A novel circular invasion assay mimics in vivo invasive behavior of cancer cell lines and distinguishes single-cell motility in vitro

When skin is compromised, or wounded, the damaged epidermal edges migrate forward to cover the wound surface [9]. Fundamental to our understanding of wound-healing, is the knowledge that wound margins proliferate and migrate onto newly laid matrix in the wound gap [10]. Wound-healing assays have been carried out in tissue culture for many years to estimate the proliferation rates and migratory behavior associated with different cells and culture conditions [11]. Migration of cells can be conveniently studied in vitro by using these classical assays, whereby confluent epithelial cells are scratched with a tool such as a razor blade to remove a linear strip of cells from a monolayer. The filling or "healing" of the remaining "wounded" area is then observed using time-lapse microscopy [12, 13]. Such a method can provide information regarding the behavior of those migratory cells that act to heal the inflicted wound, which indirectly provides additional information about cancer progression. However, as might be expected, when the initial "wounding" is not precisely controlled, this method is encumbered with problems of quantification and reproducibility [14].

When using the classical method, the wounded edges of the intact cell monolayer commonly retract on both sides of a crude, linear scratch. This suggests that many of the cells on the "wound" edge potentially lose their original morphology and function because they have been physically disrupted [15]. Additionally, since classical assays are produced using sharp objects, the migrating surface (dish or coverslip), which is often coated with extracellular protein(s) prior to monolayer growth, can also be damaged. In order to overcome these problems, a number of updated circular wound-healing assays (CWA) have been established that are less detrimental, and more standardized than the linear-scratch method [16, 17]. The CWA method involves removal of a uniform, circular portion of cells from a confluent monolayer that is then allowed to heal towards itself. This technique can be further strengthened by minimizing the surface damage inflicted to the cells by employing the use of a soft silicon tip, in place of blunt trauma [16]. Although this method is adapted to provide a more standardized model of wound-healing and migration, and overcomes some of the obstacles associated with the traditional methods, it too is fairly simplistic and limits the depth of information that can be obtained and directly linked to in vivo results.

The invasive capacity of tumor cells and the mechanisms that control contrasting types of invasion are of critical importance in metastasis. Therefore, assays that determine this measurement in reproducible, quantitative terms can be extremely useful for probing these questions in vitro. The most common method currently employed to investigate invasive potential involves a modified Boyden chamber assay using a basement membrane matrix preparation (such as Matrigel™) as the ECM barrier and conditioned tissue culture media as a chemoattractant [18, 19]. Although such a method is capable of supplying ample information about the collective migration of an entire population of cells, it fails to provide sufficient resolution for yielding precise, quantitative data for individual cell motility and invasion mechanisms. Therefore, in order to better distinguish between collective and individual movement of cells more clearly around "wounded" edges, and overcome the other problems associated with traditional assays, we have developed a novel circular invasion assay (CIA) modified from a previously established CWA technique designed by Watanabe and colleagues [16] (Figure 1A).

For our method, a stabilized, rotating, silicone-tipped drill press was used to create eight uniform circular lesions in an intact monolayer, and the wound closure recorded and calculated for each using time-lapse microscopy and advanced image analysis. In contrast to the CWA method, the CIA technique utilizes an extracellular matrix component (Matrigel™) to build a pseudo-matrix barrier on the cell-free surface or "wounds" (Figure 1B). A number of two-dimensional in vitro invasion studies have previously shown that inclusion of this basement membrane component causes cells to behave similarly as they do in vivo [19‐21]. Three-dimensional invasion assays have also been developed to include this reagent, or similar matrix components, due to an increased appreciation that their presence is necessary for demonstration of normal epithelial cell behavior [22, 23]. Consequently, an obvious advantage to using our updated CIA method is the added environmental parameter (Matrigel™) that allows us to study invasion of cells in their microenvironment more accurately, while overcoming the need for advanced three-dimensional microscopy and analysis.

In many physiological situations, such as with wound-healing or cancer metastasis, epithelium becomes motile under given stimulation [24]. In some instances, cells dissociate and individually explore their surroundings, whereas in other instances, the cells become collectively motile and enter the surrounding environment with their neighboring cells still intact. The difference between these behaviors has strong implications for our understanding of cancer-related dissemination [25].

As discussed in detail above, most classical in vitro migration/invasion assays reflect population dynamics, and are unable to provide information about individual cell behavior. Furthermore, in vivo tumor models often include "experimental" or "spontaneous" assays that assess the complete metastatic process, rather than focus its individual steps [26]. Recently, a number of both in vitro and in vivo methods have been developed which focus on illuminating the various mechanisms of motility at the single-cell level; however, the majority of these assays are performed using 3-D microscopy and fluorescent markers [27‐29]. While these assays are often very informative, they can also be quite laborious and require advanced systems. In contrast, the CIA technique gives us the ability to dissect the metastatic process in 2-D into steps or types of events, such as collective or individual cell motility, potentially enabling us to build a more detailed model of cancer invasion.

All cell lines were obtained from American Type Culture Collection (ATCC, Rockville, MD). McCoy's 5a Modified Medium was also purchased from ATCC. L-α-lysophosphatidic acid (oleoyl sodium salt, LPA 18:1) was purchased from Avanti Polar Lipids (Alabaster, AL) and used at a concentration of 2 μM. Phosphate buffered saline solution (PBS), Dulbecco's Modified Eagle's Media (DMEM), fetal bovine serum (FBS), penicillin-streptomycin antibiotics, and L-glutamine were obtained from GIBCO BRL (Carlsbad, CA). Matrigel™ Matrix Growth Factor Reduced was purchased from BD Biosciences (San Jose, CA) and used at 50% concentration (in DMEM). The optimal concentrations of these reagents were determined in previously performed dose-dependent experiments, taking concentrations given in the literature as baseline values.

We examined four cancer cell lines that have various established levels of invasiveness both in vitro and in vivo in order to test the applicability of our method to different stages of cancer progression. DLD-1 (CCL-221), a human colorectal adenocarcinoma cell line that is tumorigenic in nude mice [30, 31], MCF-7 (HTB-22), a human mammary epithelial cell line found to be nonaggressive and noninvasive in mice [32], and MDA-MB-231 (HTB-26), a human mammary epithelial cell line found to be highly aggressive and known to rapidly progress to extensive and well-vascularized metastatic lesions [32, 33], were routinely cultured and maintained in DMEM supplemented with 10% heat-inactivated FBS, 1% penicillin-streptomycin antibiotics, 1% L-glutamine, and kept in a humidified atmosphere of 5% CO₂ at 37°C. SKOV-3 (HTB-77), a human ovarian epithelial cell line found to be moderately aggressive and tumorigenic in nude mice [34], was regularly maintained in McCoy's 5a Modified Medium also supplemented with 10% FBS, 1% antibiotics, 1% L-glutamine, and grown in the same incubator conditions.

All cell lines were seeded, in sterile conditions, at a density of 1.5–2 × 10⁶ on polystyrene, 35-mm, tissue-culture treated Petri dishes (Falcon, Becton Dickinson Labware, Franklin Lakes, NJ) for 18–24 hours or until confluent. For indicated experiments (those involving LPA treatment), cells were serum-starved by incubation in DMEM in the absence of FBS for 18–24 h after monolayers reached confluence.

Uniform, circular-shaped "wounds" (1.5 – 2 mm diameter; 8 per dish) were generated using a rotating drill press (Delta Shopmaster, Type 1, Model DP200) fit with a custom-shaped silicon tip (Home Depot; manually cut down to rounded shape with approximately 1.5 mm diameter and rounded bottom, using razor blade) as seen in Figure 1A. The optimal size and shape of the "wounds", their spacing, and other parameters were established in preliminary experiments (results not shown). The silicon tip was regularly washed with 70% ethanol between "wounding" of monolayers in individual dishes. Cell debris created by "wounding" was removed from each dish by manual pipetting, and intact cells were gently washed twice more with PBS. Two ml of growth media with 10% FBS was added to each dish for the remainder of incubation. For those experiments examining LPA effects, LPA or PBS in DMEM (2 ml total volume; without FBS) was directly applied into each dish and allowed to incubate for up to 24 h.

For the novel CIA method, "wounds" were created as described above in the CWA method (8 per dish). Additionally, 50% Matrigel™ in growth media (600 μl total) was overlaid onto the "wounded" cell monolayer to create a matrix barrier against the cellular surface and allowed to polymerize for 15 min prior to imaging the original time point (Figure 1B). For those experiments examining LPA effects, LPA or PBS was added directly to Matrigel™ overlay prior to polymerization. Two ml of growth media with 10% FBS was added to each dish (LPA experiments used DMEM without FBS). "Wounded" monolayers, with fabricated matrix, were then incubated in a humidified atmosphere of 5% CO₂ at 37°C for 24 h.

Time-lapse microscopy was conducted using a Zeiss Axiovert 200 M microscope (Zeiss, Thornwood, NY; 2.5× Plan NEOFLUR objective, NA 0.075; 10× Achroplan, NA 0.25, Ph1 objective) equipped with a Hamamatsu ORCA-ER CCD camera and temperature- and CO₂-controlled chamber. Microscopy was under the control of OpenLab software (Improvision, Lexington, MA). At the beginning of each experiment (0 h), phase-contrast images were captured and microscopically accessed for standard, reproducible "wounds", with irregular outliers thrown out of the data set. Reflecting the precision of the method in creating consistent wounds by shape and size, this subpopulation of "unusable" wounds was negligible, as intra-operator variance was found to be < 3% (results not shown). Images of all "wounds" were then captured at regular time points for 24 h thereafter. The cell-free areas of each monolayer were distinguished from the surrounding intact cells by applying an automatic, software-defined threshold to each image, and pseudo-color applied to these areas using Adobe Photoshop 7.0 (Adobe Systems, Inc., San Jose, CA).

A glass coverslip was coated by incubating in PBS containing 10 μg/ml of collagen I (C8919, Sigma, St. Louis, MO) overnight at 4°C. Circular wounds were made in the DLD-1 cell layer as previously described. Cells were further incubated for 5 hr at 37°C with or without 50% Matrigel™ overlay and fixed by 3.7% formaldehyde. Cortactin (green; lamellipodial marker) and actin (red; cytoskeletal marker) were visualized by immunofluorescence staining using anti-cortactin antibody 4F11 (Upstate Biotechnology Incorporated, Lake Placid, NY) and Alexafluor 568-phalloidin (Invitrogen). Confocal images were obtained with a Zeiss LSM-510 laser scanning confocal microscope equipped with a Plan-NEOFLUAR 40×/1.3 Oil DIC lens (Zeiss, Germany).

Each cell line was sampled at least 8 times for each method (N = 8–32; Power = 0.94–1.00), over the course of 10 days (N = 1–4 days per line). Wound repair data are referenced to time point 0 h, with results presented as mean percent wound closure (out of 100%) after a given period of time ± standard deviation. To avoid confounding problems with multiple analyses along the time-response curve, final differences were only analyzed at 4, 6, 8, 12, and 24 h. Differences between cell lines were examined using Student's t-tests, and were considered significant when P < 0.05. To further compare the two methods, post-hoc analysis (ANOVA) was performed for all parameters (method, treatment, time) using SPSS, Version 16 (SPSS Inc., Chicago, IL). Post-hoc power analyses were also performed for each set of experiments using G*Power 3 (E. Erdfelder, F. Faul and A. Buchner; University of Trier).

The three cell lines used in the first set of experiments, MCF-7, SKOV-3, and MDA-MB-231, have been established to have varying degrees of invasiveness in vivo. In order to test our in vitro method's ability to translate to known physiological findings, we applied both the CWA and CIA techniques to these cell lines for comparison. Multiple, uniform wounds were created in all cell line monolayers with a custom-made silicon tip, as described in Materials and Methods (Figure 1). Starting areas of wounds (at 0 h) were highly and easily reproducible, with negligible levels of intra-operator variance experienced (< 3%; results not shown).

Figures 2A and 2B represent the image analysis techniques and quantification method that was applied to time-lapse acquired images in order to obtain results for both the standard CWA and novel CIA techniques, respectively. The first two columns of each block of images show "wounded" monolayers from each of the three cell lines tested, MCF-7, SKOV-3, and MDA-MB-231, at time point 0 h (unprocessed raw images and pseudo-color applied, respectively). Notice that original wounds are comparable in size and shape, stressing the reproducibility of our method, which involves a semi-automated process of wounding cells using a standard carving tip. Column 3 includes the "healed" wounds after a 24 h incubation period in the presence (2B) or absence (2A) of Matrigel™. Pseudo-color (red) was applied to the open, cell-free areas of each image (using an automated threshold function applied by Adobe Photoshop), in order to establish a percent wound closure measurement for each time point. Finally, column 4 was created by overlaying the image from the final time point measured (24 h) over the original wound (0 h) in order to create the difference in area (shown in blue) between the snapshots. This area represents the final calculated percent wound closure obtained for each cell line at each time point. As evident by the unequal areas (red and blue pseudo-colors) applied to each image across the different cell lines, it is apparent that each cell line tested healed (and invaded) at different rates. It is also to be noted that when comparing the wound-closure obtained from the different methods, we see that those wounds incubated in the presence of Matrigel™ (2B) heal more slowly than those in its absence (2A). This may be due, in part, to the need of proliferating cells to actively degrade the added matrix barrier prior to invasion, or due to overall constraint (more discussion later).

Figure 3 is a side-by-side comparison of the quantitative results obtained from each assay, which further contrasts the two methods' abilities to distinguish between the different invasive properties of these cell lines in vitro. Figures 3A and 3B represent the percent wound closure and invasiveness (respectively) exhibited by the cell lines tested at multiple time points (6, 12, and 24 h), in the absence or presence of Matrigel™, respectively. Using the classical CWA method (Fig. 3A), MCF-7, the non-invasive, least aggressive cell line, exhibited the lowest degree of wound closure with no matrix present (15.15 ± 7.29, 24.50 ± 4.83, and 41.08 ± 6.86%, after 6, 12, and 24 hours, respectively). SKOV-3, the moderately invasive line, showed an intermediate-to-high level of closure (31.75 ± 8. 29, 66.98 ± 12.77, and 92.99 ± 9.89%). And MDA-MB-231, the most aggressive, invasive cell line, had the highest degree of wound closure at all time points measured (49.61 ± 11.14, 76.33 ± 9.48, and 96.91 ± 4.53%). While the CWA technique is capable of picking up variations in migration across the cell lines, it does not reflect the true separation of measurement expected for each on the basis of their known in vivo behavior. Namely, the MCF-7 wounds healed considerably at all time points, which is surprising given its characterization as a nonaggressive, noninvasive line [32]. Additionally, the separation between the SKOV-3 and MDA-MB-231 cells was not significant after 24 h (P > 0.05), which is contrary to the trends exhibited by these cells in previous studies in vivo [32‐34].

In contrast, the CIA technique obtained results that were sensitive to the differences in in vivo invasive behavior of the cell lines (Figure 3B). With the addition of the Matrigel™ barrier, MCF-7 and SKOV-3 cells invaded similarly at both 6 h (4.91 ± 2.75% and 4.70 ± 2.58%, respectively) and 12 h (8.71 ± 4.07% and 10.66 ± 3.89%, respectively) time points, but were significantly different after 24 h (14.09 ± 7.88% and 20.47 ± 6.72%, respectively). Additionally, MDA-MB-231 invaded significantly (P < 0.001) more than the other cell lines at all time points measured (27.40 ± 5.09, 47.06 ± 7.64, and 66.05 ± 6.95%), as expected. From these figures, it is clear that as time progresses, the separation between levels of invasion further manifests itself, as often experienced in animal models [35]. Also, we noted that the variation and confidence levels associated with the CWA method were inferior to those of the CIA method (more discussion later). Clearly, our novel invasion assay has given us reproducible, plausible results for a variety of phenotypes, supported by previously published findings on these cell lines [32‐34].

Figure 4 includes both overhead and stacked confocal microscopy images taken from the CWA and CIA techniques. The overhead images of closing wounds in DLD-1 cells display decisively different cell morphologies and shaped protrusions (lamellipodia) across methods. In the absence of Matrigel™(CWA), cells display a spread morphology and wide, fan-like lamellipodial protrusions, a morphology consistent with a migratory phenotype. In contrast, in the presence of the overlay, cells appear to be less spread and display thinner protrusions, perhaps more consistent with an invasive morphology. Additionally, there is an apparent higher concentration of cortactin on the dorsal cell side, presumably in direct contact with Matrigel™, suggesting that the cytoskeleton is organized differently in cells undergoing migration in the CIA vs. the CWA method. Furthermore, comparison of stacked images (taken from areas indicated by green, horizontal and red, vertical lines; CWA, 37 slices (16.01 μm); CIA, 23 slices (9.78 μm)) suggests that cells remain at the plane level of the substrate for both methods, but those cells plated with Matrigel™ apparently achieve a slightly greater depth (suggesting some interaction with overlay). While more in-depth studies are necessary to characterize these distinctions, these results support the notion that the CWA and CIA assay cell migratory abilities in different ways.

Epithelial cells generally proliferate in tightly packed colonies when seeded on plastic or glass under normal culture conditions. However, our lab recently reported that the addition of lysophosphatidic acid (LPA) to these cells induced dispersal of these colonies [36]. LPA is a bioactive lipid mainly synthesized by platelets that is found at micromolar range concentrations in plasma and serum, and has several effects on cells [37]. Relevant to this paper, this agent has been reported to cause dissociation of cell contact, leading to changes in morphology and behavior that spurs individual motility [36]. We introduced LPA in our CIA method in order to determine whether this assay can also be used to distinguish and quantify individual versus collective cell motility. Recently, these two kinds of cell motility have been proposed to play distinct roles in cancer invasion [25, 38].

Figure 5A includes images of DLD-1 cells (100×) in the presence or absence of LPA. We chose this cell line because it was previously shown to be stimulated by LPA [36]. These results show that LPA treatment promotes dissolution of DLD-1 colonies into individual cells, whereas PBS had little effect on cell dissociation and cells remained in direct contact with their neighboring cells. This figure also illustrates the typical cell morphological changes brought on by treatment with LPA, namely the onset of "ginkgo leaf-shaped" cells with increased lamellipodia formation and cell-cell dissociation with tails, as previously reported [36]. Results for these experiments were obtained using the same quantification method previously described using the other cell lines. The concentration of LPA employed in our experiments was 2 μM, which is within range of physiological concentrations (0.1 to 10 μM/L) that are generally found in plasma or serum [37]. A requirement for the LPA dispersal response is that cells undergo serum-deprivation for 24 h prior to LPA addition [36]; in non-deprived colonies, LPA had no effect (results not shown).

Figure 5B is a summary of results obtained from DLD-1 cells, in the presence or absence of LPA at various times over the course of 24 h, using the CIA method. The addition of LPA treatment significantly (P < 0.001) enhanced invasion at all time points using this method (10.53 ± 0.91, 17.77 ± 1.76, and 29.62 ± 2.55 at 4, 8, and 24 h, respectively) compared to the PBS control (6.51 ± 0.74, 10.32 ± 1.00, and 22.98 ± 2.23). Also noteworthy is the fact that as incubation time is increased, the difference of rates of healing between LPA and PBS further manifests itself. No differences in DLD-1 motility were observed between the CWA and CIA techniques (not shown), presumably because cells had to be serum-starved in order to sensitize them to LPA [36].

Figure 5C includes images (100× magnified) of the invasive fronts of invading DLD-1 cells at 8 h in the presence and absence of LPA. This time point was chosen because the characteristic "LPA effect" is known to occur around this period. Results taken from the later time point (24 h) distinguishes between the different treatments, but loses some of the classic cell morphology expected, based on our previous findings. As evident by the magnified images of these cells given each treatment, DLD-1 cells migrate into the cell-free area as single cells in the presence of LPA, and collectively in its absence. Similar to cells seen in Fig. 4A, LPA stimulation induces the expected morphological changes that include the onset of "ginkgo leaf-shaped" cells with increased lamellipodia formation and cell-cell dissociation with tails. Clearly, the proposed CIA method is compatible with high-resolution imaging that detects morphological nuances of single cells and successfully differentiates between collective and individual invasion in vitro.