|Home | About | Journals | Submit | Contact Us | Français|
For decades, the work of cell and developmental biologists has demonstrated the striking ability of the mesenchyme and the stroma to instruct epithelial form and function in the mammary gland [1–3], but the role of extracellular matrix (ECM) molecules in mammary pattern specification has not been elucidated. Here, we show that stromal collagen I (Col-I) fibers in the mammary fat pad are axially oriented prior to branching morphogenesis. Upon puberty, the branching epithelium orients along these fibers, thereby adopting a similar axial bias. To establish a causal relationship from Col-I fiber to epithelial orientation, we embedded mammary organoids within axially oriented Col-I fiber gels and observed dramatic epithelial co-orientation. Whereas a constitutively active form of Rac1, a molecule implicated in cell motility, prevented a directional epithelial response to Col-I fiber orientation, inhibition of the RhoA/Rho-associated kinase (ROCK) pathway did not. However, time-lapse studies revealed that, within randomly oriented Col-I matrices, the epithelium axially aligns fibers at branch sites via RhoA/ROCK-mediated contractions. Our data provide an explanation for how the stromal ECM encodes architectural cues for branch orientation as well as how the branching epithelium interprets and reinforces these cues through distinct signaling processes.
The mammary gland is the organ responsible for milk production, and it is comprised of the ectodermally derived epithelium and mesodermally derived stroma (Figure 1A). As females enter puberty, a pulse of hormonal cues activate mammary epithelial outgrowth into the surrounding stroma, a process known as branching morphogenesis. To gain insight into how the epithelium is patterned during branching morphogenesis, we developed a method for measuring epithelial orientation.
Mammary glands were carefully isolated and mounted onto glass slides for the preservation of tissue architecture. Before imaging, we used morphological landmarks to register the samples such that the long axis of the mammary gland was specified as 0° (Figure 1A, dashed arrows). For each mammary gland, the entire epithelium was traced with ImageJ and the orientation measured in 50 μm segments along each trace with the use of a custom MATLAB script, yielding ~100–300 epithelial orientation measurements per gland. Mammary epithelial orientation was measured in this way at early (Figure 1A, Week 3) and late (Figure 1A, Week 8) postnatal stages of branching morphogenesis. We observed no stereotyped orientation pattern in early-stage mammary glands, with epithelia being randomly distributed along either the short (Figure 1A, ±90°) or long axis (Figures 1A and 1B, 0°). In contrast, the late-stage epithelial pattern was stereotyped, displaying a significant orientation along the long axis of the mammary gland (Figures 1A and 1B, 0°).
Given that epithelial orientation became axially biased after substantial outgrowth into the previously unexplored mammary stroma, we hypothesized that a patterning cue for orientation may already be present. Because Col-I is a common stromal component and because a previous study focused on macrophages showed the presence of Col-I fibers proximal to the mammary epithelium , we chose Col-I as a candidate patterning cue. Additionally, malignant cell migration occurs on—or around—Col-I fibers [5, 6], although little is known about the role of endogenously oriented Col-I bundles in interacting with the developing mam-mary epithelium.
To analyze Col-I orientation, we isolated and mounted glands as previously described with some alterations in the staining, imaging, and analysis methods. Col-I was labeled with Alexa Fluor 594 conjugated to CNA35, a probe which reported high affinity to triple helical Col-I . Confocal microscopy was performed in stromal regions distal to the epithelium of early stage mammary glands (Figure 1C, dashed box). Col-I orientation was measured via the ImageJ plugin OrientationJ . We observed heterogeneity in Col-I intensity and organization (Figure 1C, High–Low). High-intensity regions were typically proximal to the outer fascial layer, whereas medium-intensity regions formed tracks that were observed at various depths (Figure 1C, XZ). Both high- and medium-intensity regions contained fibers significantly oriented toward the long axis (Figure 1D). In contrast, low Col-I staining intensity was located in the stroma at adipocyte-rich regions containing no fibers and displayed no discernible orientation bias (Figures 1C and 1D, Low). From these data, we conclude that tracks of oriented Col-I fibers exist within the mammary stroma prior to branching morphogenesis in regions accessible to the growing epithelium.
We also examined Col-I organization at sites proximal to the epithelia in early-stage mammary glands using the previously described methodology, except that Col-I orientation was plotted relative to branch orientation (Figure 1E, dashed arrows). Near end buds, fibers persisted approximately 300 μm ahead of the epithelium (Figure 1E), demonstrating that Col-I fibers are formed prior to epithelial association in vivo. The proximal Col-I fibers were significantly co-orientated with end bud direction (Figure 1F), inferring that the fibers act as tracks for epithelial outgrowth. We also observed Col-I fibers proximal to epithelial ducts in late-stage glands (Figure S1), and Col-I labeling was around larger ducts. These data suggested that Col-I fibers may encode orientation cues within the extracellular space of the mammary gland prior to the branching of the epithelium.
To test whether there was a causal relationship between Col-I fibers and epithelial orientation, we developed a novel method for patterning an axial orientation bias in Col-I matrices (Figure 2A) before the inducing of branching in three-dimensional cultures. Malleable wells were fabricated from polydimethylsiloxane (PDMS) and uniaxially stretched (Figure 2A). A liquid mixture of Col-I and mammary epithelial cells was prepared as previously described  and added into the prestretched PDMS wells. Upon Col-I matrix polymerization, we gently released the PDMS wells to generate a modest unidirectional compression (~20% strain). Visualizing the Col-I matrix by confocal reflection microscopy, we found that compression significantly oriented Col-I fibers in comparison to the randomly oriented control (Figures 2B–2D). Supplemental experiments to assess porosity (Figure S2D) and residual stress (Figure S2K) found only minimal changes upon compression, indicating that compression in our assay primarily affected Col-I fiber orientation.
To determine whether Col-I fiber orientation was sufficient to direct branch orientation, we embedded organoids in either compressed (patterned) or uncompressed (control) Col-I matrices (Figures 2E–2F). Given that the axis of Col-I fiber orientation is determined by the compression direction and uniformity throughout the sample (Figure 2C, green arrow), branch orientation measurements were made with the axis of Col-I orientation as 0° (Figure 2F, green arrow). Whereas organoids branching in control Col-I matrices showed no orientation bias, we observed axially oriented branching in patterned Col-I matrices (Figure 2G), providing strong evidence that Col-I fibers impart directional cues to branching mammary epithelial cells. We repeated the assay in lower-density Col-I matrices (Figures S2E–S2J) where the ECM was more porous and less stiff. The preference for epithelial branches to align with oriented Col-I fibers remained (Figure S2J), suggesting that the process is not strictly dependent on the total matrix density or stiffness.
Using our patterned Col-I branching assay, we sought to identify components critical to co-orientation. Initial time-lapse experiments showed frequent and repeated cellular protrusions at branch sites proximal to oriented Col-I tracks (Movie S1), inferring an involvement of the cell's motility machinery with orientation sensing. Therefore, we chose to modulate this behavior by activating Rac1, a GTPase shown to regulate lamellipodial protrusions through local remodeling of both the actin cytoskeleton and focal adhesions . Expression of a constitutively-active form of Rac1 (Rac1-CA) lowered branch orientation of mammary epithelial aggregates significantly in comparison to the vector alone (Figures 2H–2J). Branch morphology was also altered in Rac1-CA aggregates, which increased in the number of short multicellular branches that terminated in long cellular protrusions (Figure 2I, arrow). Conversely, we tested whether the expression of fascin-1, a factor downstream of Rac1 that stabilizes some actin-based protrusions by tightly bundling actin , could increase Col-I fiber co-orientation. When aggregates expressing fascin-1 were embedded in patterned Col-I matrices (Figures 2K–2M), we observed a modest, though not significant, enhancement in sensing Col-I orientation. From these data, we infer that Rac1 activity can modulate epithelial sensitivity to oriented Col-I fibers.
During initial time-lapse experiments, we also observed significant contractions of the Col-I matrix by branching aggregates (Movie S2), suggesting that active cell contractions could play a role in local Col-I orientation. To further study this interaction, we performed time-lapse experiments on mammary organoids embedded in randomly oriented Col-I matrices and used confocal microscopy capable of simultaneously imaging epithelial outgrowth and Col-I matrix organization. We established a time window for comparing Col-I orientation at an early stage prior to branching, when the Col-I matrix is still randomly oriented (Figure 3A, 28 hr), to a late stage in which substantial branching had occurred (Figure 3B, 39 hr). Visualizing Col-I fibers by confocal reflection microscopy, we observed initial local contractions at branch sites at the early stage (Figure 3A, inset) that propagated over time, leading to an increase in Col-I fiber co-orientation by the late stage (Figures 3B–3C). These results indicate that, in the absence of global bias in the orientation of Col-I fibers, mammary epithelial cells can locally generate oriented Col-I paths via contractions.
Suspecting that cell-generated Col-I orientation involved RhoA/ROCK-mediated actomyosin contractions , we inhibited ROCK activity with Y-27632, a small molecule inhibitor that we confirmed interrupted contractions in our assay (Figure S3). Treatment of mammary epithelial aggregates with Y-27632 significantly disrupted the local orientation of Col-I fibers at epithelial branches within randomly oriented Col-I matrices in comparison to the vehicle alone (Figures 3D–3F). These data demonstrate that branching mammary epithelial cells are capable of enhancing Col-I co-orientation via ROCK-mediated contractions.
To study the role of RhoA/ROCK-mediated contractions in generating oriented Col-I tracks, we conducted the experiments above (Figures 3A–3F) in randomly oriented Col-I matrices. To determine whether such contractions were also involved in sensing Col-I fiber orientation, we instead embedded mammary cells in preoriented Col-I matrices. In this context, branch orientation was not significantly different between Y-27632 (20 μM) and vehicle treatment (Figures 3G–3I). In addition, we inhibited RhoA activity through the expression of a dominant negative form of RhoA (RhoA-DN). We found no significant change in branch orientation between aggregates expressing RhoA-DN and vector control (Figures 3J–3L), consistent with results from Y-27632 experiments. From these data, as well as from studies where we inhibited molecules downstream in the RhoA/ROCK pathway (Figures S3F–S3H), we concluded that, during branching morphogenesis, the role of actomyosin contractions is restricted to the enhancement of Col-I fiber orientation proximal to branch sites.
The question of whether patterning cues are present in the stroma prior to epithelial branching morphogenesis is unresolved, despite cumulative evidence demonstrating the significance of interactions between epithelia and stroma during development [1–3, 13–16]. Here, we show that Col-I fibers are oriented in the mammary fat pad long before the initiation of branching morphogenesis. This observation suggested the intriguing possibility that epithelial architecture may be prepatterned in the stromal microenvironment in vivo.
Published work has shown the importance of cellular interaction with Col-I during branching morphogenesis. For example, properly-regulated Col-I interactions are required for branching  as well as directing tissue —and cell— polarity . The Col-I matrix also presents physical barriers to migration  and changes to Col-I structure significantly alter the mechanical properties of the ECM [21, 22]. However, despite the reports in the literature that the majority of Col-I is generated by stromal cells [23, 24] presumably to signal to—or to shape—epithelia, to our knowledge, no conclusive study that details the generation and patterning of Col-I in the mammary gland has been published. A previous study did mention the possible involvement of macrophages in Col-I fibrillogenesis proximal to the branching epithelium . Although these studies reported that macrophage deficiency leads to aberrant terminal end bud morphology , they did not explore whether macrophage-mediated fibrillogenesis directed the epithelial orientation. We hypothesized that the prepatterned Col fibers proximal to—and ahead of— the branching epithelium were likely important contributors to directional cues for the epithelium.
Using a three-dimensional culture model in which a Col-I matrix is axially patterned by the application of mechanical strain, we demonstrate that stromal Col-I fiber orientation directs epithelial branching. Previous studies characterizing Col-I matrix structure in culture [25, 26] and in vivo  showed linear relationships between mechanical perturbations and fiber alignment, but our method shows that approximately 20% strain was sufficient to align fibers and observe significant differences in branch orientation. Our assay, in combination with studies of branching in randomly oriented Col-I matrices, was sufficient to discern “path-finding” from “path-generating” mechanisms during branching morphogenesis.
We identified Rac1 as a modulator of Col-I orientation sensing that acts during branching morphogenesis. Although other factors required for branching morphogenesis, such as integrin β1 [28, 29], could also be involved in sensing Col-I orientation, we were unable to determine an orientation phenotype when no measurable branching occurred, as in the case of Itgb1 knockdown (Figures S2A–S2C). Rac1 has been implicated in directional persistence during single- and collective-cell migration [30–32], as have several of its downstream effectors [33–35]. Whereas we found fascin-1 overexpression did not significantly increase branch orientation in our assay, knockdown of fascin-1 in human colon carcinoma cells resulted in a decrease in number of filopodia, focal adhesion turnover, and directional persistence . Other Rac1 effectors involved in directional migration regulate microtubule organization. Work on kidney epithelial cell migration found that Rac1 activation results in the recruitment of the effector IQGAP1 to the leading edge, which forms a complex with APC and CLIP-170, linking actin cytoskeleton to microtubule dynamics . An automated RNAi screen of Rac1 effectors involved in microtubule dynamics found that MAP/ microtubule affinity-regulating kinase 2 (MARK2) was necessary for microtubule orientation in lamellipodial and directional migration . Intriguingly, our analysis of a previously published gene expression profile of the branching mammary epithelium  revealed a significant increase of genes in the Rac1 axis at end buds versus ducts (fold increases: S100a6, 3.22; Stmn1, 2.68; Fscn1, 2.1; Iqgap1, 1.7; Racgap1, 2.5; Itga5, 1.7; and Sdc1, 2.7). The biased expression of these genes at the end buds supports the involvement of Rac1 signaling in branching morphogenesis at the migratory front where Col-I sensing presumably occurs.
Despite previous publications, the role of RhoA/ROCK signaling in branching morphogenesis remained unclear. Studies using culture models demonstrated the involvement of ROCK in sensing Col-I stiffness as well as tubular pattern self-assembly [12, 37], whereas studies in vivo where RhoA was inactivated within the mammary epithelium failed to inhibit branching . By separating the processes of fiber patterning from orientation sensing, we have begun to reconcile these seemingly dissonant observations. Our results demonstrate that RhoA/ROCK-mediated contractions are not necessary to sense Col-I fiber orientation. However, we observed that these contractions enhance Col-I fiber orientation proximal to the mammary epithelium. Thus, the ability of the mammary epithelium to generate actomyosin contractions could be important for reinforcing directional decisions during branching morphogenesis. Future work is necessary to confirm this possibility in vivo and to determine where along the epithelium contractions occur. It has been shown that end buds express P-190B , a factor which inhibits RhoA activity; therefore, contractions most likely occur proximal to—or within—ductal epithelium, where RhoA activity is required for repolarization.
Our study provides strong evidence that epithelial orientation is prepatterned by stromal Col-I fibers prior to the initiation of branching morphogenesis. We show that the machinery employed by the epithelium to sense Col-I fiber orientation is molecularly distinct from those that generate Col-I alignment. Our findings emphasize that stromal patterning as well as orientation sensing of the extracellular milieu are critical processes in determining the architecture of epithelial tissues. Future work will aim to characterize how such collagen fiber patterns are generated and which cell types are involved in stromal patterning.
This research was supported by grants from the US Department of Energy and the Office of Biological and Environmental Research (DE-AC02-05CH1123) to M.J.B.; by the National Cancer Institute (NCI) (R37CA064786, U54CA126552, U01CA143233, U54CA112970, and U54CA143836; Bay Area Physical Sciences–Oncology Center, University of California) to M.J.B. and D.G.B.; and by a US Department of Defense Innovator Award (W81XWH0810736) to M.J.B. The work of D.A.F. and G.V. was supported by the NCI and National Science Foundation. We are grateful to Saori Furuta as well as both the Bissell and Fletcher laboratories for discussion and critical reading of the manuscript. We thank Sanjay Kumar and Joanna MacKay for the image correlation MATLAB code.
Supplemental Information contains Supplemental Experimental Procedures, three figures, and two movies and can be found with this article online at http://dx.doi.org/10.1016/j.cub.2013.03.032.