGN concentration in the supernatant was quantified using HPLC. The suspension stability of nanoparticle suspensions (1 mg/mL) at these condi- tions was examined by measuring the hydrodynamic diameter and PDI by DLS. Daily suspension stability was also examined via DLS for these particles incu- bated at 37 or 4 ti C without solution replacement.

Assessing Nanoparticle Interactions with hMSCs Cell Culture
hMSCs (passage 4–5) were expanded in low-glucose (1000 mg/L) DMEM supplemented with 10% v/v FBS, 1% v/v penicillin–streptomycin, and 4 mM L- glutamine at 37 ti C in 5% CO2. Chondrogenic differ- entiation media was composed of high-glucose (4500 mg/L) DMEM supplemented with 100 mM so- dium pyruvate, 100 mM L-proline, 0.2 mg/mL dex- amethasone, 1% v/v penicillin–streptomycin, 1% ITS + premix, and 0.02 M ascorbate-2-phosphate.

hMSCs (passage 6) were plated at 10,000 cells/cm2 into a 96-well plate and then treated with growth media only or growth media supplemented with 0.5 mg/mL of nanoparticles with and without KGN loaded. Following a 24-h incubation, CCK-8 solution was added for 2 h at 37 tiC and absorbance of the wells was measured at 450 nm using a Cytation3 Plate Reader (BioTek, Winooski, VT). Normalized cell via- bility was calculated using Eq. (4).
to the cells at 0.1 mg/mL and incubated for varying time or incubated for 4 h at varying concentration. Follow- ing incubation, the cells were passaged, fixed, and measured using a 488 nm (FITC) laser on a 4-laser, 19- parameter BD FACSAria Illu flow cytometer (Franklin Lakes, NJ). For confocal imaging, FITC-loaded nanoparticles in serum-free media were added to the cells at 0.1 mg/mL for 2 h, fixed, and stained for nuclei (2 lM DAPI, 5 min at RT) and F-actin (alexa-fluor 594- phalloidin 1:100 in 1 9 PBS, 30 min at RT).

hMSC Differentiation
A free-floating MSC pellet was formed by first centrifuging (5009g, 5 min) 250,000 cells to form a pellet in a 15 mL canonical tube. Following overnight incubation, the cells formed a free-floating pellet in the tube, which was allowed to mature for 3 days prior to inducing differentiation. The following conditions were examined (all supplemented in chondrogenic media): chondrogenic media only, KGN only, DMSO only, TGF-b1 only, KGN and TGF-b1, water only, blank nanoparticle formulations, KGN-loaded nanoparticle formulations, or KGN-loaded nanoparticle formula- tions with TGF-b1. Pellets were fed every 3 days with fresh nanoparticles and TGF-b1 (3 ng/mL) where indicated. Nanoparticle concentration was adjusted for each formulation based on drug loading and the known nanoparticle concentration after nanoprecipi- tation to ensure cumulative release of 100 nM KGN over 3 days (i.e., 42.4, 18.4, and 11.28 lg/mL for PLGA, PLGA–PEG, and PLGA–PEG–HA nanopar- ticles supplemented within the media, respectively).
The sGAG content, indicative of chondrogenesis, was assessed using a Blyscan sGAG assay kit. Briefly, pellets were digested using papain extraction reagent and mixed with Blyscan dye. The dye-sGAG complex was pelleted, dissociated, and solubilized prior to measuring absorbance at 656 nm against chondroitin sulfate standards. The sGAG content was normalized against total wet weight of the pellets. Total DNA was

quantified using a PicoGreen assay. Briefly, extract containing DNA was added to 19 Tris–EDTA buffer. Quant-iT PicoGreen reagent was added and allowed to incubate for ~ 3 min prior to measuring at an excita- tion/emission of 480 nm/525 nm against a standard curve of k dsDNA.

Statistical Analysis
All measurements were performed in triplicate and data is presented as mean ± standard deviation unless otherwise stated. Statistical analysis was conducted using one-way or two-way analysis of variance (AN- OVA) with Tukey–Kramer post hoc analysis using GraphPad Prism. Data was considered statistically significant for p < 0.05 (note: *p < 0.05, **p < 0.01, ***p < 0.001). RESULTS Characterizing Nanoparticle Physicochemical Properties We synthesized three KGN-loaded nanoparticle formulations: PLGA, PLGA–PEG, and PLGA–PEG– HA nanoparticles (Scheme 1). We first confirmed successful conjugation of the PLGA–PEG copolymer (Fig. S1). Next, we investigated differences in nanoparticle size, PDI, and f potential (Table 2). Functionalizing the PLGA nanoparticle surface with hydrophilic PEG and HA resulted in a significant in- crease in the hydrodynamic diameter. Particle diame- ters obtained from TEM images (Fig. 1) were significantly lower than the hydrodynamic diameters obtained via DLS (Table 2), as expected. The largest difference in TEM vs. DLS diameter (64% smaller) was noted for the PLGA–PEG–HA particles. There was no significant difference in PDI between formu- lations, suggesting that all formulations were stable and monodisperse in aqueous conditions. PLGA and PLGA–PEG–HA nanoparticles exhibited negative f potentials, indicative of a negative surface charge, as 10,31,36,40 both HA and PLGA are anionic. PLGA–PEG nanoparticles exhibited positive f potential, due to the presence of free amines at the surface.11 The encapsulation efficiencies of KGN-loaded PLGA, PLGA–PEG, and PLGA–PEG–HA nanopar- ticles were ~ 62, 71, and 55%, respectively. The hydrophobic PLGA particles had the highest drug loading, with approximately 47% of the particle mass SCHEME 1. KGN-loaded nanoparticle fabrication. In step one, PLGA is activated and conjugated to PEG-bis-NH2 to form a PLGA– PEG copolymer. In step two, KGN is mixed with either PLGA or PLGA–PEG to form KGN-loaded PLGA nanoparticles or KGN- loaded PLGA–PEG nanoparticles. In step three, HA is activated and mixed with KGN-loaded PLGA–PEG nanoparticles to form KGN-loaded PLGA–PEG–HA nanoparticles. TABLE 2. KGN-loaded nanoparticle properties. Particle diameter (nm) Zeta (f) Drug loading (mg HA content (mg KGN-loaded nanoparticles DLSa TEMb PDI (AU)a potential (mV)c Encapsulation efficiency (%)d KGN/mg nanopar- ticles)d HA/mg nanoparti- cles)e PLGA 166.63 ± 4.48 84.4 ± 7.2 0.282 ± 0.023 2 33.1 ± 1.6 62.0 ± 3.6 0.467 ± 0.192 n/a PLGA–PEG 297.32 ± 4.55 164.2 ± 37.5 0.236 ± 0.014 11.2 ± 0.3 70.5 ± 4.8 0.128 ± 0.026 n/a PLGA–PEG– HA 507.01 ± 12.03 182.1 ± 44.9 0.293 ± 0.021 2 28.5 ± 0.9 55.3 ± 11.8 0.156 ± 0.033 0.357 ± 0.086 n/a not applicable. aHydrodynamic diameter determined by DLS. bDry diameter determined by TEM. cMeasured using a Zetasizer Nano-ZS equipped with a standard capillary electrophoresis cell. dDetermined by HPLC. eDetermined by CTAB assay. FIGURE 1. TEM micrographs of nanoparticle formulations. (a) Blank and (b) KGN-loaded nanoparticles. Scale = 200 nm. attributed to KGN while the hydrophilic formulations exhibited similar KGN loading of approximately 13–16% of particle mass. An HA content of ~ 36% of particle mass was observed for PLGA–PEG–HA par- ticles. Effects of Surface Functionalization on KGN Release and Nanoparticle Suspension Stability The release of KGN from these particles was investigated in vitro in PBS at 37 ti C. In the first 3 days, PLGA nanoparticles released £ 1% of the loaded KGN. Following 3 days, KGN release was below the instrument detection limit (~ 1 lg/mL). PLGA–PEG and PLGA–PEG–HA nanoparticle formulations both released significantly greater KGN than PLGA parti- cles. Cumulative KGN release profiles are shown in Fig. 2 to demonstrate the variability we observed in KGN release from different nanoparticle batches. In general, PLGA–PEG–HA particles released more of their loaded KGN over 21 days compared with PLGA–PEG nanoparticles (15–27% vs. 2–19%, respectively). Additionally, both PLGA–PEG and PLGA–PEG–HA particles exhibited a burst release ranging from 5 to 15% of the total KGN released over the first 3 days. Following 21 days, any KGN release from these particles was below the instrument detec- tion limit. The significant heterogeneity in KGN release sug- gests potential nanoparticle aggregation at the release conditions. Thus, we examined KGN-loaded nanoparticle suspension stability by DLS in conditions mimicking the KGN release study (Fig. 3). The diameter of all nanoparticle formulations generally increased to over 1 lm over the 21-day release period, suggesting significant particle aggregation. PDI also increased during this time with PLGA particles exhibiting the largest PDI at the end of the 21-day release (~ 0.7). We also examined whether changes in particle size and PDI occurred when particles were stored at 37 or 4 ti C in PBS without centrifugation and resuspension over 7 days. Figure S2 shows that there was no significant change in the hydrodynamic diam- eter and PDI of PLGA and PLGA–PEG–HA nanoparticles over 7 days at both 4 and 37 ti C. How- ever, PLGA–PEG nanoparticles showed a large in- crease in the hydrodynamic diameter (a maximum of 2–3 lm) and PDI (a maximum ~ 0.5) at these condi- tions. Interestingly, despite maintaining a consistent hydrodynamic diameter over 7 days, PLGA–PEG–HA particles at 37 tiC experienced an increase in PDI, also reaching a maximum of ~ 0.5 following 1 day of incubation. Nanoparticle Cytocompatibility and Interaction with hMSCs We examined hMSC viability in the presence of nanoparticles at concentrations above the equivalent KGN half maximal effective concentration (EC50) of 100 nM.14 Cells exposed to each of the nanoparticle formulations exhibited viability similar to the non- treated controls (p > 0.05) (Fig. S3). Next, we inves- tigated nanoparticle interactions with hMSCs by loading FITC during nanoparticle fabrication. Flow cytometry results (Fig. S4) showed an increase in hMSC fluorescence with increasing FITC-loaded nanoparticle concentration and increasing incubation time, suggesting nanoparticle association and/or up- take by hMSCs. Figure 4a shows the effect of incu- bation of FITC-loaded nanoparticles with hMSCs at an intermediate concentration (0.1 mg/mL) for varying time. An increase in fluorescence signal was observed with increasing incubation time; however, there was no discernable difference in the fluorescence signal between different nanoparticle formulations. Figure 4b shows confocal microscopy images of hMSCs incu- bated with these nanoparticles. From these images, the hydrophobic PLGA nanoparticles resulted in a more globular localization to the cell surface, compared with a more evenly distributed fluorescence signal throughout the cell volume, suggesting uptake, seen for the hydrophilic PLGA–PEG and PLGA–PEG–HA nanoparticles.

FIGURE 2. KGN percent cumulative release normalized to total KGN loaded from PLGA–PEG and PLGA–PEG–HA nanoparticles over 21 days in 13 PBS at 37 ti C. Release from four distinct nanoparticle batches per formulation type are shown indicating variability in KGN release from these particles at the release conditions examined.

KGN-Loaded Nanoparticle Mediated hMSC Chondrogenesis
The effect of nanoparticle formulations on hMSC chondrogenesis was examined over 7 days in pellet

FIGURE 3. Nanoparticle suspension stability over time in 13 PBS at 37 tiC. Changes in (a) hydrodynamic diameter and (b) PDI of KGN-loaded PLGA, PLGA–PEG, and PLGA–PEG–HA nanoparticles were analyzed. Samples were centrifuged and resuspended in fresh 13 PBS at 37 ti C every 3 days prior to taking measurements. Statistical significance (*p < 0.05) between time points is indicated using two-way ANOVA with Tukey’s post hoc analysis. FIGURE 4. Nanoparticle interaction with hMSCs. (a) Flow cytometry histograms of hMSCs incubated with FITC-loaded nanoparticles at 0.1 mg/mL for 15, 30, or 60 min. hMSC + FITC condition represents a control in which free FITC was added to the MSC incubation media. (b) Maximum intensity projection confocal images of 0.1 mg/mL FITC-loaded nanoparticles incubated with hMSCs for 2 h. Blue represents nuclei stained via DAPI, red is F-actin, and green is FITC. Both xy (top) and xz (bottom) image are shown. Scale = 100 lm and applies to all images. culture. Nanoparticle concentrations were selected for each formulation by calculating the amount of nanoparticles necessary to ensure release 100 nM KGN over 3 days based on the in vitro release data. Control conditions (including media, solvents, and supplements) were also examined (Fig. S5). Total DNA (Fig. S6) and sGAG content per pellet wet weight (w/w) (Fig. 5) were quantified. DNA content varied over a narrow range for hMSCs incubated with different formulations, indicating a relatively consis- tent number of cells in each pellet. Examining sGAG content as an indication of hMSC chondrogenesis, we first noted that all KGN-loaded nanoparticles en- hanced sGAG production compared with non KGN- loaded nanoparticles. Interestingly, the addition of TGF-b1 in the media during pellet culture for non- nanoparticle incubated controls (Fig. S5) did not in- crease sGAG, and in the case of KGN-loaded PLGA and PLGA–PEG nanoparticles (Fig. 5), addition of TGF-b1 actually led to a decrease in sGAG. DISCUSSION In this work, we sought to investigate the effects of surface chemistry on KGN-loaded nanoparticle prop- erties and interactions with hMSCs by comparing KGN-loaded PLGA, PLGA–PEG, and PLGA–PEG– HA nanoparticles. First, we studied how varying nanoparticle formulation affects nanoparticle physic- ochemical properties, including diameter, charge, and KGN EE% and DL (Table 2). We noted that PLGA formed the smallest nanoparticles, followed by PLGA– PEG, and finally PLGA–PEG–HA, with more dra- matic differences noted for hydrodynamic diameters than dry diameters measured via TEM, likely due to the highly hydrophilic nature and large MW of HA. Despite a smaller diameter, PLGA nanoparticles had the greatest drug loading of all formulations, which we suggest is due to hydrophobic–hydrophobic interac- tions between KGN and PLGA. FIGURE 5. sGAG content per pellet weight for hMSCs incubated with nanoparticle formulations over 7 days in chondrogenic media. Statistical significance (***p < 0.001) between time points is indicated using one-way ANOVA with Tukey’s post hoc analysis. Some interesting differences were noted between the KGN-loaded PLGA nanoparticles formulated here using nanoprecipitation and the only other previously reported KGN-loaded PLGA nanoparticles.30 First, our average particle size was ~ 37% smaller (~ 170 nm vs. 270 nm diameter).30 This difference in size may be attributed to the different fabrication methods and also our use of a lower MW PLGA (15.1 kDa vs. 40– 75 kDa),30 given the known MW influence on nanoparticle size.24 However, despite the difference in diameter, the EE% of our nanoparticles was similar to this previously reported formulation (62% compared to 67%).30 Release properties from our PLGA nanoparticles (< 1% release of loaded KGN) differed dramatically from these previously reported KGN- loaded PLGA particles, which exhibited ~ 60% release over the first 20 days in vitro using similar release conditions.30 These differences in release may be at- tributed to various factors including nanoparticle size variations, as well as potential differences in pore size arising from differences in polymer MW and/or fab- rication method utilized. The PLGA–PEG and PLGA–PEG–HA nanoparti- cles released greater amounts of KGN than the PLGA nanoparticles (Fig. 2), suggesting greater KGN inter- action with hydrophobic PLGA compared with the hydrophilic PEG or HA at the nanoparticle surface and potentially larger pore sizes of the larger PLGA– PEG and PLGA–PEG–HA particles (the latter being the largest particles with the greatest release of loaded KGN). The large variation in release profiles between nanoparticle batches and our studies of particle size and PDI confirmed that centrifugation and resuspen- sion of the nanoparticles during the release study causes aggregation of these particles (Fig. 3), which does not occur at all or to this extent for samples stored at the same conditions without these centrifu- gation and resuspension steps (Fig. S2). A particularly interesting observation of nanopar- ticle formulations incubated over time in PBS at 37 ti C without these centrifugation steps (Fig. S2), was the finding that PLGA–PEG nanoparticles exhibited an increase in hydrodynamic diameter over time, while PLGA and PLGA–PEG–HA nanoparticles did not, suggesting that stability was not dictated by hydrophilicity of the particle surface alone. We hypothesize that suspension stability is more impacted by surface charge. Aqueous colloidal suspensions with f potential between ~ ± 30 mV are often unstable due to the lack of electrostatic repulsion between particles, whereas suspensions exhibiting f potential ‡ ~ | ± 30| mV are stable.8,26 KGN-loaded PLGA formulations exhibited f potential (~ 2 33.1 mV) within the region of stability. The KGN-loaded PLGA–PEG nanoparticles exhibited f potential clearly within the instability regime (~ 11.2 mV), which can explain the increase in diameter over time. For PLGA– PEG–HA nanoparticles at 37 ti C, an increase in PDI but not diameter was observed. f potential of these KGN-loaded PLGA–PEG–HA nanoparticles (~ 2 28.5 mV) is close to the approximate cut off for stable suspensions but not as definitive, potentially contributing to this observation. Next we investigated the interaction of these for- mulations with hMSCs. This interaction may include cell surface association and/or particle internalization. Many factors can influence these mechanisms includ- ing size and charge. Smaller nanoparticles often demonstrate the greatest uptake by cells.29 If only size is considered, we would hypothesize the greatest up- take for PLGA nanoparticles. Positively charged nanoparticles often more favorably interact with and are taken up by cells compared with negatively charged particles, due to the negative charge of cell mem- branes.7 Considering charge, we would predict PLGA– PEG formulations to exhibit the most interaction. However, flow cytometry indicated no discernable differences between hMSCs incubated with different FITC-loaded nanoparticle formulations at a set nanoparticle concentration and incubation time (Figs. 4a and S4), suggesting similar levels of interac- tion for all nanoparticle formulations with these cells. Confocal imaging indicated that although overall interaction may be similar between formulations, differences in uptake vs. surface-association occur, with the more hydrophilic PLGA–PEG and PLGA– PEG–HA particles exhibiting greater internalization (Fig. 4b). We were also interested in exploring the contribu- tion of HA-CD44 interactions to the overall interac- tion of PLGA–PEG–HA nanoparticles with hMSCs. To do so, we first incubated cells with free HA to act as a competitive inhibitor of PLGA–PEG–HA nanopar- ticles as previously reported32 before nanoparticle incubation with hMSCs (Fig. S7). No inhibition of nanoparticle interaction with hMSCs was observed with this HA pre-incubation using flow cytometry (Fig. S7). These results suggest that HA-CD44 inter- actions are not the primary mechanism of interaction of these particles with hMSCs, but we cannot entirely discount the possibility of HA-CD44 interactions contributing to PLGA–PEG–HA nanoparticle-hMSC interaction. It is possible that these interactions may be observed with greater HA functionalization of the nanoparticle surface. However, it should be noted that HA-CD44 interactions have previously been inhibited using these methods for PLGA–PEG–HA nanoparti- cles (albeit with different cell types), at lesser nanoparticle HA functionalization (0.116 mg HA/mg NP compared to 0.357 mg HA/mg NP in our study).32 Finally, the potential influence of nanoparticle for- mulations on hMSC chondrogenic differentiation was examined by incubating these formulations with hMSCs for a 7-day period. At this time, we examined sGAG content of each hMSC pellet for every condi- tion tested. The sGAG content is a representative cartilage-specific matrix component. This quantity (sGAG mass per pellet mass) is a common metric used to assess chondrogenesis.21 As expected, all KGN- loaded formulations exhibited a greater sGAG content compared with non KGN-loaded formulations (Fig. 5). At 7 days, there were no dramatic differences observed in sGAG content between cells incubated with different KGN-loaded nanoparticle formulations (Fig. 5). This result is not surprising considering that nanoparticle mass was adjusted to provide the same KGN dosing from all formulations and that overall nanoparticle interaction with hMSCs was observed not to vary between formulations. Additionally, sGAG content was examined after 7 days; more variation between cells incubated with different formulations may be observed and can be examined in future studies 16,17 over lengthier incubation times. In conclusion, we demonstrated that nanoparticle surface functionalization influences nanoparticle physicochemical properties, drug release behavior, and suspension stability. All KGN-loaded formulations interacted with hMSCs with differences in surface-as- sociation vs. internalization observed between hydrophobic and hydrophilic nanoparticle surface properties. All KGN-loaded nanoparticles improved hMSC chondrogenesis compared with non KGN-loa- ded formulations over a short timescale. These nanoparticle formulations have the potential to be used to impact hMSC chondrogenesis for OA thera- pies. These nanoparticles can be incubated with cells during hMSC in vitro culture prior to injection, they could be used as a stand-alone therapeutic, or they may be combined with an injectable carrier such as a hydrogel for localized use in vivo. The findings of this study may also be applicable to the design of nanoparticle carriers for delivery of other hydrophobic small molecules to hMSCs or other cell types. ELECTRONIC SUPPLEMENTARY MATERIAL The online version of this article (https://doi.org/10. 1007/s10439-019-02430-x) contains supplementary material, which is available to authorized users. ACKNOWLEDGMENTS The authors acknowledge financial support from Brown University. We thank Dr. Alessia Battigelli from Brown University for her helpful guidance on the chemical syntheses presented. We also thank Kevin Carlson from the Flow Cytometry Core at Brown University and Professor Robert Hurt at Brown for use of his DLS. Finally, we thank Anthony McCor- mick from Brown University’s Electron Microscopy Core Facility for assistance with TEM. REFERENCES 1Barry, F., and M. Murphy. Mesenchymal stem cells in joint disease and repair. Nat. Rev. Rheumatol. 9:584, 2013. 2Danhier, F., E. Ansorena, J. M. Silva, R. Coco, A. Le Breton, and V. Pre´at. PLGA-based nanoparticles: an overview of biomedical applications. J. Control. Release 161:505–522, 2012. 3Decker, R. S., E. Koyama, M. Enomoto-Iwamoto, P. Maye, D. Rowe, S. Zhu, P. G. Schultz, and M. Pacifici. Mouse limb skeletal growth and synovial joint develop- ment are coordinately enhanced by Kartogenin. Dev. Biol. 395:255–267, 2014. 4Fakhari, A., and C. Berkland. Applications and emerging trends of hyaluronic acid in tissue engineering, as a dermal filler and in osteoarthritis treatment. Acta Biomater. 9:7081–7092, 2013. 5Fan, W., J. Li, L. Yuan, J. Chen, Z. Wang, Y. Wang, C. Guo, X. Mo, and Z. Yan. Intra-articular injection of Kartogenin-conjugated polyurethane nanoparticles atten- uates the progression of osteoarthritis. Drug Deliv. 25:1004–1012, 2018. 6Freitag, J., D. Bates, R. Boyd, K. Shah, A. Barnard, L. Huguenin, and A. Tenen. Mesenchymal stem cell therapy in the treatment of osteoarthritis: reparative pathways, safety and efficacy—a review. BMC Musculoskelet. Disord. 17:230, 2016. 7Fro¨hlich, E. The role of surface charge in cellular uptake and cytotoxicity of medical nanoparticles. Int. J. Nanomed. 7:5577–5591, 2012. 8Gumustas, M., C. T. Sengel-Turk, A. Gumustas, S. A. Ozkan, and B. Uslu. Chapter 5: Effect of polymer-based nanoparticles on the assay of antimicrobial drug delivery systems. In: Multifunctional Systems for Combined Delivery, Biosensing and Diagnostics, edited by A. M. Grumezescu. Amsterdam: Elsevier, 2017, pp. 67–108. 9Herrera, M. B., B. Bussolati, S. Bruno, L. Morando, G. S. Mauriello-Romanazzi, I. Stamenkovic, L. Biancone, and G. Camussi. Exogenous mesenchymal stem cells localize to the kidney by means of CD44 following acute tubular in- jury. Kidney Int. 72:430–441, 2007. 10Horkay, F., P. J. Basser, D. J. Londono, A.-M. Hecht, and E. Geissler. Ions in hyaluronic acid solutions. J. Chem. Phys. 131:184902, 2009. 11Hsu, H. J., S. Sen, R. M. Pearson, S. Uddin, P. Kra´l, and S. Hong. Poly(ethylene glycol) corona chain length controls end-group-dependent cell interactions of dendron micelles. Macromolecules 47:6911–6918, 2014. 12Hu, Q., B. Ding, X. Yan, L. Peng, J. Duan, S. Yang, L. Cheng, and D. Chen. Polyethylene glycol modified PA- MAM dendrimer delivery of Kartogenin to induce chon- drogenic differentiation of mesenchymal stem cells. Nanomed. Nanotechnol. Biol. Med. 13:2189–2198, 2017. 13Jo, C. H., Y. G. Lee, W. H. Shin, H. Kim, J. W. Chai, E. C. Jeong, J. E. Kim, H. Shim, J. S. Shim, I. S. Shin, J. C. Ra, S. Oh, and K. S. Yoon. Intra-articular injection of mes- enchymal stem cells for the treatment of osteoarthritis of the knee: a proof-of-concept clinical trial. Stem Cells 32:1254–1266, 2014. 14Johnson, K., S. Zhu, M. S. Tremblay, J. N. Payette, J. Wang, L. C. Bouchez, S. Meeusen, A. Althage, C. Y. Cho, X. Wu, and P. G. Schultz. A stem cell-based approach to cartilage repair. Science 336:717, 2012. 15Kang, M. L., S. Y. Jeong, and G. I. Im. Hyaluronic acid hydrogel functionalized with self-assembled micelles of amphiphilic PEGylated Kartogenin for the treatment of osteoarthritis. Tissue Eng. Part A 23:630–639, 2017. 16Kang, M. L., J. E. Kim, and G. I. Im. Thermoresponsive nanospheres with independent dual drug release profiles for the treatment of osteoarthritis. Acta Biomater. 39:65–78, 2016. 17Kang, M. L., J. Y. Ko, J. E. Kim, and G. I. Im. Intra- articular delivery of Kartogenin-conjugated chitosan nano/ microparticles for cartilage regeneration. Biomaterials 35:9984–9994, 2014. 18Kang, H., K. Zhang, Q. Pan, S. Lin, D. S. H. Wong, J. Li, W. Y.-W. Lee, B. Yang, F. Han, G. Li, B. Li, and L. Bian. Remote control of intracellular calcium using upconversion nanotransducers regulates stem cell differentiation in vivo. Adv. Funct. Mater. 28:1802642, 2018. 19Kong, L., L. Z. Zheng, L. Qin, and K. K. W. Ho. Role of mesenchymal stem cells in osteoarthritis treatment. J. Or- thop. Transl. 9:89–103, 2017. 20Kristjansson, B., and S. Honsawek. Current perspectives in mesenchymal stem cell therapies for osteoarthritis. Stem Cells Int. 2014. https://doi.org/10.1155/2014/194318. 21 Li, X., J. Ding, Z. Zhang, M. Yang, J. Yu, J. Wang, F. Chang, and X. Chen. Kartogenin-incorporated thermogel supports stem cells for significant cartilage regeneration. ACS Appl. Mater. Interfaces 8:5148–5159, 2016. 22Li, J., W. Y. W. Lee, T. Wu, J. Xu, K. Zhang, D. S. Hong Wong, R. Li, G. Li, and L. Bian. Near-infrared light-trig- gered release of small molecules for controlled differentia- tion and long-term tracking of stem cells in vivo using upconversion nanoparticles. Biomaterials 110:1–10, 2016. 23Makadia, H. K., and S. J. Siegel. Poly lactic-co-glycolic acid (PLGA) as biodegradable controlled drug delivery carrier. Polymers 3:1377–1397, 2011. 24Martı´nez Rivas, C. J., M. Tarhini, W. Badri, K. Miladi, H. Greige-Gerges, Q. A. Nazari, S. A. Galindo Rodrı´guez, R. A´ . Roma´n, H. Fessi, and A. Elaissari. Nanoprecipitation process: from encapsulation to drug delivery. Int. J. Pharm. 532:66–81, 2017. 25Oueslati, N., P. Leblanc, C. Harscoat-Schiavo, E. Rondags, S. Meunier, R. Kapel, and I. Marc. CTAB turbidimetric method for assaying hyaluronic acid in complex environ- ments and under cross-linked form. Carbohydr. Polym. 112:102–108, 2014. 26Pate, K., and P. Safier. Chemical metrology methods for CMP quality. In: Advances in Chemical Mechanical Planarization (CMP), edited by S. Babu. Cambridge: Woodhead Publishing, 2016, pp. 299–325. 27Pittenger, M. F., A. M. Mackay, S. C. Beck, R. K. Jaiswal, R. Douglas, J. D. Mosca, M. A. Moorman, D. W. Simonetti, S. Craig, and D. R. Marshak. Multilineage potential of adult human mesenchymal stem cells. Science 284:143–147, 1999. 28Rezvantalab, S., N. I. Drude, M. K. Moraveji, N. Gu¨vener, E. K. Koons, Y. Shi, T. Lammers, and F. Kiessling. PLGA-based nanoparticles in cancer treatment. Front. Pharmacol. 9:1260, 2018. 29Shang, L., K. Nienhaus, and G. U. Nienhaus. Engineered nanoparticles interacting with cells: size matters. J. Nanobiotechnol. 12:5, 2014. 30Shi, D., X. Xu, Y. Ye, K. Song, Y. Cheng, J. Di, Q. Hu, J. Li, H. Ju, Q. Jiang, and Z. Gu. Photo-cross-linked scaffold with Kartogenin-encapsulated nanoparticles for cartilage regeneration. ACS Nano 10:1292–1299, 2016. 31Soler, R. R., A. Munar, R. F. Soler, X. Peirau, M. Huguet, M. Alberca, A. Sanchez, S. J. Garcia, and L. Orozco. Treatment of knee osteoarthritis with autologous expanded bone marrow mesenchymal stem cells: 50 cases clinical and MRI results at one year follow-up. J. Stem Cell Res. Ther. 2015. https://doi.org/10.4172/2157-7633.1000285. 32Vangara, K. K., J. L. Liu, and S. Palakurthi. Hyaluronic acid-decorated PLGA–PEG nanoparticles for targeted delivery of SN-38 to ovarian cancer. Anticancer Res. 33:2425–2434, 2013. 33Wakitani, S., T. Goto, S. J. Pineda, R. G. Young, J. M. Mansour, A. I. Caplan, and V. M. Goldberg. Mesenchymal cell-based repar of large, full-thickness defects of articular cartilage. J. Bone Joint Surg. 76A:579–592, 1994. 34Wang, E. C., and A. Z. Wang. Nanoparticles and their applications in cell and molecular biology. Integr. Biol. 6:9– 26, 2014. 35Wittenauer, R., L. Smith, and K. Aden. Background Paper 612 Osteoarthritis. Geneva: World Health Organization, 2013. 36Xu, Q., A. Crossley, and J. Czernuszka. Preparation and characterization of negatively charged poly(lactic-co-gly- colic acid) microspheres. J. Pharm. Sci. 98:2377–2389, 2009. 37Xu, J., J. Li, S. Lin, T. Wu, H. Huang, K. Zhang, Y. Sun, K. W. K. Yeung, G. Li, and L. Bian. Nanocarrier-mediated codelivery of small molecular drugs and siRNA to enhance chondrogenic differentiation and suppress hypertrophy of human mesenchymal stem cells. Adv. Funct. Mater. 26:2463–2472, 2016. 38Yadav, A. K., A. Agarwal, G. Rai, P. Mishra, S. Jain, A. K. Mishra, H. Agrawal, and G. P. Agrawal. Development and characterization of hyaluronic acid decorated PLGA nanoparticles for delivery of 5-fluorouracil. Drug Deliv. 17:561–572, 2010. 39Yang, W., Y. Zheng, J. Chen, Q. Zhu, L. Feng, Y. Lan, P. Zhu, S. Tang, and R. Guo. Preparation and characteriza- tion of the collagen/cellulose nanocrystals/USPIO scaffolds loaded Kartogenin for cartilage regeneration. Mater. Sci. Eng. C 99:1362–1373, 2019.
40Zhu, H., N. Mitsuhashi, A. Klein, L. W. Barsky, K. Weinberg, M. L. Barr, A. Demetriou, and G. D. Wu. The role of the hyaloronan receptor CD44 in mesenchymal stem cell migration in the extracellular matrix. Stem Cells 24:928–935, 2009.

Publisher’s Note Springer Nature remains neutral with re- gard to jurisdictional claims in published maps and institu- tional affiliations.