ABSTRACT
Efficient homing of human mesenchymal stem cells (hMSCs) is likely to be dictated by a combination of physical and chemical factors present in the microenvironment. However, crosstalk between the physical and chemical cues remains incompletely understood. Here, we address this question by probing the efficiency of epidermal growth factor (EGF)-induced hMSC chemotaxis on substrates of varying stiffness (3, 30 and 600 kPa) inside a polydimethylsiloxane (PDMS) microfluidic device. Chemotactic speed was found to be the sum of a stiffness-dependent component and a chemokine concentration-dependent component. While the stiffness-dependent component scaled inversely with stiffness, the chemotactic component was independent of stiffness. Faster chemotaxis on the softest 3 kPa substrates is attributed to a combination of weaker adhesions and higher protrusion rate. While chemotaxis was mildly sensitive to contractility inhibitors, suppression of chemotaxis upon actin depolymerization demonstrates the role of actin-mediated protrusions in driving chemotaxis. In addition to highlighting the collective influence of physical and chemical cues in chemotactic migration, our results suggest that hMSC homing is more efficient on softer substrates.
INTRODUCTION
Human mesenchymal stem cells (hMSCs) are multipotent cells isolated from multiple locations within the body, including bone marrow, adipose tissues and umbilical cord, and have the potential to differentiate into bone, cartilage and fat tissues (Piersma et al., 1985; Clark and Keating, 1995; Pereira et al., 1998). Migration of hMSCs is induced by homing signals released by other cells at the site of injury and/or inflammation (Kang et al., 2012; Wu and Zhao, 2012). Apart from promoting angiogenesis, hMSC migration to the tumour microenvironment has been associated with both inhibition and enhancement of tumour growth (Khakoo et al., 2006; Karnoub et al., 2007). The potential of these cells to home or migrate to the injured tissues or sites can be exploited for applications in tissue engineering as well as cancer therapeutics (Chapel et al., 2003; Chavakis et al., 2008).
MSCs migrate in response to various chemokines and growth factors including TNF-α (tumour necrosis factor-α), IL-6 (interleukin-6), IFN-γ (interferon-γ), SDF-1 (stromal derived factor-1), MCP-1 (monocyte chemoattractant protein-1), PDGF (platelet-derived growth factor), HGF (hepatocyte growth factor), VEGF (vascular endothelial growth factor) and bFGF (basic fibroblast growth factor) via expression of multiple surface receptors including VCAM-1, CXC motif and CD44, and activation of downstream signalling pathways (Sun et al., 2014). For example, hMSC migration triggered by spinal cord injury is associated with overexpression of PDGFB (Hata et al., 2010) by activation of PI3K and Rac (Kim et al., 2008). Similarly, the inflammatory cytokine TNF-α induces hMSC migration by activating ERK and p38 MAPK pathways (Fu et al., 2009). In addition, the directional cue is reinforced by sensitization of hMSCs towards other chemokines such as SDF-1 (Schmid and Huisken, 2015), which serves as a chemoattractant for multiple types of stem cells, including MSCs, haematopoietic stem cells, neural stem cells and endothelial progenitor cells (Lau and Wang, 2011). Apart from sensitization to one growth factor, hMSC chemotaxis is enhanced by the synergy of multiple chemokines. For example, VEGF and PDGF have been demonstrated to synergistically modulate hMSC chemotaxis (Ball et al., 2007). hMSC homing to tumours is also driven by multiple factors, including IL-6 (Liu et al., 2011), TNF-α (Fu et al., 2009), TGF-β (Shinojima et al., 2013) and SDF-1 (Gao et al., 2009).
In addition to soluble cues, cells are also sensitive to physical properties of the matrix (Hou et al., 2011; Vincent et al., 2013). A plethora of studies have demonstrated the role of ECM stiffness, topography and dimensionality in regulating a wide range of cellular processes, including cell spreading, cell proliferation, cell migration and cell differentiation (Kyburz and Anseth, 2013; Hadden et al., 2017). For example, hMSC fate has been shown to be sensitive to ECM stiffness, with hMSCs differentiating into neurogenic lineages on soft neuron-like environments and into osteogenic lineages on stiff bone-like environments (Engler et al., 2006). Although many different cell types tend to spread more and migrate faster on stiffer substrates, recent reports have documented faster migration of some cells, including dermal fibroblasts and ovarian cancer cells on softer substrates (Ghosh et al., 2007). While hMSCs have been reported to migrate up rigidity gradients (durotaxis) in vitro (Vincent et al., 2013), their in vivo migration behaviour is likely to be dictated by a combination of various physical cues sensed by them. Indeed, in polydimethylsiloxane (PDMS) microfluidic devices, which provide multiple physical cues (Hou et al., 2011), hMSCs are maximally migratory on substrates of intermediate stiffness, roughness and hydrophobicity (Menon et al., 2015).
While studies have individually demonstrated the role of chemical (Yoon et al., 2016; Ball et al., 2007; Watts et al., 2016) and physical cues (Hadden et al., 2017; Hou et al., 2011; Vincent et al., 2013) in directing hMSC migration, there have been no studies that have comprehensively probed the collective influence of these two cues. Here, we address this question by studying epidermal growth factor (EGF)-induced chemotaxis of hMSCs through PDMS microchannels with varying substrate stiffness. We show that under identical chemokine gradients, hMSCs migrate fastest on 3 kPa soft PDMS substrates compared with stiffer (30 kPa and 600 kPa) PDMS substrates. We then show that faster hMSC chemotaxis on softer PDMS substrates is associated with formation of smaller adhesions. Finally, by perturbing the cytoskeleton with various drugs, we show that hMSC chemotaxis is mediated by actin-dependent polymerization. Collectively, our studies indicate that soft substrates enhance chemotaxis of hMSCs.
RESULTS
Human mesenchymal stem cells exhibit increased motility on soft PDMS substrates
Different tissues in the body possess different mechanical properties which change both during development and in different diseases. In vivo, there is wide range of stiffness from 0.5 kPa (adipose tissue) to 20 MPa (bone) (Handorf et al., 2015). Using polyacrylamide gels, many studies have demonstrated the role of substrate stiffness in regulating spreading, motility and differentiation of hMSCs. Since microfluidic devices are fabricated using PDMS, to ensure proper bonding of the top PDMS devices with the bottom substrates, we chose to make the bottom substrates using PDMS as well. To probe the effect of PDMS stiffness on hMSC spreading and motility, PDMS gels of varying stiffness were fabricated by mixing Sylgard 527 and 184 at different ratios (5:1 and 50:1, respectively), and their stiffness was quantified by using rheology (Chaudhury and Whitesides, 1991) (Fig. 1A). While the Young's Modulus (E) of gels fabricated using Sylgard 184 was estimated to be ∼600 kPa, combination of Sylgard 527 and Sylgard 184 at different ratios allowed us to fabricate gels 20 or 200 times softer at 30 kPa and 3 kPa, respectively. Independent verification of stiffness of these gels using JKR (Johnson, Kendall, Roberts) theory yielded almost identical values (Fig. S1A). The surface properties of the different PDMS gels were probed by contact angle measurements. While the PDMS gels were hydrophobic in nature, after plasma treatment, the surfaces became hydrophilic (Fig. S1B). PDMS gels were functionalized at 4 different concentrations of 1, 5, 25 and 50 µg/ml collagen, which fall within the range of values widely reported in the literature (Palchesko et al., 2012; Trappmann et al., 2012). Collagen intensity analysis revealed that independent of PDMS gel stiffness, all the gels were uniformly coated, with maximum intensity observed on 50 µg/ml collagen-coated surfaces (Fig. S1C,D). Consistent with these intensity measurements, on a given stiffness substrate, cells were spread to the same extent (Fig. S1E,F). Since the focus of this study was to probe the effect of substrate stiffness on chemotaxis, for all subsequent studies on PDMS gels, we chose to work at a collagen concentration of 50 µg/ml.
Spreading of hMSCs on PDMS gels exhibited a weak dependence on ECM stiffness (Fig. 1B,C). In comparison to spreading on PDMS gels, on 0.5, 3 and 35 kPa polyacrylamide gels functionalized with 1, 5 or 25 µg/ml collagen, cell spreading was found to increase with gel stiffness, with cells exhibiting a rounded morphology on the softest 0.5 kPa polyacrylamide gels (Fig. S2A,B). Since the PDMS substrates used here are moderately stiff, hMSC spreading on the polyacrylamide gels suggest that these cells spread much less on substrates softer than the softest, 3 kPa PDMS used here. Although hMSC spreading on PDMS gels was relatively insensitive to stiffness over the range probed in this study, hMSCs were maximally motile on the softest 3 kPa PDMS gels (Fig. 1D,E). However, no change in cell persistence was observed across the different gels (Fig. 1F). In comparison, on polyacrylamide gels functionalized with 5 or 25 µg/ml collagen, hMSCs exhibited a biphasic motility pattern (Fig. S2C) similar to that reported in other studies (Peyton and Putnam, 2005; Shebanova and Hammer, 2012; Tee et al., 2011). Maximum cell speed observed on 3 kPa polyacrylamide gels was comparable to that observed on 3 kPa PDMS gels. Together, these results suggest that in spite of differences in the nature of dependence of hMSC spreading and motility on gel stiffness, cells were maximally motile on gels with 3 kPa stiffness.
hMSCs exhibit fastest chemotaxis on the softest substrates
Our results thus far suggest that hMSCs are more motile on soft PDMS gels under 2D conditions. Next, PDMS devices were designed to provide confinement and a gradient of chemokine for understanding the combined effect of mechanical and chemical cues on cell motility. A simple design was chosen wherein two side channels were connected by multiple transverse channels (Fig. 2A). This design has been previously shown to support the formation of a linear gradient (Irimia et al., 2007). Assuming an epidermal growth factor (EGF) diffusion constant of 8×10−11m2/s as reported previously (Thorne et al., 2004), COMSOL 3D simulations were performed to assess temporal evolution of the EGF gradient. Simulations revealed a decrease in the EGF concentration gradient slope from 30 min to 10 h. The plot of normalized concentration of EGF depicts the decrease in slope from 0.91 (a.u.) to 0.43 (a.u.) between 30 min and 10 h (Fig. 2B). To compare the simulation predictions with experiments, temporal evolution of the gradient was experimentally assessed by introducing FITC-dextran at the inlet of one of the side channels and PBS at the opposite inlet of the other side channel (Fig. 2C). Intensity quantification of the channels along the transverse channel length revealed results similar to those obtained upon simulation with decrease in the slope of the concentration gradient at 10 h (Fig. 2D).
To probe the collective influence of physical and chemical cues on the migratory potential of hMSCs, the above-designed microfluidic device was bonded to different stiffness PDMS gels functionalized with 50 µg/ml collagen, hMSCs were seeded on one of the side channels and cell motility tracked for 12 h in the absence and presence of EGF (Fig. 3A). EGF was introduced at the opposite side channel at three different concentrations of 1, 5 and 50 ng/ml, respectively. Since the cell outlines were not clearly visible, experiments were performed by labelling the nuclei (blue) with Hoechst 33342 and tracking nuclear movement along the channels (yellow dotted lines). Representative trajectories (red) of cells migrating through the transverse channels in the presence of 5 ng/ml EGF suggested that cell motility is collectively dictated by PDMS stiffness and EGF concentration, with fastest migration observed on the softest substrates in the presence of EGF (Fig. 3B, Movies 1,2). The dependence of cell speed on PDMS gel stiffness and chemokine concentration remained nearly identical when experiments were performed with PDMS gels functionalized with 5 and 25 µg/ml collagen (Fig. S3A,B). Although there were instances when two cells contacted each other within the channel during migration, these encounters did not alter their motility (Fig. S3C).
Given that the EGF gradient decreases with time (Fig. 2D), it remained unclear if cells exhibit chemotaxis. To address this, average cell speed for every 3 h period was evaluated on 3 and 30 kPa substrates. In experiments performed in the absence of any EGF, the average cell speed remained unchanged until 9 h and decreased marginally in the last 3 h (Fig. S4A,B). In comparison, when experiments were performed under uniform EGF conditions where medium was supplemented with 5 ng/ml EGF in the entire microfluidic device, the temporal profile of the average cell speed remained similar to that of the ‘no EGF’ condition, but increased at all time points, indicative of a chemokinetic effect. In contrast, increase in average cell speed in the presence of an EGF gradient between 3–6 h and 6–9 h thus suggests that the cells indeed sense the EGF gradient and exhibit chemotaxis (Fig. S4A,B). This behaviour was similar on both 3 and 30 kPa substrates. Further support for hMSC chemotaxis within the devices in the presence of an EGF gradient comes from the quantification of persistence, which was identical for ‘no EGF’ and ‘uniform EGF’ conditions, but increased markedly in the presence of the EGF gradient (Fig. S4C,D). The drop in average cell speed in the last 3 h across all the conditions may be attributed to the low volume of medium (∼3 µl within the device and ∼50 µl in reservoirs) and the lack of serum.
Having confirmed that hMSCs indeed exhibit chemotaxis inside the channels, we next performed experiments at two more EGF concentrations (1 and 50 ng/ml) for all three values of substrate stiffness, quantified the average cell speed for the entire 12 h, and compared results across the different conditions. The collective dependence of average cell speed (v) on substrate stiffness (E) and EGF concentration ([C]) was well approximated by the expression vavg=vE+vC, where vE represents the substrate stiffness-dependent speed observed in the absence of any EGF (i.e. speed at [C]=0) and vC represents the chemokine concentration-dependent component. The substrate-dependent component (vE) exhibited an inverse dependence on ECM stiffness, with maximum vE observed on the softest 3 kPa substrate (Fig. 3C and inset). The chemokine concentration-dependent component (vC) was well fit using the expression vC=14/(1+e−0.6([C]−2.7)) indicative of a logistic dependence on chemokine concentration. Furthermore, the overlap in the values of vC observed at different stiffnesses (Fig. S4E) suggests that vC does not exhibit any stiffness dependence. In contrast to the dependence of average cell speed on EGF concentration, the increased persistence in the presence of EGF did not exhibit any concentration dependence (Fig. 3D). In addition to demonstrating the collective dependence of cell motility on physical and chemical cues, our results suggest that fastest chemotaxis on the softest substrates is a consequence of faster motility on these substrates.
Faster chemotaxis on softer substrates is associated with fewer focal adhesions and increased protrusion rate
Thus far, our results reveal faster migration in channels on softer substrates, both in the absence and in the presence of EGF. It is well known in the literature that cells on stiffer substrates are known to form more robust focal adhesions (Gupton and Waterman-Storer, 2006). To test the association between faster migration on softer PDMS gels and focal adhesion size or number, cells within channels were fixed and stained for the focal adhesion protein vinculin (Fig. 4A). Quantification of focal adhesions revealed that both the number and size and focal adhesions were lowest on the softest 3 kPa PDMS gels (Fig. 4B,C). Also, addition of EGF (5 ng/ml) led to further reduction in the number and size of focal adhesions across all the PDMS gels.
To further probe the link between focal adhesions and chemotaxis, experiments were performed where the number and size of focal adhesions were altered either by destabilizing integrins with RGD peptide (Russo et al., 2013) or stabilizing integrins with MnCl2 (Changede et al., 2015; Nishizaka et al., 2000) (Fig. 4D, Fig. S5A). Across all the PDMS gels, while treatment with RGD led to reduced focal adhesions, MnCl2 treatment induced the formation of robust focal adhesions (Fig. 4E, Fig. S5B). Interestingly, while these two perturbations had completely opposite effects on focal adhesions, both these treatments led to inhibition in cell motility across all the PDMS gels (Fig. 4F). Together, these results suggest that faster motility on softer PDMS gels can be attributed to fewer and smaller adhesions.
Chemotaxis is known to upregulate Rac-based signalling which, in turn, modulates actin dynamics (Vorotnikov and Tyurin-Kuzmin, 2014; Benard et al., 1999). To probe the role of the actin cytoskeleton in driving chemotaxis, we first investigated actin cytoskeletal organization by staining with Phalloidin and comparing integrated F-actin intensities across the conditions. Although F-actin content was higher in hMSCs cultured on stiff 30 kPa and 600 kPa substrates, stress fibres were also observed on the softest 3 kPa PDMS gels (Fig. 5A), as well as on 3 kPa polyacrylamide gels (Fig. S2D). EGF treatment led to a further increase in F-actin intensity across all the PDMS gels (Fig. 5B). To next probe the relationship between cell motility and protrusion rate, protrusion dynamics inside the channels was monitored and protrusion rate was quantified using kymographs (Fig. 5C,D). Interestingly, analysis of kymographs revealed fastest protrusion rate on the softest 3 kPa substrates (Fig. 5E). Additionally, while addition of EGF further increased protrusion formation on all substrates, the increase was maximal on the softest substrate. Collectively, our results suggest that faster chemotaxis on softer substrates is attributed to the combination of weaker adhesions and increased protrusion rate.
Actin polymerization-dependent protrusions are key for sustaining chemotaxis
While the above results demonstrate the contribution of increased protrusion formation on softer substrates, the exact mechanism of migration utilized by cells remains unclear. To address this, experiments were done in the presence of 5 ng/ml EGF and cytoskeletal disrupting drugs including ML-7 (ML7), blebbistatin, cytochalasin D (Cyto D) and nocodazole. While ML7 may influence protrusion dynamics via inhibition of myosin light chain kinase (MLCK) signalling (Betapudi et al., 2006), blebbistatin inhibits myosin II ATPase activity (Hung et al., 2013). Cyto D (Hayot et al., 2006) and nocodazole (Sui and Downing, 2006) disrupt the actin and microtubule cytoskeletal networks, respectively. Drugs were added 3 h after initiating the experiments to allow cells to migrate and partly enter the channels (Fig. 6A). While ML7, blebbistatin and nocodazole did not induce any change in cell morphology within the channels, Cyto D treatment led to cell rounding on all the PDMS gels (Fig. 6B). To check how cytoskeletal organization was perturbed by the drugs, F-actin staining intensity was compared across the conditions (Fig. S6A,B). Compared with controls (CTL), a ∼20–30% drop in integrated F-actin intensity was observed in ML7- and blebbistatin-treated cells. While F-actin intensity remained unchanged in Noc-treated cells, a large drop (∼75%) was observed in Cyto-D-treated cells. Furthermore, compared with controls, a ∼50% drop in tubulin staining intensity was observed in nocodazole-treated cells (Fig. S6C,D). Tracking of cell trajectories allowed us to quantitatively determine the effect of each drug on cell motility (Fig. 6C, Movie 3). While Cyto D-treated cells exhibited minimal migration, the effects of the other drugs on cell motility was much less severe. Quantification of cell trajectories revealed a ∼5–35% drop in average cell speed in cells treated with ML-7, blebbistatin and nocodazole, and a ∼90% drop in Cyto D-treated cells (Fig. 6D). Collectively, these results highlight the role of actin-based protrusion in driving chemotaxis.
DISCUSSION
In this paper, we have probed the collective influence of substrate stiffness and chemokines in dictating hMSC chemotaxis. To address this, we designed and validated a microfluidic device that allows the formation of a chemical gradient for ∼10 h. By tuning PDMS composition, we fabricated substrates of stiffness varying over three orders of magnitude. Then, by using EGF as a chemokine and tracking hMSC migration through channels of varying base stiffness, we demonstrated the role of ECM stiffness in regulating hMSC chemotaxis. Based on our results, we hypothesize that chemotactic speed is determined by a substrate stiffness-dependent component and a chemokine concentration-dependent component. While the substrate stiffness-dependent component scales negatively with ECM stiffness, the chemokine-dependent component exhibits a logistic dependence on the chemokine concentration. Based on analysis of focal adhesions, leading edge protrusion and drug treatment experiments, we propose a model wherein ECM softness enhances chemotactic efficiency through a combination of higher focal adhesion turnover and faster leading-edge protrusion mediated by actin polymerization (Fig. 7).
Among the several hydrogel systems that are used for studying mechanobiology (Caliari and Burdick, 2016; Jansen et al., 2015), polyacrylamide gels and PDMS substrates are among the most popular. Although both polyacrylamide and PDMS gels are used extensively for 2D studies, microfluidic devices are fabricated almost exclusively with PDMS. However, in contrast to polyacrylamide gels where it is easy to fabricate soft gels (∼100 Pa in stiffness), the high viscosity of PDMS poses a significant challenge in fabricating soft PDMS gels (∼100 Pa to 1000 kPa) using Sylgard 184 – the most widely used PDMS formulation. However, Palchesko et al. (2012) were successful in fabricating 5 kPa stiff PDMS gels using Sylgard 527. Taking a cue from this study, we combined Sylgard 527 and Sylgard 184 at ratios of 5:1 and 50:1 to fabricate PDMS of stiffness 30 kPa and 3 kPa, respectively. It is likely that this approach of combining two different Sylgard preparations at different ratios can be further exploited to fabricate even softer PDMS gels.
Using PA gels, a plethora of studies have demonstrated the role of substrate stiffness in regulating a range of cellular processes including cell spreading, cell division, cell motility and stem cell differentiation across various cell types (Engler et al., 2006; Charest et al., 2012). Most cell types including hMSCs spread much more robustly on stiffer polyacrylamide gels compared with softer polyacrylamide gels, and exhibit a bi-phasic dependence of cell motility on gel stiffness (Chen et al., 2009; Shebanova and Hammer, 2012; Peyton and Putnam, 2005). Our observations of a stiffness-dependent increase in hMSC spreading and bi-phasic dependence of cell motility on PA gel stiffness are consistent with these reports. Given the comparable spreading observed on 3 kPa PDMS and polyacrylamide gels, as well as on 30 kPa PDMS and 35 kPa polyacrylamide gels, it is likely that the nature of spreading is similar across these gel systems. Additionally, hMSC motility was maximum on 3 kPa PDMS and PA gels, and were of comparable magnitudes. In addition to stiffness, ligand type has also been shown to play a key role in cell migration, with cells being more responsive on collagen compared with fibronectin and laminin (Nelson et al., 1996).
hMSC homing to various tissues is driven by a plethora of growth factors and inflammatory cytokines, including EGF (Nakamizo et al., 2005; Wadhawan et al., 2012; Barrientos et al., 2008). While our observation of EGF-induced increase in hMSC motility is consistent with earlier reports (Vertelov et al., 2013; Ponte et al., 2007), our findings suggest that this increase is independent of stiffness in hMSCs. This increase in cell motility is mediated by EGF binding to EGF receptors (EGFRs) at the cell surface. Recently, in fibroblasts and Cos-7 cells, EGFRs have been demonstrated to drive cell spreading and rigidity sensing on rigid, but not soft substrates (Saxena et al., 2017). Interestingly, the authors observed no increase in cell motility on soft substrates when EGF was added, which contrasts with our results. This suggests that EGF might have cell-specific roles.
Given the responsiveness of hMSCs to both physical and chemical cues, it is likely that hMSC homing efficiency is determined by the combination of physical and chemical cues present in the microenvironment. Therefore, we performed a microfluidic study (Wadhawan et al., 2012; Irimia et al., 2007; Boneschansker et al., 2014) to investigate the combined effect of stiffness and chemokine (EGF). While numerous studies have probed the individual effects of physical or chemical factors on cell behaviour (Shukla et al., 2016; Schultz et al., 2015; Takebayashi et al., 2013; Menon et al., 2015; Ogura et al., 2004; Engler et al., 2006), to the best of our knowledge, this work represents the first study to probe the collective influence of these two factors on hMSC migration. Our findings suggested that the basal stiffness-dependent component of cell speed (vE) was modulated further by the chemokine concentration, and revealed that a gradient generated by an EGF concentration of 1 ng/ml increased cell directionality in a stiffness-independent manner. It would be interesting to explore if the migration behaviour of other cell types is similar to that of hMSCs.
The first step of mesenchymal cell migration is leading edge protrusion, which is stabilized by focal adhesion complexes to transmit contractile forces to the substrate, leading to forward movement of the cell. While weak adhesions fail to stabilize the cell front, thereby leading to impaired motility, adhesions that are too strong also lead to impaired motility due to inability of the cell to break adhesions at the rear (Prager-Khoutorsky et al., 2011). Thus, optimal focal adhesion turnover (i.e. assembly and disassembly of adhesions) is needed for fast cell motility. The fastest motility of hMSCs observed on the softest substrates may be enabled by small, but firm adhesions at the cell front that allow transmission of contractile forces to the substrate (Ilić et al., 1995), while at the rear, the focal adhesions undergo turnover, which leads to cell migration. Consistent with this notion, both adhesion weakening by RGD and adhesion strengthening by MnCl2 led to a drop in cell motility across all PDMS gels. Further enhancement in cell motility induced by EGF treatment was associated with a drop in the number and size of focal adhesions across all PDMS gels, as reported elsewhere (Xie et al., 1998).
In addition to optimal focal adhesion size, our findings suggest that faster motility on softer substrates is driven by an increased actin protrusion rate. This increase in actin protrusion may be driven by EGF-induced Rac activation at the cell front, leading to faster actin polymerization (Weiss-Haljiti et al., 2004). In line with an actin-dominated mechanism of migration, a maximal drop in motility was observed upon depolymerization of actin filaments by Cyto D treatment. In comparison to actin, retardation of cell motility by inhibition of myosin activity by ML7 and blebbistatin was much less pronounced. The effect of ML7 on cell motility was much more potent than that of blebbistatin. Consistent with the role of MLCK during cell protrusion and lamellar extension in assisting cell migration, an ML7-induced drop in cell migration may be associated with reduction in the tension required for focal adhesion maturation (Betapudi et al., 2006; Webb et al., 2004). In contrast to published results showing that blebbistatin increases cell motility through Rac1 activation (Liu et al., 2010; Pathak and Kumar, 2012), in our studies we observed a marginal drop in motility in blebbistatin-treated cells. These contrasting observations may be partly attributed to the dimensionality of migration. In contrast to 2D migration, where blebbistatin causes faster motility, in 1D migration, blebbistatin has been shown to significantly inhibit cell motility (Doyle et al., 2012). Although our channels were 35 µm in width, hMSCs migrated by maintaining continuous contact with one of the channel walls, thus exhibiting 1D-like motility. In our studies, despite a ∼50% drop in tubulin staining intensity, its effect on cell motility was blunted. The relative insensitivity of cell motility to disruption of the microtubule cytoskeleton, which maintains polarity, may be due to maintenance of polarity by a combination of wall-mediated contact guidance and continuous actin polymerization. The nocodazole-induced drop in motility was maximal on the stiffest 600 kPa gels. Similar to the effect of MnCl2, this may be due to stabilization of focal adhesions induced by increased cell contractility upon treatment with nocodazole (Sen and Kumar, 2009).
In conclusion, we investigated the collective dependence of hMSC motility on physical and chemical cues, and our results suggest that fast chemotaxis on soft substrates is attributed to the combination of fewer focal adhesions and faster actin protrusions. Given that hMSCs exhibit durotaxis (Tse and Engler, 2011), i.e. migrate from softer to stiffer side when cultured on a substrate with a stiffness gradient, it remains to be seen how hMSCs respond when placed in an environment providing both durotactic and chemotactic cues.
MATERIALS AND METHODS
Fabrication and characterization of PDMS gels
Polydimethylsiloxane (PDMS) gels were prepared using Sylgard 527 and Sylgard 184 (Dow Corning). Sylgard 527 was prepared as per the manufacturer's directions by vigorously mixing equal mass of part A and part B and degassing using a vacuum dessicator to remove bubbles. Sylgard 184 was similarly prepared by mixing 10 parts of base to one part curing agent and subsequent degassing. Next, Sylgard 527 and Sylgard 184 were mixed at two different ratios of 5:1 and 50:1, respectively. For fabricating 2D PDMS gels, the prepared mixtures were degassed and spin coated onto glass coverslips using a spin coater (spinNXG-P2, ApexIndia) at 600 rpm for 30 s. The glass coverslips were then cured at 70°C overnight in a hot air oven.
Stiffness of PDMS gels was quantified using rheology as well as with JKR theory as described elsewhere (Chaudhury and Whitesides, 1991). For rheology measurements, cylindrical plugs, 2.5 cm in diameter and 2 mm thickness, were prepared using different combinations of PDMS prepared above. Small-strain oscillatory shear experiments were carried out at room temperature on a rheometer with plate-plate geometry (MCR-301 Anton Paar, Germany) to measure the storage and loss moduli. Storage modulus (G′), loss modulus (G″) and phase angle (δ) were measured as a function of frequency (from 0.05 to 10 Hz) in constant-strain mode (γ=0.05) to determine the mechanical spectra. Young's modulus (E) of PDMS substrates was calculated using the expression E=3G′ using the value of G′ at a frequency of 1 Hz.
After fabrication of PDMS gels, hydrophobicity of the gels was assessed by doing contact angle measurements with a goniometer (Apex Instruments). For these measurements, 5 µl water was placed on PDMS surfaces before and after plasma oxidation and the angle between the drop and the surface before plasma treatment was measured. For coating PDMS gels with collagen, substrates were cleaned with Scotch tape and autoclaved. After plasma oxidization of substrates for 30 s, PDMS gels were incubated with rat-tail collagen I (C3867, Sigma) overnight at 4°C. On the following day, substrates were washed with PBS at least three times to remove excess collagen. The extent of collagen coating was assessed by incubating the functionalized PDMS gels with anti-collagen antibody (Merck, custom-made antibody raised against rat-tail collagen I, 1:400 dilution) overnight at 4°C. Gels were then washed with PBS three times and incubated with secondary antibody (Life Technologies, 1:1000 dilution) for 2 h at room temperature (RT). The samples were then imaged using a laser scanning confocal microscope under identical exposure and gain settings (LSM510, Zeiss, 10× magnification). For quantifying the extent of collagen adsorbed onto PDMS gels, mean intensities were measured after correcting for the background intensities (observed in secondary antibody only control samples). Finally, normalized mean intensities were calculated from the absolute intensities by normalizing with respect to mean intensity observed on 600 kPa gels coated with 1 µg/ml collagen. For cell culture experiments, the substrates were incubated with complete or serum-free medium at 37°C for at least 15 min prior to seeding cells.
Polyacrylamide gel fabrication
Polyacrylamide gels of varying stiffness (0.5, 3 and 35 kPa) were made by using 40% acrylamide and 2% bis-acrylamide in the ratios reported in the study by Tse and Engler (2010). The gels were prepared on glass coverslips functionalized with 3-APTMS (Sigma). After pouring 50 µl gel solution onto hydrophobic glass slides [treated with octadecyle-trichlorosilane (Sigma-Aldrich)], functionalized hydrophilic glass coverslips were placed on top of the polymerizing solution. After gel formation, the gels were removed from the hydrophobic glass slides and washed with ethanol and PBS. After treating the gels with the hetero bifunctional crosslinker Sulfo-SANPAH (Thermo Scientific), the gels were treated with type-I collagen at 4°C overnight. Gels were then washed with PBS to remove unbound collagen and incubated with medium for 30 min prior to seeding cells.
Design and fabrication of microfluidic devices
For studying chemotaxis, a simple device design was chosen wherein two 500-µm-wide channels spaced 500 µm apart were connected via multiple 35-µm-wide transverse channels (Fig. 2). Each of the wide channels had one inlet and one outlet, with the left-most points serving as inlets and the right-most points serving as outlets. Gradient generation in these devices was simulated using COMSOL (version 5.2) assuming laminar flow conditions and assigning concentrations of 1 and 0 respectively to the two inlets. The device design was printed on to a transparency mask using a printer with at least 600 dpi resolution. The design was patterned on a silicon wafer by photolithography with SU-8 (2050) photoresist (MicroChem) to obtain a thickness of ∼70 µm. PDMS devices were then fabricated by pouring PDMS (Sylgard 184) onto the master, and curing them in the oven. After peeling the PDMS substrates, inlets and outlets were punched using a 3 mm biopsy punch. Finally, devices were fabricated by bonding the PDMS devices to PDMS gels of varying basal stiffness. Gradient generation within the device was experimentally assessed by adding 10 µM FITC-dextran (Sigma, cat. no. FD10s-100) in one of the inlets and PBS in the other one, and doing time-lapse imaging over a period of 10 h. Images were then processed using ImageJ (NIH) to determine the temporal evolution of the gradient.
Cell experiments
Primary human mesenchymal stem cells (hMSCs) were obtained from Texas A&M Health Science Center, College of Medicine, Institute for Regenerative Medicine at Scott & White, and cultured in MEM-α (Thermo Scientific, cat. no. A1049001) supplemented with 16% MSC certified fetal bovine serum (FBS, Thermo Scientific, cat. no. 12662029), 1% Glutamax (Thermo Scientific, cat. no. 35050061) and 1% Anti-Anti (Thermo Scientific, cat. no. 15240062). For 2D experiments on PDMS gels, cells were seeded at a density of 400 cells/cm2. For experiments inside channels, 10 µl medium containing ∼10,000 cells were introduced through one of the channel inlets. For chemotaxis experiments, once the cells adhered to the substrates (after ∼6 h), EGF (Sigma, cat. no. E9644) was introduced as a chemokine at the opposite channel inlet. For probing the effect of gradient strength, experiments were performed using three different EGF concentrations (1 ng/ml, 5 ng/ml and 50 ng/ml, respectively). For measuring cell motility, time-lapse microscopy was performed for 12 h using an inverted microscope equipped with an onstage incubator (Evos FL Auto, Life Technologies). Images were acquired every 15 min. Cell motility was quantified using the manual cell tracker plugin in ImageJ (NIH). For experiments with RGD and MnCl2, 2.5 h after starting time-lapse movies, RGD (2 µM) and MnCl2 (1 mM) were added and imaging was performed for the next 3 h in the presence of these compounds. For drug experiments, 3 h after starting to record time-lapse movies, ML-7 (10 µM, Calbiochem, cat. no. 475880), blebbistatin (10 µM), cytochalasin D (1 µM, Calbiochem, cat. no. 250255) and nocodazole (0.1 µM, Calbiochem, cat. no. 487928) were added and imaging was performed in the presence of the drugs for another 9 h.
For immunostaining, hMSCs cultured on different substrates and inside channels were stained with vinculin antibody for probing focal adhesions (FAs), and Phalloidin or β-tubulin for visualizing the actin and microtubule cytoskeletons before and after drug treatment. hMSCs were first fixed in the presence of permeabilization buffer (0.5% Triton X-100) for 1 min to remove soluble cytoplasmic proteins. Cells were fixed with ice-cold 4% paraformaldehyde (pH 7) for vinculin and F-actin staining; for tubulin staining, fixation was performed with 0.5% glutaraldehyde for 15 min at 37°C. After fixation, cells were washed with cytoskeleton stabilizing buffer (CSB) (60 mM PIPES, 27 mM HEPES, 10 mM EGTA, 4 mM magnesium sulphate heptahydrate, pH 7) three times, and then incubated with blocking buffer for 30 min at 4°C. Cells were then incubated with anti-vinculin antibody (cat. no. V9131, mouse monoclonal, Sigma, 1:400 dilution), anti-β-tubulin antibody (cat. no. T4026, Sigma, 1:300 dilution) overnight at 4°C. Cells were then washed with CSB three times and then incubated with secondary antibody (cat. no. A11061, Life Technologies, 1:1000 dilution) and with Alexa Fluor-conjugated Phalloidin 488 (cat. no. A12379, Life Technologies, 1:400 dilution) for 2 h at room temperature (RT). The cells were then imaged using laser-scanning confocal microscope (LSM510, Zeiss, 20× magnification for vinculin and F-actin imaging, and 40× magnification for tubulin imaging). Size distribution of focal adhesions was obtained by appropriately thresholding the acquired images as described elsewhere (Horzum et al., 2014). For quantification of protrusion rate, cells in the devices were imaged for 15 min at 15 s intervals in the presence and absence of EGF. Kymographs generated from the obtained movies were analysed using ImageJ (NIH) for calculating protrusion rate, as described elsewhere (Maheshwari et al., 1999).
Statistical analysis
All statistical analysis was performed using Origin 9.1 with P<0.05 considered to be statistically significant. Based on the normality of data assessed using Kolmogorov–Smirnov normality test, one-way or two-way ANOVA was performed to assess statistical significance, and Fisher post hoc test was used to compare the means.
Acknowledgements
The authors thank Dr Anirban Banerjee for sharing FITC-Dextran, Centre for Nanoelectronics, IIT, Bombay for providing lithography facility, and the Confocal Microscopy Central Facility. hMSCs used in this work were provided by the Texas A&M Health Science Center, College of Medicine, Institute for Regenerative Medicine at Scott and White through a grant from NCRR of the National Institutes of Health (P40RR017447). We thank IRCC, IIT Bombay for use of the live cell microscopy facility (14IRCCSG002).
Footnotes
Author contributions
Conceptualization: N.S., S.J., S.S.; Methodology: N.S., P.M., S.D., A.M., S.J., S.S.; Formal analysis: N.S.; Investigation: N.S.; Resources: P.M., S.D., A.M.; Data curation: N.S.; Writing - original draft: N.S., S.J., S.S.; Writing - review & editing: N.S., S.J., S.S.; Supervision: S.J., S.S.; Project administration: S.J., S.S.; Funding acquisition: S.J., S.S.
Funding
The authors acknowledge financial support from Department of Biotechnology, Ministry of Science and Technology (Govt. of India) (BT/PR7741/MED/32/275/2013). N.S. was supported by an Inspire Fellowship from the Department of Science and Technology (Govt. of India) (DST/INSPIRE Fellowship/2013/1033). P.M. was supported by a fellowship from the Industrial Research and Consultancy Centre (IRCC), Indian Institute of Technology Bombay (IIT).
References
Competing interests
The authors declare no competing or financial interests.