Enhancing Chondrogenesis and Mechanical Strength Retention in Physiologically Relevant Hydrogels with Incorporation of Hyaluronic Acid and Direct Loading of TGF-β

Abstract

Cell-loaded hydrogels are frequently applied in cartilage tissue engineering for their biocompatibility, ease of application, and ability to conform to various defect sites. As a bioactive adjunct to the biomaterial, transforming growth factor beta (TGF-β) has been shown to be essential for cell differentiation into a chondrocyte phenotype and maintenance thereof, but the low amounts of endogenous TGF-β in the in vivo joint microenvironment necessitate a mechanism for controlled delivery and release of this growth factor. In this study, TGF-β3 was directly loaded with human bone marrow-derived mesenchymal stem cells (MSCs) into poly-D,L-lactic acid/polyethylene glycol/poly-D,L-lactic acid (PDLLA-PEG) hydrogel, or PDLLA-PEG with the addition of hyaluronic acid (PDLLA/HA), and cultured in vitro. We hypothesize that the inclusion of HA within PDLLA-PEG would result in a controlled release of loaded TGF-β3 and lead to a robust cartilage formation without the use of TGF-β3 in the culture medium. ELISA analysis showed that TGF-β3 release was effectively slowed by HA incorporation, and retention of TGF-β3 in the PDLLA/HA scaffold was detected by immunohistochemistry for up to 3 weeks. By means of both in vitro culture and in vivo implantation, we found that sulfated glycosaminoglycan production was higher in PDLLA/HA groups with homogenous distribution throughout the scaffold than PDLLA groups. Finally, with an optimal loading of TGF-β3 at 10 μg/mL, as determined by RT-PCR and glycosaminoglycan production, an almost two-fold increase in Young’s modulus of the construct was seen over a 4-week period compared to TGF-β3 delivery in culture medium. Taken together, our results indicate that direct loading of TGF-β3 and stem cells in PDLLA/HA has the potential to be a one-step point-of-care treatment for cartilage injury.

1. Introduction

Articular cartilage is a vital component for the smooth, painless, and functional movement of joints. However, it remains one of the few tissues in the body with minimal ability to self-repair after traumatic damage, disease, or aging-induced thinning.[1] This characteristic is especially important given the rates of osteoarthritis (OA) in the general population – 27 million people are affected each year with a cost of approximately 89.1 billion yearly to the healthcare system in the U.S.[2,3] This debilitating disease is accompanied by pain, compromised mobility, and a decrease in functional activity. While the pathogenesis of this degenerative joint disease is still unclear, effective early treatment of focal chondral defects, which are likely to progress to OA, would be a large step forward in tackling this disease.[4,5]

Several current surgical techniques, such as autograft, allograft, and microfracture, exist to address cartilage defects, but each has its own limitations.[6] Autografts rely on the transplantation of cartilage or an osteochondral plug from a non-weight-bearing region of cartilage to fill in the defect site, which leads to donor site morbidity and limited tissue sources. On the other hand, allograft utilizes the same method with tissues from another donor, but is complicated by immune rejection, potential disease transfer, and limited integration to host tissue. Lastly, microfracture involves drilling into the subchondral bone to release marrow to speed endogenous healing, but this method generally results in the formation of fibrocartilage that has functionally inferior biological and mechanical properties and can be painful for the patient.[7]

In addition to these surgical approaches, regenerative techniques such as autologous chondrocyte implantation (ACI) and matrix-assisted ACI (MACI) have been employed to effect cartilage repair.[8,9] These techniques both utilize the harvest of autologous chondrocytes from non-weight-bearing regions of articular cartilage, followed by expansion ex vivo and subsequent implantation into the defect site. In the case of ACI, the implant is secured with a periosteal flap, while in MACI a collagen sponge is used to provide the scaffolding for the cells. These techniques have been met with some success, but major limitations involve the de-differentiation of chondrocytes during in vitro culture as well as the formation of an inferior fibrocartilaginous tissue.

More recently, attention has shifted to the use of tissue engineered constructs to effect cartilage regeneration.[1,10] Within these designs, three major components are utilized: cells, growth factors, and scaffolds. On the cellular side, mesenchymal stem cells (MSCs) in particular have garnered considerable interest in view of their multi-lineage differentiation potential, including cartilage, bone, and adipose, among others.[11–13] Of the available scaffolds, hydrogels have been the most popular, given their biocompatibility, ease of application, and ability to conform to defect sites.[14,15] Lastly, among growth factors, transforming growth factor beta (TGF-β) is known to be essential for the induction and maintenance of a chondrocyte phenotype.[16] Thus, the application of MSCs in hydrogels utilizing TGF-β has been widely studied in a variety of approaches to create neo-tissue.

Among current designs, challenges still remain in allowing for point-of-care repair of cartilage defects. It has been shown that material stiffness aligning with the properties of native tissue allows for enhanced differentiation into the desired lineages.[17–20] Hydrogels are generally limited in mechanical strength and are not suitable to bear load instantaneously after implantation, which leads to a mismatch between the physical and physiological properties of the hydrogel and the surrounding native tissue.[14,15] We recently reported on the development of a hydrogel, poly-D,L-lactide/poly(ethylene glycol)/poly-D,L-lactide (PDLLA-PEG 1000), referred to simply as PDLLA here, that possessed physiologically relevant mechanical properties while retaining a chondroinductive environment.[21] However, we found that despite the high compressive moduli of the original construct, cells were not able to secrete matrix in a way that allowed for mechanical strength retention over time. Indeed, in a high cell-density culture, the periphery of the scaffold showed stronger matrix deposition than the center, likely due to the low diffusion of TGF-β through the entire scaffold.[21] In addition, there was no facile method to supply the necessary growth factor, TGFβ, for implantation in vivo.

In order to address these issues, we turned our attention towards hyaluronic acid (HA), a major component of the native cartilage matrix. HA is the only non-sulfated glycosaminoglycan (GAG) found in cartilage, and its presence has been shown to promote GAG deposition within hydrogels.[22,23] It is noteworthy that HA, by virtue of its many polar groups, binds and interacts with multiple growth factors, [24,25] and possesses slow release capabilities for a number of bound growth factors.[26,27] Thus, we hypothesized that addition of HA into the PDLLA, coupled with one-step loading of both TGF-β3 and cells, would represent a more physiologically relevant, tissue-like construct, exhibiting controlled release of TGF-β3 and robust chondrogenic differentiation of the encapsulated cells without the use of exogenous TGF-β3 supplemented in the culture medium. We also hypothesized that such a fabrication approach would lead to retention of scaffold mechanical strength by means of cellular GAG deposition. In this study, we tested the incorporation of HA into our PDLLA material, and compared the efficacy of direct loading of TGF-β3 versus supplementation in culture medium. Overall, our results showed that direct loading of TGF-β3 in a physiologically relevant PDLLA/HA hydrogel was a promising method to for in vivo cartilage tissue engineering.

2. Materials and Methods

All chemicals were purchased from Sigma Aldrich (St. Louis, MO, USA) unless otherwise specified. All procedures were performed with the approval form the Institutional Animal Care and Use Committee (IACUC) at the University of Pittsburgh.

2.1. Human bone marrow-derived mesenchymal stem cells (hBMSCs)

Isolation of hBMSCs from human specimens was approved by the Institutional Review Board (IRB) at our institutions (University of Pittsburgh and University of Washington – IRB# PRO-13030393).[28] First, trabecular bone was cored out from the femoral head using a curette or rongeur and sieved through 40-μm mesh screens to remove the remaining debris. Cells were then pelleted by centrifugation (300g, 6 min). After rinsing, cells were re-suspended in MSC growth medium (GM, α-MEM containing 10% fetal bovine serum (FBS, Invitrogen, Carlsbad, CA), 1.5 ng/mL FGF-2 (RayBiotech, Norcross, GA), and 1% antibiotics-antimycotics (Life Technologies, Carlsbad, CA)), and plated into 150 cm² tissue culture flasks. After 4 days, cells were washed with phosphate-buffered saline (PBS) and fresh GM was added. The medium was changed every 3 to 4 days. Once 70 to 80% confluence was reached, cells were detached with 0.25% trypsin-EDTA (Life Technologies) and passaged. MSC populations isolated from individual patients were routinely validated as capable of osteogenic, adipogenic, and chondrogenic differentiation (data not shown). Cells at passage 4 were used.

2.2. Synthesis of methacrylated PDLLA-PEG 1000 and hyaluronic acid (HA)

Preparation of mPDLLA-PEG 1000 was performed in the following steps.[21] Briefly, 50 g of PEG (1 kDa molecular weight) was placed into a 250 mL beaker and subjected to 600W microwave irradiation for 3 min. Subsequently, 3.5 g (2.80 mL) of stannous octoate [Sn(Oct)₂] was added to the molten PEG followed by addition of 7.2 g D,L-lactide. The mixture was briefly swirled to mix the contents and then subjected to 600W microwave irradiation for 1 min. The initial PDLLA-PEG 1000 polymer was precipitated in 500 mL cold isopropanol and dried under vacuum for 2 days. The dry polymer was placed into a dry 500 mL round bottom flask and dissolved in 100 mL dichloromethane (DCM) followed by addition of three equivalents of trimethylamine (TEA, ~5.25 mL) and three equivalents of methacrylic anhydride (MA, ~5.60 mL). The reaction mixture was placed under Argon gas and allowed to stir at room temperature for 7 days. After completion of the reaction, the mixture was precipitated into diethyl ether. For further purification, the macromer was re-dissolved in minimal amounts of chloroform and re-precipitated in diethyl ether. In addition, the two precipitation steps (isopropanol and diethyl ether) included 1 hour of cooling at −20°C after initial precipitation to allow for the product to completely precipitate out.

Methacrylated HA (MeHA) macromer was synthesized by adding methacrylic anhydride (Sigma, St. Louis, MO, USA) to HA solution (1% w/v, MW~60kDa, Lifecore, Chaska, MN), and incubated in ice overnight. The pH was adjusted regularly to 8.0, using 10 N sodium hydroxide. Macromer solution was purified via dialysis (3.5K MWCO) against deionized water for 48–72 h with more than 10 changes of water. The final product was dried by lyophilization and stored in a desiccator before use.

2.3. Fabrication of hBMSCs seeded hydrogel constructs

Solutions of polymer were prepared in HBSS (HyClone, Logan, Utah) in the following concentrations: (1) mPDLLA-PEG 1000 + mHA. mPDLLA-PEG 1000 was 22% (w/v), mHA was 3% (w/v). (2) mPDLLA-PEG 1000. mPDLLA-PEG 1000 was 25% (w/v). Therefore, final mass concentration was the same for both groups. MSCs at passage 4 at 90% confluency were trypsinized and pelleted. An appropriate amount of uncured hydrogel solution was used to resuspend cell pellets at 50×10⁶ cells/mL density. TGF-β3 (Peprotech, Rocky Hill, NJ, USA) was loaded into hydrogel/cell suspension at concentrations of 2, 5, 10 and 20 μg/mL (80, 200, 400 and 800 ng total TGF-β3 for each construct). The cell/hydrogel suspension was then transferred into a silicone mold (5 mm diameter; 2 mm thickness), and photopolymerized with UV light illumination (395nm) for 2 minutes, with 0.15% lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) added as the photoinitiator.[29]

For medium-supplemented groups, the final constructs were subjected to a rotating culture in bioreactors described previously by Radtke et al.[30] with chondrogenic medium (CM, high-glucose Dulbecco’s modified Eagle medium supplemented with 1% antibiotics-antimycotics; 0.1μM dexamethasone (Sigma); 50 μg/mL ascorbate 2-phosphate (Sigma); 40μg/mL L-proline (Sigma); 10 μg/mL ITS+ (Thermo Fisher, Waltham, MA, USA); TGF-β3 (10 ng/mL)) for up to 4 weeks. Pre-loaded groups (identified by S# in the manuscript) were subjected to rotating culture with basal medium (chondrogenic medium without TGF-β3). For in vivo study, constructs were first cultured in chondrogenic medium for 2 weeks and then intramuscularly implanted into hind legs of female Severe Combined Immunodeficiency mice (7–8 weeks old, CB17/Icr-Prkdc^scid/IcrIcoCrl, Charles River Laboratories; Wilmington, MA).[31] All experiments were performed in accordance with relevant guidelines and regulations, under the protocols approved by University of Pittsburgh Institutional Animal Care and Use Committee.

2.4. TGF-β3 release study

TGF-β3-loaded scaffolds (described above without cells) were incubated in 0.5 mL of 0.5% BSA solution in 1.5 mL pre-coated tubes (5% BSA for 1 hour) and placed on a microplate shaker (PMS-1000, Grant-Bio, Cambridge, UK) within an incubator set at 37 °C. Solution was collected at different time points for enzyme linked immunosorbent assay (ELISA) followed by adding fresh 0.5 mL 0.5% BSA into the tubes. The ELISA assay was carried out according to the manufacturer’s instructions (Human TGF-β3 Pre-Coated ELISA Kit, BioGems, Westlake Village, CA, USA). Using a microtiter plate reader (Synergy HT, BioTek, Winooski, VT, USA) the well plate was then read at an absorbance of 450 nm with correction at 570 nm. The readings were converted to a concentration using a standard curve generated.

2.5. Mechanical testing

Mechanical testing of constructs was conducted with a mechanical tester (Bose Electroforce model 3230 Series II). Briefly, the cylindrical scaffolds were placed between the compressive motor and the load cell and subjected to 10% compression (0.2 mm) at 0.01 mm/s. The stress-strain curve was then plotted, and the linear area was used to calculate the compressive modulus of the constructs.

2.6. Biochemical analyses

To determine the GAG content in constructs, samples were homogenized and digested in papain solution (125 μg/mL papain, 50mM sodium phosphate buffer, 2 mM N-acetyl cysteine (Sigma), pH 6.5), and incubated overnight at 60°C. Each construct (around 40–50 μl) was treated with 500 μl of papain buffer. Afterwards, the suspension was cleared by centrifugation (12000g for 10 minutes) and the supernatant was collected. The 1,9-dimethylmethylene blue dye-binding assay (Blyscan, Biocolor, United Kingdom) was used to determine GAG content against a standard curve of chondroitin-6-sulphate. Picogreen dsDNA assay (Molecular Probes, Tarrytown, NY, USA) was used to measure DNA content. Cell-free PDLLA and PDLLA/HA samples were used as blanks (baseline measurements for normalization) for the respective groups. Of note, the cell-free PDLLA/HA samples also had similar readings to process control blanks (samples derived from only digestion buffer). As such, HA has negligible to no contribution to the GAG readings obtained through our assay.

2.7. RNA isolation and gene expression analysis

Total RNA was obtained by homogenizing all samples in Trizol Reagent (Invitrogen) and following the protocol of the Rneasy Plus Mini Kit (QIAGEN, Germantown, MD, USA). Total RNA concentration was determined using a Nanodrop 2000c Spectrophotometer (Thermo Fisher). Reverse transcription was performed by following the instructions of SuperScript™ IV VILO™ Master Mix (Invitrogen). Polymerase chain reaction (PCR) was performed on an Applied Biosystems real-time PCR system using SYBR Green Reaction Mix (Applied Biosystems, Foster City, CA, USA). Ribosomal protein L13a (RPL13a) was set as the housekeeping gene. Typical chondrogenic and hypertrophic genes, including collagen type II (COL2), aggrecan (AGG), collagen type X (COL10), matrix metallopeptidase 13 (MMP-13), and SOX9, were analyzed with data normalized to the housekeeping gene (RPL13a). The sequences of primers for each gene were listed in read Supplementary Table 1. The relative gene expression was calculated using the ^ΔΔCt method.

2.8. Histology and immunostaining

All hydrogels samples were fixed in 10% paraformaldehyde, then dehydrated using a graded ethanol series up to 100%, embedded in paraffin, and sectioned at a thickness of 6 μm following a standard histological procedure. Sections were stained with Alcian Blue dye (Rowley Biochemical, Danver, MA). For immunohistochemistry (IHC), sections were stained for targets of interest using the Vectastain ABC kit and NovaRED peroxidase substrate kit (Vector Labs, Burlingame, CA, USA). Primary antibodies against TGF-β3 (Abcam, Cambridge, MA) were used in this study. Incubation with primary antibodies was carried out at 4°C overnight followed by incubation with appropriate secondary antibodies. Images were acquired by OLYMPUS CKX41. For immunofluorescence (IF), primary antibody against human collagen type II (Abcam) was used to incubate sections at 4°C overnight, followed by incubation with appropriate secondary antibody. Vectashield Mounting Medium with DAPI (Vector Labs, Burlingame, CA, USA) was used to stain nuclei. Native human articular cartilage was used as the positive control (Figure S1), with the approval from Committee for Oversight of Research and Clinical Training Involving Decedents (CORID, #878). Images were acquired with an Olympus IX2-USB microscope.

2.9. Statistical analysis

All data were expressed as mean ± standard deviation. Statistical analysis was performed in Prism 7. Statistical analysis was performed using ANOVA with Tukey’s HSD post-hoc testing or one-tailed t-test where appropriate. A threshold of p < 0.05 was used to determine statistical significance.

3. Results

3.1. Hyaluronic acid incorporation in PDLLA scaffolds allows for slow release of pre-loaded TGFβ3

In order to achieve a facile one-time application of TGF-β3 for cartilage repair, we tested supplementation of hyaluronic acid in PDLLA, which was expected to release TGF-β3 in a controlled manner and support chondrogenic differentiation of hBMSCs. Figure 1A illustrates the fabrication of PDLLA (25% w/v) and PDLLA/HA (22%/3% w/v) with pre-loaded TGFβ3. Cells, polymer, and concentrated TGF-β3 were homogeneously mixed and photopolymerized in a single process. The results in Figure 1B showed the cumulative release of TGF-β3 over a 21-day course in PDLLA/HA and PDLLA scaffolds, with the culture medium removed completely at each sampling. PDLLA/HA had a significantly lower burst release within the first 24 hours (6.43 ± 0.38 ng versus 28.05 ± 6.69 ng, p = 0.0126) and plateaued at a lower total release (22.59 ± 2.02 ng versus 49.34 ± 12.75, p = 0.0009). In addition, PDLLA/HA displayed a continuous TGF-β3 release profile throughout the time course of 21 days, whereas PDLLA unloaded most of the TGF-β3 cargo in the first 7 days. Interestingly, the variability in the PDLLA/HA group was also lower.

Figure 1. Construction of scaffolds and release profile of pre-loaded TGF-β3 over 21 days.

(A) Depiction of fabrication and culture of cell-loaded constructs. The top row represents the “with HA (w/HA)” (PDLLA/HA) group and the bottom is the “without HA (w/o HA)” (PDLLA) group (B) Cumulative release profile pre-loaded TGF-β3 over 21 days from scaffolds with or without HA. PDLLA/HA exhibited a significantly lower burst release over the first day (6.43 ± 0.38 versus 28.050 ± 6.69 ng, p = 0.0126) and plateaued at a significantly lower value at 21 days (22.59 ± 2.021 versus 49.34 ± 12.75 ng, p = 0.0009). N=3.

To further assess the persistence and localization of TGF-β3 in these scaffolds, immunohistochemical staining for TGF-β3 was performed on pre-loaded scaffolds between day 0 and week 3 post-fabrication. As shown in Figure 2A–D, at day 0, TGF-β3 staining (brown) was seen to cluster around cells to a greater degree in PDLLA versus PDLLA/HA, i.e., TGF-β3 distributed around the cells as well as throughout the material in the latter scaffold, whereas there was an absence of TGF-β3 staining within the material for the former.[21] Results from Figure S2 further showed that endogenous TGF-β3 production by cells did not contribute to the positive staining in the polymer. With longer culture time, from week 1 to week 3 (Figures 2E–J), staining was only weakly positive in the PDLLA group at week 1, and largely disappeared by weeks 2 and 3. In contrast, PDLLA/HA maintained strong staining at week 1 followed by weakly positive at week 2, and by week 3 displayed scarce staining. Taken together, the results in Figures 1B & 2 demonstrated that TGF-β3 was better retained in, as well as released at a slower rate from, the hyaluronic acid-incorporated PDLLA.

Figure 2. Immunostaining of TGF-β3 directly pre-loaded into PDLLA and PDLLA/HA scaffolds.

(A-D) High magnification images of PDLLA (A,B) and PDLLA/HA (C,D) scaffolds stained for TGF-β3 post-fabrication at Day 0. Note the positive staining within the scaffold material in the PDLLA/HA group. (E-J) Low magnification images of PDLLA (E-G) and PDLLA/HA (H-J) scaffolds stained for TGF-β3 from 1 to 3 weeks post-fabrication. Note the presence of strong TGF-β3 staining in PDLLA/HA group at up to 2 weeks versus very weak staining at 2 weeks for the PDLLA group. Scale bars: (A-D) 50 μm; (E-J) 100 μm.

3.2. Hyaluronic acid-incorporated PDLLA scaffolds display enhanced chondrogenesis of encapsulated hBMSCs

We next examined the biological and functional outcomes of the TGF-β3 pre-loaded scaffolds on the encapsulated hBMSCs. TGF-β3 pre-loading was performed at a final concentration of 5 μg/mL with simultaneous seeding of hBMSCs using PDLLA (w/o HA) or PDLLA/HA (w/HA) scaffolds, and the hydrogel constructs were then cultured for 28 days in chondrogenic medium, without supplementation of TGFβ3. Relative gene expression levels of AGG, COL2, and SOX9 (all normalized to RPL13A) are shown in Figure 3A–C. By a one-tailed t-test, significance was only observed for AGG, but COL2 was also close to the defined significance of p = 0.05 (p = 0.040, 0.057, and 0.25 for AGG, COL2, and SOX9, respectively). For GAG production and the mechanical property of Young’s modulus, shown in Figure 3D and 3E, PDLLA/HA was found to be significantly higher in both parameters (p = 0.0001 and p = 0.0026 with approximately 2.5-fold and 75 kPa difference, respectively). Indeed, PDLLA/HA Young’s modulus increased over the course of 4 weeks from day 0 (165 ± 3.6 kPa) (Figure 4B), as opposed to PDLLA which decreased from day 0 [21]. These data at 4 weeks demonstrated that the chondrogenic response of hBMSCs was greater in pre-loaded PDLLA/HA versus PDLLA, likely owing to the incorporation of HA and slowed release of TGFβ3. In fact, histology showed that GAG staining appeared to be distributed throughout the scaffold material in HA-containing scaffolds (Figure S3), which was not seen in PDLLA-only scaffolds.

Figure 3. RT-PCR, GAG, and mechanical analyses of hBMSC-seeded, TGFβ3-preloaded hydrogel constructs at 4 weeks of culture.

TGF-β3 was pre-loaded at a concentration of 5 μg/mL. (A-C) RT-PCR analysis of gene expression levels of AGG, COL2, and SOX9, normalized to that of RPL13A. (D) GAG production and (E) Young’s modulus of constructs after 4 weeks culture. *, p<0.05. *, p<0.005. ***, p<0.001. PDLLA/HA (w/HA) group displayed significantly higher COL2 expression (p = 0.040), GAG production (p = 0.0001) and Young’s modulus (p = 0.0026) than PDLLA only (w/o HA) group. N=3 in Figure 3A–D. N=4 in Figure 3E.

Figure 4. GAG production and Young’s modulus of PDLLA/HA scaffolds with varying TGF-β3 pre-load concentrations.

(A) GAG production and (B) Young’s modulus after 4 weeks of culture for hBMSC-seeded scaffolds pre-loaded with TGF-β3 at concentrations ranging from 2 μg/mL to 20 μg/mL (S-2 to S-20), compared to those exposed to TGF-β3 supplementation in the medium at the standard concentration of 10 ng/mL (M). Note that maximal GAG production and Young’s modulus values are observed at TGF-β3 pre-loading of 10 μg/mL, and are higher than those achieved through medium supplementation of TGF-β3 alone. D0, day 0 post-fabrication mechanical properties; M, TGF-β3 supplemented only in medium (none in scaffold); S-2, TGF-β3 pre-loaded in scaffold at concentration of 2 μg/mL; S-5, 5 μg/mL; S-10, 10 μg/mL; and S-20, 20 μg/mL. **, p<0.005 for S-10 compared to all groups except S-20. ***, p<0.0001 for S-10 compared to all groups except S-5. N=3.

3.3. TGF-β3 pre-loaded in PDLLA/HA displays a dose-dependent pro-chondrogenic effect and is superior to medium supplementation

The optimal dose of TGF-β3 pre-loading in PDLLA/HA was next determined, with the control being supplementation of TGF-β3 in culture medium at 10 ng/mL, the most commonly used method of chondrogenic stimulation of hBMSCs. [32] Of note, S# groups did not receive exogenous TGF-β3 supplementation in the medium. As shown in Figure 4A and 4B, GAG production and Young’s modulus outcomes of preloading TGF-β3 at concentrations ranging from 2 μg/mL to 20 μg/mL (designated S-2 to S-20) were compared to those for medium TGF-β3 supplementation (M). For both outcomes, the maximum pro-chondrogenic effect peaked at a concentration of 10 μg/mL. For GAG production, S-10 was significantly higher (p < 0.005) than all groups except S-20 (p = 0.21). For Young’s modulus, S-10 was significantly higher (p < 0.001) than all groups except S-5 (p = 0.058). For gene expression levels (Figure 5A–E), S-10 was again the optimal TGF-β3 loading concentration among those tested (p = 0.014, S-10 versus M for AGG; p = 0.003 and 0.0038 for S-10 versus M and S-10 versus S-20 for SOX9, respectively). Non-significant differences were found for the hypertrophy-associated genes MMP-13 and COL10 among all groups as well as for the chondrogenic gene COL2. Overall, the increased functional outcomes (GAG production and Young’s modulus) coupled with significantly increased chondrogenic gene expression (AGG and SOX9) and non-significant changes in hypertrophy-associated genes (MMP-13 and COL10) established PDLLA/HA with 10 μg/mL pre-loaded TGF-β3 as a superior alternative to the standard medium supplementation protocol of 10 ng/mL TGFβ3.

Figure 5. Gene expression profiles of hBMSC-seeded PDLLA/HA constructs with varying concentrations of pre-loaded TGFβ3.

(A-E) RT-PCR analysis of gene expression levels of AGG, COL2, and SOX9, MMP-13, and COL10, after 4 weeks of culture. Relative expression levels are normalized to that of RPL13A, as well as that of a control group (PDLLA with no TGF-β3 stimulation, not shown). M, TGF-β3 supplemented only in medium (none in scaffold); S-2, TGF-β3 pre-loaded in scaffold at concentration of 2 μg/mL; S-5, 5 μg/mL; S-10, 10 μg/mL; S-20, 20 μg/mL. *, p<0.05, and **, p<0.005. N=3.

3.4. PDLLA/HA with pre-loaded TGF-β3 promotes uniform GAG distribution in vivo

Lastly, we examined the chondrogenic capabilities of TGF-β3 pre-loaded scaffolds with encapsulated hBMSCs implanted in vivo. After fabrication, the hydrogel constructs were cultured for 2 weeks in vitro followed by 2 weeks of subcutaneous implantation in vivo for a total duration of 4 weeks. This time frame was established to allow the hydrogel-encapsulated hBMSCs treated with TGF-β3 supplemented in medium to undergo chondrogenesis before implantation (the pre-culture period and the total time period were kept the same for all groups). The retrieved implants were examined histologically with Alcian blue staining and collagen type II immunostaining (Figure 6). As seen in Figure 6A,B, in the TGF-β3 pre-loaded PDLLA construct, Alcian blue positive GAG staining was localized primarily as clusters solely around cells, and not present in the biomaterial. In Figure 6C,D, the TGF-β3 pre-loaded PDLLA/HA groups showed staining clustering around cells as well as distributed within the biomaterial, with a homogeneous pattern from the margin to the center of the implant. In contrast, when PDLLA/HA constructs were treated only with TGF-β3 supplemented in culture medium, shown in Figure 6E,F, Alcian blue GAG staining was strong in the peripheral margins but weak in the center of the scaffold, with some staining throughout the biomaterial. These observations were also seen in vitro (Figure S3). Corresponding higher magnification views of the strong versus weak staining and GAG deposition patterns within the construct are shown in Figures 6G–L.

Figure 6. Alcian blue staining of in vivo implants of hBMSC-seeded PDLLA and PDLLA/HA constructs pre-loaded with TGFβ3.

Scaffolds were cultured for two weeks in vitro followed by subcutaneous implantation in vivo and subsequently harvested at two weeks. (A-F) Low magnification images of scaffold margins and centers for PDLLA (A,B), PDLLA/HA (C,D), and PDLLA/HA constructs exposed to TGF-β3 supplemented in medium during the in vitro culture phase (E,F). Bar = 100 μm. (G-L) Corresponding higher magnification images of constructs shown in (A-F). The TGF-β3 pre-loaded PDLLA/HA group exhibit uniform Alcian blue staining throughout the construct (white arrows), while the PDLLA/HA group exposed to TGF-β3 in the culture medium showed much stronger staining at the margins compared to the center (black arrows). The PDLLA group shows Alcian blue staining only around cell clusters. Bar = 50 μm.

Collagen type II immunofluorescent staining (green) shown in Figure 7 further demonstrated the advantage of pre-loading versus medium supplementation of TGF-β3 in terms of promoting hBMSC chondrogenesis. Low magnification and corresponding high magnification views of the collagen type II immunostaining are shown in Figure 7A–F and 7G–L, respectively. Similar to the distribution of Alcian blue GAG staining, with TGF-β3 pre-loading, collagen type II was found clustered around cells in PDLLA (Figure 7A,B), while in PDLLA/HA (Figure 7C,D), it was more uniform and widespread in the centers. The immunostaining patterns in both TGF-β3 pre-loaded groups were remarkably different from that in the TGF-β3 medium supplementation group (Figure 7E,F). The continuous TGF-β3 stimulation with TGF-β3 pre-loading allowed for deposition of collagen type II, while the withdrawal of TGF-β3 in the medium supplementation group resulted in stunted collagen type II production by the loaded cells. This finding strongly suggests that in the subcutaneous in vivo environment, collagen type II production did not occur without exogenous TGF-β3 supplementation. Taken together, the findings presented in Figures 6 and 7, clearly demonstrate that the pro-chondrogenic effects of pre-loaded TGF-β3 continued under in vivo conditions, making this method suitable for applications in vivo.

Figure 7. Collagen type II immunofluorescence of in vivo implants of hBMSC-seeded PDLLA and PDLLA/HA constructs pre-loaded with TGFβ3.

Scaffolds were cultured for two weeks in vitro followed by subcutaneous implantation in vivo and subsequently harvested two weeks later. Staining: collagen type II, green; nucleus, blue. (A-F) Low magnification images of scaffold margins and centers for (A,B) PDLLA, (C,D) PDLLA/HA, and (E,F) PDLLA/HA with TGFβ3, after 2-week in vitro phase culture. (G-L) Corresponding higher magnification images of scaffolds from (A-F). The TGF-β3 pre-loaded PDLLA and PDLLA/HA groups demonstrate uniform collagen type II immunostaining throughout the construct while the PDLLA/HA group exposed to TGF-β3 only in the culture medium is largely devoid of collagen type II staining, likely due to the absence of TGF-β3 after its withdrawal from the culture medium and lack of the molecule in the in vivo tissue environment. Bar = 100 μm (A-F), or 50 μm (G-L).

4. Discussion

The aim of this study was to develop a potential point-of-care biomaterial scaffold possessing physiologically relevant mechanical and cell-friendly properties for cartilage tissue engineering, while having the ability to retain mechanical strength over time. In addition, we sought to overcome the challenge of limited TGF-β3 diffusion at high cell densities that we observed in our previous study.[21] To address these issues, we have specifically examined the effect of adding hyaluronic acid to physiologically relevant PDLLA scaffolds. HA inclusion was found to slow release of pre-loaded TGF-β3 and also promote the deposition of GAGs by the encapsulated hBMSCs in a uniform manner throughout the biomaterial matrix. In contrast, in PDLLA scaffolds without HA, pre-loaded TGF-β3 was released quickly, and resulted in GAG deposition clustered only around the encapsulated cells. On the other hand, treatment of the cell-loaded construct with TGF-β3 only as a medium supplement led to uneven GAG distribution more prominent in the margins of the material, even in PDLLA/HA constructs. Overall, pre-loading TGF-β3 in HA-containing PDLLA hydrogel allowed for increased GAG deposition and mechanical strength retention superior to TGF-β3 supplementation in the medium. These results strongly suggest the utility of this strategy for the production of a mechanically strong scaffold suitable for point-of-care cartilage tissue engineering.

HA contains many hydrophilic groups that can interact with charged amino acid residues on growth factors via electrostatic interactions.[24,25] As such, we observed that, with HA inclusion, pre-loaded TGF-β3 was retained in PDLLA for up to three weeks, whereas in PDLLA alone, TGF-β3 release displayed a burst profile, with complete release from the scaffold within a week (Figure 1). This effect was in accordance with previous studies that used forms of HA to achieve slow release of BMP-2 and FGF-2, both highly charged growth factors, over a course of 4 weeks.[27,26] Indeed, retention of TGF-β3 within the scaffold was observable by immunohistochemistry (Figure 2), which was not due to endogenous TGF-β3 production (Figure S2). Of note, post-fabrication and over the course of 3 weeks, TGF-β3 was characteristically absent within the material itself in the PDLLA group. Instead, the growth factor was observed to solely cluster around cells as opposed to observations in the PDLLA/HA group where TGF-β3 both clustered around cells and distributed within the material. This is likely due to the fact that poly(ethylene glycol) (PEG) is thought to repel proteins, thus cells outcompete the polymer for binding to TGFβ3.[33] Interestingly, without the introduction of exogenous TGF-β3, single MSCs did not express TGF-β3 (Figure S2B). However, when they formed a cluster, positive staining was observed (Figure S2A, B). Therefore, the high TGF-β3 surrounding the cells and clusters in Figure 2A–D may be due to the combination of both exogenous and endogenous TGF-β3.

Although PDLLA/HA retained TGF-β3 over a longer period than PDLLA alone, at 4 weeks chondrogenic gene expression levels of AGG and COL2 were higher (approximately 2- and 6-fold, respectively), while that of SOX9 was unchanged (Figure 3). Sox9 is an early marker for chondrogenesis, and the similar levels suggest that by 4 weeks TGF-β3 signaling is absent in both groups. However, interestingly, GAG production and Young’s modulus were both significantly higher (2.5× and 75 kPa higher with p<0.001 and p<0.005, respectively). HA is known to promote chondrogenesis via interaction with the CD44 receptor,[22,34] which could explain the increased GAG production. On the other hand, HA could also be binding charged GAGs more effectively, thus allowing GAGs to distribute throughout the material as opposed to clustering around cells (Figure 6A,B).[25,35] This effect has also been reported by Levett et al. who found collagen type II and aggrecan staining to be more homogeneously distributed throughout HA-loaded material when compared to chondroitin sulfate or gelatin.[36] The accumulation of GAGs is also likely the cause for increased compressive moduli, as the deposited GAGs can reinforce the degrading scaffold matrix. Indeed, chondrocyte-encapsulated in HA hydrogels are able to significantly increase the compressive modulus of the construct over time.[37]

This increased chondrogenic effect with HA was also found to be dependent on the initial loading concentration of TGF-β3 (Figures 4,5). Interestingly, GAG production and Young’s modulus were found to achieve a peak value at loading of TGF-β3 10 μg/mL. This peaking effect has been seen in a previous study where cellular proliferation was found to peak at 5 μg/mL after 6 days in vitro culture, indicating that the dose response of chondrogenesis is not linear.[38] Perhaps the most surprising finding, however, was that both GAG and Young’s modulus were significantly higher in the S-10 group (PDLLA/HA with 10 μg/mL pre-loaded TGFβ3) compared to the M group (TGF-β3 supplemented in medium), with approximately a 2-fold and 1.67-fold increase, respectively. We hypothesized this outcome to be based on the incomplete diffusion of TGF-β3 into the center of the scaffold at high cell densities, as we have previously reported.[21] Indeed, results in Figures 6 and 7 demonstrated that after 2 weeks in vitro culture followed by in vivo implantation, PDLLA/HA with pre-loaded TGF-β3 and encapsulated hBMSCs displayed uniformly strong Alcian blue staining for GAGs and immunofluorescent staining for collagen type II in both the margin and center of the scaffold, while PDLLA/HA treated only with TGF-β3 as a medium supplement displayed strong Alcian blue staining in the margin but weak staining in the center, as well as absence of collagen type II staining. This dissimilarity in GAG distribution likely led to the disparity in GAG and Young’s modulus seen in Figure 4. Gene expression profiling (Figure 5) confirmed this peaking effect in AGG and SOX9, but failed to reach significance in most cases. Importantly, we noted that pre-loading TGF-β3 did not increase the expression levels of hypertrophy-associated genes (MMP13 and COL10), compared to TGF-β3 medium supplementation, nor were gene expression levels of chondrogenic genes (AGG, COL2, SOX9) lower in pre-loaded samples. Thus, pre-loading TGF-β3 in PDLLA/HA scaffolds represents a viable alternative to medium supplementation of TGFβ3.

We recognize that the loading concentration of TGF-β3 in our method (2–10μg/ml) is many times higher than that found in medium supplementation (10ng/ml). However, a major challenge of the in vivo microenvironment is the fact that exogenous or continuous supplementation of TGF-β3 is difficult and undesirable. As such, there are numerous methods to allow for controlled delivery of growth factors for cartilage tissue engineering.[39] We have shown that a simple one-time pre-load of TGF-β3 during fabrication was able to result in induction of chondrogenesis to a similar or greater degree than the widely utilized 10 ng/mL TGF-β3 supplementation in culture medium over a course of 4 weeks (Figures 4–7), although the total TGF-β3 usage might be higher. [32] In addition, mechanical strength in the physiological range of cartilage stiffness was maintained in a PEG-based hydrogel, as opposed to previous studies where mechanically strong hydrogels significantly degraded over the culture period.[21,40] We have shown here that this strategy thus has high potential for point-of-care in vivo cartilage tissue engineering to allow for a faster return to weight bearing. Future studies will now focus on further refining this one-step approach and examining metrics of hyaline cartilage formation after long term culture, such as chondrocyte hypertrophy (a well-known consequence of TGF-β3 treatment that hinders the formation of articular cartilage),[41–43] and integration with the surrounding tissues in vivo.

5. Conclusion

In this study, we reported that inclusion of HA into hBMSC-seeded PDLLA-PEG 1000 hydrogel with TGF-β3 pre-loaded in the constructs could better control the release of TGF-β3 compared to hydrogels without HA inclusion. In addition, under optimal conditions, HA constructs with pre-loaded TGF-β3 enhanced hBMSC chondrogenesis, in terms of both chondrogenic gene expression and GAG production, and achieved significantly higher compressive moduli compared to medium TGF-β3 supplemented counterparts. Implantation in vivo confirmed that pre-loaded HA constructs could also maintain GAG and collagen type II deposition, demonstrating the potential of this method for in vivo use. Future studies will focus on evaluating the integration of such neocartilage constructs with native cartilage in vivo. Taken together, our results strongly suggest that direct loading of TGF-β3 and stem cells in PDLLA/HA has the potential to be a one-step point-of-care treatment for cartilage injury.

Statement of Significance.

Stem cell-seeded hydrogels are commonly used in cell-based cartilage tissue engineering, but they generally fail to possess physiologically relevant mechanical properties suitable for loading. Moreover, degradation of the hydrogel in vivo over time further decreases mechanical suitability of the hydrogel due in part to lack of TGF-β3 signaling. In this study we demonstrated that incorporation of hyaluronic acid (HA) into a physiologically stiff PDLLA-PEG hydrogel allowed for slow release of one-time pre-loaded TGFβ3, and when loaded with adult mesenchymal stem cells and cultured in vitro, resulted in higher chondrogenic gene expression and constructs of significantly higher mechanical strength, compared to constructs cultured in conventional TGFβ3-supplemented medium. Similar effects were also observed in constructs implanted in vivo. Our results indicate that direct loading of TGF-β3 combined with HA in physiologically stiff PDLLA-PEG hydrogel has the potential to be used for one-step point-of-care treatment of cartilage injury.