Adhesive and mechanical regulation of mesenchymal stem cell differentiation in human bone marrow and periosteum-derived progenitor cells

Mesenchymal Stem/Stromal Cells (MSCs) are cell populations found in the stroma of mesenchymal tissues such as bone, bone marrow, cartilage, ligaments, and adipose tissue. Because these cell populations can be readily isolated from patients for expansion and differentiation in vitro into at least three different lineages (Augello and De Bari, 2010; Dominici et al., 2006; Pittenger et al., 1999), MSCs are of great interest for clinical therapies. Indeed, protocols for injections of autologous MSCs are already in clinical trials not only for various musculoskeletal tissue replacement therapies including bone, cartilage, and intervertebral discs, but also to treat organ failure (cardiac, lung, liver, pancreas among others) and autoimmune diseases (for reviews, see Gómez-Barrena et al., 2011; Trounson et al., 2011; Tyndall and Gratwohl, 2009). Moreover, MSCs are being developed as a critical cell source in tissue engineering, which involves the ex vivo creation of biological implants intended eventually to replace tissues or functional organs (Marcacci et al., 2007). However, the molecular mechanisms regulating MSC differentiation into the desired terminal lineages are still incompletely understood, impeding efforts to generate useful clinical products from primary cells obtained from patients.

To study MSC differentiation, an array of in vitro assays has been developed. To drive osteogenic differentiation, MSCs are cultured in a serum-containing medium supplemented with dexamethasone, ascorbic acid, and beta-glycerophosphate (Jaiswal et al., 1997; Pittenger et al., 1999). Adipogenesis is induced by insulin, isobutyl-methylxanthine, dexamethasone, and indomethacin (Sekiya et al., 2004), and chondrogenesis is induced in serum-free medium supplemented with TGFβ (Johnstone et al., 1998). Utilizing these media, the differentiation capacities of MSCs derived from different tissues and species have been compared. As virtually all of these can undergo in vitro multilineage differentiation, differences between MSCs have been reported mostly in terms of differentiation efficiency in response to differentiation cocktails for the different lineages (Boeuf and Richter, 2010; Noël et al., 2008; Sakaguchi et al., 2005). However, these comparisons are predicated on the notion that the primary driving factor for MSC differentiation is soluble factors, which may not be true.

Biophysical cues from the microenvironment and neighboring cells may also contribute to the lineage fate and differentiation efficiency of MSCs. For example, seeding density has been shown to impact the efficiency of adipogenesis (Lu et al., 2009) or chondrogenesis in vitro (Nakahara et al., 1991a; Seghatoleslami and Tuan, 2002), and our group and others have demonstrated that lineage commitment of human bone marrow-derived cells (hBMCs) to osteogenesis/adipogenesis (Kilian et al., 2010; McBeath et al., 2004) or myofibroblastogenesis/chondrogenesis (Gao et al., 2010) is in part regulated by cell shape and/or RhoA-mediated cytoskeletal tension. Despite the central role of cell shape and cytoskeletal forces in regulating hBMCs, these findings have not been extended to MSCs from other tissue sources. Thus, the reported differences in differentiation efficiency between various MSC types may partially originate from differences in mechanotransduction in these other MSC types. It is even possible that poor differentiation of varying MSC types could be rescued by manipulating adhesive or mechanical parameters.

Similar to hBMCs, human periosteum-derived cells (hPDCs) display MSC-like multipotency from single cell-derived clonal populations (De Bari et al., 2006) and contribute to robust bone and cartilage growth and repair in vivo (Colnot, 2009; De Bari et al., 2006; Eyckmans and Luyten, 2006). hPDCs are distinct from hBMCs in their tissue of origin, but both cell types arise from mesoderm-derived populations during embryonic development. Because periosteal cells, not bone marrow cells, predominantly contribute to fracture healing in postnatal life (Colnot, 2009; Maes et al., 2010), hPDCs may even be a more suitable cell population for bone engineering applications (Agata et al., 2007; Zhu et al., 2006). Thus hPDCs are a clinically relevant source of primary mesenchymal progenitors that is currently understudied in regard to its molecular regulation of differentiation (Mahajan, 2012), particularly as hPDC-based bone grafts are already being used in the clinic to treat patients (Trautvetter et al., 2011).

Although in vitro differentiation assays for hPDCs were successfully adopted from media conditions used for hBMCs, in this paper, we report a comparison of the effects of biophysical conditions on differentiation of these two primary human MSCs. In particular, we examine the role of cell seeding density, cell shape, and RhoA-mediated cytoskeletal tension on mesenchymal stem cell differentiation in vitro to osteogenic, adipogenic, and chondrogenic lineages.

Cell seeding density, or the number of cells plated per square cm, impacts cell behavior. Our previous studies showed a link between cell seeding density and proliferation rates in multiple cell types (Nelson and Chen, 2002), and between seeding density and differentiation efficiency in MSCs (McBeath et al., 2004). Thus, we first compared the proliferation rates of hBMCs and hPDCs under different seeding densities. Independently of the cell culture vessel used, hPDCs proliferated faster than hBMCs and reached full confluence after 10 days (Fig. 1A) when initially seeded at low seeding density (5000 cells/cm²). In addition, the cell density of hPDCs obtained at confluence was consistently higher as compared to hBMCs, which suggested that hPDCs are smaller than hBMCs. Indeed, side-by-side comparison of adherent hBMC and hPDC populations under phase contrast microscopy suggested that hBMCs were more spread on tissue culture plastic with a broader cytoplasmic body than hPDCs (Fig. 1B, white arrows), whereas hPDCs displayed a more spindle-shaped phenotype (Fig. 1B, black arrows).

To examine how initial seeding density impacted hBMC and hPDC populations, we empirically chose three seeding densities, 5000 (5 k/cm²), 25000 (25 k/cm²), and 85000 (85 k/cm²). When seeded at 5 k/cm², both hPDCs and hBMCs attached efficiently as mostly isolated cells with few if any cell–cell contacts. At 25 k/cm², both cell types arranged in a nearly confluent, but not densely packed, monolayer of cells in which cells remained well spread against the substrate but also displayed augmented cell–cell contact. At 85 k/cm², both cell types appeared smaller, highly packed, and extended processes over each other (Fig. 1B), but a significant number of hBMCs did not attach after seeding at 85 k/cm². Because differentiation experiments last one to two weeks, we quantified the extent to which cell proliferation affected cell density over time at the different seeding densities. hBMCs seeded at 5 k/cm² or 25 k/cm² reached a cell density of 35 k/cm² (±2.6 k/cm²) and 41 k/cm² (±4.1 k/cm²), respectively, whereas hPDCs at those seeding densities expanded to 55 k/cm² (±5 k/cm²) and 64 k/cm² (±4.6 k/cm²). Cell density did not change over time for both cell types when seeded at 85 k/cm², though fewer hBMCs than hPDCs attached at this density (Fig. 1C).

To test whether seeding density modulates differentiation, hBMCs and hPDCs were seeded at 5 k/cm², 25 k/cm², and 85 k/cm² and subsequently treated with chondrogenic, adipogenic, or osteogenic media. Regardless of seeding density, cells cultured exclusively in growth medium did not undergo differentiation, as determined by upregulation of markers for the various lineages (data not shown). For chondrogenic differentiation, cells were plated as both monolayer and as micromass cultures (400 k cells/20 µl droplet). Micromass pellet cultures were included for comparison because it is considered to be the gold standard format for chondrogenesis. Indeed, increasing seeding density promoted expression of chondrogenic marker genes such SOX9 (Fig. 2A), and collagen type IIa1 (COL2A1) (Fig. 2B) for both cell types with the highest levels of gene expression in the micromass cultures. Interestingly, COL2A1 expression was significantly higher in hPDCs as compared to hBMCs, suggesting that hPDCs may be more prone to chondrogenic differentiation than hBMCs. Of note, cells seeded in monolayer at 85 k/cm² detached from tissue culture plastic after three days of treatment with chondrogenic medium. This seeding condition was therefore omitted from further analysis for chondrogenic assays (data not shown).

Increasing seeding density also promoted adipogenic differentiation of both cell types (Fig. 2C–F). After two weeks of treatment, all conditions displayed lipid droplets that stained positive for Oil Red O (Fig. 2C; insets display higher magnification of lipid droplets within differentiated cells). Quantification of Oil Red O positive cells demonstrated an increased number of adipogenic cells at 85 k/cm² as compared to 5 k/cm² for both cell types. However, there was a tenfold higher efficiency of adipogenic differentiation in hBMCs versus hPDCs (Fig. 2D). Similar results were obtained when measuring gene expression of PPARG2, a key transcription factor for adipogenic differentiation (Fig. 2E). Transcription of lipoprotein lipase (LPL), an enzyme involved in lipid metabolism, was upregulated by increased seeding density in hBMCs but not in hPDCs (Fig. 2F).

Unlike chondrogenesis and adipogenesis, the comparisons between hBMCs and hPDCs with respect to osteogenic differentiation were less clear. First, osteogenic medium induced transient alkaline phosphatase (ALP) activity in hBMCs by one week after stimulation, whereas two weeks of osteogenic induction were required to induce ALP activity in hPDCs (Fig. 3A). This difference in ALP kinetics between hBMCs and hPDCs is in line with previous reports (Chai et al., 2011; Eyckmans and Luyten, 2006; Jaiswal et al., 1997). Second, hBMCs but not hPDCs displayed spontaneous ALP activity in growth medium when seeded at high density. Third, seeding hBMCs but not hPDCs at 25 k/cm² resulted in more ALP positive cells (Fig. 3B) and increased mineralization (Fig. 3A) as compared to seeding at 5 k/cm². These findings at the level of protein activity were further supported by gene expression analysis for the osteogenic markers RUNX2, ALP, and osteocalcin (OCN). After two weeks of stimulation with osteogenic medium, hBMCs plated very densely at 85 k/cm² displayed higher levels of RUNX2 and OCN in comparison with lower seeding densities, whereas ALP gene expression peaked at one week in the intermediate seeding density 25 k/cm² condition (Fig. 3C). By the third week, extensive calcification of the cell-secreted matrix prevented efficient RNA extraction from hBMCs (data not shown). In contrast to hBMCs, seeding density did not affect expression of the osteogenic marker genes in hPDCs during the first two weeks of stimulation. By three weeks after stimulation, however, hPDCs seeded at low density tended to express higher mRNA levels of ALP and OCN (Fig. 3C). The absence of an effect of cell density on RUNX2 and the relatively mild effect on ALP and OCN expression in hPDCs suggest that osteogenic differentiation in hPDCs may be less sensitive than the hBMCs to cues affected by cell density.

The augmented differentiation response observed at higher seeding densities could be attributed to multiple co-varying factors that inevitably change when increasing the number of cells in a culture vessel, such as cell-extracellular matrix adhesion, cell–cell contact, paracrine signaling, and cell shape. In contrast to epithelial cells that form a monolayer upon full confluency, primary MSCs can expand on top of each other, giving rise to multilayered cell sheets. By consequence, the surface area to which a cell can adhere is not exclusively restricted by the number of seeded cells. To gain insight into how seeding density impacts cell shape in hPDCs and hBMCs, we tagged a fraction of the cells with cytoplasmic GFP and mixed them with non-fluorescent cells prior to seeding at different cell densities. The next day, the medium was replaced by growth, osteogenic, or adipogenic medium for two more days. Measurements of projected area of cells at 20 hours after seeding confirmed our initial observation that hPDCs were less spread than hBMCs when seeded at 5 k/cm², and that cell area was significantly reduced in conditions seeded at 25 k/cm² (hBMCs) and 85 k/cm² (hBMCs and hPDCs) (Fig. 4A,B). Furthermore, spread area remained inversely correlated with initial seeding density up to day 3. For both cell types, the variability in cell size was consistently reduced at high seeding densities (Fig. 4B,C), suggesting that cell shape is more homogenous in high cell density cultures. In addition to the effects of seeding density on cell area, treatment with adipogenic medium also reduced cell area at higher seeding densities (Fig. 4C), demonstrating that cell shape is in part regulated by soluble factors.

Because cell shape appeared to be influenced by cell density, we utilized fibronectin micropatterning techniques to directly assess the impact of cell shape on MSC differentiation without the complex effects of seeding density. Briefly, 25μm×25µm (625 µm²) or 100μm×100µm (10000 µm²) fibronectin islands were stamped on a polydimethylsiloxane surface, followed by seeding of hBMCs and hPDCs. We used smaller fibronectin patterns (625 µm²) here than previously utilized for hBMCs (1024 µm²) (Gao et al., 2010; McBeath et al., 2004) since hPDCs are smaller than hBMCs (Fig. 4B). Cells that adhered to 625 µm² fibronectin islands adopted a round, unspread cell shape, while the cells on the larger fibronectin patterns were able to spread out to a surface area of 10000 µm² (Fig. 5A). As such cell shape is altered without modulating seeding density.

Under these conditions, treatment of hBMCs and hPDCs with chondrogenic medium for two weeks resulted in an increased percentage of cells that stained positive for collagen type II when adherent on small 625 µm² versus large 10000 µm² islands (Fig. 5A). Similarly, hBMCs seeded on small patterns versus large patterns displayed higher levels of mRNA expression for the chondrogenic marker SOX9 (Fig. 5B). Although a similar trend was observed in hPDCs for COL2A1, the effect was not significant due to the high variability between experiments. COL2A1 expression in hBMCs was only detected in small patterned cells in one of four experiments (Fig. 5B). These micropatterned data show that chondrogenic gene expression in hPDCs appears to be less regulated by cell shape than that in hBMCs, whereas chondrogenic expression at the protein level is similarly regulated by shape in both cell types. Thus the promotion of chondrogenic gene expression by increased cell density in hPDCs may be more heavily influenced by density-induced cues other than cell spreading, such as increased cell–cell signaling. Nonetheless, the trend toward improved chondrogenesis in small versus large patterns still suggests some contribution of cell spreading to the observed increase in chondrogenesis with high seeding densities.

Comparable to chondrogenesis, directly restricting cell spreading also promoted adipogenic differentiation. The percentage of Oil Red O positive hBMCs and hPDCs was higher in cells constrained on small fibronectin islands as compared to cells patterned on large islands (Fig. 5C). These data were supported by gene expression analysis of PPARG2 and LPL, which tended toward higher levels of expression in unspread as compared to spread hBMCs (Fig. 5D). Although similar trends were obtained for unspread and spread hPDCs within single experiments, there was considerable variability across experiments (Fig. 5D).

In contrast to the adipogenic and chondrogenic lineages, the osteogenic marker ALP was expressed at lower levels in unspread versus spread cells. When spread on large fibronectin islands, more hBMCs and hPDCs displayed ALP activity (Fig. 5E) and increased ALP gene expression (Fig. 5F) as compared to size-restricted cells. However, cells adherent to small patterns showed significantly higher levels of RUNX2 mRNA transcripts (Fig. 5F). As RUNX2 is an earlier marker of osteogenesis than ALP (Otto et al., 1997), these findings are consistent with a possibility that different degrees of cell spreading may be optimal for different stages of differentiation. Interestingly, the observation that increased cell spreading enhanced ALP expression via micropatterning, and that increasing cell seeding density (which decreases cell spreading) either enhanced (hBMCs) or decreased ALP expression (hPDCs), suggests that increased hBMC seeding density introduces an additional pro-osteogenic cue that is independent of cell shape.

Because changes in cell shape alter cytoskeletal tension and RhoA/Rho Kinase (ROCK) activity (Bhadriraju et al., 2007; McBeath et al., 2004), it has been hypothesized that RhoA/ROCK signaling may regulate MSC differentiation. Indeed, we have previously demonstrated in hBMCs that manipulation of RhoA/ROCK signaling modulates osteogenic versus adipogenic commitment when cells are treated with a mixed osteogenic/adipogenic stimulation medium (McBeath et al., 2004). In addition, hBMC osteogenesis is inhibited by treatment with the ROCK inhibitor Y27632 (Wang et al., 2011). Others have shown that RhoA/ROCK signaling also regulates chondrogenic differentiation in adipose tissue-derived MSCs (Lu et al., 2010) and gingiva-derived MSCs (Hsu et al., 2012). No data have yet been reported, however, regarding a role for RhoA/ROCK signaling in hPDCs. Here we examined whether inhibition of RhoA/ROCK signaling, by addition of the ROCK inhibitor Y27632, alters osteogenic, adipogenic, and chondrogenic differentiation in hPDCs seeded at low and high density.

Y27632 treatment caused micromasses to remain adherent to tissue culture plastic as compared with non-treated controls indicating drug efficacy (Fig. 6A, bright field image inset). However, Y27632 treatment did not significantly influence chondrogenic differentiation for any conditions as measured by alcian blue uptake, even though measurements trended toward significance for Y27632-treated micromasses (Fig. 6A). At the gene expression level, chondrogenesis for hPDCs seeded as micromasses was modestly enhanced by Y27632 treatment, but not for hPDCs seeded at low density (Fig. 6B).

For adipogenesis, Y27632 treatment increased the percentage of Oil Red O positive cells (Fig. 6C) and upregulated gene transcription of PPARG2 and LPL (Fig. 6D) in hPDCs seeded at low density. The additional fraction of lipid droplet-forming adipogenic cells induced by Y27632 at low seeding density did not reach the level of adipogenic cells without Y27632 at high seeding density. These trends did not reach statistical significance for cells seeded at high density.

In contrast, ALP activity (Fig. 6E) and ALP and RUNX2 gene expression (Fig. 6F) were significantly downregulated in the presence of Y27632 for hPDCs seeded at low density. In hPDCs seeded at high density, osteogenic gene expression but not ALP activity was suppressed by Y27632. Interestingly, Y27632 treatment even in osteogenic differentiation medium resulted in a small but detectable number of cells that produced lipid droplets staining positive for Oil Red O (supplementary material Fig. S1). These results in hPDCs are consistent with reports of Y27632-induced effects on mesenchymal differentiation in hBMCs (McBeath et al., 2004), suggesting that cytoskeletal tension-mediated regulation of differentiation is conserved for hPDCs. Of note, addition of Y27632 did not significantly affect proliferation of hPDCs cultured in any of the differentiation media as compared to untreated controls over the duration of the differentiation experiments (data not shown), indicating that Y27632 was not toxic to the cells.

Cells undergo dynamic changes in cell shape during embryonic morphogenesis in vivo (Mammoto and Ingber, 2010), but only recently have scientists had the tools to manipulate cell shape in vitro, for instance by micropatterning (Eyckmans et al., 2011). With the ability to test the hypothesis in vitro that cell shape influences stem cell fate, we and other groups have recently demonstrated that cell shape can modulate the multipotential lineage commitment of hBMCs (Gao et al., 2010; Kilian et al., 2010; McBeath et al., 2004). Micropatterned hBMCs in mixed osteogenic/adipogenic media give rise to either fate depending on the initial spread cell area, and contractility inhibitors promote “round” cell fates like adipogenesis while inhibiting osteogenic differentiation (McBeath et al., 2004). In addition, increases in cell density impose physical constraints on cell shape, leading to decreased projected cell area (McBeath et al., 2004; Nelson and Chen, 2002; Xu et al., 2010). In this study, we extend these findings to hPDCs and show that differentiation efficiency in hBMC and hPDC populations can be guided by changing seeding density or cell shape. In both cell types, increasing cell seeding density or reducing cell spreading yielded better adipogenic and chondrogenic differentiation. In addition, the effect of inhibiting RhoA/ROCK signaling in hPDCs partially mimicked cell density/shape dependent differentiation effects, which is consistent with what has already been published for hBMCs (Gao et al., 2010; McBeath et al., 2004; Wang et al., 2011). Moreover, this study is in line with prior work suggesting that ROCK acts downstream of the effects of cell shape on differentiation (McBeath et al., 2004), since despite the fact that Y27632 induces a differentiation phenotype of a less spread cell, morphologically Y27632 either does not impact cell shape or enhances cell spreading.

Interestingly, while our data showed consistent trends for regulation of chondrogenic and adipogenic differentiation by cell adhesion, shape, and mechanics in both hBMCs and hPDCs, mechanoregulation of osteogenic differentiation was more complex. Increasing seeding density significantly decreased (hPDCs) or increased (hBMCs) the fraction of MSCs expressing active ALP protein, but ALP gene expression and protein activity were higher in MSCs on spread versus small patterns. Because higher cell density not only decreased cell spreading, but also would be expected to increase the level of paracrine (diffusible) and juxtacrine (cell–cell contact) signaling from neighboring cells, these data suggest the presence of a pro-osteogenic signal with increasing seeding density that opposes and masks the underlying spreading-dependent osteogenic signal. That is, the biomechanical effect of cell spreading on promoting osteogenic differentiation may be concealed by pro-osteogenic cell–cell signaling effects in high density cultures but exposed by the micropatterned cultures. Cell type-specific differences in sensitivity to this phenomenon may explain why osteogenesis in hPDCs was less affected by cell density manipulations than in hBMCs, while both cell types responded to differences in spread cell area through micropatterning. In addition, we observed that confining hPDCs on small fibronectin islands yielded a higher percentage of adipogenic differentiation in comparison to hPDCs with the same spread area in high density cultures, suggesting that these paracrine/juxtacrine signals also suppress hPDC adipogenesis. Paracrine signaling may also strongly influence chondrogenic differentiation in hPDCs, since chondrogenic gene expression was promoted by high density cultures to a greater degree than by small, unspread patterned cultures. Indeed, paracrine signals have been shown to impact differentiation. For instance, changes in paracrine PDGF signaling can modulate the cardiomyogenic potential of mouse BMCs (Pallante et al., 2007). In another study using a microfluidic platform designed to continuously wash away cell-secreted factors, investigators found a requirement for paracrine signaling in mouse embryonic stem cells committing to the neuroectodermal lineage (Blagovic et al., 2011). Hence increased cell density may partially mediate differentiation effects by enhancing paracrine signaling, via both the greater numbers of cells able to secrete signaling factors and the decreased diffusion distance between neighboring cells. In fact, a critical mass of MSCs in osteogenic scaffolds is required for bone formation to occur in vivo (Eyckmans et al., 2010; Mankani et al., 2007), indicating that cell–cell signaling is indispensible in certain differentiation assays. Our results show that effects of both shape and cell–cell signaling are important for chondrogenesis and osteogenesis in hPDCs, suggesting that these cues are relevant to cartilage and bone formation and repair. These findings are consistent with literature showing that hPDCs implanted at high cell density, but not low density, in vivo appear to pass through an early chondrogenic phase (Nakahara et al., 1991b).

It is clear from our data and other reports, however, that differences in cell shape can affect differentiation independently of cell–cell contact. The mechanisms by which cell shape regulates MSC differentiation and other cell behaviors remain to be elucidated. There is mounting evidence for the involvement of focal adhesion signaling and cytoskeletal tension as key factors. Focal adhesions serve dual functions as biophysical anchorage points between the cytoskeleton and the extracellular matrix and as active signaling sites. Altering adhesive conditions may influence the availability or activity of various signaling molecules that localize to adhesions and have been implicated in differentiation, including FAK, Src, β-catenin, and MAPK proteins, among others (Blagovic et al., 2011; Kuo et al., 2011; Miyamoto et al., 1995; Schiller et al., 2011). Additionally, focal adhesion maturation is correlated with increased RhoA-mediated cytoskeletal tension and the formation of stress fibers (Ridley and Hall, 1992; Riveline et al., 2001), both of which contribute to cell shape (Chen et al., 2003; Tan et al., 2003). Inhibition of cytoskeletal tension via drugs like the ROCK inhibitor Y27632 has been reported to regulate mesenchymal differentiation in hBMCs and adipose-derived MSCs (Lu et al., 2010; McBeath et al., 2004; Wang et al., 2011; Xu et al., 2010). Similarly in our experiments, treatment of hPDCs with Y27632 promoted adipogenic differentiation but reduced osteogenic differentiation. Although cytoskeletal tension is required for stabilization of adhesions, it also impacts cellular signaling at the transcriptional level. Actomyosin-mediated contractility regulates nucleocytoplasmic transport of mRNA binding proteins (Lee et al., 2004; Oleynikov and Singer, 2003) and transcription factors such as Yes-associated protein (YAP) and transcriptional coactivator with PDZ binding motif (TAZ), both of which are involved in osteogenic/adipogenic differentiation (Dupont et al., 2011). Actin polymerization also drives nuclear localization of serum response factor (SRF) and its co-factor myocardin-related transcription factor (MRTF) (Vartiainen et al., 2007), which can drive terminal differentiation of epidermal stem cells (Connelly et al., 2010) and epithelial-mesenchymal transition (Gomez et al., 2010). Thus cytoskeletal tension and mechanochemical signaling at adhesions contribute to the role of cell shape in regulation of stem cell differentiation.

Taken together, these studies suggest that mechanoadhesive regulation of differentiation extends beyond hBMCs to hPDCs, and that this process likely plays a role in many if not all MSCs. As such, adhesive and mechanical cues will need to be optimized for each MSC source in order to fairly compare differentiation of different stem cells. For instance, in our experiments, the best condition for adipogenic differentiation of hBMCs was at high seeding density, whereas hPDC adipogenic commitment was better on small micropatterns with minimized cell–cell signals. Another example is that because hPDCs are smaller and spread less than hBMCs, the seeding densities and micropattern area that impacted differentiation were shifted. Thus, the general mechanisms by which biophysical cues regulate behavior may be conserved across different stem cells, but the relative importance and optimal conditions may differ for each cell type. Given this, comparison studies between cell types should carefully control for and describe biophysical parameters for accurate replication in future work, which may also reduce some of the reported variability in MSC differentiation studies.

hBMCs from two donors were purchased from Lonza (Walkersville, MD, USA) and expanded in Dulbecco's Modified Eagle Medium (DMEM, 1 g/L glucose; Mediatech Inc, Manassas, VA, USA) supplemented with 10% Fetal Bovine Serum (FBS) (Atlanta Biologicals, Lawrenceville, GA, USA). hPDCs were isolated from four patients and expanded in growth medium (DMEM, 4.5 g/L glucose + 10% FBS) as described previously (Eyckmans et al., 2010). In brief, periosteal biopsies were harvested from the proximal tibia of male and female patients (mean: 26 (±14) years of age), digested in collagenase (Invitrogen, Grand Island, NY, USA) in growth medium, collected by centrifugation, and seeded in a T25 flask in growth medium. Non-adherent cells were removed after 5 days by changing the medium. The ethical committee for Human Medical Research (Katholieke Universiteit Leuven) approved all procedures, and patient informed consent forms were obtained.

To establish growth curves, hBMCs and hPDCs were seeded in 12 well plates at 5 k/cm². At different time points, cells were stained with DAPI and imaged with a SPOT CCD camera (Diagnostic Instruments, Sterling Heights, MI, USA) mounted on a Nikon TE200 microscope at 100× magnification. Nuclei were counted with Cell Profiler (Carpenter et al., 2006).

To assess cell shape, cells infected with a lentiviral vector encoding for enhanced Green Fluorescent Protein (eGFP) were mixed with non-infected cells and seeded at 5 k/cm², 25 k/cm², or 85 k/cm² in growth medium. The next day, the cultures were imaged at 100× and medium was changed to either growth medium, osteogenic, or adipogenic medium (Lonza). Two days later, GFP-positive cells were imaged and analyzed in ImageJ (Abramoff et al., 2004). For each condition, 30 to 60 cells were manually outlined and projected cell area was measured.

To control cell shape, 25μm×25µm (625 µm²) or 100μm×100µm (10000 µm²) adhesive islands of fibronectin were stamped on Polydimethylsiloxane (PDMS) (Dow Corning Sylgard 184, Midland, MI, USA) substrates as described previously (Tan et al., 2002). Briefly, glass coverslips were spin-coated with PDMS, baked to cure the PDMS, cleaned in 70% ethanol, rinsed with PBS, and treated with UV-ozone (UVO Cleaner, Jelight Company Inc, Irvine, CA, USA) for 5 minutes. Fibronectin (BD Biosciences, San Jose, CA, USA) was loaded on PDMS positive-featured stamps at 50 µg/ml for 1 hour, rinsed in sterile water, dried with a nitrogen gun, and stamped onto the PDMS-glass substrates. Subsequently, the stamped substrates were blocked with pluronic F127 (BASF, Mt. Olive, NJ, USA) to prevent cell attachment outside the fibronectin islands, and rinsed with PBS several times before cell seeding.

For differentiation assays, passage 4 to passage 7 cells were seeded in 24 or 12 well plates at 5 k/cm², 25 k/cm², 85 k cells/cm², in micromasses (400 k cells in a 20 µl droplet), or on patterned fibronectin substrates; all in growth medium (DMEM + 10% FBS (Atlanta Biologicals, Lawrenceville, GA, USA). After 24 hours, the medium was replaced with osteogenic, adipogenic, or chondrogenic induction media (Lonza Poetics^TM: hMSC Osteogenic Differentiation BulletKit^TM (osteo basal medium, dexamethasone, ascorbate, mesenchymal cell growth supplement, L-glutamine, penicillin, streptomycin, β-glycerophosphate), hMSC Adipogenic Differentiation BulletKit^TM (adipo basal medium, indomethacin, isobutylmethylxanthine, recombinant human insulin, dexamethasone, GA-1000 (gentamicin, amphotericin B), mesenchymal cell growth supplement, L-glutamine), or hMSC Chondrogenic Differentiation BulletKit^TM [chondro basal medium, dexamethasone, ascorbate, ITS+ supplement (insulin, human transferrin, selenous acid), pyruvate, proline, GA-1000, L-glutamine) supplemented with 10 ng/ml TGFβ1 (StemRD, Burlingame, CA, USA)]. Differentiation induction media were changed every other day for up to three weeks unless otherwise stated. To reduce cytoskeletal tension during differentiation, a Rho Kinase (ROCK) inhibitor, Y27632 (Tocris Bioscience, R&D systems Inc, Minneapolis, MN, USA), was added to differentiation medium at a final concentration of 10 µM or 25 µM.

Cells were fixed with 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) and stained for alkaline phosphatase with Fast Blue RR Salt/naphthol (Sigma–Aldrich, St. Louis, MO, USA), lipid droplets with Oil Red O (Sigma–Aldrich) in 60% isopropanol, and nuclei with DAPI or hematoxylin. Highly sulphated mucins, characteristic for cartilage-like matrix, were stained with alcian blue 8GX (in 0.5 M HCl, pH 1.5) (Sigma–Aldrich) for 24 hours. After washing, dye bound to the matrix was extracted with 6 M guanidium chloride (Sigma–Aldrich) and absorbance was measured at 592 nm with a GENios microplate reader (Tecan, Durham, NC, USA). Mineralization was stained by Alizarin Red S (Acros Organics, Fair Lawn, NJ) as a 2% solution in water with pH adjusted to 4.2 with 10% ammonium hydroxide for 15 min followed by washes. Immunohistochemistry for collagen type II was performed with a monoclonal mouse anti-human collagen type II antibody (1:300 dilution) (MP Biomedicals, Solon, OH, USA) as a primary antibody and a goat anti-mouse alexa 488 conjugated (Invitrogen) as secondary antibody, as described previously (Gao et al., 2010).

RNA was extracted with the RNA micro/minikit from Qiagen (Qiagen Sciences, Germantown, MD, USA) according to the manufacturer's protocol. Complementary DNA (cDNA) was obtained by reverse transcription of 1 µg of total RNA with qScript^TM cDNA Supermix (Quanta Sciences, Gaithersburg, MD, USA). Realtime PCR was performed in 20 µl reactions using the PerfeCTa^TM mastermix (Quanta Sciences) and a 7300 Real Time PCR System (Applied Biosystems, Foster City, CA, USA) cycling at 95°C for 3 seconds and 60°C for 20 seconds for 40 cycles. The following Taqman® primer/probe sets (Applied Biosystems) were utilized for marker gene expression: RUNX2: Hs00231692_m1, ALP: Hs00758162_m1; OCN: Hs01587814_g1; SOX9: Hs00165814_m1; COL2a1: Hs00264051_m1; PPARG2: Hs00234592; LPL: Hs00173425; GAPDH: Hs99999905_m1. Relative gene expression is shown either as (1) 2^−ΔCT with ΔCT = [CT value for Gene of interest−CT value for Glyceraldehyde 3-phosphate dehydrogenase (GAPDH)] where the Gene of Interest is a differentiation marker, or (2) 2^−ΔΔCT as normalized to the 5 k/cm² control condition in the same media condition, as detailed in each figure legend.

For each graph, the mean represents the average of three to six experiments. Statistical comparisons between experimental conditions were performed utilizing a Mann-Whitney U test or an unpaired two-tailed t-test. Cell projected areas under different density and media conditions were compared with a one-way ANOVA followed by Tukey post hoc test. For all tests, statistical significance was assigned at p-value ≤ 0.05.

This work was supported by National Institutes of Health [grant numbers GM60692, EB 00262]. J.E. is a postdoctoral fellow of the Research Foundation – Flanders (FWO).