Evaluation of β1-integrin expression on chondrogenically differentiating human adipose-derived stem cells using atomic force microscopy

Abstract

The expression of β1-integrin on human adipose-derived stem cells, differentiating toward a chondrogenic lineage, is hypothesized to decrease when cells are grown under in vivo-like environments due to sufficient extracellular matrix (ECM) buildup in the engineered tissues. The opposite is true when cells are grown in static cultures such as in pellet or micromass. To probe β1-integrin distribution on cellular surfaces, atomic force microscopy cantilevers modified with anti-β1-integrin antibodies were used. Specific antibody–antigen adhesion forces were identified and indicated the locations of β1-integrins on cells. ECM properties were assessed by estimating the Young's modulus of the matrix. Specific single antibody–antigen interactions averaged 78 ± 10 pN with multiple bindings occurring at approximate multiples of 78 pN. The author's results show that upregulated β1-integrin expression coincided with a less robust ECM as assessed by mechanical properties of tissues. In micromass and pellet cultures, transforming growth factor β3(TGF-β3) elicited a decrease in Young's modulus by 3.7- and 4.4-fold while eliciting an increase in β1-integrin count by 1.1- and 1.3-fold, respectively. β1-integrin counts on cells grown in the presence of TGF-β3 with oscillating hydrostatic pressure decreased by a 1.1-fold while the Young's modulus increased by a 1.9-fold. Collectively, our results suggest that cells in insufficiently robust ECM express more integrin perhaps to facilitate cell–ECM adhesion and compensate for a looser less robust ECM.

I. INTRODUCTION

Articular cartilage (AC) has a limited capacity for self-repair, and therefore, when lesions are present, damage often progresses to a state of a degenerative joint disease known as osteoarthritis (OA). Current OA therapies such as lavage and arthroscopy, debridement, Pridie drilling, and microfracture are incapable of providing long-term benefits.^1,2 Therefore, research into AC tissue engineering is taking precedence to provide a future remedy for OA. In functional AC tissue engineering, cells, scaffolds, and appropriate growth factors are typically combined in vitro in a bioreactor that mimics in vivo conditions to develop tissues with properties similar to native cartilage. This tissue is then expected to be grafted in vivo to repair a cartilage lesion. The use of growth factors is required for proper differentiation of cells, improved extracellular matrix (ECM) production, and attainment of tissue mechanical properties comparable to those of native cartilage. In vivo, chondrocytes sense mechanical stimulation and respond to it by regulating the rates of ECM synthesis and degradation. Through mechanotransduction, chondrocytes can thus control the ECM composition and tissue mechanical properties.³ Engineering functional AC tissues requires a combination of appropriate growth factors as well as application of mechanical stimuli. To enhance the properties of engineered ECM and AC tissues, a fundamental understanding of mechanotransduction pathways is needed. Better understanding of how signals are transduced from the ECM to the cell can help researchers make strides in developing pharmaceutical therapies for degenerated cartilage.

It is known that interactions between the extracellular environment and cells are mediated by transmembrane glycoprotein receptors called integrins.⁴ Integrins are a family of heterodimeric transmembrane glycoproteins involved in cell adhesion to ECM and in cell–cell adhesion.^5,6 They interact with intracellular proteins upon ligand binding and are thereby involved in signal transduction through the actin cytoskeleton. Because of their ability to interact with ECM and to mediate intracellular signaling, they are suspected of being important in transmitting information about mechanical loads on the cartilage to the cell so that the cell can respond to mechanical cues. This would allow the chondrocytes to detect changes in the ECM with integrins acting as mechanotransducers.⁶ A functional integrin heterodimer is made up of α and β subunits. The β1 integrin dimer will be studied here because among cartilage heterodimers, it is the most prominent β dimer and it interacts with many different α dimers.⁷

With cyclic pressure induced strain, human articular chondrocytes respond with membrane hyperpolarization facilitated by a flux of K⁺ ions across the cell membrane. Blocking α5β1 integrins with antibodies or arginylglycylaspartic acid peptides inhibits this response and thereby suggests that α5β1 integrins are a major mechanoreceptor in chondrocytes.⁸ Furthermore, cyclic pressure induces tyrosine phosphorylation of focal adhesion kinase and paxillin and induces the β1-integrin-dependent association of protein kinase C-α with receptor for activated C kinase 1 and β1-integrin.⁹ The activation of these intracellular signaling molecules stimulates signaling cascades that can be important in mechanotransduction.

Because of this postulated role of β1-integrin in mechanotransduction, we investigated how mechanical stimuli and growth factors affect the expression and distribution of β1-integrin on surfaces of cells isolated from engineered AC tissues. To engineer AC tissues, human adipose-derived stem cells (hASCs) were allowed to differentiate toward chondrogenic lineages that formed AC tissues in a novel centrifugal bioreactor (CBR) operating with or without an oscillating hydrostatic pressure (OHP) and in the presence or absence of transforming growth factor beta 3 (TGF-β3). The density and distribution of β1-integrins on surfaces of cells, the messenger ribonucleic acid (mRNA) expression of β1-integrin, and the mechanical properties of the engineered AC tissue were investigated.

Quantifying the location, density, and adhesive properties of β1-integrins was accomplished using an atomic force microscopy (AFM). AFM is unique in its ability to quantify biomolecular forces with a pico-Newton resolution in fluids mimicking the physiological conditions with minimal sample preparation and alteration.^10–13 By comparing AFM images to AFM force maps, a structure–function relationship at the molecular scale can be obtained in a way that traditional methods cannot provide.^14–17 AFM has been used to study the interaction forces of ligand–receptor pairs^{10,14,18–25} and x-ray and electron crystallography confirmed that probing ligand–receptor interactions with AFM does not disrupt the protein structure.¹⁷ Using AFM to probe the surface localization of β1-integrins on differentiated hASCs, we show, for what may be the first instance, the location, density, and adhesive characteristics of β1-integrins on live cells and correlate the expression of β1-integrins with mechanical and growth factor stimulation.

Single-molecule force microscopy (SMFS) has been used to map interactions and locations of native proteins on membranes of cells such as TRA-1-81 on human embryonic stem cells,²⁴ CD20 on lymphocytes,²³ β protein 1 on breast cancer cells,²⁶ and the Met receptor on hippocampal neurons.¹⁰ However, mapping the location of proteins on chondrogenically differentiating adipose-derived stem cells has never been investigated in the literature before except in our own work where we mapped N-cadherins on these cells.²⁷ N-cadherins relate to the pathways of cellular condensation of differentiating stem cells. Here, the locations of β1-integrins on the chondrogenically differentiating adipose-derived stem cells are probed due to their importance in mechanotransduction pathways. These pathways are expected to serve to translate the cyclic hydrostatic pressure applied to the AC tissue engineered in the CBR to intracellular signals that initiate cellular responses and adaptations. To the authors' best knowledge, mapping β1-integrins on cells has been done in only one study previously in the literature where cells were mapped using SMFS and confocal laser scanning microscopy (CLSM) on osteosarcoma MG-63 cells.²⁸ In this study, CLSM and SMFS results indicated how TGF-β1 affected the β1-integrin expression while lateral force microscopy was used to quantify the forces necessary to displace a cell from the surface. Although this study is interesting, researchers only investigated the effects of one treatment group (TGF-β1) on the expression of β1-integrin, and the expression was not further related to the function of the tissue. In comparison, we investigate the expression of β1-integrins on different cells using six different treatment groups and three culturing techniques. In addition, the role of β1-integrins in mechanotransduction is investigated by comparing the expression of β1-integrins with the elastic modulus of the tissue and the state of chondrogenesis of the cell. Unlike any other study, we are able to show, for the first time, how applying cyclic hydrostatic pressure to chondrogenically differentiating stem cells affects β1-integrin expression, chondrogenesis, and the elasticity of the resulting tissue.

II. EXPERIMENTAL PROCEDURES

Cells and cell culture supplies were purchased from Invitrogen-Gibco^®, Grand Island, NY, USA, unless otherwise specified.

A. Cell culture

hASCs isolated from a lipoaspirate tissue of a 33-year-old female were cultured in an expansion medium (EM) containing high-glucose Dulbecco's modified Eagle's medium (F12) supplemented with 10% fetal bovine serum, 100 U/ml penicillin, 100 μg/ml streptomycin (Sigma-Aldrich, St. Louis, MO), and 5 μg/ml Gentamicin. Cells were maintained under standard conditions (37 °C in a humidified incubator with 5% CO₂) with medium changed three times a week. Upon 80%–90% confluency, cells were passaged using Gibco^® TrypLE™ Select and used at passage 7 for the following experiments.

B. Chondrogenic differentiation

Chondrogenesis was induced in micromass, pellet, and in samples in our novel CBR as described previously.²⁹ Briefly, for micromass differentiation, 10 μl droplets from a 1.6 × 10⁷ cells/ml suspension were placed in the center of each well in a 24-well plate. After allowing cells to adhere for 2 h under standard conditions, 500 μl of fresh EM was added. After a day, 250 μl of EM was removed and replaced with either a base medium consisting of DMEM/F12 supplemented with 1 mM sodium pyruvate, 2 mM l-glutamine, 5 μg/ml gentamicin, 1% insulin–transferrin–selenium, 50 μM l-proline (Alfa Aesar, Ward Hill, MA), and 1% penicillin–streptomycin (Sigma-Aldrich), or by a chondrogenic medium consisting of a base medium with 100 nM dexamethasone, 50 μg/ml l-ascorbic acid (both from Sigma-Aldrich), and 10 ng/ml TGF-β3 (PeproTech, Ward Hill, NJ). Cells that received the base medium were considered to be the negative controls (NC), and those that received the chondrogenic medium were considered to be the positive controls (PC). For pellet culturing, aliquots of 5 × 10⁵ cells in 500 μl EM were centrifuged at 600 × g for 5 min in 15-ml polypropylene conical tubes. After incubation in standard conditions for a day, half of the medium was replaced by either the base or the chondrogenic medium. For the CBR study, after sterilizing the system by pumping 70% ethanol for 24 h, 6 × 10⁶ cells were injected into each reactor. The reactors were mounted on a COBE Spectra™ Apheresis System (TERUMO BCT, Lakewood, CO) and centrifuged at 500 rpm for 15 min. A base or a chondrogenic medium was continuously pumped for one week into the reactors. Additionally, 20 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid buffer was added on a daily basis to control the pH. To expose cells to OHP, medium flow was stopped in the reactors, the corresponding shut off valves were closed, and cells within the bioreactors were pressurized for 2 h a day for a week (cells were exposed to 290 psi for 2 s intervals). After pressurization, medium pumping was reestablished. Reactor samples were found to be free of contamination as tested on tryptic soy agar plates. For micromass, pellet, and bioreactor cultures, half the medium was exchanged three times a week. Micromass and pellet cultures served as static controls. Schematics of the reactor and oscillating pressure patterns have been presented elsewhere.²⁹

C. Cell digestion and immobilization

The bioengineered AC tissue was digested to isolate cells using collagenase type II. Briefly, tissues were transferred into tubes with the base medium supplemented with 10,000 U/g of tissue of collagenase type II (Worthington Biochemical, Co). Tubes were incubated overnight at 37 °C at 5% CO₂ and filtered through a 100 μm nylon cell strainer. Filtrate was centrifuged at 200–300 × g for 10 min, and pelleted cells were collected and seeded onto poly-l-lysine (Sigma-Aldrich, St. Louis, MO) coated glass slides for later use in AFM experiments.

D. AFM experiments and tip functionalization with antibodies

AFM experiments were conducted with a PicoForce™ scanning probe microscope with a Nanoscope IIIa controller and extender module (Bruker Corp., Bruker AXS, Santa Barbara, CA). Before each experiment, the spring constant of each cantilever was determined using the spectral density of thermal noise fluctuations.^30,31 The average spring constant of functionalized cantilevers was 0.0655 ± 0.0033 N/m, which is near the manufacturer's reported spring constant of 0.06 N/m.

To study the distribution of β1-integrins on membranes of cells, monoclonal anti-Integrin β1 (Abgent, San Diego, CA) antibodies were chemically bonded to a gold-coated silicon nitride (Si₃N₄) AFM cantilever. Si₃N₄ cantilevers (Bruker Corp., Bruker AXS, Inc., Santa Barbara, CA) were sputter-coated with a 5 nm Cr adhesive layer followed by a 40 nm Au layer using an Edwards Auto 306 physical vapor deposition sputtering machine. The gold-coated cantilevers were then cleaned in ethanol and deionized water (18 MΩ·cm) before deposition of the desired self-assembled monolayer (SAM) onto them. Cantilevers were then incubated in a 1 mM solution of 16-mercaptohexadecanoic acid (MHA) (Sigma-Aldrich, St. Louis, MO) dissolved in degassed acetonitrile for 30 min to create thiol-gold linkages, leaving the terminal carboxylic (COOH) groups exposed. These cantilevers were then washed with ethanol to remove excess MHA that did not form the SAM. To activate the terminal COOH groups, the cantilevers were incubated in 0.1 M N-hydroxysuccinimide (Sigma-Aldrich, St. Louis, MO) and 0.4 M 1-[3-(dimethylamino)propyl]-3-ethylcarbodiimide (Sigma-Aldrich, St. Louis, MO) in deionized water for 30 min. This facilitated the amine coupling of the anti-β1-integrin antibody to the AFM cantilever upon incubation of cantilevers in 25 μg/ml monoclonal antibody for 1 h at room temperature.¹⁰ The method of antibody attachment to cantilevers was tested by rinsing the cantilevers in phosphate-buffered saline and incubation in Alexafluor 350 donkey-anti-rabbit secondary antibody (Invitrogen, Carlsbad, CA) for 2 h. Using fluorescent imaging, the cantilevers with attached primary antibodies showed green fluorescence, whereas gold cantilevers only incubated in the fluorescent secondary antibody showed little fluorescence (not shown).

Force–volume (FV) imaging was used to collect force curves over a 16 × 16 grid of equally spaced points over a 10 $\times$ 10 μm area covering the cell [Fig. 1(a)]. The cantilever moved along the area at 1 Hz with a trigger threshold of 3.9 nN. Based on the indentation depth and geometry of the tip, the contact area of the AFM tip with the cell is calculated to be 0.0066 μm². The area of contact between the antibodies on the tip and the cell will actually be larger than that calculated since the steric interactions associated with the antibodies attached to the tip will increase the contact area. Therefore, with this technique, we are able to contact over a 100 molecules in a single approach-retraction cycle. Since the area of each pixel is approximately 0.39 μm², the area of the pixel that the AFM tip covers each time it contacts the surface is larger than 59%. At least, three cells were probed per treatment group. Details of the calculations above are included in the supplementary material.³²

Fig. 1.

Methodology. (a) A 16 × 16 pixelated height image of the cell with squares indicating the number and location of specific adhesion peaks for each pixel. The gray circle indicates the location of the cell. Adhesion at each pixel was determined from the retraction curve captured at that pixel. (b) Example retraction curves showing [(b), i] no specific adhesion, [(b), ii] a single specific adhesion peak, and [(b), iii] a curve containing both specific (gray) and nonspecific (black with arrows) adhesion peaks. The wormlike chain model [Eq. (1)] was fit to the specific adhesion peaks as shown by the black dotted lines.

Approach and retraction curves were collected for each pixel. The retraction data were used to extract specific antibody–antigen adhesion peaks and generate specific adhesion force maps demonstrating the location of β1-integrins on cell surfaces [Fig. 1(a)].

E. Analysis of retraction curves

Retraction curves from each force-indentation profile were individually analyzed using an in-house matlab-based software (InciSpec 5.1). Retraction curves are expected to display a variety of interaction types including:

• (1)no interactions between the functionalized cantilever and the cell surface [Fig. 1(b), i]
• (2)specific antibody–antigen interactions [Fig. 1(b), ii]
• (3)specific (gray) and nonspecific (black with arrows) interactions between the functionalized cantilever and the cell surface [Fig. 1(b), iii].^10,11,33

The ratio of curves with specific adhesion to curves without specific adhesion was 1:3.2. Nonspecific interaction forces exist when other forces, besides the antibody–antigen interactions, pull the AFM cantilever to the cell surface. These forces are most often van der Waals or electrostatic interactions.³⁴ For our study, we are only interested in the specific antibody–antigen binding events. To decouple these binding events from other adhesion events measured in the retraction curves, we first selected peaks with shapes that theoretically fit a wormlike chain (WLC) statistical model (see description below), which is representative of antibody–antigen binding events,^10,35 with examples plotted in Fig. 1(b). As mentioned above, the AFM tip can contact on the order of 100 integrins, therefore we expect multiple antibody–antigen interactions to occur in a single approach–retraction cycle. With multiple-bindings in mind, the specific adhesion forces were plotted as histograms with bin widths varying from 8 to 26 pN. These plots show that the distribution of specific adhesion forces were independent of the bin size used to construct the histogram of force distributions.²⁰ Inherent peaks, or areas where specific adhesion peaks were clustered, were visually chosen, and the shaded areas behind the plot indicate these selected ranges for bindings (Fig. 2). This was done for each treatment group separately, and the results for most-probable specific adhesions are represented in Fig. 3 for all treatment groups. In order to count the number of β1-integrins per treatment group, the magnitude of the specific adhesion force was considered as well as the area of the cell. Because the size and area scanned for each cell vary, the nominal count of β1-integrins was normalized to the number of pixels representing the area of the cell surface in a force–volume (FV) image. Areas of low height and high Young's moduli were considered the substrate and removed from analysis. For example, the cell in Fig. 1(a) takes up 77% of the scanned area.

Fig. 2.

Distribution of adhesive forces measured between a functionalized AFM probe with anti-β-1-integrin antibodies and β1-integrins for each treatment group. The force distribution data are plotted using bin widths, which varied from 8 to 26 pN. This way, we located peaks or locations where the majority of adhesion forces are with less bias. These ranges of most-probable adhesion forces are indicated by the shaded areas behind the chosen peaks. The ranges for these areas are represented in Fig. 3.

Fig. 3.

Ranges of locations of most-probable adhesion forces from Fig. 2 are visually represented in this box plot. Each box plot represents the range in which the majority of adhesion forces were found for each treatment group. Horizontal solid lines represent the force when forces from all treatment groups were averaged. These means were found to be 78, 158, and 240 pN, respectively, which are approximate multiples of 78 pN.

F. Determination of protein elasticity—WLC model

The WLC model, also known as the Kratky–Porod chain model has been successfully fit to stretched force-distance profiles of deoxyribonucleic acid (DNA) proteins, polysaccharides, and other molecules.³⁰ The WLC model can be applied to polymer chains that have a continuous curvature that is random at any point in the chain. It takes into account the local stiffness of the chain in terms of the persistence length $(L_P)$ and long-range flexibility. The force required to stretch a wormlike chain $(F_{\text{chain}})$ in a solvent to a length $D$ is given by

$$F_{\text{chain}}=\frac{-k_{B}T}{L_P}\left[\frac{D}{L_c}+\frac{1}{4{\left(1-\frac{D}{L_c}\right)}^2}-\frac{1}{4}\right],$$ (1)

where $k_B$ is the Boltzmann constant, $T$ is the absolute temperature, and $L_c$ is the polymer contour length. The appropriate fit was chosen by the method of least squares in which the overall solution minimizes the sum of the squares of the errors.

G. Determination of Young's moduli of the tissue—Hertz model

AFM force-indentation approach curves were obtained as previously described.^27,29,36,37 Force-indentation profiles were then fit to the Hertz model of contact mechanics, which assumes an infinitely hard sphere indenting a flat, deformable elastic substrate as described in Eq. (2)

$$F=\frac{4}{3}\times \frac{E_Y{R}^{1/2}{\delta }^{3/2}}{1-{\upsilon }^2},$$ (2)

where F is the applied force, E_Y is the Young's modulus, R is the relative radius of the indenter, ν is Poisson's ratio set at a constant of 0.5, and δ is the indentation depth. The Hertz model was fit to the force-indentation data to get an R² value closest to 1.0 using our in-house built matlab software. We believe that the use of the Hertz model on our biological cells is justified for the following three reasons: the indenter is not deformable; the model was only applied to approach curves without attractive forces; and the model was only applied to the linearly elastic portion of the indentation profile.

H. Quantitative real-time polymerase chain reaction

Quantitative real time polymerase chain reaction was used to quantify gene expression in harvested samples. Total RNA was isolated with TRIzol, and chloroform was used for phase separation. The aqueous phase, containing total RNA, was purified using the MagMAX™-96 for Microarrays Total RNA Isolation Kit (Life Technologies) according to the manufacturer's specifications. Genomic DNA was removed using MagMAX™ Turbo™ DNase Buffer and TURBO DNase from the MagMAX kit (Ambion by Life Technologies). Total mRNA (up to 2.5 μg) was reverse-transcribed into complementary DNA (cDNA) using SuperScript^® VILO™ Master Mix, which includes: SuperScript III RT, RNaseOUT™, Recombinant Ribonuclease Inhibitor, a proprietary helper protein, random primers, MgCl2, and dNTPs (Life Technologies). cDNA was amplified with the TaqMan^® Gene Expression Master Mix (Applied Biosystems by Life Technologies) on an ABI 7900HT Sequence Detection System (Applied Biosystems) and probes specific for GAPDH (a housekeeping gene), sex-determining region Y (SRY)-box 9 (SOX9; Hs00165814_m1), and integrin β1 (ITGB1; Hs01127536_m1) were used. The relative gene expression was calculated using the ΔΔC_T method, where a fold difference was determined using the expression $2^{-\Delta {\Delta }_T}$ as described previously.³⁸ mRNA values from hASCs prior to differentiation assays were considered as the reference. For both pellet and micromass controls, the materials of four constructs of the same type and culture condition were pooled for each biological replicate. A total of three technical replicates were analyzed per treatment.

I. Statistical analysis

Statistical analysis was performed using the SigmaPlot 11.0 (Systat Software Inc., San Jose, CA) software package. One-way analysis of variance (ANOVA) was used with the Dunn test to determine whether significant differences existed between treatment groups, with statistical significance reported at the 95% confidence level (P < 0.05). When comparing two samples, a t-test was used. When fitting data to the WLC model and the Hertz model, the R² value was used to judge the quality of the fit. The R² fit closest to 1 was chosen.

III. RESULTS AND DISCUSSION

AFM was utilized in the SMFS mode to assess the distribution of β1-integrins on cellular surfaces and in the force-indentation mode to quantify the Young's moduli of the tissues as an assessment of the quality and robustness of the ECM produced. AFM measurements were done with or without TGF-β3 for static controls as well as for CBR samples. AFM results are listed in Table I.

Table I.

Summary of the count, range, mean, standard deviation, and standard error of the mean of the specific antibody–antigen interactions found for different treatments and culturing techniques.

	Count	Range (pN)	Mean ± σ (pN)	Std error (pN)
NC micromass	353	956	188 ± 138	7.33
PC micromass	340	3360	260 ± 347	18.8
NC pellet	78	373	116 ± 68.0	7.70
PC pellet	100	1180	164 ± 164	16.4
NC OHP	336	2580	266 ± 362	19.7
PC OHP	433	1350	190 ± 154	7.41

A. Strength of a single β1-integrin antibody–antigen interaction force

Our results indicate that the average strength of a single β1-integrin antibody–antigen binding is 78 ± 10 pN (Fig. 3). Using ANOVA, the strength of β1-integrin antibody–antigen binding was found to be independent of treatment groups (P = 0.170).

The estimated 78 pN for a single β1-integrin antibody–antigen interaction is consistent with literature forces reported for single antibody–antigen interactions. Ranges between 40 and 100 pN (Refs. 18, 20, 21, 28, 39, and 40) with their multiples apparent at forces larger than 100 pN, depending on the pull-off rate used in AFM measurements were previously reported.

B. Forces involved in multiple antibody–antigen interactions

In addition to single antibody–antigen interactions, larger adhesion forces formed observable specific peaks as indicated in Fig. 2. The average for the second range of specific forces was found to be 158 ± 16 pN, and the average for the third range of specific forces was found to be 240 ± 23 pN. When compared to the 78 pN, which is representative of a single binding. These higher order bindings are consistent with expectations of multiple bindings. We can expect that binding of two β1-integrins to two antigens will have a strength of 156 pN and three bindings to have a strength of 234 pN. By comparing these expected values to the means of the experimental data obtained for the second and third ranges, we can conclude that specific adhesion around 158 pN represents two antigens binding two antibodies and a specific adhesion around 240 pN represents three antigens binding three antibodies.^10,20 The actual values chosen for these three ranges for each treatment group are depicted in Fig. 3. Although no statistical differences between treatment groups were observed for single bindings, the multiple bindings show some statistical differences. This could be attributed to incomplete antibody–integrin matching due to steric hindrances.⁴¹

C. Effect of culture method on the distribution of β1-integrin proteins on cell surfaces

Specific adhesion peaks measured between β1-integrins and the β1-integrin antibody-functionalized cantilevers are reported in adhesion maps (supplementary Fig. 1). These example adhesion maps show the number of specific adhesion peaks detected at each pixel. The pie charts (Fig. 4) show the probability of each of those detected peaks to be a single or a multibinding event based on the magnitude of its adhesion forces as established in Sec. III B. For example, if the adhesion force is around 78 pN, then it is considered a single-binding event. If the adhesion force is around 240 pN, then it would be considered a triple-binding event. The distirbution of binding magnitudes tells us whether the integrins are evenly spread across the cell membrane or if they tend to cluster. We found that supplementation with TGF-β3 did not affect the clustering of integrins. This was based on observing more single binding events than multiple binding events for this treatment. However, the different culturing formats, micromass, pellet, and CBR, were dramatically different from each other in terms of the integrin clustering characteristics. For example, the pellet cultures had integrins more evenly spread on the membrane with single integrin binding events making up 56.5% (61.7% for NC and 51.2% for PC) of the total binding events (Table I and Fig. 4).

Fig. 4.

Pie charts that indicate the probabilities of observing specific adhesion peak magnitudes found to be multiples of 78 pN. Magnitudes around each average binding adhesion force shown in Fig. 3 were counted and plotted here. Values under a 120 pN were considered to be a single binding, those between 120 and 200 pN were considered to be a double binding event, and values between 200 and 280 pN were considered to be triple binding events between the anti-β1-integrin antibody and its antigen.

The literature shows that integrin receptors placed 100 nm apart have significantly reduced tension per ligand and diminished capacity for focal adhesion formation compared to integrin receptors spaced 50 nm apart.⁴² When the pellet culture results are viewed with this in mind, the data support the premise that hACSs grown in pellet culture have single integrin molecules spread across the surface. Their abundance, however, leads to higher surface density and a greater likelihood of forming more integrin clusters and focal adhesion complexes compared to other treatments.

D. β1-integrin protein count correlation to β1-integrin mRNA expression

Experimental results show a strong linear correlation between AFM determined β1-integrin protein surface expression and β1-integrin mRNA upregulation [Fig. 5(a)]. Our data also show that increased β1-integrin mRNA expression correlates with more abundant protein manufacture and integration into the membrane surface [Fig. 5(b)]. This further validates the use of AFM as an important tool for characterizing the density of cell–surface receptors on a cell membrane. Hence, AFM measurements and analyses complement those for mRNA analyses. mRNA Taqman analyses aim at understanding the role of mechanotransduction and growth factor supplementation on gene regulation. In comparison, AFM identifies the distribution and clustering of surface membrane proteins on cellular surfaces. As such, when taken in concert with mRNA data, knowledge on whether gene upregulation truly results in protein production and membrane surface incorporation of proteins that are important to ECM adhesion and cellular responsiveness can be gained.

Fig. 5.

(a) Scatter plot that compares β1-integrin mRNA expression for various treatment groups to the specific β1-integrin-antigen specific adhesion events per cell area as estimated from AFM measurements. The solid line is the linear regression described by y = 1.24x + 0.468, r²= 0.767. (b) A scatter plot that shows the relationship between SOX9 mRNA expression and β1-integrin-antigen specific adhesion events for each treatment group. The solid line is the linear regression described by y = 1.65x − 0.178, r²= 0.811.

E. Relationship between β1-integrin expression and tissue mechanical properties

The importance of investigating tissue mechanical properties, in this case assessed by estimating the Young's moduli of tissues engineered using variable treatments, is to evaluate the integrity of the resulting tissues. Tissues with higher Young's moduli have mechanical properties closer to those of native tissues, and therefore, the culture methods used to produce these tissues are considered more ideal. When analyzing the data displayed in Table II, the β1-integrin count was at least two times lower, and the average Young's modulus at least 45 times higher in OHP samples compared to static controls. This inverse relationship between β1-integrin count and tissue Young's modulus persisted in moving from one treatment group to another; if the Young's modulus is increased, β1-integrin count decreased and vice versa. Furthermore, the use of culture medium containing TGF-β3 in static cultures decreased the Young's modulus by 3.7- and 4.4-fold but increased the β1-integrin count by 1.12- and 1.27-fold for micromass and pellet samples, respectively. Samples grown with OHP showed that culturing in the presence of TGF-β3 leads to a 1.9-fold increase in Young's modulus and barely a change, a 1.1-fold decrease, in β1-integrin count. The relationship found between β1-integrin count and tissue Young's modulus indicates that β1-integrin counts increase when the ECM is mechanically insufficient as indicated by a lower Young's modulus. On the contrary, samples with a more robust ECM, as evidenced by having a higher Young's modulus, have less of a need for β1-integrin.

Table II.

Summary of results from three analyses for tissue Young's modulus, SOX9 mRNA expression, and β1-integrin count and their implications on chondrogenesis.

	Young's modulus most probable (kPa ± std. err. mean)		SOX9 mRNA expression		β1-integrin count (normalized)		Conclusions
NC micromass	1.9 ± 0.01	Soft	0.57 ± 0.15	Moderate SOX9 expression suggests moderate chondrogenesis.	0.42	Medium	Cells are not chondrogenically differentiating, and the resulting tissue is soft and has mechanical properties that are far from those of a native tissue. The β1-integrin count is relatively high because cells need it to attach to the weak ECM present in the tissue.
PC Micromass	0.50 ± 0.00		1.0 ± 0.72		0.47
NC pellet	3.8 ± 0.00	Soft	0.88 ± 0.23	Highest relative SOX9 expression suggests the most amount of chondrogenesis is occurring at this point in time.	0.62	Highest	Tissues have significantly lower Young's moduli compared to native tissues but are chondrogenically differentiating. The β1-integrin count is relatively high because cells need to attach to the weak ECM present in the tissue.
PC pellet	0.87 ± 0.00		1.4 ± 0.60		0.79
NC OHP	171 ± 0.34	Hard	0.0 ± 0.0	Lowest relative SOX9 expression suggests the least amount of chondrogenesis is occurring at this point in time	0.19	Low	Tissues are already nearing the Young's modulus of native tissue and therefore most likely mature, which is indicated by the low chondrogenic differentiation at this point in time. The β1-integrin count is low because cells can easily attach to the sufficient ECM matrix present in the tissue.
PC OHP	318 ± 0.36		0.53 ± 0.74		0.18

F. Relationship between β1-integrin protein count and culture method

β1-integrin expression information also may be used to infer important information about matrix production. For example, integrin counts were two and three times higher in micromass and pellet static cultures, respectively, compared to OHP samples (Fig. 5). This can be explained if we consider that micromass and static samples, due to the lack of stimulation, produce ECM that is insufficient mechanically and therefore the cells need more integrins to facilitate cell–ECM interactions. With ongoing dedifferentiation, the β1-integrin gene and the associated surface protein expression levels increase. Ongoing dedifferentiation causes a decrease in collagen II which prompts an increase in β1-integrin to facilitate the adhesion of the cell to the remaining collagen II. This way, the cell is generating an environment suitable for meeting physiological needs.⁴³ This has also been shown to be true in the work of Loeser et al.⁴⁴ for osteoarthritic cartilage and cells in monolayer culture, which have more β1-integrins compared to normal cartilage. In other words, when the matrix is deficient, a greater amount of β1-integrin is needed and the corresponding expression increases. Our AFM results demonstrate a clear link between chondrogenically enhancing conditions, protein expression on cellular surfaces, and the resulting ECM. When the above AFM-unique functions are combined with genotypical data on β1-integrin mRNA upregulation obtained using PCR, phenotypical impacts can be assessed more thoroughly. For example, our results show that the use of OHP in our new CBR system is indeed helpful in stimulating chondrogenesis because it led to reduced β1-integrin cell surface count by 3.1-fold on average, increased Young's modulus by 140-fold on average, and reduced SOX9 mRNA expression by 3.6-fold on average, all of which indicate improved chondrogenesis (Table II).

G. Relationship between β1-integrin protein counts and chondrogenic differentiation

Because SOX9 is known as the master transcription factor in chondrogenesis, its mRNA levels are important as a measure of chondrogenic differentiation. The relationship between SOX9 mRNA expression and β1-integrin protein counts had a generally positive correlation when plotted against β1-integrin protein count [Fig. 5(b)]. The large standard deviations observed in the SOX9 mRNA expression may be due to the typical heterogeneity of biological tissues.⁴⁵ As described earlier, integrins are upregulated when the ECM is deficient,⁴⁴ which is greatly affected by the differentiated state of the cell.⁶ Since mRNA expression only gives us a snapshot of mRNA expression at a single point of time and not the accumulative expression over the entire culture period, it is difficult to say what the SOX9 mRNA expression at day 21 means. Nevertheless, we suggest that chondrogenically differentiating constructs, samples with high SOX9 expression, may be initiating chondrogenesis and therefore have not secreted sufficient ECM. These less differentiated constructs tend to have insufficient ECM, such as in osteoarthritic cartilage,⁴⁶ and need more β1-integrins to compensate for the loss of cell–ECM adhesion. β1-integrins are also important in differentiation and are suspected of having a role in chondrocyte phenotype maintenance.⁴ The samples with low SOX9 expression are the CBR samples. These samples may have expressed sufficient SOX9 earlier in culture and are now chondrogenically differentiated with a robust ECM. With more ECM available, fewer β1-integrins are necessary to facilitate cell–ECM attachment. This is further supported by prior work that showed that β1-integrins are upregulated during chondrogenic differentiation of ASCs and subsequently down-regulated when cells mature.⁴⁷

Although confirmatory studies are needed to tighten the error bars observed in Fig. 5(b) and to determine if observed trends persist, combined AFM and q-PCR studies are important. The advent of ECM interrogation with AFM analysis of Young's moduli and β1-integrin levels in concert with determination of SOX9 mRNA expression is crucial to our understanding of how chondrocytes respond to mechanical and chemical stimulation with regard to differentiation.

H. Effect of TGF-β3 on chondrogenic differentiation

Since TGF-β3 is implicated in stimulating chondrogenesis,^48,49 we would expect to see an upregulation in SOX9 mRNA when cultures are supplemented with TGF-β3. Indeed this is what the data show. SOX9 expression increased by 0.43, 0.55, and 0.53 arbitrary units for, micromass, pellet, and CBR cultures, respectively, when supplemented with TGF-β3. In other words, all TGF-β3 treatment groups (PC) exhibited an increase in chondrogenesis, as evidenced by SOX9 upregulation. These results further validate the use of TGF-β3 as an important growth factor for inducing chondrogenesis.

Our results are not surprising as TGFβ3 is well known to regulate cell differentiation by its activation of several signaling cascades,⁵⁰ and several observations^51,52 suggest that SOX9 expression is under direct control of TGFβ/BMP activity during chondrogenesis. In fact, SOX9 is the critical transcription factor in multiple stages of chondrogenesis as it increases the expression of SOX5 and SOX6 transcription factors and with that enhances the expression of chondrocyte-specific markers such as Col2a1. Supplementation with TGF-β3 especially augments SOX9 expression, even more so than the inhibition of the Wnt signaling pathway.⁵³ Research also shows that ASCs, genetically modified to have enhanced TGF-β3 expression, also undergo enhanced chondrogenesis based on the detection of collagen II and glycosaminoglycans.⁵⁴

With the above, it is clear that TGF-β3 has a positive impact on chondrogenesis, but what is less straightforward is how integrins may be related to chondrogenic events. First, several growth factors, including TGFβ, IGF-I, and IL-1, have been found to regulate integrin expression in chondrocytes.^6,55 Second, chondrocytes with blocked integrin α₁β₁ are hypersensitive to TGF-β1; that is, when such cells are treated with the same concentration of TGF-β1, significantly more (P < 0.0001) phosphorylated Smad2 and Smad3, indicators of TGF-β's conical pathway progression, were found in α₁-null chondrocytes in comparison to wild type cells.⁵⁶ The conclusion of this section, however, is that β1-integrin mapping as afforded by the techniques used in this study will add a critical component to future studies to elucidate how TGF-β affects chondrogenesis by correlating extracellular protein expression to genetic data.

I. Implications regarding articular cartilage tissue engineering

The combination of results including the tissue Young's moduli, SOX9 mRNA expression, and β1-integrin count provides us with information about the differentiation and feasibility of using varied culture methods for enhancing articular cartilage tissue engineering. A summary of these results per culture method can be found in Table II. The micromass samples had very low Young's moduli and SOX9 mRNA expression, meaning that the samples did not have nearly the properties of native human articular cartilage. The pellet samples had slightly higher Young's moduli compared to micromass but were still more than 190 orders of magnitude lower than the native articular cartilage. Despite the pellet cultures not having near the mechanical properties of native cartilage, the high SOX9 expression tells us that the cells are differentiating toward a chondrogenic lineage. The bioreactor samples grown with oscillating pressure had the highest Young's moduli and were nearing the magnitude of native articular cartilage elasticity. Because the SOX9 expression was low, we believe that the OHP samples are not differentiating and may have already matured based on their high Young's moduli. SOX9 mRNA expression is just a snapshot of the chondrogenic differentiation at the time of sampling (day 21) and does not reflect the chondrogenic differentiation that occurred during the culture period. As referred to earlier, the OHP samples had the lowest number of β1-integrins present on the dorsal side of their cells which may be caused by less of a need for cell–ECM attachment in the mature tissues.

IV. CONCLUSIONS

Integrins play an important role in cell–ECM adhesion, and it is suspected that they are responsible for transducing mechanical signals from the ECM into the cell.⁶ In this study, we examined the relationship between β1-integrin expression, chondrogenic differentiation, and tissue mechanical properties. Integrin expression on the surface of hASCs was investigated using AFM in response to TGF-β3 stimulation and/or OHP. We found that the specific adhesion force for a single β1-integrin antibody–antigen interaction was 78 pN, independent of treatment. This suggests that the addition of TGF-β3 or mechanical stimulation does not change the fundamental structure or function of a single β1-integrin. Higher specific interaction forces were found and represented multiple antibody–antigen bindings. In static controls, TGF-β3 caused an increase in β1-integrin expression and this increase was consistent with an increase in SOX9 mRNA expression. In the CBR, TGF-β3 still increased SOX9 expression, but a small, 1.06-fold, decrease in β1-integrin expression was observed. Because bioreactor samples have a stronger ECM, as supported by the 139-fold average increase in Young's moduli compared to static controls, and because β1-integrins are responsible for cell–ECM adhesion, we suspect that static controls express more β1-integrins in order to attach to what little ECM is available. As we would predict, the bioreactor samples have higher Young's moduli, indicating that they have a stronger ECM. In all treatment groups, β1-integrin expression trends were opposite to those of the tissue Young's moduli. In other words, as integrin expression increased, Young's moduli decreased. This again supports our hypothesis that integrin expression is upregulated in tissues with insufficient matrix as shown in the literature on the culture of osteoarthritic cells.