A mechanically active heterotypic E-cadherin/N-cadherin adhesion enables fibroblasts to drive cancer cell invasion

Abstract

Cancer-associated fibroblasts (CAFs) promote tumour invasion and metastasis. We show that CAFs exert a physical force on cancer cells that enables their collective invasion. Force transmission is mediated by a heterophilic adhesion involving N-cadherin at the CAF membrane and E-cadherin at the cancer cell membrane. This adhesion is mechanically active; when subjected to force it triggers β-catenin recruitment and adhesion reinforcement dependent on α-catenin/vinculin interaction. Impairment of E-cadherin/N-cadherin adhesion abrogates the ability of CAFs to guide collective cell migration and blocks cancer cell invasion. N-cadherin also mediates repolarization of the CAFs away from the cancer cells. In parallel, nectins and afadin are recruited to the cancer cell/CAF interface and CAF repolarization is afadin dependent. Heterotypic junctions between CAFs and cancer cells are observed in patient-derived material. Together, our findings show that a mechanically active heterophilic adhesion between CAFs and cancer cells enables cooperative tumour invasion.

Figure 1. CAFs exert pulling forces on cancer cells

(a) Illustration of the 3D invasion assay (CAF, red; A431, green). (b) Confocal image of a spheroid (1:1 mixture of CAFs and A431 cells) after 60 hours of invasion. CAFs (red) led collective strands of A431 cells (green). Image representative of 6 samples. Scale bar, 100µm. (c) Magnified view of the strand highlighted in b. Scale bar, 20µm. (d) Illustration of the 2D migration assay. (e) A spheroid of A431 cells (unlabeled) 10 hours after CAF seeding (red). Image representative of >10 samples. Scale bar, 100µm. (f) Magnified view of the strand highlighted in panel e. (g) The incident angle (α) is defined as the angle between the longest axis of the CAF (black line) and the tangent to the spheroid edge (yellow line) at the first time of contact. By symmetry, α is taken in range 0-90º. (h) The escape angle (β) is defined as the angle between the longest axis of the CAF (black line) and the tangent to the spheroid after contact (yellow line). (i) Distribution of the incident angle α (n=46 CAFs from 3 independent experiments). (j) Distribution of the escape angle β (n=47 CAFs from 3 independent experiments). (k) Cancer cell velocity at the spheroid edge in the presence/absence of contact with CAFs. (l) Spheroid edge curvature in regions where CAFs contacted the spheroid (+CAF) and in regions of the same spheroid in which CAFs were absent (-CAF). In k and l, n=177 image fields without CAFs (-CAFs) and n=40 image fields with CAFs (+CAFs) from 5 independent experiments. (m,n) Merged image of a CAF (red) before (m) and 235 min after (n) contact with the spheroid edge (unlabled). Scale bars, 10µm. (o, p) Traction force maps of the CAFs shown in panels m and n, respectively. The purple vector indicates the magnitude and direction of the force transmitted at the interface between the CAF and the following A431 cell. The total traction force generated by the CAF is indicated in white and the force transmitted to the cancer cells is indicated in purple. Images and maps in m,o and n,p are representative of 12 and 13 samples, respectively. In o, the transmitted force falls within background noise levels (15.3±9.4 nN, mean±s.d.). Scale bars, 10µm. Error bars represent s.e.m. *** indicates P<0.0001, Mann-Whitney test.

Figure 2. CAFs and A431 cells form heterophilic E-cadherin/N-cadherin junctions

(a) TEM image of contact (white arrows) between a CAF and a A431 cell. Image representative of 20 contacts from 3 independent experiments. Scale bar 100nm. (b) mRNA expression levels of E-,N- and P-cadherin in CAFs and A431 cells measured using QRT-PCR. Bars show average of technical triplicates. (c) Confocal immunofluorescence images of N-cadherin (red), E-cadherin (green), and CAGAP-mcherry (constitutively expressed by CAFs as a marker) in a co-culture of CAFs and A431 cells. Image representative of >4 samples. Scale bar, 5 µm. (d) Confocal immunofluorescence images of N-cadherin, P-cadherin, and CAGAP-mcherry (CAFs) in a co-culture of CAFs and A431 cells. Image representative of >4 samples. Scale bar, 5 µm. (e) SIM immunofluorescence images of N-cadherin (green), E-cadherin (yellow), β-catenin (red) and F-actin (blue) at a contact between CAF and A431 cell. Image representative of 15 samples. Scale bar is 1μm for zoomed areas, 10μm for merged overview projection. (f) STORM image of N-cadherin/E-cadherin localization at the contact between CAF and A431 cell. Image representative of 3 samples. Scale bar, 500nm. (g) Time-lapse images of a CAF expressing N-cadherin-GFP contacting A431 cells expressing E-cadherin-WT (red) (upper panels) or A431 cells expressing E-cadherin-W2A mutant (red) (lower panels), scale bars, 20µm. (h) Stacked histogram of life-time of the E-cadherin/N-cadherin junction (based on the E-Cadherin and N-cadherin fluorescent signals) at the contact between CAFs and A431 cells, for CAFs mixed with A431-EcadWT cells (rescue control, n=14 contacts from 3 independent experiments) and A431-EcadW2A mutant cells (n=28 contacts from 3 independent experiments). Data are pooled in three categories of contact life-time, from 0 to 30 min, from 30 to 60 min, and longer than 60 min duration. *** indicates p=0.0007, Chi-squared test.

Figure 3. Evidence of E-cadherin/N-cadherin junctions in lung adenocarcinoma and vulval squamous cell carcinoma

(a,b) Co-cultures of CAFs from two patients with lung adenocarcinoma and H1437 cells show E-cadherin/N-cadherin junctions and β-catenin colocalization. Images representative of 2 samples for each panel. Scale bars, 5µm (see Supplementary Figure 5 for a third patient). (c) Immunostaining of the contact between cancer cells and CAFs both isolated from one patient with vulval squamous cell carcinoma. N-cadherin (red), E-cadherin (green), F-actin (blue). Images representative of 2 patient samples. Scale bar, 5µm. (d) Panels show intravital imaging xz and xy sections of a tumor growing in the mouse ear: A431 (green), CAF (red), and collagen second harmonic (magenta), arrows highlight the different tumor components. Images representative of 3 samples. Scale bar is 20μm. (e) Panels show intravital imaging xz and xy sections of a tumor growing in the mouse ear: A431-Ecad-Ruby and vulval CAF-Ncad-GFP. White arrow highlights heterotypic contact. Images representative of 3 samples. Scale bar is 20μm. (f) Images show staining of F-actin (blue), E-cadherin (green), and αSMA (red) in normal human oral mucosa and oral squamous cell carcinoma. White arrow highlights heterotypic contact between CAF and cancer cell, v - vessel. Images representative of 5 samples. Scale bar, 10μm. (g) Staining of fibronectin (magenta), active integrin β1 (green), and β-catenin (red) in normal human oral mucosa and oral squamous cell carcinoma. White arrow highlights heterotypic contact between CAF and cancer cell, yellow arrow highlights integrin/ECM contact by CAF, BM – basement membrane. Images representative of 5 samples. Scale bar, 10μm.

Figure 4. Heterophilic E-cadherin/N-cadherin junctions withstand forces and trigger mechanotransduction

(a) Illustration of the magnetic tweezers experimental setup. (b) Bead detachment data in A431 cells (CT), A431-EcadKO cells (EKO) and A431 cells pre-treated with E-cadherin blocking antibody (AbE). Percentage of beads coated with N-, E-, and P-cadherin that remained attached to A431 cells after application of a force pulse. (c) Bead detachment data in CAFs transfected with siRNA Control (CT) and CAF-siNcad (siN). Percentage of beads coated with N-, E-, and P-cadherin that remained attached to CAFs after application of a force pulse. (d) Illustration of the magnetic twisting experimental setup. (e) Representative fluorescence (top) and bright field (bottom) images showing the recruitment of β-catenin, P-cadherin, and E-cadherin in A431 cells subjected to magnetic stimulation using N-cadherin-coated magnetic beads. Yellow asterisks indicate the location of the beads. Scale bars, 5µm. (f) Quantification of the recruitment of β-catenin, P-cadherin and N-cadherin mediated by N-cadherin coated beads with/without (+/- Force) mechanical stimulation. (g) Representative bead traces for A431 cells and CAFs in response to a series of force pulses applied to beads coated with N-cadherin (red), E-cadherin (blue), P-cadherin (green) or uncoated (black). Vertical bars, 200nm. (h) Stiffening of the A431 cell-bead contact defined as the time evolution of the ratio between applied force and bead displacement relative to baseline (N-,E-,P-cadherin coated beads, and uncoated beads). (i) Stiffening of the CAF-bead contact. (j) Stiffening of the cell/E-cadherin-coated bead contact for control A431 cells (A431-WT) and α-catenin mutants. (k) Stiffening of the cell/N-cadherin-coated bead contact for A431-WT cells and α-catenin mutants. (l) Stacked histogram of life-time of the E-cadherin/N-cadherin junction at the contact between CAFs and A431 cells, for CAFs mixed with A431-αcatWT cells and A431-αcatΔVBS cells. Data are pooled in three categories of contact life-time, from 0 to 30 min, from 30 to 60 min and above 60 min duration. See Supplementary Table 1 for sample numbers and statistical analysis. Error bars represent s.e.m.

Figure 5. E-cadherin is required for force transmission between CAFs and A431 cells

(a) Net force transmitted between CAFs and A431 cells before the onset of contact and during contact. Experiments were performed under control conditions and after depletion of E-cadherin in the A431 cells using CRISPR/Cas9. The white bar indicates background noise levels. (-/+) n=12 CAFs from 9 independent experiments, (+/+) n=13 CAFs from 9 independent experiments, (-/-) n=13 CAFs from 2 independent experiments, (+/-) n=17 CAFs from 2 independent experiments, n=13 image regions from 8 independent experiments (noise level). Error bars represent s.e.m. *** indicates p<0.0001, * indicates P=0.0409, t-test. (b) Time evolution of the transmitted force between a CAF and the follower cancer cell for control A431 (red open symbols) and A431-EcadKO (black open symbols). The dashed line indicates the noise floor. (c-e) Snap shots of the collective migration of control A431 cells led by one CAF (CAGAP-mcherry) corresponding to the three time points labeled in b. The green vector indicates the magnitude and direction of the net transmitted force. Data representative of 5 time lapse experiments. See Supplemental Video 10 for the full time-lapse. (f-h) Snap shots of the collective migration of control A431-EcadKO cells led by one CAF corresponding to the three time points labeled in b. See supplemental Video 11 for the full time-lapse. Data representative of 5 time lapse experiments. Scale bars, 50 µm.

Figure 6. A mechanically active heterotypic adhesion regulates cell trajectories, leader/follower patterns, and CAF polarization.

(a) Illustration of the two modes of 2D collective invasion. CAFs (red) were classified either as “leaders” if their invasion was followed by a strand of A431 cells or as “loners” if they migrated away from the spheroid (gray) as individual cells. (b) Fraction of “leaders” vs “loners” in dermal CAFs compared to normal dermal fibroblasts (NF, Skin) paired with A431 cell spheroids. n=57 CAFs and 194 NFs from 3 independent experiments; *** indicates P<0.0001, Mann-Whitney test (c) Fraction of “leaders” vs “loners” in lung CAFs compared to normal lung fibroblasts paired with H1437 cell spheroids. n=202 CAFs and 192 NFs from 3 independent experiments; *** indicates P<0.0001, Mann-Whitney test. (d) Fraction of “leaders” vs “loners” upon depletion of the E-cadherin/N-cadherin junction (“+/+” n=86 CAFs from 4 independent experiments, “-/+” n=30 CAFs from 4 independent experiments (P=0.006), “+/-” n=67 CAFs from 3 independent experiments (P<0.0001), “-/-” n=76 CAFs from 3 independent experiments (P=0.001), Mann-Whitney test. (e) Distribution of the incident angle α for control CAFs (n=46 CAFs), CAF-siNcad (n=46 CAFs) and A431-EcadKO (n=42 CAFs), all pooled from 3 independent experiments. Distributions were not significantly different from each other, Kolmogorov-Smirnov test. (f) Distribution of the escape angle β for control CAFs (n=44 CAFs), CAF-siNcad (n=53 CAFs) and A431-EcadKO (n=35 CAFs), all pooled from 3 independent experiments. The escape angle distribution for CAF-siNcad and A431-EcadKO was significantly different from that of CAF control (P<0.001 for CAF-siNcad and P<0.05 for A431-EcadKO, Kolmogorov-Smirnov test). (g) Fraction of “leaders” vs “loners” in CAFs paired with a A431-CT (n=106 CAFs), A431-αcatWT (n=194 CAFs, P=0.098), and A431-αcatΔVBS (n=248 CAFs, P<0.0001). Data pooled from 3 independent experiments, Mann-Whitney test. (h) Distribution of the incident angle α for control CAFs paired with A431-CT (n=36 CAFs), A431-αcatWT (n=60 CAFs) and A431-αcatΔVBS (n=62 CAFs), all pooled from 3 independent experiments. Distributions were not significantly different from each other, Kolmogorov-Smirnov test. (i) Distribution of the escape angle β for A431-CT (n=36 CAFs), A431-αcatWT (n=60 CAFs) and A431-αcatΔVBS (n=62 CAFs), all pooled from 3 independent experiments. The escape angle distribution for A431-αcatΔVBS was significantly different from that of A431-CT (P<0.0001 Kolmogorov-Smirnov test). All other distributions were not significantly different from each other. *** indicates P<0.001, ** indicates P<0.01, n.s. indicates not significantly different.

Figure 7. Afadin and nectins 2 and 3 are required for CAF-led migration of cancer cells and for CAF polarization.

(a) Confocal images of nectin-3 (blue), N-cadherin (green), E-cadherin (red) in a co-culture of CAFs and A431 cells (upper panels); nectin-2 (blue), N-cadherin (green), E-cadherin (red) (middle panels); afadin (blue), N-cadherin (green), E-cadherin (red) (lower panels). Yellow arrows show the localization of the CAF/A431 cell contact. Images representative of 2 samples. Scale bars, 5 µm. (b) Staining of afadin (green), and p120catenin (red) in normal human oral mucosa and oral squamous cell carcinoma. White arrow highlights heterotypic contact between CAF and cancer cell. Images representative of 5 samples. Scale bar is 10μm. (c) Fraction of “leaders” vs “loners” in CAF-siCT (n=90 CAFs) and CAF-siAF (n=95 CAFs). Data pooled from 3 independent experiments. *** indicates P<0.001, t-test. (d) Distribution of the incident angle α, and escape angle β for CAF-siCT (n=30, 32 CAFs for α and β, respectively) and CAF-siAF (n=29 CAFs for α and β), all pooled from 3 independent experiments. Escape angle distribution of CAF-siAF was significantly different from that of CAF-siCT (P=0.0001 Kolmogorov-Smirnov test). All other distributions were not significantly different from each other. (e) Confocal fluorescence images of afadin staining in CAFs 3 days after siRNA transfection with siRNA control or siRNA targetting afadin. Images representative of 3 samples. Scale bars, 20 µm. (f) Western blot of afadin and α-tubulin for CAFs-siCT and CAFs-siAF.

Figure 8. The E-cadherin/N-cadherin junction enables collective cancer cell invasion in 3D

(a-e) Fluorescence images of spheroids containing different mixtures of CAFs and A431 cells after 60 hours of invasion in an organotypic ECM. (a) 1:1 mixture of control A431 (YPet) and control CAFs (KEIMA). (b) 1:1 mixture of A431-EcadKO (mCherry) and control CAFs (KEIMA). (c) 1:1:2 mixture of A431 control (YPet), A431-EcadKO (mCherry), and control CAFs (KEIMA). Arrow points to one A431-EcadKO cell in the invasive strand. (d) 1:1 mixture of A431 control (YPet) and CAFs-siRNA (KEIMA). (e) 1:1 mixture of A431-αcatΔVBS cells (mcherry) and CAFs-siRNA (KEIMA). See Supplementary Figure 7 for additional spheroid conditions. Scale bars, 100 µm. (f) Average number of strands per spheroid in the conditions shown in (a-e), and CAF-siCT and A431-αcatWT. Number of spheroids measured: n=24 (control), n=18 (EKO, P<0.0001), n=18 (PKO, P>0.999), n=19 (siCT), n=19 (siN, P<0.0001), n=11 (αcat WT), 6 (αcat ΔVBS, P=0.016), from 3 independent experiments. One-way Anova with Dunn’s multiple comparison test. (g) Percentage of the A431 cells found immediately after the CAFs in spheroids containing a triple mixture of two distinct populations of A431 (YPet, mCherry) and one population of CAFs (KEIMA). The three combinations of A431 cells are: A431-control (Ypet)/A431-control (mCherry) (n=66 strands measured), A431-control (Y-Pet)/A431-EcadKO (mCherry) (n=41 strands measured, P=0.0024), and A431-control (YPet)/A431-PcadKO (mCherry) (n=56 strands measured, P=0.826), from 3 independent experiments. These results show that when A431 control cells and A431-EcadKO cells are mixed, the probability of finding an A431-EcadKO behind the leading CAF is negligible. Error bars are s.e.m; n.s, indicates not significantly different when compared with controls. One-way Anova with Dunn’s multiple comparison test.