MCF-7 cells were cultured in DMEM/F12 (Gibco, Grand Island, NY, USA) medium supplemented with 5% fetal bovine serum (FBS; Hyclone, Logan, UT, USA) and 1% penicillin/streptomycin (P/S; Gibco, Grand Island, NY, USA). MCF-7 cells were passaged at 80% confluence.

Human MSC isolation and expansion: Approximately 10 mL of bone marrow aspirate were taken from the iliac crest of fully informed healthy volunteer donors in the Mater hospital (Brisbane, Australia). Ethical approval for this study was granted through the Mater Health Services Human Research Ethics Committee and the Queensland University of Technology in accordance with the Australian National Health and Medical Research Council’s Statement on Ethical Conduct in Research Involving Humans. Samples were diluted with phosphate-buffered saline (PBS) and mononuclear cells were enriched via density gradient centrifugation as described previously [23,24]. MSC were expanded in low-glucose DMEM (Gibco) containing 10% FBS and 1% P/S, and then sub-cultured in DMEM/F12 supplemented with 5% FBS and 1% P/S for one passage prior to use in migration assays to enable cell adaption to the base medium used in the MCF-7 monocultures and planned migration assays. MSC were used at passages 2 to 4.

Figure 1 depicts the design and dimensions of the microchannel migration device. Figure 1A shows the silica wafer (100 mm diameter) used to cast the bodies of three microchannel migration devices simultaneously. The features on the silicon wafer (3 groups of 14 rectangular features, each 1 mm long, 500 µm wide, and 100 µm in height) were formed using SU-8 photo resist (SU8-2025, MicroChem, Newton, MA, USA) as per the manufacturer’s instructions (and as described previously) [25]. A 4 mm deep layer of polydimethylsiloxane (PDMS, Sylgard, Dow Corning, Midland, TX, USA) was poured over the wafer surface and cured at 80 °C for 30 min. The cast patterned PDMS was removed from the silicon wafer and trimmed to make two opposing reservoirs for cell culture as shown in Figure 1B. The PDMS shaped device was then plasma treated for 30 s (high frequency plasma generator Model BD-20, Electro-Technic Products, Chicago, IL, USA), and bonded to a 2 mm thick sheet of glass (Proscitech, Brisbane, Australia) and then placed on a hot plate (80 °C) overnight.

Prior to use, the device was left dry at room temperature for at least two days. Over this time the plasma treated PDMS surface became hydrophobic once again [26]. This is a necessary step, as the microchannels must be hydrophobic for the device to function. The device was sterilized by immersion in 70% ethanol for 2 h, with care to ensure that the ethanol solution entered the microchannels. When necessary, a syringe with needle was used to displace air bubbles. Following sterilization, the ethanol solution was removed, and the device was left in an open sterile petri dish in a biosafety cabinet for two days to dry. Again, it is important that the PDMS surface is not wetted by the ethanol solution prior to use.

To establish opposing monolayer cell populations, or control medium volumes, 1 mL of cell suspension and/or control media was pipetted into the appropriate reservoirs. For MCF-7 cells, suspensions of 500,000 cells mL⁻¹ were loaded, whilst for MSC, 60,000 cells mL⁻¹ were loaded via pipette. The difference in cell seeding densities reflects the relative spread area of the individual cell types and the cell numbers required to generate a confluent monolayer. Each reservoir was loaded in rapid succession to ensure that the air bubbles were trapped within the microchannels (Figure 1C). The hydrophobic nature of the material creates surface tension in the medium that deters immediate medium flow into the small cross-sectional opening of the microchannels. Once both reservoirs were filled with medium, the air bubbles in the microchannels were trapped, and the air-valve was stabilized. Thus, cell attachment and the formation of a monolayer were restricted to the respective reservoirs, and cells and media were excluded from entering the microchannels. Once cell monolayers were established by overnight incubation, the air bubbles were displaced from the microchannels to create a source-sink configuration. Bubbles were displaced from the microchannels by gently increasing the air pressure disproportionately over one of the reservoirs, thus forcing flow into the opposing chamber (Figure 2). This action was performed using a custom PDMS lid, made freshly before use to enhance adhesiveness, which was fitted with a tube connected to an air-filled syringe over one of the reservoirs, and then gently expelling air from the syringe. The lid was a rectangular PDMS piece large enough to cover the whole reservoir surface. As both the lid and migration device were fabricated from PDMS they formed a strong, but reversible, electrostatic seal. Once a seal had been formed, approximately 1 mL of air was expelled from the syringe to displace the bubbles in the microchannels.

Immediately following displacement of the air bubbles from the microchannels, the medium in each reservoir was exchanged with fresh medium of the appropriate composition. Time zero phase contrast images were recorded (Nikon Eclipse TE2000-U microscope, Nikon Coolpix 4500 camera (not shown); Tokyo, Japan) Migration cultures were returned to the incubator (37 °C, 5% CO₂), and phase contrast images were recorded again at days 1, 3 and 5. Parallel cultures were also terminated, cells were fixed in 4% paraformaldehyde and nuclei stained with DAPI (Sigma, St. Louis, MO, USA) Fluorescent images were recorded using the previously described microscope, and cell migration distances were quantified using Image J software (Version #1.47; NIH: Bethesda, MD, USA) Cell migration was characterized by measuring the distance of the cell wave-front in the microchannels from the edge of the channel for cancer cell lines, and by averaging the migrated distance of each single cell inside the channel for MSC populations (see Figure 3). Tracking the movement of individual MSC was required, as MSC did not migrate as a uniform multi-cellular wave-front.

In all cases the migration of MCF-7 cells and MSC towards each other was quantified, whilst also quantifying the migration of cells towards cell-free medium of similar serum supplementation (5% FBS = negative control), and cell-free medium supplemented with double the serum content (10% FBS = positive control). Although it is possible to conduct similar experiments where the negative control and culture medium are both serum-free, we favored maintaining a serum fraction in all cultures, as serum deprivation is known to dramatically influence cell behavior [27].

Figure 3A,B show the migration of MCF-7 cells and MSC, respectively, through a microchannel and illustrate how migration distances were estimated. The migration behavior of the two cell types differed dramatically, with MCF-7 cells moving as a communal wave front, and MSC migrating independently.

The migration of MCF-7 cells and MSC towards each other, towards equivalent control medium, and towards medium supplemented with 10% FBS was quantified (Figure 4A). Over the first 24 h, MCF-7 cell migration rates towards medium of higher serum content (10% FBS) were greater than towards MSC populations or towards control medium (5% FBS). However, at day 3 and day 5, migration rates towards MSC and higher serum content medium were similar, and with both rates statistically greater (p < 0.05) than towards control medium. Maximal MCF-7 wave front migration was observed between day 1 and day 3, and was estimated to be 137 ± 26 μm per day. The migration of MSC populations towards medium of higher serum content was significantly greater than towards either control medium or towards MCF-7 cell cultures (Figure 4B, p < 0.05 for day 1 and p < 10⁻⁷ for days 3 and 5). The migration rate of MSC towards higher serum content medium was estimated to be 210 ± 29 μm per day. In contrast, the migration rates of MSC towards MCF-7 populations and control medium were similar, but significantly slower than towards the serum-rich control medium (estimated to be 84 ± 14 and 83 ± 19 μm per day). Thus, MSC populations did not demonstrate preferential active migration towards MCF-7 cell cultures, unlike the MCF-7 populations, which did actively migrate towards MSC cultures.

Herein, we describe the fabrication and operation of a novel cell migration device, and use it to characterize MCF-7 breast cancer epithelial cells and bone marrow-derived MSC migration towards each other in interesting source-sink type assays. Various microchannel devices have been used in studies regarding the migration of cancer cells and other cell types [28,29], however the unique feature of the presented device is its array of 14 microchannels that connects source and sink culture reservoirs. During set-up, the entry of cells and/or culture medium into the microchannels was prevented by air bubbles (air-valves) that occupied each hydrophobic microchannel. Once monolayer cultures had been established in the isolated reservoirs, the air-valve bubbles were displaced, connecting the source and sink via the microchannel array. Various source-sink configurations were used to determine if MCF-7 populations or MSC populations preferentially migrated towards each other relative to media having varying serum supplementation.

Our results indicated that MCF-7 cells actively migrated both towards higher FBS concentrations and towards MSC populations (Figure 4A). This result is rational given the wealth of published data demonstrating that MSC populations secrete a number of chemokines including CXC ligand 12 (CXCL12), CX3CL1, CXCL16, CC chemokine ligand 3 (CCL3), and CCL19 [30]. Among these CXCL12 plays a pivotal role in modulation of mobilization and homing of different cell types [31]. Breast cancer cells are known to respond to CXCL12 [32,33,34,35]. The MCF-7 cell response to MSC paracrine signals is also consistent with recent published data indicating that MSC secretions may play a key role in tumor development in vivo [21,36]. Contrary to the concept that MSC might be used to combat cancer, evidence suggests that MSC can support further tumor progression when infused into animal models by supporting the growth of cancer and possibly cancer stem cells [21]. The result may depend on the particular microenvironment in which the cells are found as well as on the chemokine pairing between different cell types within that environment. Cumulatively, these reports suggest that MCF-7 would respond to MSC paracrine signals. The migration device we describe here might be useful in characterizing cancer cell responses to the various panels of MSC secretions.

MSC have been reported to migrate towards cancer lesions in vivo [14,18,37,38,39,40], and in vitro studies have shown that MSC actively migrate towards the MCF-7/Ras mutant cells [21]. Stem cells derived from adipose tissue also actively migrate towards breast cancer cells in vitro, likely as a response to tumor cell-derived platelet-derived growth factor BB (PDGF-BB) [41]. However, our results in Figure 4B indicate that bone marrow-derived MSC populations do not necessarily preferentially migrate towards MCF-7 cells in vitro. MSC migration within cancer lesions is believed to be largely in response to the localized inflammatory response [42], and in vitro cultures do not accurately mimic an inflammatory response. Therefore, it may not be logical to expect MSC to migrate towards MCF-7 cells in vitro. Nevertheless, the rationalization that MSC may migrate only in response to paracrine signals generated by immune cells prompts us to query if “death-signals” can be expected to reach their intended destination, or will they more likely be delivered to these by-stander immune cells? Additionally, will an alternative and perhaps unappreciated inflammatory response elsewhere in the body simply divert delivered MSC from their intended path? These are interesting questions that are beyond the scope of this paper, but we believe that improved migration assays like the one described here will facilitate characterization of cell migration behavior in response to chemotactic gradients.