Mechanical plasticity of collagen directs branch elongation in human mammary gland organoids

Latest Research

Human mammary gland organoids invade the ECM by non-continuous contractions

For the generation of human mammary organoids, single EpCAM+/CD49f+/CD10+ basal mammary epithelial cells were isolated from human reduction mammoplasties and seeded at clonal density into freely floating collagen type I gels22 (Fig. 1a, Supplementary Fig. S1A). The development of the mammary organoids can be classified in three stages (Fig. 1b): the establishment phase (day 1–6), the branch elongation phase (day 7–9) and the alveologenesis phase (day 10–14). Characteristic for the establishment phase were small, rod-like cell clusters with lengths of around 120 µm (Fig. 1c), that showed only rudimentary branches. During the elongation phase, these branches invaded into the ECM and developed primary and secondary side branches. During branch elongation, we detected proliferative cells throughout the whole organoid (Supplementary Fig. S1B) and observed the formation of filopodia-like cellular protrusions at the leading edge of the branches, hinting towards an invasive branching process (Supplementary Fig. S1C). Ultimately, during the alveologenesis phase, the organoids reached a size of around 1 mm in diameter (Fig. 1c). Moreover, the branches of these organoids thickened and formed rounded end buds allowing the formation of a lumen, resulting in the formation of TDLU-like structures. This third phase also coincided with polarized expression of the previously described markers for the two major lineages within the mammary gland22: p63 for basal cells, expressed within the outer cell layer adjacent to the ECM, and GATA3 for luminal cells, expressed within the cell layer adjacent to the forming lumen (Supplementary Fig. S1D). The expression of these specific markers showed that the mammary organoids generated in 3D floating collagen I gels resemble the bilayered architecture of the human mammary gland.

Fig. 1: Human mammary gland organoids invade the ECM by non-continuous contractions.
figure1

a Schematic overview of 3D culture: single primary human basal mammary epithelial cells are cultured in floating collagen gels. b Characteristic organoid morphology at three developmental stages (Establishment n = 36 organoids, Branch elongation n = 75 organoids, Alveologenesis n = 111 organoids). Nuclei are visualized using sirDNA. c The organoid diameter of the long axis during the different stages reveals an increase in diameter during the elongation phase (Establishment n = 36 organoids, Branch elongation n = 75 organoids, Alveologenesis n = 111 organoids). Box plots indicate median (red line), 25th, 75th percentile (blue box) and 5th and 95th percentile (whiskers) as well as outliers (single points). d Live-cell imaging reveals an anisotropic deformation field with strong deformations in front of the branches and no deformation at the sides of the branches (n = 24 organoids). The near field (n.f.) is defined as area between the branch tip and the ECM 300 µm away from it. e The deformation is decreasing with increasing angle (theta) to the branch (n = 14 organoids). Error bars, mean ± s.d. f The bead displacement is non-continuous over time with contractions towards the branches and relaxations into the opposite direction (n = 14 organoids). g ECM contractions and relaxations slowly diminish with increasing distance to the organoid. Between each line 25 min passed, highlighting the alternations between contractions of the ECM towards the branches and relaxations away from them (n = 23 organoids). h The cumulative bead displacement in front of branches is increasing over time (red, n = 5 organoids), while the branch elongation is discontinuous in time (gray, n = 7 organoids). Scale bars, 200 µm (b), 70 µm (d). Organoids were derived from three biologically independent donors (Supplementary Table S1). P values are from a two-tailed Mann–Whitney test and provided in Supplementary Table S5. Source data are provided as a Source Data file.

The expansion of the developing organoids was observed by live-confocal-microscopy over extended time periods, from hours to days (Supplementary Video 1). As observed by means of embedded tracer particles, the elongation of each branch induced a significant long-range deformation field within the ECM directed towards the branch (Fig. 1d). After 14 days of culture these local deformations added up to millimeters, which led to a macroscopic shrinkage of the collagen gel to about half its diameter22. In the near-field, the observed strains exhibited local heterogeneities and were highly anisotropic: deformations were the strongest directly at the tip of the branch in extension to the direction of elongation and steadily decreased with increasing angle from the branch tip (Fig. 1e/ Supplementary Fig. S1E). Moreover, we observed a periodic displacement of the embedded tracer particles towards the branch with periods in the hour scale (Fig. 1f). Specifically, bead displacements towards the branch were followed by relaxations in the opposite direction. These periodic displacements were not restricted to the close proximity of the branches but could also be detected further away from the organoid (Fig. 1g). Computing the cumulative displacement of the beads over time showed a steady increase in bead displacement towards the branches (Fig. 1h, red line).

We concluded that the endogenous contractility of the myoepithelial cell sheet specifically induces large anisotropic deformation fields in the ECM in front of the elongating branch. Indeed, already during the organoid establishment phase, we observed that small clusters consisting of just a few basal cells were already able to induce considerable deformation of the surrounding ECM (Supplementary Fig. S1F, H), whereas non-contractile luminal mammary epithelial cells grew as multicellular spheres driven by proliferative pressure without resolvable ECM deformation (Supplementary Fig. S1G, H). Occasionally, only short-ranged and localized elastic deformations appeared. To test directly whether endogenous contractility of basal cells is required for branching initiation, the Rho kinase (ROCK)-inhibitor Y-27632 was added to the culture medium from day 1 on. Indeed, branch formation was inhibited, leading mainly to the emergence of unstructured and dense cell clusters as observed previously22 (Supplementary Fig. S1I). Moreover, the inhibition of cellular contractility during the branch elongation phase on day 10 prevented further branch elongation as well as the global anisotropic contraction of the ECM as observed before. Whereas the formation of filopodia-like-protrusions was only observed in leading cells in control organoids (Supplementary Fig. S1J), ROCK-inhibitor treated organoids displayed cellular protrusions in stalk cells all along the branch axis (Supplementary Fig. S1K, Supplementary Video 2). This observation correlated with deformations of just a few microns perpendicular to the branch in contrast to the large oriented deformations observed in control organoids that emanated from the leading front in direction of branch elongation.

Taken together, we observed that during branch elongation a highly anisotropic deformation field was generated as a result of endogenous contractility. Importantly, ECM deformations were found to be non-continuous and long-ranging. In addition, branch elongation occurred discontinuously in time, showing a back-and-forth movement by the leading cells (Fig. 1h, gray line). Together, these observations hinted towards dynamic cellular rearrangements within elongating branches.

Collective cell migration facilitates ECM deformations

In order to identify the mechanism driving the discontinuous branch elongation, we studied cellular dynamics within elongating branches by live-cell imaging of organoids expressing LifeAct-GFP. Thereby, we could observe that the tip of the branch was led by a few cells, which formed filopodia-like protrusions and appeared to actively invade into the collagen network (Fig. 2a, Supplementary Video 3). Those leader cells were followed by stalk cells, which were reported previously to support branch elongation via intercalation processes as observed in epithelial multilayers24, during endothelial sprouting25 or tube elongation of the Drosophila trachea26. However, nuclei labeling using sirDNA revealed a dynamic exchange of leader cells within elongating organoid branches. More specifically, leading cells were regularly overtaken by cells behind them which thereafter established the new tip of the branch, a phenomenon previously described during endothelial sprouting27,28 (Fig. 2b, Supplementary Video 4). More than 70% of the observed tip exchanges were initiated by a deceleration of the tip cells, while follower cells were continuously migrating outward (Supplementary Fig. S2A, B). The former leading cells became part of the stalk cells and either stayed behind the new leading cell or, in rare cases, integrated into the cellular motion within the branch (n = 3). In a time course of 24 h, in 51% of analyzed branches (n = 47) at least one exchange of a leader cell was observed (Fig. 2c). Moreover, we observed those tip cell exchanges to happen more likely in shorter and narrower branches (Supplementary Fig. S1C, D).

Fig. 2: Collective cell migration facilitates ECM deformations.
figure2

a LifeAct-GFP staining reveals dynamic remodeling of the actin network in the leading cells. Invadopodia of leader cells show dynamic interaction with the ECM throughout the invasion process. b During branch elongation leading cells (red cell, 2) exchange places with cell behind them (cyan cell, 1). c Within 24 h tip cell exchange was observed in about half of the branches analyzed (n = 47 organoids). d Internal collective cell migration (n = 16 organoids). Top panel: Nuclei channel representing the organoid morphology (left). Total velocity measurement of the cells inside the organoid reveals highly dynamic cells (right). Low panel: Velocity in parallel ({v}_{{||}}) (left) and orthogonal ({v}_{perp }) direction (right) shows clusters of cells collectively moving in the same direction. e Cell velocity distribution within one branch of an organoid at day 12 over time. Only the velocity parallel to the branch is plotted. Signs are defined as depicted in the according nuclei channel. f Schematic overview of observed collective cell migration phases. g Bead and cell motion are both discontinuous in time and show periods of correlated phases, during which their direction is pointing in the same direction (n = 11 organoids). During highly correlated phases, outward pointing cell migration correlates with relaxations of the beads in front of the branch away from it. During inward pointing cell migration beads get pulled towards the organoid. Confocal shows a representative organoid at day 8. Scale bars, 100 µm (a, d, e, g), 10 µm (b). Organoids were derived from three biologically independent donors (Supplementary Table S1). Source data are provided as a Source Data file.

We next analyzed cellular dynamics in the whole organoid by nuclear labeling. Cells throughout the whole organoid were highly motile and migrated in a collective manner, leading to cells migrating in cohorts (Supplementary Video 1). We could observe that those collectively migrating cohorts changed their size over time from only a few cells migrating in the same direction to persistent movements within whole branches. Specifically, cells exhibited a movement predominantly parallel to the branch axis with only smaller movements in the orthogonal direction (Fig. 2d). Particularly within single branches, stalk cells migrated with speeds of up to 1 µm/min, but frequently changed their direction or paused their movement (Fig. 2e). Cells within single branches exhibited different phases of cell migration. Phases of cells moving collectively in the same direction within whole branches were followed by phases in which cells moved individually, resulting in a temporary loss of coordination. Ultimately, the uncoordinated movement changed to a collective migration phase again (Fig. 2f).

Importantly, the direction of the net movement of the cells was synchronous with the deformation field within the ECM in front of the branch tips (Fig. 2g). Thus, we could observe phases in which cell movement and ECM deformation were highly correlated. When collective cell motion was pointing outward, mainly relaxations in the ECM were observed. In contrast, when stalk cells within the whole branch moved away from the tip towards the organoid center, large deformations within the ECM towards the tip appeared. No directed collective cell movements and ECM deformations were observed perpendicular to the long axis of the branches (Supplementary Figs. S1E and S2E). Taken together, these observations strongly suggested that the non-continuous ECM deformations in front of elongating branches result from the collective nature of cell movements within the branch and did not emanate from the tip cells only. Indeed, such long-range deformation fields could not be recapitulated in experiments with single cells (Supplementary Fig. S3A).

ECM deformations are enabled through tension equilibrium

To further investigate the mechanism of the underlying force buildup within the branch, we performed immunofluorescence staining of alpha smooth muscle actin. Indeed, we observed high expression of alpha smooth muscle actin in basal cells at the outer cell layer of branches adjacent to the ECM (Supplementary Fig. S3B). Moreover, phalloidin staining revealed thick actin cables connecting neighboring cells (Supplementary Fig. S1C). In addition, we detected strong cell coupling of the cells within the branches via E-cadherin, supporting that tension buildup observed during live cell imaging resulted from a cell-collective effort (Supplementary Fig. S3C). To investigate this further, the actin network was disrupted by addition of Cytochalasin D. This led to a loss of tension, resulting in an instantaneous relaxation of the organoid branches. Due to the elastic counterforce of the surrounding ECM, relaxation was reflected in concomitant branch expansion (Fig. 3a, b, Supplementary Video 5). Yet, the ECM structure in front of the invading branches was unaffected by the Cytochalasin D treatment (Fig. 3c). In order to further investigate the origin of the tensile forces inside the ECM, UV laser ablation of the collagen matrix in front of invading branches was conducted (Fig. 3d). Such ablation was followed by an instantaneous relaxation of the whole branch towards the organoid body (Fig. 3e, Supplementary Video 6). Accordingly, the ECM relaxed in the opposite direction of the initial deformation field, suggesting that the tension in the ECM originated from tension buildup within the organoid branches.

Fig. 3: ECM deformations are enabled through a tension equilibrium.
figure3

a Upon treatment with Cytochalasin D organoid branches relax. Beads retract in the opposite direction of the initial deformation field (white arrow, n = 31 organoids). b Representative behavior of cumulative displacement in front of the branch during branch elongation (green) and relaxation upon treatment with Cytochalasin D (1 µg/ml, purple). c Fiber alignment in front of the branch is conserved after treatment with Cytochalasin D. d UV-cuts of the ECM in front of growing branches and following contraction of the branches towards the organoid body (n = 121 organoids). e Tracking the tip of a branch after a cut reveals a fast contraction towards the organoid body. Error bars, mean ± s.d. f After disruption of the actin network via Cytochalasin D the restoring forces of the ECM dominate, leading to branch elongation (n = 31 organoids). Contrary, after laser ablation, the dominating tension of the branch leads to branch shrinkage (n = 21 organoids). Box plots indicate median (red line), 25th, 75th percentile (blue box) and 5th and 95th percentile (whiskers) as well as outliers (single points). g Drug screening reveals loss of potential to grow TDLU-like structures upon Y-27632 (n = 31 organoids) and HECD1 (n = 10 organoids) treatment. h Tensile forces of the branches and the restoring forces of the aligned ECM lead to a tension equilibrium stabilizing the branches. Scale bars, 50 µm (a, inset a), 20 µm (c), 15 µm (d). Organoids were derived from three biologically independent donors (Supplementary Table S1). Source data are provided as a Source Data file.

To prove the necessity of collective tension buildup, we lowered cell-cell adhesion by addition of a function-blocking anti-E-cadherin antibody (HECD1 clone) throughout the entire period of organoid culture. As a result, thin stick-like branches evolved, that completely lacked alveologenesis. (Fig. 3g, Supplementary Fig. S3E–G). Further, the initial collective cell migration observed in control organoids was completely abolished, resulting in only individual short-ranging cellular movements (Supplementary Fig. S3H, Supplementary Video 7). Complementary, inhibiting endogenous contractility of the basal cells by the addition of the ROCK-inhibitor Y-27632 throughout the entire period of organoid culture led to the formation of disorganized star-like multicellular structures, which did not resemble the branched architecture of control organoids (Fig. 3g, Supplementary Fig. S1I). Taken together, these results suggested that tension is generated by a collective of contractile cells within the branch and is equilibrated by the surrounding ECM (Fig. 3h).

Collagen is plastically remodeled by the invading epithelium

During the relaxation experiments we noticed that only a small fraction of the total deformation which accumulated during the growth process was released. These findings indicated that the observed deformations in the collagen network were predominantly plastic in nature (Fig. 3b).

As a consequence of these stable ECM deformations, we observed highly aligned collagen fibers and bundles in front of the invasive branches. Alongside the branches, fiber alignment was significantly reduced compared to the invasive tip, mimicking the initial anisotropic deformation field (Supplementary Fig. S4A, C). By contrast, the collagen network far away from organoids was fully isotropic with randomly oriented collagen fibers, indicating that collagen fiber alignment in front of elongating branches was generated by the expanding branches of the organoid (Supplementary Fig. S4B, C).

The plastic nature of the collagen network was further underscored by the observation that orientational order of the collagen network was kept in this alignment upon Cytochalasin D treatment (Fig. 3c, Supplementary Fig. S4C), thus maintaining deformation even after loss of tension. This plastic mechanical response of collagen can be observed clearly in cyclic shear rheology (Fig. 4a, Supplementary Fig. S5). Here, the stress–strain response depends on the number of initially applied shear cycles. Specifically, already after the first cycle (I) the slope of the stress–strain curve, which corresponds to the tangential modulus, diminishes significantly up to the previously maximal applied strain, where a high stiffening is observed. This strain memory effect in polymeric materials is also known as Mullins softening and is based on structural changes inside collagen networks29,30. Each cycle leads to plastic fiber elongation31 and the formation of weak crosslinks between approaching collagen fibers32, which gradually changes the structural properties of the fibrillar matrix, similar to what has been demonstrated for crosslinked actin networks33.

Fig. 4: Collagen is plastically remodeled by the invading epithelium resulting in a stable collagen cage.
figure4

a Cyclic shear rheology of collagen shows plastic remodeling. b Collagen accumulation around growing organoids visualized by fluorescent collagen. A high-resolution reconstruction can be seen in Supplementary Video 9. c Distribution of the collagen cage width (n = 153 organoids). d Collagen intensity in dependency of position (Side n = 25 organoids, Front n = 25 organoids, Far field n = 25 organoids). Box plots indicate median (red line), 25th, 75th percentile (blue box) and 5th and 95th percentile (whiskers) as well as outliers (single points). e Organoids treated with Triton X. Upper panel: The collagen cage retains its structure even after the collapse of the branch. Lower panel: The cage is visible in the bright field image. f During ECM invasion leader cells squeeze through pores at the invasive front. g Immunostaining of laminin at day 11 of organoid growth showed localized expression at the cell-ECM interface. h Schematic representation of the collagen cage and the invasion of the tip. Scale bars, 30 µm (b xy), 10 µm (b xz and yz, e), 20 µm (f), 50 µm (g). Organoids were derived from three biologically independent donors (Supplementary Table S1). P values are from a two-tailed Mann–Whitney test and provided in Supplementary Table S5. Source data are provided as a Source Data file.

Live-cell imaging revealed that fiber alignment was induced by the collective and dynamic mechanical tension produced by the collective motion of cells within the elongating branches, mimicking a cyclic strain application (Supplementary Video 8). In addition, small range deformations were observable due to the invasion dynamics of the leading cells. These cells did not continuously attach to the same collagen fibers, but frequently changed their attachment sites within the collagen network, leading to only a localized but spatiotemporally inhomogeneous deformation field (Supplementary Fig. S4D). These short-ranging deformations occurred asynchronously to the large deformation field which resulted from the collective cell motion and thereby caused small deviations from the correlation between collective motion and ECM deformation. Thus, the observed fiber alignment was a direct consequence of the observed contractile deformation field and captured its history due to the plastic properties of the collagen network.

By the use of fluorescently labeled collagen, we observed that this tension-induced collagen fiber alignment ultimately resulted in an enrichment of collagen along the branch axis leading to the formation of a continuous collagen cage (Fig. 4b). High-resolution microscopy revealed that this collagen cage had a porous structure with holes of an approximate size of 1 up to 3 µm and a thickness of up to 12 µm (Fig. 4c, Supplementary Video 9). Towards the tip of the branch the cage thinned out with larger holes at the invasion site, which are still significantly smaller than in the far field (Fig. 4d). Washing out the epithelial cell layer by addition of Triton X left an empty collagen cage behind, lining the borders of the vanished structure (Fig. 4e). Thus, the plastic deformation of the collagen generated a collagen cage which encased the organoid and, once formed, remained mechanically stable even in the absence of cells. Together, these observations suggested that mechanical stability and anisotropy of the collagen cage guide the further elongation process of branch. While at the side the dense cage prevented further outgrowth34, the front of the cage at the tip of the branch was porous and weak enough for individual cells to squeeze through (Fig. 4f, Supplementary Fig. S4E). Indeed, these results may also explain why formation of a new branch was observed exclusively through bifurcation, rather than side branching (Supplementary Fig. S6). In addition to accumulated collagen around the epithelial branches, we detected expression of laminin, a major component of the basement membrane (Fig. 4g).

To further determine this anisotropic accumulation of collagen around the branches we performed immunofluorescence staining of MMP9, a metalloproteinase previously described to have a dominant role in matrix remodeling in the mammary gland35. MMP9 staining revealed a localized expression at the tip of the invading branches (Supplementary Fig. S4F). In order to test the interplay of plastic deformation and local degradation of the matrix, we inhibited the activity of Metalloproteinases (MMPs) by the addition of Marimastat. At the beginning of the organoid establishment phase, addition of 10 µM Marimastat led to the formation of very short and thin branches, thus may prevent efficient branching morphogenesis altogether (Supplementary Fig. S4G). The inhibition of MMP activity during the branch elongation phase induced an arrest of invasion by the branch tip cells into the collagen, although formation of filopodia was still observed (Supplementary Fig. S4H, I, blue line). Moreover, generation of tension by the branches was not significantly impeded, as shown by large strains in the collagen and further plastic deformation of the surrounding matrix (Supplementary Video 10). However, MMP-inhibition impaired elongation of the branch. Since proliferation of cells within the branches continued, the reduction in branch elongation observed upon MMP-inhibition resulted in increased cell density and a concomitant slowing of collective cell motility, eventually leading to thickening of the branches (Supplementary Fig. S4H, orange line). Based on these results, we concluded that the formation of the collagen cage resulted from a combination of mechanically induced accumulation of collagen in front of the branch and single-cell invasion of the tip cells with active degradation of the ECM (Fig. 4h, Supplementary Video 11).

Products You May Like

Leave a Reply

Your email address will not be published.