siRNA Nanoparticle Functionalization of Nanostructured Scaffolds Enables Controlled Multilineage Differentiation of Stem Cells

Introduction

Tissue engineering has the potential to alleviate disease by producing abundant and tolerated replacement organs.¹ The standard approach is to seed patient-derived terminally differentiated cells on porous three-dimensional cell supports (scaffolds). When it comes to generating tissues with multiple cell types, however, this method is limited to scaffold structures that can be loaded with cells in physically separated locations such as spheres (bladders²) and tubes (larynx³). A similar rapid prototyping approach is cell printing where different cell populations are deposited into three-dimensional shapes.^4,5 Unfortunately, this approach presents limitations⁶ including cell alterations from induced mechanotransduction during processing. Moreover, reconstruction with a priori differentiated cells is problematic as cells perform important functions such as preparing extracellular matrix while undergoing differentiation and loose this ability when fully differentiated.⁷ Consequently, we explored the seeding of stem cells onto scaffolds before cell specialization. A limitation with this approach is that conventional global provision of differentiation cues fails to differentiate stem cells into multiple cell types in discrete locations. Here, we present a novel strategy where nanostructured scaffolds are coated with different nanoparticles in spatially discrete parts directing the differentiation pathway of one homogenously seeded stem cell population into multiple cell types in situ.

Several different drugs have been used to steer differentiation on scaffolds including proteins,⁸ plasmids,^9,10 and viruses.¹¹ We believe a more versatile drug type to be small-interfering RNA (siRNA). Once introduced into cells, siRNAs can silence synthesis of a specific protein by base pairing with its mRNA sequence.¹² When applied to cultured monolayers of mesenchymal stem cells (hMSCs), commonly used in tissue engineering,¹³ siRNAs were able to enhance their differentiation into bone,¹⁴ cartilage,¹⁵ fat,¹⁶ muscle,¹⁷ liver,¹⁸ and nerve cells.¹⁹ A nanoparticle-based implant coating, capable of delivering siRNAs into stem cells situated in a scaffold, therefore, has broad applications in tissue engineering.^20,21

Here, we present a novel technology, siRNA-enhanced scaffolds (siRESs), composed of biodegradable nanostructured poly-ε-caprolactone (PCL) scaffolds functionalized with a lyophilized polymer/lipid-based nanoparticulate siRNA delivery system.²² We investigate, the particle localization and retention properties as well as the silencing and in vitro and in vivo enhancement of differentiation of the system, using hMSCs as progenitor cells and siRNAs targeted to enhanced green fluorescent protein (EGFP), tribbles homolog 2 (TRIB2, also known as TRB2), and BCL2 like 2 (BCL2L2, also known as BCL-w). We demonstrate successful enhancement of differentiation, and importantly, that tailored cell specialization can be affected differently in discrete locations within a composite scaffold by controlled deposition of BCL2L2 siRNA and TRIB2 siRNA containing nanoparticles.

Results

Discussion

This work presents a nanoparticle decorated nanostructured scaffold, capable of retaining and delivering siRNA, with broad applications for controlling stem cell differentiation in vitro and in vivo. When functionalized with two different siRNAs such a scaffold was shown to promote two alternative differentiation pathways in specific locations, both in vitro and in vivo.

siRNA has previously been delivered to cells growing within a three-dimensional matrix either by embedding the siRNA in flexible hydrogels²⁷ or by adding the siRNA to the culturing medium^28,29 alternatively, cells have been transfected with siRNA before seeding onto implants.³⁰ None of these methods, however, is applicable to delivery of siRNA in specific locations in order to generate multiple cell types within a matrix. This necessitates stable binding of siRNA to scaffold walls until subsequent delivery directly to attaching cells.

DNA delivery from scaffolds has been explored more extensively than siRNA. These studies have been performed by adsorbing naked plasmid DNA,³¹ DNA containing nanoparticles¹⁰ and viral vectors¹¹ onto scaffolds. The transfection efficiency of naked nucleic acids is usually very low³² and naked siRNA delivery is further complicated by instability to RNAses.³³ As viral vectors have disadvantages due to oncogenicity³⁴ and immunogenicity,³⁵ nonviral vectors offer an attractive delivery solution. Lyoprotected, dried siRNA nanoparticles can normally achieve high silencing²² but common lyoprotectants, such as glucose, can interfere with differentiation.³⁶ Importantly, TransIT-TKO/siRNA complexes can be freeze-dried without a lyoprotectant, stored dry for prolonged periods and still retain high knockdown efficiency in serum promoting their use for scaffold delivery.²²

When lyophilized onto the scaffolds, the nanoparticles retained their prelyophilization morphology and located to the smaller cavities within the scaffold walls. NaOH-treated PCL is negatively charged because of exposed carboxylic acid²⁵ and we found TransIT-TKO/siRNA nanoparticles to be positively charged. The observed interaction is therefore most likely electrostatic in nature. The nanostructured holes presumably provide a greatly increased surface area with which the cationic TransIT-TKO component can interact through multiple ionic interactions. Furthermore, these nanopores can be expected to protect against fluid flow displacement. The stability of the interaction is indicated by the continued retention of the nanoparticles in the walls after 24 hours in serum containing medium. This stable adherence enables localization of siRNA nanoparticles and the regional knockdown they induce.

In vitro we observed that BCL2L2 silencing enhanced osteogenic differentiation when combined with osteogenic medium. This confirms a previous study done with in vitro monolayer cuture.¹⁴ This study showed that silencing the antiapoptotic factors BCL2L2, BCL2, TBX3, and BIRC4 increased osteogenic differentiation, although the exact mechanism remains unknown. In vivo, we found no evidence of mineralization, confirming previously published in vitro findings¹⁴ that BCL2L2 knockdown alone does not induce mineralization. We did, however, find that BCL2L2 knockdown alone enhanced early osteogenic differentiation with regard to collagen type I deposition and organization in vivo.

TRIB2 silencing increased adipogenic differentiation in vitro when combined with adipogenic medium. This confirms several in vitro monolayer studies demonstrating that the TRIB family members repress adipogenic differentiation.^16,37,38 TRIB2 inhibits adipogenesis by inhibiting Akt1 activation while increasing C/EBPβ degradation. EGFP-specific siRNA appeared to increase adipogenic differentiation, although not as much as TRIB2 siRNA. This confirms an earlier study showing that siRNAs induce weak sequence independent adipogenesis.³⁹ The in vivo study showed that TRIB2 knockdown alone was enough to abolish collagen birefringence while allowing a subset to differentiate into S100 positive adipocytes after 8 weeks.

A single siRNA sequence has previously been shown to stimulate terminal MSC differentiation,^18,19 but our results indicate that silencing BCL2L2 or TRIB2, while enhancing differentiation, is insufficient to drive terminal osteogenic and adipogenic specialization, respectively, without further stimuli. The few S100 positive adipocytes observed in vivo, may be a result of a weak adipogenic stimulus from the scaffolds or the implantation site. To achieve more effective differentiation, siRNA cocktails may be used.¹⁴ Alternatively, microRNAs (miRNAs) or miRNA inhibitors (antimirs) could be used. miRNAs are endogenous RNA molecules related to siRNAs, and they simultaneous regulate multiple genes. miRNAs increasing osteogenic⁴⁰ and adipogenic⁴¹ differentiation have been found. Considering that the TransIT-TKO delivery system can deliver miRNA⁴² and antimirs, the system presented here should also be directly applicable for scaffold delivery of miRNA regulators.

We were able to develop a nanostructured scaffold that preferentially stimulated adipocytic and osteoblastic differentiation in distinct regions using localized coating with siRNAs against TRIB2 and BCL2L2. This was confirmed in vitro by the differential expression of the markers AP2 and ALP. In the same experiment RUNX2, PPARγ2, adiponectin, and OC expression was equal between the two sides at day 7. Adipogenesis and osteogenesis are two oppositely regulated pathways and differentiation enhancers are known to downregulate genes relating to the opposite pathway.⁴³ We find it conceivable that this effect may delay the upregulation of certain markers and account for their equal expression in the two sides. It does not, however, seem to affect the development of distinct tissues in vivo. In the in vivo incubated scaffolds, the adjacent parts developed markedly different morphologies and collagen deposition and organization was highly increased in the siBCL2L2-coated side. Based on these results, we can imagine engineering tissues containing multiple cell types, by assembling differentially coated building blocks. Such cocultures are of great interest in tissue engineering⁴⁴ as they form the basis of most organs. Although we have focused on siRNA nanoparticles due to known versatility, spatial localization of other differentiation cues could possibly be used to achieve the same result. In addition, other assembly techniques such as rapid prototyping of different drugs to distinct locations could conceivably be used instead of manually assembling building blocks.

To our knowledge, we have developed the first example of an siRNA nanoparticle functionalized scaffold capable of modulating stem cell differentiation. We demonstrate that the siRNA is delivered into seeded cells and induces sequence specific gene silencing. As clinically relevant examples, we show that specific targeting of TRIB2 and BCL2L2 on implants leads to enhanced adipogenic and osteogenic differentiation, respectively. Scaffold coating with a single type of siRNA is thus a versatile and effective method for enhancing the development of single cell type tissues and represents a method for conducting gene knockout studies in three dimension. Importantly, the nanostructured nature of the scaffolds enables nanoparticle retention and localization of different siRNAs to distinct parts of an implant. This made it possible to guide stem cells into alternate differentiation in specified locations. We believe this method will provide a realistic strategy to engineer tissues and organs that contain multiple cell types in the future.

Materials and Methods

Materials. Phosphate-buffered saline, minimum essential medium, fetal bovine serum, antibiotics, Trizol, MMLV RT, and SYBR Green qPCR kits were from Invitrogen (Carlsbad, CA). TransIT-TKO was from Mirus Bio (Madison, WI). Primers were from DNA Technology (Risskov, Denmark). PCL (grade 6506) was from Solvay (Warrington, UK). 1,4-Dioxane, insulin, 3-isobutyl-1-methylxanthine, dexamethasone, paraformaldehyde, vitamin D3, vitamin C, and β-glycerophosphate were from Sigma (St Louis, MO). MTS assay was from Promega (Madison, WI). Ultra-low cell adhesion plates were from Corning (Corning, NY). EGFP siRNAs were from Dharmacon (Lafayette, CO). siRNAs against BCL2L2 and TRIB2 were from DNA Technology or Ribotask (Odense, Denmark). Sequences can be found in Supplementary Tables S1 and S2.

Complex formulation. TransIT-TKO/siRNA complexes were made by mixing 3 µl TransIT-TKO with 25 µl minimum essential medium, and after 5 minutes 3 µl of 5 µmol/l siRNA was added. After 15 minutes, the complexes were used or evaluated for hydrodynamic diameter and ζ potential. All measurements were performed in triplicates on a Zetasizer Nano ZS (Malvern, Malvern, UK).

Scaffold production. PCL scaffolds were prepared by thermally induced phase separation of 41.8 mg PCL in 1 g 1,4-dioxane and 100 ng H₂O. After lyophilization, the scaffolds were frozen in liquid N₂, and scaffold discs (Ø = 8 mm, h = 4 mm) were cut. The discs were washed with EtOH and double-distilled H₂O. The scaffolds were then etched in 0.25 mol/l NaOH for 16 hours and 0.25 mol/l HCl for 1 hour. Afterward they were neutralized, dehydrated, and soaked with TransIT-TKO/siRNA solution followed by lyophilization at −20 °C for 24 hours.

Cell culture. The hMSCs were maintained in minimum essential medium with 10% fetal bovine serum and 1% penicillin/streptomycin at 37 °C, 5% CO₂, and 100% humidity, and were split when confluent. The hMSCs used had been passaged between 59 and 62 times. Coated scaffolds were placed in ultra-low cell adhesion plates and were soaked in maintenance medium with 4,000 cells/µl. After 15 minutes, 2 ml medium was added. The medium was replaced after 24 hours and twice every week. The adipogenic medium contained 100 nmol/l dexamethasone, 450 nmol/l 3-isobutyl-1-methylxanthine, and 10 nmol/l insulin. The osteogenic medium contained 10 mmol/l β-glycerophosphate, 10 nmol/l dexamethasone, 290 nmol/l vitamin C, and 10 nmol/l vitamin D3. The complex medium contained 27.5 nmol/l dexamethasone, 112.5 nmol/l 3-isobutyl-1-methylxanthine, 2.5 nmol/l insulin, 2.5mmol/l β-glycerophosphate, 2.5 nmol/l vitamin D3, and 290 nmol/l vitamin C.

Monolayer transfection. Monolayer reverse transfection experiments were performed by adding 25,000 cells/cm² in 125 µl/cm² to wells precoated with 12.5 µl/cm² lyophilized siRNA reagent giving an active well concentration of 25 nmol/l siRNA. After 24 hours, the medium was removed and replaced with 250 µl/cm² fresh medium. The medium was changed twice every week.

Flow cytometry. The flow cytometry experiments were performed with 24-well plates. The cells were harvested from wells using a standard phosphate-buffered saline and trypsin procedure. The harvested cells were then centrifuged for 5 minutes at 1,500 r.p.m., washed with phosphate-buffered saline, centrifuged again, and finally fixed in 1% paraformaldehyde in phosphate-buffered saline. The cells were processed on a FACScalibur (BD Biosciences, San Jose, CA). The cell population was selected using forward and side scatter, the geometric mean of this population in the FL1 channel was used as a measure of EGFP fluorescence. Each sample type was performed in three biological replicates.

Viability assay. The viability was assessed using a tetrazolium based MTS-based assay in 96-well plates. After 14 days of cell culture, wells were added fresh maintenance medium with 20% cytotoxicity assay solution, the wells were then incubated at 37 °C until color development had taken place (~30 minutes). OD_490nm and OD_650nm were then measured for each well, OD_490nm was then corrected for OD_650nm and was subtracted the average value of four wells without cells. The resulting number was then used as a measure of viability. All samples were run in four biological replicates.

Scanning electron microscopy. For scanning electron microscopy, the scaffolds were placed on aluminum stubs and visualized using a NovaSEM system (low vacuum detector, spot size: 3, voltage: 3 kV, working distance <5 mm).

Confocal laser scanning microscopy. Fluorescence microscopy was performed using an inverted Zeiss LSM510 META NLO laser scanning confocal microscope system. Images were acquired with a ×40 NA 1.2 water objective with Z-directional slice thickness of 1 µm. The scaffolds were visualized by means of two-photon excitation at 760 nm using a femtosecond laser (MaiTai Broadband; Spectra Physics, Santa Clara, CA). The scaffold autofluorescence signal was detected in the 405–490 nm interval using the META detector. Fluorescence signals from EGFP and Cy5 were imaged on the other detectors using 488 and 633 nm excitation, respectively, and appropriate bandpass filters. Images were acquired sequentially (multitrack mode) with pin holes set 1 airy unit for the EGFP and Cy5 detection channels but left fully open for the scaffold detection channel.

qPCR. RNA was extracted from scaffolds using Trizol. Complementary DNA was produced using an MMLV RT kit. qPCR was performed on an mxpro3005 system, Stratagene (La Jolla, CA), using a SYBR Green kit. CT values were found using the dR method. The PCR efficiency was measured using standard curves. The relative mRNA levels were calculated using the following equation:

The mRNA levels were normalized to GAPDH or β-2 microglobulin. GAPDH is upregulated in hypoxic hMSCs⁴⁵ and was not used for differentiation experiments as deposited extracellular matrix could interfere with oxygen diffusion. The qPCR experiments were conducted on at least biological triplicates in technical duplicates. A two-tailed t-test was performed to test for difference of means. Primer sequences can be found in Supplementary Table S2.

Immunohistochemistry and staining. The scaffolds were fixed in 4% formaldehyde at 4 °C for 24 hours and stored at 4 °C in 0.9% NaCl. Samples were paraffin embedded and 4-µm thick sections were mounted on slides, deparaffinized with xylene, and rehydrated by stepwise washing in ethanol with increasing quantities of water. Slides were demasked by boiling in a microwave oven. Staining was performed, following the EnVision+ protocol (Dako, Glostrup, Denmark). See Supplementary Table S3 for target proteins, primary antibodies, dilutions, and suppliers used. Slides were incubated with primary antibody for 1 hour, washed with TNT-buffer (0.1 mol/l Tris HCl, 0.15 mol/l NaCl, 0.05% Tween-20, pH 7.5), and incubated with secondary antibody; peroxidase-labeled EnVision+ polymer, for 30 minutes. After washing with TNT-buffer, slides were incubated with 3,3-diaminobenzidine for 10 minutes to produce the stain. Slides were counterstained with Mayer's hematoxylin and mounted on a cover slip with Harleco Aquatex mounting medium (EMD Chemicals, Gibbstown, NJ).

Von Kossa staining was performed on deparaffinized sections in 10% silver nitrate for 20 minutes in the dark. Slides were then placed in sunlight for 45 minutes, rinsed in 5% pyragallol for 2 minutes, and immersed in 5% sodium thiosulphate for 5 minutes before counterstaining in eosin and mounting. Sections were stained with picro-Sirius red and visualized under circular polarized light. Whole-section images were obtained at ×100 magnification using a Leica microscope, DFC480 camera, and Survey 5.5.4.3 automated specimen scanning software.

Micro CT scanning. The scaffold discs were fixed with 4% formaldehyde for 24 hours at 4 °C and washed for four times with double-distilled H₂O before being dehydrated and stacked in a cylinder. Tomographic reconstructions (6-µm voxel resolution) were performed on the cylinder using a µCT-40 from Scanco Medical (Brüttisellen, Switzerland). The maximum signal strength from a scaffold without cells was used as threshold X-ray absorbance value for “bone” tissue. The threshold was then applied equally to all slices in all samples to establish the degree of mineralization (bone volume/total volume).

Implantation. The in vivo implantation was modified from previous.⁴⁶ Half-discs containing either BCL2L2 or TRIB2 siRNA were sutured (Ethilon EH7446 monofilament) together at each end of their flat diameter to generate a combined scaffold disc with a juxtaposed central diameter. The BCL2L2 side received an additional suture for identification. Composite scaffolds or single siRNA-coated scaffolds were seeded with 800,000 cells and maintained overnight in maintenance medium. Scaffolds were implanted subcutaneously in 8-week-old female NOD/LTSz-Prkdc^scid mice.

SUPPLEMENTARY MATERIAL Figure S1. Monolayer transfections of hMSCs (EGFP+) were carried out with EGFP match or mismatch siRNA for 2 days as detailed in Figure 1a. EGFP fluorescence was investigated using a histogram. Representative pictures are shown. Figure S2. Alkaline phosphatase (a), oil red O (b) and alizarin red (c) staining of hMSCs (EGFP-) grown for 2 days in maintenance medium in non-coated wells (NC) or wells coated with siEGFP, siBCL2L2 #1 or siTRIB2 #1 followed by 12 days in maintenance or differentiation medium. Positive staining appears red, counter staining appears blue (H&E staining, only applied to a & b). MM, OM and AM denote maintenance, osteogenic and adipogenic media, respectively. All samples were performed in triplicate. Note that the transfected samples contain fewer cells than the non-transfected controls. Figure S3. In vitro osteogenic (a) and adipogenic (b) differentiation of hMSCs (EGFP-) seeded on non-coated scaffolds and scaffolds coated with TransIT-TKO nanoparticles containing siRNA targeted to two regions of BCL2L2 or TRIB2, respectively. After 3 days of culturing in maintenance medium, differentiation was induced with osteogenic (a) or adipogenic (b) medium for 7 days. The ratio of ALP, COL1, RUNX2 and OC (a) or AP2 and PPARγ2 (b) mRNA to GAPDH mRNA was characterized. Averages are shown with standard deviation indicated by error bars. The samples represent n = 5, the values were normalized to the non-coated control scaffolds which were set to 100, “*” denotes that the sample average is significantly different (p < 0.05) from the non-coated control group. Table S1. siRNA sequences used. Table S2. Primers used for quantitative real time PCR. Table S3. Antibodies for immunohistochemistry.

Supplementary Material