Collective Cell Movement in Primary Melanoma Explants: Plasticity of Cell-Cell Interaction, β1-Integrin Function, and Migration Strategies^1,2

Tumor cell invasion into interstitial tissue is a complex process that includes the migration of individual cells as well as multicellular, collective cell movement (1, 2). Collective cell movement, i.e., the migration of cell sheets, aggregates, or clusters, is relatively common in higher eukaryotes, as described previously for two-dimensional sheets of keratinocytes (3), fish melanocytes (4), and adenocarcinoma cells (5); the in vivo movement of cell groups in embryological development (6, 7, 8); and the migration of cell clusters from tumor explants cultured in three-dimensional collagen lattices (9).

To understand the development of different migration programs in tumor cell invasion and motility, it is important to dissect cellular and molecular similarities and differences among individual and collective cell movements (10, 11). Both migration strategies are dependent on front-rear asymmetry driven by a dynamic leading edge and coordinated detachment of the trailing edge (1, 9). In two- and three-dimensional models for individual tumor cell migration, anterior force generation is provided by adhesion receptors of the integrin family interacting with extracellular matrix components (12, 13, 14, 15), supporting the concept of haptokinetic adhesion-dependent migration (1, 12).

Collective cell movement requires that several cells, grouped together by adhesive cell-cell contacts, generate front-rear asymmetry via unipolar ruffling in cells at the leading edge, whereas cells located at the trailing edge remain largely nonmotile (3, 4, 9). Such multicellular units constitute a larger-sized unit with reduced deformability, compared with individually migrating cells, and utilize yet unknown pathways of cell-cell communication. Polarized substrate interaction in tumor cell collectives has been shown for colon adenocarcinoma sheets, which exhibit preferential expression of matrix metalloproteinases MMP-2 and MT1-MMP and focalized proteolytic degradation of gelatin substrate at the outward edges (16). However, how polarized cell-substrate interactions are established and which adhesion receptors are required for force generation in collective versus single-cell migration are unknown.

In the present study using primary melanoma explants, we show that collective tumor cell movement depends on β1-integrin-mediated cell-matrix adhesion, force generation, and migration. Intriguingly, impairment of cluster cohesion resulted in the detachment and dissemination of amoeboid single cells, highlighting unexpected plasticity in tumor cell migration.

Primary melanoma specimens were divided in the operation theater. The pathological diagnosis was established from paraffin sections by routine histology combined with immunohistochemistry for S100 protein, gp100, and MART-1. For primary explant culture, the malignant region of the explant was confirmed by frozen section, and immediately adjacent tissue portions were used for further explant culture in three-dimensional collagen lattices. In collagen, melanocytic cells within multicellular invasion zones and detached clusters were detected by immunohistochemistry using mAb⁴ HMB45 (Dako, Hamburg, Germany) and anti-S100 mAb (Dako).

Tumor samples were carefully dissected (maximum size, 1 × 1 × 1 mm) and incorporated into three-dimensional collagen lattices (1.67 mg/ml; Vitrogen bovine dermal collagen; Collagen Corp., Palo Alto, CA), as described previously (9). Explant cultures were monitored by bright-field microscopy for up to 6 weeks. Invasion zones or migrating cell clusters were defined as areas with more than four coherent cells and further monitored by time-lapse videomicroscopy for the exclusion of passively scattered nonmobile material or debris (9).

For function blocking, adhesion- perturbing anti-β1 mAb 4B4 (Coulter, Hamburg, Germany; Ref. 17) and anti-E-cadherin mAb HECD-1 (Zymed, Berlin, Germany; Ref. 18) were used. Antibody 4B4 abrogates invasive melanoma cell migration within collagen lattices (15), and HECD-1 blocks E-cadherin-mediated cell-cell interactions, resulting in the scattering of epithelial cells from multicellular aggregates (16).

Locomotion parameters were obtained by computer-assisted cell tracking and reconstruction of cell paths, as described previously (15, 19). In brief, the X and Y coordinates of six individual cells at different positions within the cluster were obtained for each step (step interval, 2–15 min) and reconstructed. Speed represented the total length of the calculated mean path of a cell group divided by time. Relative step angles of individual cell paths within a collective (corresponding to directional migration persistence) were calculated as relative deviation from each previous step and ranged from 0 (migration in a straight line) to 180 degrees (backward movement).

For analysis of cluster interaction with collagen fibers and matrix remodeling, confocal reflection contrast and fluorescence microscopy (Leica TCS-4D; Bensheim, Germany) were performed as described previously (14, 20).

As monitored by explant culture in three-dimensional collagen lattices, 6 of 13 primary melanoma samples from different patients developed collective cell movement, as represented by multicellular invasion zones that progressed into migrating cell clusters. This migration type combined the maintenance of adhesive cell-cell junctions with invasive cell dissemination into tissue, as is frequently seen in melanoma histopathology (21, 22). Collective movement from melanoma explants was most prominent in samples from high-risk melanoma (tumor thickness >3.6 mm) as well as in samples from nodular and acro-lentiginous melanoma (Table 1). Cluster frequency, onset of migration (3–14 days), and morphological and migratory phenotypes were similar to clusters from rhabdomyosarcoma and epithelial cancer (9).

We first investigated how cluster polarity and movement corresponded to adhesive interactions with collagen fibers. The migratory direction was obtained by time-lapse videomicroscopy (Fig. 1,A, inset, large arrow), followed by fixation and three-dimensional confocal reconstruction. As an indicator of high traction force, fiber alignment to thicker bundles converging toward the cluster body was preferentially detected at the leading edge and, less extensively, at the trailing edge (Fig. 1, A–C; compare with Fig. 2,B). In contrast to bundles, individual fiber insertions were equally distributed at leading and trailing edges (Fig. 1, A–C). Although the three-dimensional nature of the collagen did not permit us to deduce traction forces, cluster-induced reorientation of collagen fibers (ranging up to 150–300 μm in the forward direction) greatly exceeded fiber alignment generated by individual melanoma cells from previous studies (maximum of 50–100 μm in the forward direction; Refs. 14, 15, 20).

As was apparent from time-lapse videorecordings (video 1) and reinforced by the structure of the actin cytoskeleton, migrating cell clusters behaved as asymmetric structural and functional units. The most prominent F-actin staining (Fig. 1,D, white arrowhead) was located close to attachment and bundling sites at the leading edge, whereas more diffuse F-actin staining was present at the trailing edge. Cortical actin rims lacking stress fibers were present at most cell-cell junctions along traction lines over several adjacent cell bodies (Fig. 1,D, black arrowheads), suggesting some degree of supracellular cytoskeletal organization. The presence of melanocytic cells within multicellular invasion zones and detached clusters was confirmed by the presence of dark pigmented cells (Fig. 1,E) as well as by positive staining for melanocytic marker HMB45 (Fig. 1 F).

During cluster migration, β1 integrins comprising the major collagen receptors (i.e., α1β1, α2β1, α3β1) were redistributed and clustered in ruffling microspikes along the leading edge(s) (Fig. 2,A, white arrowheads). In contrast, nonclustered, linear β1-integrin distribution was detected at some, but not all, cell-cell interactions (Fig. 2,A, asterisk). In some clusters, β1-integrin staining intensity was highly heterogeneous up to the exclusive expression of β1 integrins in a subset of cells in association with the leading fiber traction zone (Fig. 2,B, black arrowheads). In contrast to β1 integrins, E-cadherin showed linear staining at cell-cell junctions only (Fig. 2,C, arrowheads), in accordance with previous data on adenocarcinoma cell sheets (16). The extension of clusters clearly surpassed preformed matrix gaps, and a circumscribed matrix defect was detected at the trailing edge (Fig. 2 B, asterisk), supporting the concept of proteolytic removal of matrix barriers during collective movement (16).

To assess the function of β1 integrins in migratory force generation, migrating clusters were monitored before (endogenous baseline control) and after the addition of adhesion-perturbing anti-β1-intergin mAb 4B4. In two different explants, the addition of antibody 4B4 led to nearly complete abrogation of cluster migration (Fig. 3, A and B), whereas anti-E-cadherin mAb HECD-1 had no effect (Fig. 3,C). In the absence of mAb, spontaneous baseline migration was maintained for at least 36 h (endogenous control; Fig. 3 B). Together, polarized integrin engagement at the leading edge and mAb 4B4-induced abrogation of collective movement indicate an integrin-dependent, traction-driven mode of force generation.

It became clear from the videorecordings that after the action of the β1 integrins was blocked, impaired cluster migration involved complex changes in both cell-matrix and cell-cell interaction (video 1 in the supplementary material).² Collective baseline migration (Fig. 4,A, before) resulted from stringent path coordination between individual cells at different locations within the cluster and a highly persistent mean path of the cluster (Fig. 4,B), also shown by low directional deviation from step to step (relative step angles; Fig. 4,C). As a first event 8–10 h after the addition of mAb 4B4, a second leading edge was developed by the former trailing edge, which pulled in the direction opposite to the previous direction of migration (Fig. 4,A and video 2).² Consequently, path persistence was impaired (Fig. 4,B), resulting in strongly oscillating relative step angles (Fig. 4,C). Approximately 20 h after the addition of mAb 4B4, disruption of the cluster into two separate aggregates was complete (Fig. 4,A and video 3).² Hence, after β1-integrin function was disturbed and front-rear asymmetry was lost, previously nonmigratory subsets at the trailing edge were able to initiate active migration within hours. As an end point, net cluster migration was lost (Fig. 4,A and video 4),² resulting in scrambled uncoordinated cytoskeletal oscillations (“running on the spot”), whereas the residual capacity of cells to change position within the cluster (Fig. 4 B, top right panel) and to interact with the collagen substrate (via β1-integrin-independent mechanisms) remained intact.

The sessile yet oscillatory state of the cluster gave rise to nonpolar outward ruffling and detachment of single cells, which migrated within the collagen at speeds of 0.2–0.6 μm/min (Fig. 5, A and B; videos 3 and 4). Hence, previously nonmigrating cells located within the cluster retained a basic capacity to locomote individually, which however, appeared to have remained silenced as long as the cluster was intact.

Detached cells developed linear path segments alternating with sharp and tortuous changes in direction (Fig. 5,C) and maintained an elliptoid shape coupled to morphological adaptation along matrix structures (Fig. 5,D, panel a, asterisk). Constriction rings developed at locations of narrow matrix constraints (Fig. 5,D, panel b, white arrowheads), whereas no signs of prominent fiber bundling or proteolytic matrix remodeling were detected (Fig. 5 D, panel a, black arrowheads). Hence, by altering the adhesive strength of cell-cell and/or cell-matrix interactions, neoplastic cells might be able to switch from one migration type to another.

One major difference between individual and collective tumor cell movement is that collective migration requires a promigratory subset of cells at the leading edge. In morphogenic multicellular migration, cells at the leading margin of the zebrafish blastoderm were designated “forerunner cells” (8). In clusters, cells at the leading edge appear to function as “path generating” and “guiding” cells that are prevented from “forerunning” by stringent cell-cell contact within the cluster. Cells that form the leading edge may retain structural and functional properties not necessarily present in cells located at inner regions of the cluster. In principle, the functional specificity of guiding cells could result from a genetic program, as detected in E-cadherin-mediated migration of ovarian follicle cells (slow border cells) within cell-rich tissue in embryonic Drosophila melanogaster (7). Alternatively, promigratory guiding activity could be a function of polarized integrin expression and engagement, as shown in the present study, as part of a temporary functional state maintained by graded cell-matrix interactions (13) and concomitant silencing of neighboring cells via cell-to-cell signaling. Clearly, more detailed studies are required on the integrin subsets, cell-cell adhesion, and cell-cell communication involved in collective movement. Although primary melanoma explants, as used in the present study, offer a natural source of highly invasive samples, some important restrictions reside in the limited availability of fresh source material as well as considerable donor-dependent heterogeneity in biological behavior. Hence, for reasons of efficiency and reproducibility, future studies will additionally require the development of improved three-dimensional models using solid spheroids from established cell lines.

In melanoma clusters, the blocking of β1 integrins initiated a striking transition from multicellular migration to single-cell crawling. This process is reminiscent of the epithelial-to-mesenchymal transition that occurs when single epithelial cells scatter after treatment with SF/HGF or anti-E-cadherin antibody (16, 23). Blockage of the β1 integrins could directly interfere with integrin-mediated cell-cell cohesion by reducing homophilic integrin-integrin binding (24), by disturbing integrin binding to extracellular matrix components, such as fibronectin, contained along cell interphases (25), or by inducing changes in intracellular signaling toward other receptors that maintain homotypic cell-cell interaction in melanoma cells, such as N-cadherin (26) or L1 (27).

The transition of β1-integrin-dependent collective movement to β1-integrin-independent migration of single cells suggests that the forces involved in single-cell movement range below those required for migration of cell groups. The features of crawling cells after detachment from the cluster, i.e., elliptoid morphology, oscillatory path structure, the lack of matrix remodeling, and apparently low adhesive β1-integrin-independent movement, are reminiscent of the amoeboid migration of T lymphocytes, which squeeze and crawl through matrix gaps present in three-dimensional collagen matrices (19, 28). A similar transition from fibroblast-like morphology toward amoeboid locomotor behavior was reported previously for neural crest cells after blockage of fibronectin-binding integrins (29), supporting the concept of residual migratory capacity after functional abrogation of certain integrin subsets. In conclusion, by altering the adhesive strength of cell-cell and/or cell matrix interactions, neoplastic cells may undergo transition from one migration program to another.

These results support a concept of diversity and adaptation in tumor cell migration within a three-dimensional tissue matrix. Collective cell movement in primary melanoma explants represents an efficient migration strategy that allows both active and passive translocation of heterogeneous sets of cells, thereby potentially promoting the dissemination of cells of different clonality and function within one functional unit. Importantly, once cell-substrate and/or cell-cell contacts are weakened, multicellular migration may convert to single-cell movement, representing the transition toward a secondary “salvage” migration strategy. Because cell migration may comprise diverse cellular and molecular mechanisms, future therapeutic targeting of the invasion cascade will require taking such migratory plasticity into consideration, i.e., to further determine similarities and differences in collective versus single-cell movement as well as related transition stages.

We acknowledge Martina Jossberger for excellent technical assistance, Christian Rose for expert histological assessment, and Katarina Wolf for critical reading of the manuscript.