Collective cell migration is spatiotemporally regulated during mammary epithelial bifurcation

Abstract

Branched epithelial networks are generated through an iterative process of elongation and bifurcation. We sought to understand bifurcation of the mammary epithelium. To visualize this process, we utilized three-dimensional (3D) organotypic culture and time-lapse confocal microscopy. We tracked cell migration during bifurcation and observed local reductions in cell speed at the nascent bifurcation cleft. This effect was proximity dependent, as individual cells approaching the cleft reduced speed, whereas cells exiting the cleft increased speed. As the cells slow down, they orient both migration and protrusions towards the nascent cleft, while cells in the adjacent branches orient towards the elongating tips. We next tested the hypothesis that TGF-β signaling controls mammary branching by regulating cell migration. We first validated that addition of TGF-β1 (TGFB1) protein increased cleft number, whereas inhibition of TGF-β signaling reduced cleft number. Then, consistent with our hypothesis, we observed that pharmacological inhibition of TGF-β1 signaling acutely decreased epithelial migration speed. Our data suggest a model for mammary epithelial bifurcation in which TGF-β signaling regulates cell migration to determine the local sites of bifurcation and the global pattern of the tubular network.

Keywords: Bifurcation; Branching morphogenesis; Collective cell migration; Epithelial development; Mammary gland; TGF-β.

Fig. 1.

Epithelial cell speed decreases within clefts during mammary ductal bifurcation. (A) Confocal projection of an organoid branch undergoing bifurcation, expressing membrane-targeted tdTomato (red, top). Three representative phases of bifurcation are shown: initial, rounded; bud flattening; and cleft formation (bifurcate). In total, 28 of 38 organoids (73.7%) were observed to undergo a bud flattening phase. 38 organoids from r=10 replicate experiments. (B) Schematic of the three phases of mammary branch bifurcation. (C) Confocal projection of an organoid branch undergoing bifurcation expressing H2B–GFP (green) and membrane-targeted tdTomato (red). Time points are as indicated in C′. (C′) Nuclei trajectories for the organoid branch in C are shown for branch cells (green), cleft cells (red) and stalk cells (blue). Tracks represent the cell path over the previous 5 h. (D) Mean cell speeds (µm/h) were calculated from nuclei trajectories as track length divided by time for branch cells (8.79±2.15 µm/h, 147 cells), cleft cells (7.42±1.60 µm/h, 54 cells), and stalk cells (6.79±1.53 µm/h, 71 cells). Mean±s.d. of cells in seven organoids from r=4 replicate experiments. Kruskal–Wallis ANOVA reached significance (****P<0.0001; ns, not significant). (E) Confocal projection of an organoid branch undergoing bifurcation expressing H2B–GFP (green) and membrane-targeted tdTomato (red). Nucleus trajectories are shown for an individual cell migrating in the branch (green, +4 h) to the cleft (red, +8 h) and then returning to the branch (green, +13.5 h). Tracks represent the entirety of the path length, with path lengths for each stage indicated. Image on the right shows the complete cell track. (E′) Images showing the nucleus trajectories described in E. (E″) Magnified views of the trajectories shown in E′. (F) Paired mean cell speeds (µm/h) were calculated from nuclei trajectories as track length divided by time for individual cells that were migrating in a branch (9.03±1.36 µm/h, mean±s.d.; pre-cleft) to a cleft (6.75±1.15 µm/h) and returning to the branch (9.65±3.04 µm/h; post-cleft). Data is shown for 16 cells from eight organoids imaged in r=4 replicate experiments. Paired Friedman's ANOVA with Dunn's multiple comparisons reached significance (****P<0.0001; ns, not significant). Dashed lines in A,C,C′ and E′ indicate the organoid branch outline.

Fig. 2.

Epithelial cells are anisotropically protrusive towards bifurcation clefts. (A) Schematic illustrating overlay method and orientation for assigning protrusions to 45° bins. Branch cells are aligned with the direction of elongation. Stalk and cleft cells are aligned with the bisection line between the two elongating branches. (B,B′) A representative confocal projection (B) and 3D reconstruction (B′) of a stalk cell with isotropic protrusions, imaged during active bifurcation. (C) Polar histogram showing protrusions per bin quantified from organoid stalk cells (1343 protrusions from 12 cells in eight organoids imaged in r=3 replicate experiments). Two-way ANOVA did not reach significance (NS, P>0.05). (D,D′) A representative confocal projection (D) and 3D reconstruction (D′) of a branch cell with anisotropic protrusions, imaged during active bifurcation. (E) Polar histogram showing protrusions per bin quantified from organoid branch cells (1139 protrusions from 12 cells in six organoids imaged in r=3 replicate experiments). Two-way ANOVA reached significance (P<1×10⁻¹³). Two-way MANOVA reached significance for the comparison between stalk and branch cells (P<5×10⁻⁸). (F,F′) A representative confocal projection (F) and 3D reconstruction (F′) of a cleft cell with anisotropic protrusions, imaged during active bifurcation. (G) Polar histogram showing protrusions per bin quantified from organoid cleft cells (1030 protrusions from 12 cells in seven organoids imaged in r=3 replicate experiments). Two-way ANOVA reached significance (P<1×10⁻³⁸). Two-way MANOVA reached significance for the comparison between stalk and cleft cells (P<1×10⁻¹⁰). Branch cells were not compared to the cleft cells as they have different axes of orientation. Cells in B,D and F are labeled with H2B–GFP (green) and membrane-targeted tdTomato (red).

Fig. 3.

Inhibition of TGF-β signaling results in hyperbranching of organoids. (A) DIC images taken at day 7 showing organoids that were cultured from day 0 in the presence of vehicle (DMSO), 5 ng/ml exogenous TGF-β1 or 50 µM TGF-βR1 inhibitor (LY364947). Yellow boxes indicate regions shown in magnified images (bottom). (B) DIC images taken at day 7 showing organoids that were cultured from day 4 (active branch elongation) in the presence of vehicle (DMSO), 5 ng/ml exogenous TGF-β1 or 50 µM TGF-βR1 inhibitor (LY364947). Yellow boxes indicate regions shown in magnified images (bottom). (C) Mean±s.d. percentage of organoids forming branches at day 7 following treatment with the indicated concentrations TGF-β1, TGF-βR1 inhibitor or vehicle (Veh; DMSO) from day 0. Vehicle, 81.3±17.1% (269 organoids). Exogenous TGF-β1: 0.05 ng/ml, 73.4±10.6% (368 organoids); 0.5 ng/ml, 49.0±1.1% (310 organoids); 2.0 ng/ml, 25.7±7.7% (386 organoids); 5.0 ng/ml, 3.6±1.7% (274 organoids). TGF-βR1 inhibitor (LY364947): 0.1 µM, 76.0±7.8% (426 organoids); 1.0 µM, 79.1±1.5% (399 organoids); 10 µM, 83.9±2.5% (398 organoids); 50 µM, 81.6±9.6% (409 organoids). Data are from r=3 replicate experiments. Ordinary two-way ANOVA with Tukey's multiple comparison's test reached significance (ns, P>0.05; **P<0.01; ****P<0.0001). (D) Mean±s.d. number of branches per area for organoids at day 7 following treatment with vehicle (Veh; DMSO) or the indicated concentration of TGF-βR1 inhibitor (LY364947) from day 0. Vehicle, 1.41±0.10; 0.1 µM, 1.67±0.24; 1 µM, 1.84±0.34; 10 µM, 2.81±0.58; and 50 µM, 3.60±0.66. Data are from 10 organoids per condition, r=3 replicate experiments. Ordinary one-way ANOVA with Holm–Sidak multiple comparison test with a single-pooled variable reached significance (ns, P>0.05; ****P<0.0001).

Fig. 4.

TGF-β1 signaling is acutely required for epithelial cell migration. (A) Confocal projections of an organoid branch undergoing elongation, expressing H2B–GFP (green) and membrane-targeted tdTomato (red). (A′) Nuclei trajectories for the branch shown in A, with tracks in multiple colors to allow identification at different time points. Tracks represent the cell path over the previous 5 h. Scale bar: 10 μm. (B) Confocal projections showing the same organoid branch as depicted in A following treatment with 5 µM TGF-βR1 inhibitor (LY364947). Times are shown relative to the start of imaging pre-inhibition. (B′) Nuclei trajectories from the branch shown in B, with tracks in multiple colors to allow identification at different time points. The last 6 h of imaging are displayed. Tracks represent the cell path over the previous 5 h. Scale bar: 10 μm. (C) Mean cell speeds (µm/h) in organoid branches were calculated from nuclei trajectories as track length divided by time for cells before (Pre; median, 6.55 µm/h; interquartile range, 7.72–9.17 µm/h; 2535 cells) and after (Post; median, 4.66 µm/h; interquartile range, 5.44–6.32 µm/h; 3821 cells) treatment with 5 µM TGF-βR1 inhibitor (LY364947). Cells were from 69 organoids imaged in r=4 replicate experiments. Two-tailed unpaired t-test reached significance (****P<0.0001). Boxplots show median and interquartile range, with whiskers marking the minimum and maximum cell speeds. (D) Persistence was calculated from nuclei trajectories as displacement length divided by total track length for cells in organoid branches before (0.55±0.14, 10 cells) and after (0.21±0.15, 10 cells) treatment with 5 µM TGF-βR1 inhibitor (LY364947). Mean±s.d. of cells from four organoids imaged in r=3 replicate experiments. Two-tailed unpaired t-test reached significance (****P<0.0001).