Glucose Availability Affects Extracellular Matrix Synthesis During Chondrogenesis In Vitro

Abstract

Understanding in vitro chondrogenesis of human mesenchymal stem cells (hMSCs) is important as it holds great promise for cartilage tissue engineering and other applications. The current technology produces the end tissue quality that is highly variable and dependent on culture conditions. We investigated the effect of nutrient availability on hMSC chondrogenesis in a static aggregate culture system by varying the medium-change frequency together with starting glucose levels. Glucose uptake and lactate secretion profiles were obtained to monitor the metabolism change during hMSC chondrogenesis with different culture conditions. Higher medium-change frequency led to increases in cumulative glucose uptake for all starting glucose levels. Furthermore, increase in glucose uptake by aggregates led to increased end tissue glycosaminoglycan (GAG) and hydroxyproline (HYP) content. The results suggest that increased glucose availability either through increased medium-change frequency or higher initial glucose levels lead to improved chondrogenesis. Also, cumulative glucose uptake and lactate secretion were found to correlate well with GAG and HYP content, indicating both molecules are promising biomarkers for noninvasive assessment of hMSC chondrogenesis. Collectively, our results can be used to design optimal culture conditions and develop dynamic assessment strategies for cartilage tissue engineering applications.

Impact statement

In this study, we investigated how culture conditions, medium-change frequency and glucose levels, affect chondrogenesis of human mesenchymal stem cells in an aggregate culture model. Doubling the medium-change frequency significantly increased the biochemical quality of the resultant tissue aggregates, as measured by their glycosaminoglycan and hydroxyproline content. We attribute this to increased glucose uptake through the glycolysis pathway, as secretion of lactate, a key endpoint product of the glycolysis pathway, increased concurrently. These findings can be used to design optimal culture conditions for tissue engineering and regenerative medicine applications.

Introduction

Human mesenchymal stem cells (hMSCs) are a promising cell source for cartilage tissue engineering.^1–3 Diseases like osteoarthritis can be treated if the therapeutic potential of hMSCs is exploited to regenerate patient-specific cartilage tissue.^4–6 However, the end tissue quality in hMSC chondrogenesis is highly variable and dependent on culture conditions such as media composition, mass transport, and donor variability.^7–9

During chondrogenesis, hMSCs undergo several phenotypic changes: condensation, differentiation, maturation, and terminal differentiation, to become chondrocytes. Various growth and differentiation factors, transcription factors, and extracellular matrix (ECM) proteins are involved in each stage.^10,11 In vivo, hMSCs and chondrocytes are located in different metabolic niches. While hMSCs reside as perivascular cells in bone marrow,¹² articular chondrocytes reside in hyaline cartilage, where nutrient mass transport is very limited due to the avascularity of the tissue.^13,14 Therefore, it is highly likely that there are changes to cell metabolism during hMSC chondrogenesis. Understanding culture condition requirements to address the metabolic demands of hMSCs during chondrogenesis can be important to facilitate the differentiation process and improve the end quality of the tissue. However, current perfusion^15–17 as well as static^18–22 culture conditions are not designed to address such demands. More importantly, the optimal exogenous environment for in vitro chondrogenesis is still underinvestigated.²³

In the classic aggregate model of chondrogenesis, starting medium composition and frequency of medium change are two major culture conditions that can easily be manipulated to provide a better environment for hMSC chondrogenesis. We and others^8,24–26 have shown that glucose plays an important role in hMSC chondrogenesis as it provides both the sources for energy in ATP and the carbon skeletons for biosynthesis of glycosaminoglycans (GAGs), which are key macromolecules in cartilage ECM. Not surprisingly, due to the large ECM content of cartilage, amino acids are also crucial biomolecules in hMSC chondrogenesis; they play a key role in the synthesis of cartilage ECM proteins such as type II collagen.

In addition to nutrients and nutrient levels in culture medium, for static culture conditions, frequency of medium change can affect the end-tissue quality as well. A change of medium change can provide fresh nutrients (i.e., glucose and amino acids) and remove waste (i.e., lactate); however, endogenous soluble factors, which can serve autocrine factors for chondrogenesis as well as immature ECM molecules (e.g., procollagen), may also be removed. Therefore, balancing the nutrient supply with disposal of waste and autocrine and immature ECM molecules is important for hMSC chondrogenesis.

In this study, we investigated the effects of different initial glucose levels as well as different medium-change methods (continuous, static with different medium-change frequency) on hMSC chondrogenesis using an aggregate culture system. We monitored levels of glucose, lactate, amino acids and proteins during chondrogenesis. Our results show that glucose uptake and lactate secretion increased, and amino acid and protein uptake/secretion varied during chondrogenesis. Furthermore, static culture systems with more frequent medium changes led to significantly greater GAG and hydroxyproline (HYP) content; this occurred concurrently with enhanced glucose uptake and lactate secretion denoting increased availability of nutrients as the key reason for the improvement. Increasing the frequency of medium change led to prevention of failure—as defined histologically—of chondrogenesis. Collectively, these results can be used to design optimal culture conditions for engineered cartilage.

Materials and Methods

Materials

Cell culture medium (Dulbecco's modified Eagle's medium [DMEM] with 4.5 g/L—DMEM-HG, 1.0 g/L—DMEM-LG, and 0 g/L—DMEM-NG), trypsin, l-glutamine, antibiotic/antimycotic solution (10⁵ U/mL penicillin G sodium, 10 mg/mL streptomycin sulfate, and 25 μg/mL amphotericin B in 0.85% saline), nonessential amino acids, dexamethasone, sodium pyruvate, phosphate-buffered saline (PBS), and nuclease-free water were purchased from Invitrogen (Carlsbad, CA). Fetal bovine serum (FBS; Life Technologies, Carlsbad, CA) was lot selected (Lot# 1805387).²⁷ Transforming growth factor-β1 (TGF-β1) and human fibroblast growth factor-2 (FGF-2) were obtained from Peprotech (Rocky Hill, NJ). Insulin-transferrin-selenium (ITS)+ Premix (20 mL solution of human recombinant insulin, human transferrin [12.5 mg each], selenous acid [12.5 μg], BSA [2.5 g], and linoleic acid [10.7 mg]) was a product of Becton Dickinson (Franklin Lakes, NJ). Ethylenediaminetetraacetic acid (EDTA), hydrochloric acid (HCl), sodium hydroxide (NaOH), sodium phosphate, sodium chloride, disodium phosphate, sodium acetate, isopropanol, bovine serum, 16% phosphate-buffered formaldehyde solution, optimal cutting temperature (OCT), and clear 96-well plates were obtained from Fisher Scientific (Pittsburgh, PA). Ascorbate-2-phosphate (A2P) was purchased from Wako (Richmond, VA). Safranin-O, HYP, papain, cysteine, 4-(dimethylamino) benzaldehyde, perchloric acid 70% reagent grade, hydrogen peroxide (H₂O₂), sulfuric acid (H₂SO₄), calf thymus DNA standard, cetylpyridinium chloride, copper(II) sulfate pentahydrate (CuSO₄), and Hoechst 33258 were obtained from Sigma-Aldrich (St. Louis, MO). Chondroitin sulfate C (from shark cartilage) was from Seikagaku America (East Falmouth, MA). A Contour Next USB Blood Glucose Monitoring System and glucose test strips were acquired from Bayer AG (Leverkusen, Germany). Nitrocellulose membrane and dot-blot apparatus were bought from Bio-Rad (Hercules, CA). Ninety-six-well plates with ultraviolet (UV) transparent flat bottom and black wells (for DNA assay) were purchased from Corning (Kennebunk, ME). Hemocytometer was from Hausser Scientific (Horsham, PA). Ninety-six-well polypropylene plates with conical bottom were purchased from Evergreen Scientific (Los Angeles, CA). A spectrophotometer was obtained from Molecular Devices (Sunnyvale, CA). Type II collagen antibody was purchased from the Developmental Studies Hybridoma Bank (Iowa City, IA) and FITC (fluorescein isothiocyanate)-conjugated secondary antibody was obtained from MP Biomedicals (Solon, OH).

Cell culture

hMSCs were obtained from bone marrow. Bone marrow was procured from human donors by the Stem Cell Core Facility of the Case Comprehensive Cancer Center after obtaining informed consent as per an Institutional Review Board-approved protocol. The hMSCs were isolated using previously published methods.^22,28,29 Briefly, bone marrow isolates were diluted in complete medium (DMEM-LG supplemented with 10% FBS) and centrifuged to remove fat layer and supernatant. The cell pellet was resuspended and layered on top of a Percoll gradient and centrifuged. The dense MSCs at the bottom of the tube were plated in tissue culture dishes at a density of 1.7 million/cm². After primary culture, the cells were grown from a plating density of 15 × 10⁵ cells per T150 flask in complete medium supplemented with 10 ng/mL FGF-2 in a cell culture incubator kept at a temperature of 37°C and a humidified 95% air and 5% CO₂ environment. The medium was changed twice a week. After a week, the cells were typically 80–90% confluent.

Aggregate culture and chondrogenic induction

All the experiments in this work used passage 2 cells. For chondrogenesis induction, hMSCs from T-150 flasks were detached and made into aggregates as described before.^18–22 Briefly, cells were detached using 0.05% trypsin, neutralized in bovine serum, and counted using a hemocytometer. Chondrogenic differentiation medium was prepared with different glucose concentrations of 1, 2, 3, and 4.5 g/L by mixing DMEM-HG (high glucose) and DMEM-NG (no glucose) in proper proportion. The media were supplemented with ITS+ Premix (10 mL/1000 mL medium), 146 μM A2P, 2 mM l-glutamine, 0.1 mM nonessential amino acids, 1 mM sodium pyruvate, 10⁻⁷ M dexamethasone, 1% antibiotic/antimycotic, and 10 ng/mL TGF-β. Twenty-five thousand cells were seeded in 96-well, conical bottom, 300-μL polypropylene microplates at a volume of 200 μL/well. The plates were gently centrifuged for 5 min at 500 g to promote aggregation and cell–cell interaction.²¹ These plates were placed in a cell culture incubator kept at a temperature of 37°C in humidified 95% air and 5% CO₂ environment. The medium was replaced every 24 (medium-change frequency = 1 day⁻¹) or 48 h (medium-change frequency = 0.5 day⁻¹). Twenty-microliter samples were collected every day and frozen at −20°C for biomolecular analyses (see below).

Glucose analysis

Contour Next USB Blood Glucose Monitoring System (Bayer AG) was used to measure glucose content in the medium samples. Calibration curve was generated from DMEM of known glucose concentrations. The samples were thawed at room temperature for 1 h, vortexed, and centrifuged before analysis. A 3-μL sample was pipetted into a Petri dish one by one for analysis using test strips to minimize evaporation. Glucose uptake rates were calculated from the difference in glucose levels in the medium divided by the time interval and normalized to cell numbers.⁸

Lactate analysis

Lactate contents in the collected culture medium samples were analyzed according to a previously published method with some modifications.⁸ Frozen medium samples were thawed at room temperature, spun down using microcentrifuge, and vortexed. Five-microliter samples plus 5 μL of deionized water were added per well to a standard clear bottom 96-well plate in duplicates. One hundred ninety microliters assay cocktail containing 320 mM glycine, 3 mM NAD⁺, 320 mM hydrazine, and 20 U/mL fresh-made lactate dehydrogenase solution in water was added to each well. The plates were then sealed by low-evaporation films to prevent evaporation of hydrazine, and incubated at 37°C for 30 min. Absorbance values were read using a spectrophotometer at 340 nm. Calibration curves were generated from known lactate standards (0–10 mM) and used to obtain lactate levels in samples from the absorbance values. Lactate secretion rates were calculated from the lactate contents in the medium divided by time interval and normalized to cell numbers.⁸

Sample harvest and processing

At various time points during chondrogenic induction, aggregates were harvested for biochemical and histological analysis. Immediately after the harvest, the aggregates were washed in PBS and imaged using a stereo microscope (Leica Microsystems GmbH, Wetzlar, Germany). The images were analyzed using ImageJ to obtain area (A_P) and perimeter (P_P) data of aggregate projections. Sphericity (2D) values were calculated using the formula $4\pi A_P∕P_P^2$ Some aggregates were frozen at −80°C until analysis of DNA, GAG, and HYP. Some aggregates were processed for histology and immunohistochemistry (see below).

Histology

Qualitative measurement of GAG/proteoglycan content and distribution³⁰ was done on 4% formaldehyde-fixed, paraffin-embedded, and sectioned aggregates stained with Safranin-O and toluidine blue using standard protocols. Type II collagen content distribution in aggregates was shown using immunohistochemical staining according to previously published method.^22,31 Unfixed aggregates were embedded using OCT compound at −80°C and then processed for cryosectioning. Cold (4°C) acetone was used to fix the sections for 10 min and 10% normal goat serum in 1 × PBS was applied to block for 30 min. Primary antibody against type II collagen diluted in 1% normal goat serum in 1 × PBS with ratio of 1:50 was incubated with the sections for 1 h, after which FITC-conjugated goat anti-mouse IgG, IgA, and IgM diluted in 1% normal goat serum in 1 × PBS at 1:400 were applied for 45 min. The sections in parallel treated without primary antibody were used as negative controls. For type X collagen immunostaining, pellet sections were deparaffinized/rehydrated followed by antigen retrieval through incubation with 10 mg/mL proteinase-K (Roche) in Tris-EDTA buffer for 20 min at 37°C, and endogenous peroxidase activity blocked by incubating with 3% H₂O₂ during 30 min at RT. Nonspecific binding sites were blocked with 2.5% normal horse serum, and sections were incubated overnight with antibodies. Sections were incubated with anti-mouse Ig (ImmPRESS™ polymerized reporter enzyme staining system; Vector Laboratories, CA) for 30 min, followed by a short incubation with ImmPACT™ red peroxidase substrate (Vector Laboratories). The sections in parallel treated without primary antibody were used as negative controls. Stained sections were imaged using a Leica camera with a compound microscope (10 × objective; Leica Microsystems GmbH). In some cases, a mosaic of multiple images was compiled for aggregates that were too large.

GAG and DNA assay

Previously published methods^22,29,32,33 were followed to determine GAG and DNA contents of the aggregates quantitatively. For GAG analysis, Safranin-O reagent was combined with the digests of each separated aggregate. Vacuum was applied to the dot-blot apparatus and precipitates were collected on a nitrocellulose membrane. Individual dots were cut out from the membrane, and cetylpyridinium chloride was used to elute the dye. Absorbance of the eluted dye was analyzed at a wavelength of 536 nm. A standard curve was generated from known concentrations of purified chondroitin sulfate. GAG measurements were normalized with the DNA content of the aggregates. For DNA measurements, papain buffer (25 μg/mL papain, 2 mM cysteine, 50 mM sodium phosphate, and 2 mM EDTA, pH 6.5) was used for digestion and eventually combined with Hoechst 33258 for fluorescent measurement of each sample. Excitation wavelength and emission wavelength used were 340 and 465 nm, respectively. Known concentrations of calf thymus DNA were used to generate standard curves.

HYP assay

HYP content is an indicator of the total collagen content and was determined by previously published method.^34–36 During GAG/DNA assay, samples were separated after papain digestion step for HYP analysis. First, the samples were hydrolyzed using 6 N HCl at 110°C overnight in capped microcentrifuge tubes. After the hydrolysis, the microcentrifuge tubes were uncapped and the HCl was evaporated by heating the samples overnight at 60°C, and the dry samples were redissolved in water. Aliquots of 20 μL of these were then transferred into a nontissue culture 96-well plate with UV transparent flat bottom in quadruplicate to minimize errors. 0.15 M CuSO₄ and 2.5 N NaOH were added to each well of sample and incubated in an oven at ∼50°C for 5 min. Twenty microliters of 6% H₂O₂ was then added to each of the sample wells before incubating the plate for another 10 min at ∼50°C. After this, 80 μL of 3N H₂SO₄ and 40 μL of 10% p-dimethyl-amino-benzaldehyde were added to each sample before a final incubation period of 16 min at about 70°C. Between all these incubation periods, the samples were allowed to cool to room temperature. The same procedure was done simultaneously for standards of HYP between 0 and 500 μg/mL. The absorbance was read at a wavelength of 505 nm as soon as these cycles of incubation-cooling were completed.

Statistical analysis

Statistical analysis was performed using the Origin 2018 software package from Origin Lab (Northampton, MA), Minitab (State College, PA), and Microsoft Office Excel (Microsoft, WA). Student's t-test was used for select pairwise comparisons. Tukey's test was employed for multiple comparisons. Analysis of variance (ANOVA) was performed to determine significance for time-dependent data. Data are represented as mean ± standard deviation (SD). Differences were considered significant when p-value <0.05.

Glucose transport model

A mathematical model was developed to explain the dynamics of glucose levels in the culture medium in a microwell of a 96-well plate containing a spherical hMSC aggregate (Fig. 11). The model is pseudo-steady state model for glucose transport within a spherical construct, while glucose transport in the medium is based on volume-averaged glucose transport equation. The resultant mathematical model based on glucose mass conservation is³⁷

FIG. 11.

Schematic of the aggregate culture for mathematical modeling development to explain glucose uptake. Culture medium contents are assumed to be well mixed and tissue construct is assumed to be spherical. Glucose is transported across the surface of the construct, which leads to unsteady glucose depletion in the culture medium resulting in the first equation of the mathematical model (see Materials and Methods section). Glucose transport inside the construct is due to diffusion and uptake by the cells; this process is assumed to occur relatively faster than the depletion process in the culture medium, thus leading to the pseudo-steady state equation in the mathematical model. Both equations are coupled through the boundary conditions.

Cell culture medium: $V^M\frac{d{C}_G^M}{dt}={\left.-4\pi r^2{D}_G^T\frac{d{C}_G^T}{dr}\right|}_{r=R\left(T\right)},att=0,C_G^M=C_{G0}^M$

Tissue construct: $\frac{1}{r^2}\frac{d}{dr}\left(r^2{D}_G^T\frac{d{C}_G^T}{dr}\right)=\frac{V_{max}{C}_X{C}_G^T}{K_M+C_G^T},atr=0,\frac{d{C}_G^T}{dr}=0;atr=R\left(T\right),C_G^T=\lambda C_G^M$

Here, t is culture time, either 0–24 h (medium-change frequency = 1 day⁻¹) or 0–48 h (medium-change frequency = 0.5 day⁻¹), $C_G^M$ and $C_G^T$ are glucose concentrations in medium and tissue respectively, V^M is the volume of the culture medium (2 × 10⁻⁷ m³), $D_G^T$ is the diffusivity of glucose in the tissue (2.217 × 10⁻¹⁰ m²/s,⁸), R(T) is radius of the spherical construct at different times obtained by fitting a power model to the experimental data in Figure 4, $C_{G0}^M$ is the value of the starting glucose level (1, 2, 3, and 4.5 g/L), V_max is the intrinsic consumption rate of glucose (mole of glucose/g DNA/s), K_M is the Michaelis constant, C_X is the uniform DNA density (g DNA/m³) in the construct obtained from experimental data from Figures 4 and 5, and λ is the partition coefficient for glucose between the medium and aggregate (0.643). The coupled equations were solved in MATLAB using ode and bvp4c functions. Within the simulation period of 0–24 h (80 simulations over a culture period of 20 days and 4 starting glucose levels) or 0–48 h (40 simulations over a culture period of 20 days and 4 starting glucose levels), respectively, for 0.5 and 1 day⁻¹ medium-change frequencies, R(T) and C_X were assumed to be constant. V_max and K_M were assumed to be constant for the entire 21-day culture time; their averaged values were obtained from Zhong et al.⁸

FIG. 4.

Aggregate size as project areas (mm²) of images of aggregates harvested on day 1, 7, 14, and 21 during chondrogenesis for different starting glucose levels: 1 g/L (black squares), 2 g/L (red circles), 3 g/L (green triangles), and 4.5 g/L (inverted blue triangles). Medium-change frequencies were either 0.5 day⁻¹ (A) or 1 day⁻¹ (B). Error bars represent mean ± SD (N = 8 per condition). Color images are available online.

FIG. 5.

DNA content (μg) of aggregates harvested on day 1, 7, 14n and 21 during chondrogenesis. Day 0 data were obtained from initial cell samples frozen on day 0 with same number (250,000) of cells as aggregates. Initial glucose levels were varied: 1 g/L (black squares), 2 g/L (red circles), 3 g/L (green triangles), and 4.5 g/L (inverted blue triangles). Medium-change frequencies were either 0.5 day⁻¹ (A) or 1 day⁻¹ (B). Error bars represent mean ± SD (N varied from 4 to 8 per condition). Color images are available online.

Results

The primary purpose of this project is to determine whether culture conditions have an effect on tissue quality during chondrogenesis. We examined whether doubling the medium-change frequency that is typically used by researchers in the field affected nutrient availability and end-tissue quality. In addition, we also varied the glucose levels in the culture medium to mimic nutrient limitations.

Glucose uptake and lactate secretion

Figure 1 shows daily glucose levels in the culture medium during chondrogenesis over a period of 21 days for different initial glucose levels: 1, 2, 3, and 4.5 g/L. Figure 1A shows results from experiments with a medium-change frequency of 0.5 day⁻¹ (control), whereas Figure 1B shows results from experiments with a medium-change frequency of 1 day⁻¹. In both conditions, as expected, glucose levels decreased over time between medium changes. Increasing the medium-change frequency led to aggregates exposed to higher levels of glucose throughout the duration of chondrogenesis for all glucose levels Average glucose levels were about 2, 6, 11, and 19 mol/m³, respectively, for 1, 2, 3, and 4.5 g/L initial glucose levels for low-frequency treatment versus 3, 8 12, and 21 mol/m³, respectively, for 1, 2, 3, and 4.5 g/L initial glucose levels for high-frequency treatment. Supplementary Figure S1 Glucose uptake during chondrogenesis shown as cumulative uptake as a function of time (Fig. 2) shows that for all initial glucose levels, increasing medium-change frequency led to increased cumulative glucose consumption. The increase was dramatic for 1 g/L initial glucose, where on day 21, higher medium-change frequency resulted in about 89% increase in cumulative glucose uptake (Fig. 2B, black markers). When daily uptake of supplied glucose was analyzed, increasing medium-change frequency from 0.5 to 1 day⁻¹ led to significantly increased average daily utilization per aggregate for 1 and 2 g/L initial glucose levels (Supplementary Fig. S1, 0.47 [mean] ± 0.21 [SD] vs. 0.91 ± 0.06 μmol for 1 g/L and 0.89 ± 0.09 vs. 1.15 ± 0.09 μmol for 2 g/L). When analyzed as uptake rates (Supplementary Fig. S2), increasing medium-change frequency from 0.5 to 1 day⁻¹ led to relatively less variations in rates versus culture time.

FIG. 1.

Dynamics of glucose concentrations (mol/m³) in media during hMSC chondrogenesis for different starting glucose levels: 1 g/L (black squares), 2 g/L (red circles), 3 g/L (green triangles), and 4.5 g/L (inverted blue triangles). (A) Shows results for a medium-change frequency of 0.5 day⁻¹. Data for 11 medium exchanges are represented as 11 sets of 3 connected markers for each initial glucose level. (B) Shows results for a medium-change frequency of 1 day⁻¹. Data for 21 medium exchanges are represented as 21 sets of 2 connected markers for each initial glucose level. Error bars represent mean ± SD (N varied from 8 to 32 per condition due to serial harvesting of aggregates for morphological, biochemical, and histological analysis). hMSC, human mesenchymal stem cell; SD, standard deviation. Color images are available online.

FIG. 2.

Cumulative glucose uptake (μmol) of hMSCs during chondrogenesis as a function of time for different starting glucose levels: 1 g/L (black squares), 2 g/L (red circles), 3 g/L (green triangles), and 4.5 g/L (inverted blue triangles). Medium-change frequencies were either 0.5 day⁻¹ (A) or 1 day⁻¹ (B). Error bars represent mean ± SD (N varied from 8 to 32 per condition due to serial harvesting of aggregates for morphological, biochemical, and histological analysis). Color images are available online.

Lactate is a byproduct of glycolysis. During chondrogenesis, increasing medium-change frequency and increasing glucose levels led to increase in lactate secretion (Fig. 3). On day 20, cumulative lactate secretion per aggregate increased by 43% (1 g/L initial glucose), 11% (2 g/L), 10% (3 g/L), and 7% (4.5 g/L) when the medium-change frequency was doubled.

FIG. 3.

Cumulative lactate secretion (μmol) of hMSCs during chondrogenesis as a function of time for different starting glucose levels: 1 g/L (black squares), 2 g/L (red circles), 3 g/L (green triangles), and 4.5 g/L (inverted blue triangles). Medium-change frequencies were either 0.5 day⁻¹ (A) or 1 day⁻¹ (B). Error bars represent mean ± SD (N varied from 8 to 32 per condition due to serial harvesting of aggregates for morphological, biochemical, and histological analysis). Color images are available online.

Aggregate size and DNA

Aggregate size increased during chondrogenesis for all conditions; increasing glucose levels in culture medium, generally, led to increase in aggregate size (Fig. 4). For a medium-change frequency of 0.5 day⁻¹, however, aggregate size increased only for aggregates exposed to 2, 3, and 4.5 g/L initial glucose levels, whereas size decreased slightly for those exposed to 1 g/L initial glucose level (Fig. 4A). For a medium frequency change of 1 day⁻¹, aggregate size increased for all initial glucose levels (Fig. 4B). On day 21, increasing medium-change frequency increased the aggregate size significantly (p < 0.001) for all initial glucose levels. The largest increase (79%) was observed for aggregates treated with 1 g/L initial glucose level and the lowest increase (11%) was observed for aggregates treated with 4.5 g/L. Two-dimensional sphericity of the aggregates ranged from 0.94 to 1 for all conditions. There were no statistically significant trends or differences with respect to initial glucose levels or medium-change frequencies (data not shown).

In general, DNA content of the aggregate (Fig. 5) mostly remained unchanged during chondrogenesis under all culture conditions. For a medium frequency change of 0.5 day⁻¹, DNA content decreased slightly from day 1 to 21 for aggregates treated with 1 and 2 g/L, whereas a slight increase in DNA was noted for 4.5 g/L (Fig. 5A). For a medium frequency change of 1 day⁻¹, DNA content did not change significantly during the course of chondrogenesis for aggregates exposed to 2, 3, and 4.5 g/L, whereas it decreased slightly for 1 g/L (Fig. 5B).

Tissue biochemical content

Glycosaminoglycan

For a medium-change frequency of 0.5 day⁻¹, compared to levels on day 7, GAG content of aggregates on day 21 increased significantly during chondrogenesis for all starting glucose levels, except 1 g/L. The increases were 84%, 96% and 116%, respectively, for 2, 3, and 4.5 g/L starting glucose levels (Fig. 6A). For a higher medium-change frequency of 1 day⁻¹, all day-21 GAG levels significantly increased compared to day-7 levels. The increases were 82%, 184%, 198%, and 215%, respectively, for 1, 2, 3, and 4.5 g/L (Fig. 6B). For all initial glucose levels, increasing medium-change frequency from 0.5 to 1 day⁻¹ increased GAG levels of tissue significantly on day 21. The increases were 70%, 58%, 89%, and 105%, respectively, for 1, 2, 3, and 4.5 g/L. Similar trends were observed for DNA (used as a proxy for cell number)-normalized GAG content of aggregates (Supplementary Fig. S3).

FIG. 6.

GAG content (μg per aggregate) of aggregates harvested on day 1, 7, 14, and 21 during chondrogenesis for different starting glucose levels: 1 g/L (black), 2 g/L (red), 3 g/L (green), and 4.5 g/L (blue). Medium-change frequencies were varied either 0.5 day⁻¹ (A) or 1 day⁻¹ (B). Error bars represent mean ± SD (N varied from 3 to 8 per condition). GAG, glycosaminoglycan. Color images are available online.

Hydroxyproline

HYP is a widely used indicator of total collagen content. Using day-7 results as reference, for a medium frequency change of 0.5 day⁻¹, HYP content of aggregates increased significantly during chondrogenesis for all initial glucose levels (Fig. 7A): 1 (59% increase on day 21 compared to day 7), 2 (108%), 3 (156%), and 4.5 g/L (170%). For a higher medium frequency change of 1 day⁻¹, the increases were 217%, 234%, 248%, and 244%, respectively for 1, 2, 3, and 4.5 g/L (Fig. 7B). For all initial glucose levels, increasing medium-change frequency from 0.5 to 1 day⁻¹ increased HYP levels of tissue significantly on day 21. The increases were 104%, 67%, 59%, and 53%, respectively for 1, 2, 3, and 4.5 g/L. Similar trends were observed for DNA-normalized HYP content of aggregates (Supplementary Fig. S4).

FIG. 7.

HYP content (μg per aggregate) of aggregates harvested on day 1, 7, 14, and 21 during chondrogenesis for different starting glucose levels: 1 g/L (black), 2 g/L (red), 3 g/L (green), and 4.5 g/L (blue). Medium-change frequencies were either 0.5 day⁻¹ (A) or 1 day⁻¹ (B). Error bars represent mean ± SD (N varied from 3 to 8 per condition). HYP, hydroxyproline. Color images are available online.

Histology and immunohistochemistry

Images of Safranin O-stained tissue sections show that all aggregates, except the ones treated with 1 g/L and a medium-change frequency of 0.5 day⁻¹ (Fig. 8A), had GAG presence as demonstrated by pink/red regions in the sections (Fig. 8B–H). Compared to corresponding sections from the lower medium-change frequency treatment (Fig. 8A–D), sections from higher medium-change frequency had brighter pink/red regions (Fig. 8E–H) indicating higher GAG levels. Peripheral regions of sections showed staining for low or no GAG (arrows, Fig. 8). Toluidine blue (which strongly binds sulfur)-stained sections showed GAG was present in the entire section for all conditions (Supplementary Fig. S5) Type II collagen immunostaining results showed the presence of type II collagen in all aggregates (Fig. 9A–H), although the aggregate treated with 1 g/L and a medium-change frequency of 0.5 day⁻¹ had a lower level of fluorescence signal (Fig. 9A) compared to others. Also, organized collagen was seen in most aggregates, although a better outcome can be seen with aggregates exposed to a higher medium-change frequency (Supplementary Fig. S6). To evaluate hypertrophic induction, we carried out col X immunostaining. The results (Supplementary Fig. S7A–H) show that there was staining (brown-red regions) for col X in sections of aggregates for both low and high medium-change frequencies as well as at different glucose levels with no apparent trends.

FIG. 8.

Safranin O-stained section images of day 21 harvest aggregates cultured in media with different starting glucose levels (1 g/L: A, E; 2 g/L: B, F; 3 g/L: C, G; 4.5 g/L: D, H) and medium-change frequencies (0.5 day⁻¹ [A–D] and 1 day⁻¹ [E–H]). Orangish red-stained regions indicate the presence of sulfated GAGs. Scale bar: 500 μm. Color images are available online.

FIG. 9.

Type II collagen immunostaining results of day-21 harvest aggregates cultured in media with different starting glucose levels (1 g/L: A, E; 2 g/L: B, F; 3 g/L: C, G; 4.5 g/L: D, H) and medium-change frequencies (0.5 day⁻¹ [A–D] and 1 day⁻¹ [E–H]). Dashed square boxes in (A–H) show locations of collagen organization typically found in native tissue.⁵⁴ Higher magnification images of organized collagen are shown in Supplementary Figure S6. Scale bar: 500 μm. Color images are available online.

Noninvasive monitoring

Analysis of glucose uptake and lactate secretion showed that GAG and HYP contents in tissue are correlated with cumulative glucose uptake levels (Fig. 10A) and cumulative lactate secretion levels (Fig. 10B) at all times (day 7, 14, and 21), initial glucose levels (1, 2, 3, and 4.5 g/L), and medium-change frequencies (0.5 and 1 day⁻¹). The glucose uptake and lactate secretion were slightly better correlated to HYP content (Pearson's r = 0.95) compared to GAG content (Pearson's r = 0.9).

FIG. 10.

GAG and HYP contents in aggregates are correlated with cumulative glucose uptake (A) and cumulative lactate secretion (B). Markers represent experimental data and lines (red dashed: GAG, blue solid: HYP) represent linear fits. Error bars represent mean ± SD (N varied from 3 to 4 per condition). (A) GAG (red squares, Pearson's r = 0.907) and HYP (blue squares, Pearson's r = 0.954) contents of aggregates exposed to individual starting glucose levels and medium-change frequency levels (a total of eight conditions) and harvested on day 7, 14, and 21 as a function of cumulative glucose uptake. (B) GAG (red squares, Pearson's r = 0.903) and HYP (blue squares, Pearson's r = 0.941) contents of aggregates exposed to individual starting glucose levels and medium-change frequency levels (a total of eight conditions) and harvested on day 7, 14, and 21 as a function of cumulative lactate secretion. Color images are available online.

Discussion

Chondrogenesis is an important biological process with key applications in wound healing and regenerative medicine. In vitro, chondrogenesis of human bone marrow-derived MSCs holds a lot of promise in fabrication of engineered cartilage. In this research, we investigated the effect of culture condition changes such as glucose levels and frequency of medium change on chondrogenesis in vitro. Our results demonstrate that glucose availability is critically important during chondrogenesis and culture conditions that enhance glucose availability lead to improved tissue contents of HYP and GAG.

Cartilage ECM is dominated by type II collagen and GAGs. GAG synthesis during chondrogenesis is a key indicator of successful cartilage tissue formation and glucose is an important precursor of GAG synthesis in cells. During cell culture, glucose levels in culture medium decrease as it is consumed by the cells. This decrease is dependent on its rate of consumption and mass transport. Here, uptake/secretion rate is used to denote the aggregate as a whole, whereas intrinsic consumption/synthesis rate is used to denote the per cell/DNA rate. In static culture conditions, both starting glucose levels and medium-change frequency significantly affect transport and consumption rates. Mass transport is dependent on concentration gradient, whereas glucose consumption rate is dependent on local glucose concentration (e.g., Michaelis-Menten kinetics). In addition, as the aggregate increases in size during culture due to matrix accumulation (Fig. 4), the diffusional resistance increases; however, since the DNA content approximately remained constant (Fig. 5), the intrinsic consumption rate per volume decreases during culture. These two competing mechanisms can lead to differences in glucose uptake rates. For aggregates cultured with medium-change frequency of 0.5 day⁻¹ (Figs. 2A and 3A), the aggregate size varied for different initial glucose levels (Fig. 4A) with the order of increase in size as follows: 1 g/L ≤ 2 g/L ≤ 3 g/L ≤ 4.5 g/L. Such difference appeared from day 7 to 21. DNA levels, a proxy for cell numbers, were mostly constant (Fig. 5A). Distribution of same number of cells, but in different spherically shaped aggregates suggests that the glucose uptake into aggregates will be inversely proportional to the radius of the aggregate; this means that the transport resistance is the largest in 4.5 g/L and smallest in 1 g/L aggregates. And the glucose availability is highest in 4.5 g/L, and lowest in 1 g/L. The competing effects of glucose availability and transport limitation on glucose uptake likely led to similar cumulative glucose uptake for 2, 3, and 4.5 g/L. For 1 g/L, the aggregates consumed nearly all glucose in culture medium on the second day, leading to significantly lower cumulative glucose uptake compared to 2, 3, and 4.5 g/L aggregates. For aggregates cultured with medium-change frequency of 1 day⁻¹ (Figs. 2B and 3B), the aggregate size did not change appreciably for 2, 3, and 4.5 g/L glucose levels. Utilizing a mathematical model of well-mixed culture medium, incorporating these two mechanisms and published intrinsic rates,⁸ we can show that these mechanisms reasonably explain the changes in culture medium glucose levels due to changes in uptake rates (Fig. 1 and Supplementary Fig. S8) for different initial glucose levels and medium-change frequencies. This suggests that both glucose transport and glucose consumption rate (by Michaelis-Menten Kinetics) by the cells are affected by these two key design parameters. The model developed in this study is simpler to use as the external medium glucose concentration was volume averaged, and faster to implement than the previously published model,⁸ which was implemented in COMSOL^® and used primarily to determine kinetic parameters of glucose.

Increasing medium frequency replenishes glucose in the culture medium to starting glucose levels. This increases the glucose availability or average glucose level in the culture medium. This leads to enhanced glucose transport in general, but for conditions that lead to increased consumption rate (smaller sized constructs or higher DNA, which is related to higher cell number), this will have a bigger impact. Aggregates exposed to 1 g/L starting glucose levels were much smaller in size compared to those exposed to higher starting glucose levels, and showed a greater increase in cumulative glucose uptake as a result of doubling medium-change frequency (Fig. 2 and Supplementary Fig. S1). Aggregate exposed to higher starting glucose levels had a more modest increase when medium-change frequency was doubled (Fig. 2B). Collectively, these results show that local glucose consumption rate per volume plays a key role in overall uptake of glucose.

Lactate can be synthesized from glucose through glycolysis as well as directly from pyruvate in the medium; the reaction of pyruvate to form lactate is a reversible reaction. Increasing medium-change frequency decreases the average lactate level in the medium. Since chondrocytes have been known to secrete lactate even against a strong concentration gradient,³⁸ lactate secretion in the construct is likely independent of medium replenishment. This confirms our observation that doubling medium frequency had a very small effect on lactate secretion and its rate (Fig. 3 and Supplementary Fig. S9).

The significant increase in overall glucose uptake by the aggregates by doubling the medium-change frequency had a sequence of positive results. Glucose is a key nutrient for ATP production through glycolysis and indirectly involved in GAG synthesis.³⁹ Increased uptake of glucose with correspondingly increased synthesis of GAG and HYP suggests that these processes are glycolysis ATP dependent. For aggregates exposed to starting glucose level 4.5 g/L, the GAG content doubled at the end of day 21 by increasing the medium-change frequency, while HYP content increased by 50% (Figs. 6 and 7). Type II collagen immunohistostaining results suggest that enhancement in HYP content, an indicator of total collagen, is likely due to enhanced synthesis spread over a larger volume (Fig. 9). Overall, these results suggest that in addition to enhancement in GAG synthesis, additional glucose uptake also leads to enhanced protein synthesis. Protein content, especially type II collagen, of tissue-engineered cartilage constructs has been shown to be significantly reduced when compared to native issue.^40–42 The enhanced GAG and HYP also led to increased size of the constructs (Fig. 4). Both the increase in size and ECM content for approximately the same DNA (Fig. 5) suggests that with increased glucose availability through, for example, increased medium replenishment, it may be possible to drive in vitro chondrogenesis to a tissue-engineered cartilage with improved biochemical properties. It should be noted that during chondrogenesis of hMSCs, DNA content remains approximately constant as there is no cell division. While increasing medium-change frequency has been shown to be beneficial, it is likely that this has an upper limit for optimal outcomes, as there has been a number of autocrine factors produced by the cells during chondrogenesis.^43–45 Removing these signals more frequently from culture medium will likely have a negative effect on chondrogenesis.

Interestingly, uptake of glucose and secretion of lactate were shown to correlate well with GAG and HYP content of the resultant aggregates regardless of starting glucose levels or medium-change frequency (Fig. 10). This suggests that both molecules are important metabolic markers and monitoring them during chondrogenesis can lead to a method to noninvasively assess chondrogenesis.

ECM synthesis is an important process during chondrogenesis. During development, chondrogenesis is influenced by transcription factors, growth factors, hormones, and morphogens.⁴⁶ In vitro chondrogenesis using hMSCs falls short in de nova matrix synthesis compared to in vivo chondrogenesis. Specifically, GAG and type II collagen syntheses upon chondrogenesis induction in vitro have been under par for generating engineered tissues to replace native articular cartilage function as these biomolecules play a significant role in providing mechanical strength to the tissue.^47–51 To address this shortfall, a variety of scaffold technologies are developed with the aim of providing functional support in vivo until de nova synthesis of these biomolecules. Results from this article show that culture conditions play a key role to improving synthesis of key biomolecules during chondrogenesis in vitro. Both higher glucose levels and medium-change frequency led to enhanced GAG/DNA and HYP/DNA contents in hMSC aggregates. Human femoral condyle's GAG/DNA content ranges from about 100 μg/μg (superficial layer) to 1000 μg/μg (deep layer), whereas HYP/DNA content ranges from about 230 μg/μg (superficial layer) to 560 μg/μg (deep layer).⁵² While GAG/DNA content in this work was close to that in the native cartilage GAG content, HYP/DNA content was still significantly lower. A future strategy for increasing HYP/DNA synthesis can involve modulating key amino acid uptake during chondrogenesis. In addition to engineered cartilage, such strategies to enhance de nova protein synthesis are important in a variety of structural tissue engineering applications such as bone, skin, tendon, and ligaments.

The work presented here has a few limitations. One limitation of this article is that the quantitative assay used to measure collagen synthesis—while commonly used by many researchers in the field—is nonspecific to type II collagen; both elastin and type I collagen molecules can interfere with the assay. Another limitation of the article is that mechanical properties of aggregates were not assessed to show that the enhanced syntheses of GAG and type II collagen lead to improvement in mechanical properties.⁵³

In summary, results from this work showed that glucose availability is an important parameter during chondrogenesis. Enhancing glucose availability by increasing medium-change frequency and starting glucose levels in an aggregate culture system led to larger construct with increased GAG and HYP content. These results can be used to design optimal culture conditions in a tissue-engineered cartilage application.

Disclosure Statement

No competing financial interests exist.

Funding Information

This work is funded by a grant from the National Institutes of Health (EB021911) and the Virginia and David Baldwin fund.

Supplementary Material

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Figure S4

Supplementary Figure S5

Supplementary Figure S6

Supplementary Figure S7

Supplementary Figure S8

Supplementary Figure S9