3D migration/invasion assay

Two tumor explants (described above) or two MDA-MD-231 cell-seeded collagen gels (CSCGs) were organized within a 3D collagen gel as depicted in Fig. 1 A. To make CSCGs, 5 × 10⁵ cells were suspended within a 1.0 mg/mL rat-tail collagen solution (BD Biosciences, San Jose, CA), which was neutralized with an equal volume of 100 mM HEPES in 2× concentrated phosphate buffered saline. CSCGs were formed by placing a 250 μL volume of the cell/collagen solution into wells of 96-well culture plates and allowing them to polymerize at 37°C for at least 2 h. CSCGs (or TEs) were briefly floated in DMEM containing 10% FBS before being placed into the larger collagen matrix. To generate the embedding collagen matrix, an acellular 2.0 mg/mL (neutralized) collagen solution was added to culture dishes and allowed to partially polymerize for 15 min at room temperature. Then using a custom grid reference that is placed under the culture dish and identifies proper explant location along the major axis of the culture well, TEs or CSCGs were floated ~7 mm apart along the maximal well diameter and the TEs or CSCGs encased in a 2.0 mg/mL collagen gel. After two hours, DMEM containing 5% or 10% (v/v) FBS was added to the assay for TEs or CSCGs, respectively, and changed every 3–4 days over the course of the assays.

Open in a separate window

FIGURE 1

Development of the 3D migration/invasion assay. (A) Diagram illustrating the 3D migration/invasion assay used in this study. In the assay, two TEs or CSCGs are placed at a fixed distance and encased in a randomly organized collagen matrix. Note that Region 1 (R1) consists of the volume in which the cells align the matrix and describes the width, height, and distance between the two TEs or CSCGs, which is defined as the primary axis. Regions 2 (R2) are “above” and “below” the explants and are regions in which the matrix orientation distributes around 0°, due to growth induced explant volume expansion pushing the matrix, and as such these regions do not substantially realign the matrix perpendicular to the explant boundary and are designated “low alignment regions.” Regions 3 (R3) are aligned regions due to force balance within the matrix that act as a control relative to the low alignment region (R4) to separate out possible chemoattraction effects of one explant on the other. (B) Phase contrast microscopy of the dominant axis of 3D invasion (Region R1) for PyVT tumor explants over 10 days. Note the early collective migration (highlighted with an arrowhead in the upper left panel of B) that progresses and leads to lines of collective invasion and single cell invasion. Results shown are representative of 10 assays. (C) MPSLM of the TE-collagen gel boundary at day 14, showing collective (outlined) and single cell (examples marked with arrows) migration into cell-aligned collagen matrix (R1), consistent with the in vivo condition (10). Note that in low alignment regions (R2), the few cells that move into the collagen gel lack guidance from the aligned matrix and do not continue to invade through the matrix. (D) Quantification of cell-mediated matrix alignment in regions R1 and R2 using SHG imaging of the collagen matrix at the explant-collagen gel boundary (n = 3–4 assays). Collagen alignment was quantified by measuring the fiber angle relative to the tumor boundary and shown as a population distribution of the measured angles. The fiber angle was significantly different between regions R1 and R2 (p < 0.00001).

Generation of aligned collagen matrices

Collagen was magnetically aligned following the procedures outlined in Guo and Kaufman (15). Briefly, 1.5 μm diameter streptavidin coated iron oxide beads (Bangs Labs: BM551) were added to 2.0 mg/mL collagen solution at a concentration of 0.1 mg/mL. Collagen solution was pipetted into a rectangular mold and collagen aligned by placing a cylindrical magnet (~2G; McMaster: 5862K32) under the mold during collagen gel polymerization (i.e., collagen fibrillogenesis). Collagen alignment was confirmed by second harmonic imaging of the collagen matrix. To study cell migration in 3D, as a function of matrix alignment, the assay depicted in Fig. 5 A was utilized. Aligned collagen was arranged either perpendicular or parallel to a rectangle CSCG and encased within a 2.0 mg/mL collagen matrix. The migration assay was performed in the presence of DMEM containing 10% FBS.

Open in a separate window

FIGURE 5

Rho and ROCK are required for contractility-mediated matrix reorganization to produce migration promoting contact guidance (A) Phase contrast microscopy images showing 3D invasion into region R1 after Rho (10 μg/mL), ROCK (2.5 μM), and myosin-driven contractility (10 μM) inhibition. (B) MPLSM images of the CSCG-collagen gel interface under control, Rho inhibited (C3; 10 μg/mL), ROCK inhibited (H1152; 2.5 μM), and myosin contractility inhibited (Blebbistatin; 10 μM) conditions, showing decreased matrix reorganization in region R1 after 14 days. (C–F) Quantification of matrix alignment in region R1 from second harmonic imaging of the collagen matrix at the CSCG-collagen gel boundary at day 14 (n = 3–4 assays) when cells were treated with vehicle (C), Rho inhibitor C3 (10 μg/mL; D), ROCK inhibitor (2.5 μM; E), or the myosin inhibitor blebbistatin (10 μM; F). Note that all three inhibitors cause a loss in collagen fiber alignment, as evidenced by a distribution of fibers around 0° in the presence of inhibitors. (G) Quantification of the number of cells (at day 14) invading a defined distance into region R1 after matrix realignment was abolished after treatment with C3 (10 μg/mL), H1152 (2.5 μM), or blebbistatin (10 μM).

Transmitted light microscopy

Assays were imaged with a 10× objective on a TE300 Nikon inverted microscope (Nikon, Melville, NY) controlled by Slidebook software (Intelligent Imaging Innovations, Denver, CO). Composite images were created by imaging adjacent regions of the assay and reconstructing the image with Adobe Photoshop (Adobe Systems, San Jose, CA). Additionally, transmitted light images were acquired concurrently with multiphoton images using the system described below.

Multiphoton laser-scanning microscopy (MPLSM)

Live cell and collagen imaging was performed with MPLSM to generate multiphoton excitation (MPE) and second harmonic generation (SHG) as described previously (10). Briefly, the custom system utilizes a mode-locked Ti:sapphire laser (Millennium/Tsunami, Spectra-Physics, Mountain View, CA) excitation source producing around 100 fs pulse widths, which was tuned to 890 nm to generate both MPE autofluorescence signals from cells or green fluorescent protein (GFP) excitation and SHG signals from collagen. The beam was focused onto the sample with a 20× Plan Apo Multi-immersion lens (Nikon, NA = 0.75, WD 0.17 for water). The presence of collagen was confirmed using fluorescence lifetime imaging microscopy (FLIM) on the same system, since the SHG from collagen has no lifetime. Furthermore, because of the fundamentally different physical properties of MPE and SHG, signals could be discriminated by filtering the emission signal. We used a 464 nm (cut-on) long pass filter and a 480–550 nm (bandpass) filter to isolate the MPE emission from autofluorescence or GFP, respectively, from the conserved 445 nm SHG emission. A 445 nm (narrow bandpass) filter was therefore used to isolate the SHG emission (all filters: TFI Technologies, Greenfield, MA). The number of cells invading into the collagen gel was collected within three focal plains (center ± 30 μm) over the entire length between the TEs or CSCGs, or up to 250 μm into prealigned collagen matrices. Quantification of the collagen fiber angle relative to the explant or CSCG boundary was performed with ImageJ (National Institutes of Health) as previously described (10).

Metastatic tumor cells invade through aligned collagen fibers

Using late-stage tumors from the polyoma-middle-T (PyVT) transgenic model of human breast cancer, which is reliably invasive and metastatic with a 100% incidence of pulmonary metastasis (21–23), we examined 3D migration/invasion in vitro. When tumor explants from late-stage PyVT tumors were seeded into the 3D migration assay shown in Fig. 1 A, collective sheet-like volumes of cells invaded predominantly into Region 1 at early time points (days 1–5; see arrow in Fig. 1 B). This was followed (days 5–10) by continuation of the sheet-like collective migration as well as tracks of collectively invading cells (as seen in Fig. 1, B and C) and individual cell migration/invasion, which is consistent with previous reports demonstrating both individual and collective migration associated with carcinoma cell invasion (4,10,24).

To further investigate how organization of the matrix influenced 3D migration, we used live imaging of invading cells and the collagen matrix. To address the inherent challenges associated with imaging both cells and the ECM in live 3D microenvironments, live-cell and 3D matrix imaging was performed using MPLSM (25), which produces MPE of engineered or endogenous fluorophores (26,27) and SHG from noncentrosymmetric structures such as collagen (28,29). This approach allows visualization of invading cells within the matrix and quantification of matrix alignment within the planes of invasion. Using an MPLSM excitation wavelength of 890 nm, which excites endogenous fluorescence from flavin adenine dinucleotide (FAD) (30) while simultaneously producing a strong collagen SHG signal to image the matrix, we observed that the dominant region of invasion was into highly aligned collagen (Regions R1: Fig. 1, C and D and R3: data not shown) relative to regions of low alignment (R2 and R4, respectively). Quantification of fibrillar collagen organization from SHG images adjacent to the explant in Region R1 revealed a collagen alignment distributed around ~90° perpendicular to the tumor-stromal boundary, whereas low alignment regions (R2 in Fig. 1 A) were significantly different (p < 0.00001) and displayed a wide distribution around 0°, or parallel to the tumor boundary (Fig. 1 D). Hence, 3D cell invasion was enhanced in regions where the cells had perpendicularly aligned the collagen matrix, whereas few cells invaded into low alignment regions (Fig. 1 C). These observations are consistent with our previous report that when a single tumor explant is seeded into a 3D collagen matrix, groups of epithelial cells directly interact with collagen fibers and begin to organize the matrix to support invasion into the collagen gel at early (24 h) time points (10). Importantly, the invasion of metastatic cells followed interaction with, and reorganization of, the collagen matrix.

Collagen matrix alignment promotes 3D migration of human breast carcinoma cells

Using the highly migratory and invasive human breast carcinoma MDA-MB-231 cell line as a model of metastasis, we seeded cells into collagen plugs to form the CSCGs depicted in Fig. 1 A. Consistent with our findings from tumor explants, the human breast carcinoma cells underwent 3D migration/invasion predominantly into aligned regions of the matrix (Fig. 2, A–F). After 14 days, cells had invaded over 2500 μm into the aligned matrix (Fig. 2 G), resulting in an average migration speed of ~7.44 μm/h, which is consistent with data demonstrating 3D migration speeds in the range of ~4–12 μm/h in the DU-145 prostate cancer cell line (2). Importantly, significantly fewer cells had invaded into regions of low realignment (p < 0.009 for all distances ≥ 500 μm; Fig. 2, D and G; for similar results in the chemoattraction control regions R3 vs. R4 see Fig. S2). This finding suggests that radially aligned collagen fibers provide contact guidance via differential alignment, or anisotropy, of the collagen matrix (7,8). With further analysis we found that due to structural anisotropy in the collagen matrix, the degree of migration into the aligned region (M_R1) was significantly greater than migration for cells that invaded into the regions of low alignment (M_R2; Fig. 2 H). If the matrix structure was isotropic (i.e., had the same organization in all directions) then the migration ratio would be unity (Fig. 2 H), which further supports our conclusion that the degree of tumor cell invasion is a function of matrix organization. Hence, we propose that the matrix anisotropy provides contact guidance that is sensed by the cells and utilized to regulate 3D migration.

Interestingly, on 2D surfaces, substrate stiffness is known to regulate cell migration, a behavior known as durotaxis (31). Although studies have explored the mechanisms associated with durotaxis on 2D substrates (32–34), the mechanisms of durotaxis in 3D microenvironments are even less understood; but computational modeling predicts that the speed of cell migration will exhibit a matrix stiffness-dependent biphasic response (35). Importantly, in collagen matrices in vivo, matrix alignment significantly increases the strength and stiffness of the matrix relative to a matrix that is not organized along the principal axis or a disordered matrix (36,37). We interpret this to mean that in our assay the collagen matrix in the aligned region has greater stiffness due to better structural order. As such, cells probing the matrix and contracting along the primary axis may be encountering a stiffer environment, which implies that part of the physical mechanism of the contact guidance cues is, by definition, a 3D durotaxis response. However, the complex balance of a biphasic response to matrix stiffness reported on 2D substrates (34) and predicted in 3D (35) requires further study in 3D collagen matrices, and the likely influence of a haptotactic component should not be understated. Differing organization of collagen likely results in differential presentation or changes in the local density of adhesive ligands, which are known to influence cell motility on 2D substrates and within 3D matrices (32,34,35). Hence, it seems reasonable to speculate that contact guidance cues are influencing cell behavior by a coupled durotactic-haptactic response.

Matrix reorganization and 3D invasion is protease independent in vitro

Proteases are recognized to play a role in tumor progression (4,12,13). However, in vitro, tumor cells remain capable of matrix deformation (6) and invasion (3,5) when proteases are inhibited. Furthermore, recent work using a mouse model heterozygous for collagen that is resistant to MMP degradation demonstrated no inhibition of collagen realignment during local invasion (14). Therefore, to investigate whether or not extracellular proteases are required for collagen matrix realignment and 3D migration in vitro, we treated cultures with GM6001, a broad spectrum MMP inhibitor, or GM6001 + Aprotinin + Leupeptin, which additionally block a broad range of serine and cysteine proteases. Inhibition of protease activity did not inhibit invasion of MDA-MB-231 cells into collagen gels (Fig. 3 A). Furthermore, collagen matrix realignment was not inhibited by protease inhibition (Fig. 3, B–D), and quantitative analysis of 3D cell migration showed no difference in the number of cells migrating into the aligned collagen matrix (Fig. 3 E). Hence, in vitro, MDA-MB-231 cells remain capable of matrix reorganization and 3D migration in the presence of the protease inhibitors used in this study. However, the in vivo condition may be more complex with the presence of a developmentally established and highly organized ECM that must be “broken-down” to facilitate substantial reorganization. In collagen gels the fiber lengths are considerably shorter and the collagen less organized than the collagen structure surrounding a normal gland before tumor growth and progression (see mammary collagen anatomy in Provenzano et al. (10).). As such, tumor growth may require ECM degradation to accommodate early tumor progression. In fact we suspect that this loosening of the collagen matrix may help set in motion the collagen reorganization associated with local invasion. Future in vivo studies examining the temporal aspects of protease activation and reorganization over the time course of tumor progression will provide further insight into these open questions.

Open in a separate window

FIGURE 3

3D Invasion and matrix reorganization are protease independent in vitro. (A) Transmitted light microscopy of MDA-MB-231 cells invading into region R1 from CSCGs (located to the left of the images) exposed to vehicle (top), 10 μM GM6001 (middle), or the protease inhibitor (P.I.) cocktail containing 10 μM GM6001 plus 2 μg/mL aprotinin and 100 μM leupeptin (bottom) at day 10, representative of n ≥ 5. (B–D) Quantification of matrix alignment in region R1 from second harmonic imaging of the collagen matrix at the CSCG-collagen gel boundary at day 10 (n ≥ 5 assays). Cells were treated with vehicle (B), GM6001 (C), or the P.I. cocktail (D). (E) 3D migration/invasion of cells treated with vehicle alone (control), GM6001, or the P.I. cocktail was not significantly different for the number of cells that had invaded up to 1 mm into the collagen gel after 10 days.

Rho/ROCK dependent matrix reorganization regulates contact guidance driven 3D invasion in tumor cells

The Rho pathway effector, ROCK, is required for breast epithelial cell contractility to locally deform the collagen matrix in a protease-independent manner (6), and Rho and ROCK regulate whole gel contraction by breast carcinoma cells (18). We therefore hypothesized that the force needed to deform and reorganize the collagen matrix was generated by cell contractility events governed by the Rho pathway. Addition of the Rho inhibitor, exoenzyme C3 transferase (C3), substantially reduced the degree of matrix reorganization (Fig. ​(Fig.4).4). Associated with this decreased matrix alignment (i.e., loss of anisotropy; Fig. 4, G and H), tumor cells were not aligned along the primary axis of invasion (Fig. 4 A), and invasion into the matrix between the explants (Region R1 in Fig. 1 A) was decreased (Fig. 4, B and C). Matrix realignment at the explant-matrix interface was significantly reduced (p < 0.0001) in region R1 after Rho inhibition such that the matrix alignment frequency distribution resembled Region 2 in control explants (compare Fig. 1 D Region 2 with Fig. 4 H). Thus, Rho inhibition caused a loss of anisotropy that extended through the entire volume between explants (Fig. 4, I and J).

Open in a separate window

FIGURE 4

Rho is Required for matrix reorganization that facilitates 3D migration. (A and B) Transmitted light microscopy of cells invading into region R1 from PyVT tumor explants (located to the left of the images) exposed to 10 μg/mL of the Rho inhibitor C3 exoenzyme (B) or vehicle alone (A) at 14 days. (C–F) Combined (C and D) and MPE/SHG signal separated (E and F) MPLSM images near the TE-collagen interface at region R1 showing decreased cell-induced matrix reorganization when explants were treated with C3 at 10 μg/mL (D and F) when compared to control conditions (C and E). (G and H) Quantification of matrix alignment in region R1 using second harmonic imaging of the collagen matrix at the CSCG-collagen gel boundary at day 14 (n = 3–4 assays) when cells were treated with the Rho inhibitor C3 at 10 μg/mL (H) or vehicle (G). (I) SHG image of collagen alignment at the midpoint between explants after 14 days under control conditions. (J) SHG image of collagen alignment at the midpoint between explants after 14 days when cells are treated with 10 μg/mL concentration of C3 exoenzyme.

To gain further insight into Rho pathway regulation of matrix reorganization and to further confirm that the process of reorganization requires contractility-based force generation, we treated human MDA-MB-231 cells with C3, the ROCK inhibitor H1152 (a more potent and selective inhibitor than Y27632), and blebbistatin, a nonmuscle myosin ATPase inhibitor. Inhibition of Rho with C3 significantly reduced 3D migration into the collagen matrix (Fig. 5, A, B, and G), and simultaneously decreased matrix organization (Fig. 5 D) when compared to control values (Fig. 5 C), which is consistent with our observations in tumor explants (Fig. 4).

Cellular contractility is regulated by myosin-light chain (MLC), which when phosphorylated increases myosin ATPase activity to promote actin–myosin interactions and force generation. ROCK, an effector of Rho, promotes cellular contractility by increasing phosphorylation of MLC and inhibiting MLC phosphatase activity (38–40). Inhibition of ROCK resulted in a similar inhibition of both invasion (Fig. 5, A, B, and G) and alignment (i.e., loss of contact guidance anisotropy; Fig. 5 E), which suggests it is the effects of Rho on this pathway that mediate these processes.

To further establish a role for contractility in matrix reorganization, myosin-based contractility was inhibited with blebbistatin. As observed above, inhibition of actin-myosin based contractility inhibited matrix reorganization (Fig. 5 F) and 3D migration (Fig. 5, A, B, and G). These results suggest that Rho/ROCK-mediated contractility is important to establish matrix alignment and raise the possibility that the contribution of Rho/ROCK in this 3D migration/invasion assay is due primarily to the role of cellular contractility on creating an anisotropic matrix that mediates 3D migration via contact guidance.

We thank members of the Keely Laboratory and Laboratory for Optical and Computational Instrumentation for their helpful discussions regarding this work.

This work was supported by grants from the U.S. Department of Defense-CDMRP/BCRP: W81XWH-04-1-0428 (P.P.P.), National Institutes of Health-National Cancer Institute: R01-CA076537 and American Cancer Society: RSG-00-339CSM (P.J.K.), and National Institutes of Health-National Institute of Biomedical Imaging and BioEngineering : R01-EB000184 (K.W.E.).