Collagen polarization promotes epithelial elongation by stimulating locoregional cell proliferation

  1. Hiroko Katsuno-Kambe
  2. Jessica L Teo
  3. Robert J Ju
  4. James Hudson
  5. Samantha J Stehbens
  6. Alpha S Yap  Is a corresponding author
  1. Division of Cell and Developmental Biology, Institute for Molecular Bioscience, The University of Queensland, Australia
  2. QIMR Berghofer Medical Research Institute, Australia

Abstract

Epithelial networks are commonly generated by processes where multicellular aggregates elongate and branch. Here, we focus on understanding cellular mechanisms for elongation using an organotypic culture system as a model of mammary epithelial anlage. Isotropic cell aggregates broke symmetry and slowly elongated when transplanted into collagen 1 gels. The elongating regions of aggregates displayed enhanced cell proliferation that was necessary for elongation to occur. Strikingly, this locoregional increase in cell proliferation occurred where collagen 1 fibrils reorganized into bundles that were polarized with the elongating aggregates. Applying external stretch as a cell-independent way to reorganize the extracellular matrix, we found that collagen polarization stimulated regional cell proliferation to precipitate symmetry breaking and elongation. This required β1-integrin and ERK signaling. We propose that collagen polarization supports epithelial anlagen elongation by stimulating locoregional cell proliferation. This could provide a long-lasting structural memory of the initial axis that is generated when anlage break symmetry.

Introduction

Branched tubules represent one of the archetypal modes of epithelial organization (Iruela-Arispe and Beitel, 2013; Varner and Nelson, 2014). In organs such as the mammary gland and lungs, networks of hollow epithelial tubes mediate the physiological exchange of gases, nutrients, and solutes between the body and its external environment (Chung and Andrew, 2008). These definitive networks are established by a complex morphogenetic process, where tubules grow outward from their precursors until they are instructed to branch, after which outgrowth continues until the network is completed (Affolter et al., 2009; Andrew and Ewald, 2010). The branching and elongation of multicellular aggregates can thus be considered as fundamental processes in the generation of tubular networks.

Interestingly, different organs also use distinct strategies for tubulogenesis. For example, in the trachea, lumens appear early and accompany the growth of tubules (Schottenfeld et al., 2010), whereas in the salivary and mammary glands tubules first begin as non-polarized cellular aggregates (or anlage), which then elongate as solid, multicellular cords before eventually forming lumens (Bastidas-Ponce et al., 2017; Nerger and Nelson, 2019; Tucker, 2007). In the present study, we use the mammary epithelium as a model to analyze what guides the elongation of multicellular anlage.

Tubulogenesis is a highly regulated phenomenon. The decision to branch is recognized to be a critical checkpoint that is controlled by developmental signals and cell-cell and cell-extracellular matrix (cell-ECM) interactions (Goodwin and Nelson, 2020). The elongation of tubule precursors is also thought to be a regulated process controlled by receptor tyrosine kinases and other signaling pathways (Costantini and Kopan, 2010; Gjorevski and Nelson, 2010; Sternlicht et al., 2006). However, elongation occurs over hours to days, time scales that are much longer than the underlying cellular processes and the signaling pathways that guide them. This raises the question of whether there may be mechanisms that can help guide elongation over longer time scales, effectively serving as a bridge between the rapidity of cell signaling and the slow progression of macroscopic anlage elongation. In this study, we show how collagen 1 within the ECM can provide such a bridge by stimulating cell proliferation.

The ECM comprises complex mixtures of proteins, glycosaminoglycans, and glycoconjugates that fill the extracellular spaces of tissues and organs (Frantz et al., 2010). One of the main components of ECM is collagen, particularly type 1 collagen (also called collagen 1), which is often the dominant form during epithelial tubulogenesis (Graham et al., 1988; Keely et al., 1995; Llacua et al., 2018; Nakanishi et al., 1986; Simon-Assmann et al., 1995). Type 1 collagen exerts a diverse range of effects that can potentially influence epithelial elongation. Fibrillar collagen helps scaffold other molecules, such other ECM proteins and growth factors (Kanematsu et al., 2004; Wipff and Hinz, 2008). Adhesion between cells and ECM allows the chemical and mechanical properties of the ECM to regulate cell signaling and gene regulation (Shi et al., 2011). These cell-ECM adhesions also allow cell-based forces to reorganize the ECM. In particular, cells can rearrange collagen fibrils to influence cell migration (Buchmann et al., 2021; Gjorevski et al., 2015; Guo et al., 2012; Shi et al., 2014).

Importantly, fibrillar components of the ECM, such as collagen 1, are relatively long-lived (Price and Spiro, 1977; Verzijl et al., 2000), making them attractive candidates to bridge time scales during the elongation process. Indeed, rearrangement of collagen has been implicated in patterning epithelial branching (Brownfield et al., 2013; Guo et al., 2012; Harunaga et al., 2011; Ingman et al., 2006; Patel et al., 2006). However, it is difficult to elucidate the specific contribution for collagen in the complex environment of an organ, where there are many additional contributions from other cell types and various chemotactic signals. Therefore, in this study we used three-dimensional (3D) organotypic cultures to test how the ECM regulates epithelial elongation. We report that a collagen 1 matrix induces the elongation of mammary epithelial anlage. This is accompanied by the polarization of the matrix itself, an event that sustains anlagen elongation by stimulating cell proliferation.

Results

An experimental system to capture symmetry breaking and elongation of epithelial aggregates

To model the elongation of epithelial anlage, we induced multicellular aggregates to break symmetry by manipulating the ECM environment in which the cells were grown. When MCF10A cells are grown on Matrigel substrates and in media supplemented with soluble Matrigel (Debnath et al., 2003), they proliferate from single isolated cells to form a multicellular aggregate, which then polarizes and clears the central mass of apoptotic cells, generating a central lumen in a process known as cavitation (Figure 1—figure supplement 1A). In contrast, when isolated cells were embedded in a gel of type 1 collagen, which is the major ECM component in the stromal environment during mammary tubulogenesis (Schedin and Keely, 2011), they proliferated to form elongated, solid cords of non-polarized cells (Figure 1—figure supplement 1B; Krause et al., 2008). This indicated that some properties of the collagen 1 environment might provide an instructive cue for elongation.

In order to capture the process by which isotropic aggregates broke symmetry and then elongated, cells were first seeded in Matrigel until proliferation arrested at 10 days (Figure 1A); aggregates were then isolated and embedded into collagen 1 (acid-solubilized rat type I collagen, 37℃ for 30 min). Lumens were apparent in some Matrigel-embedded aggregates, but most remained as solid spheres, resembling anlage. Live-cell imaging revealed that the transplanted aggregates displayed small jiggling motions for several hours, then spontaneously broke symmetry and elongated to form cord-like structures similar to those formed when cultured in collagen from the outset (Figure 1B, Video 1). These elongating aggregates were generally solid, with no evident lumens (Figure 2A). Commonly, elongation began with a group of cells that protruded away from the more spherical original aggregate (Figure 1C and E). In contrast, when aggregates were transplanted from Matrigel back into Matrigel, they grew slightly but did not elongate (Figure 1—figure supplement 1C). Therefore, transplantation into a collagen 1 matrix effectively caused isotropic MCF10A aggregates to break symmetry and elongate.

Figure 1 with 3 supplements see all
Type 1 collagen induces elongation of MCF10A anlage via cell proliferation.

(A) Cartoon: transplantation of MCF10A cell aggregates from Matrigel into collagen gel. Single isolated cells are cultured on Matrigel and overlayed with Matrigel-containing medium. Matrigel is washed out after 10 days and the aggregates re-embedded into type 1 collagen gel. (B) Time-lapse images of MCF10A aggregates after transfer into a collagen gel. Red arrowheads: elongation from aggregates. (C) Fluorescence image of an elongated aggregate cultured for 8 days after transfer into collagen gel and stained with anti-Ki67 antibody (green), phalloidin (magenta), and DAPI (blue). Blue parenthesis: non-elongating area; red parenthesis: elongating area. (D) Percentage of Ki67-positive cells in elongating and non-elongating areas of the elongated aggregates (n = 15 aggregates). (E) Time-lapse images of elongating aggregates expressing NLS-mCherry. Yellow arrowheads: dividing nucleus. (F) Frequency of cell division in elongating and non-elongating regions of aggregates (n = 12 aggregates). (G) Percentage of Ki67-positive cells in aggregates cultured for 0–3 days (n = 102 aggregates). (H) Symmetry ratio of aggregates cultured for 0–3 days (n = 102 aggregates). (I) Percentage of Ki67-positive cells in elongating area and non-elongating areas of aggregates that broke symmetry (defined as symmetry ratio >1.5) in the first three days of culture (n = 18 aggregates). (J) MCF10A aggregates co-stained with anti-Ki67 antibody (green) and DAPI (blue). Aggregates were cultured for 0 day (i) or 8 days (ii–iv) after treatment with vehicle (ii), mitomycin C (iii), or aphidicolin (iv). (K) Percentage of Ki67-positive cells in aggregates cultured for 1 day or 8 days with or without mitomycin C or aphidicolin (n = 63 aggregates). (L) Symmetry ratio of aggregates cultured for 1 day or 8 days with or without mitomycin C or aphidicolin (n = 237 aggregates). (M) Effect of delayed inhibition of proliferation on aggregate elongation. Aggregates were cultured for 3 days before treatment with mitomycin C or aphidicolin. Data are fold change of elongation in control and drug-treated cultures (n = 57 aggregates). All data are means ± SEM, *p<0.05, **p<0.01, ***p<0.001. Data in (D, F, I) were analyzed by unpaired Student’s t-test. Data in (G, H, K–M) were analyzed by one-way ANOVA Tukey’s multiple comparisons test.

Video 1
Time-lapse images of MCF10A aggregates after transferred into collagen gel.

Images were taken every 10 min.

Figure 2 with 2 supplements see all
Mammary cell aggregates polarize the extracellular matrix (ECM) as they elongate.

(A) Fluorescent image of collagen fiber alignment and elongated aggregates expressing the cell membrane marker GFP-HRasC20. Collagen fibers are labeled with mCherry-CNA35 peptide (magenta). (B) Cartoon of symmetry ratio of aggregates (i) and gel coherency (ii). (C) Scatter plot of aggregate symmetry ratio and collagen fiber coherency (n = 75 aggregates). (D) Regional analysis of collagen coherency around aggregates. Cartoon illustrates the approach: for elongating aggregates, coherency was measured in regions of interest (ROIs) placed both at the tips of elongations and proximate to their non-rounded areas. For rounded aggregates, ROIs were placed orthogonally. Regional differences in coherency were measured as the fold difference, measured around rounded aggregates (n = 19 aggregates) and elongated aggregates (n = 57 aggregates). Elongated aggregates were defined as symmetry ratio >1.5. Double-headed arrow: elongating axis. (E) Fold difference of collagen fiber coherency around aggregates measured at early stages of elongation (first three days of culture) subdivided based on symmetry ratio (n = 98 aggregates). (F) Fluorescent images of collagen fibers labeled with mCherry-CNA35 (magenta) and aggregates expressing GFP-HRasC20 cultured for 8 days with mitomycin C or aphidicolin. (G) Coherency of collagen fibers surrounding the aggregates treated with mitomycin C and aphidicolin (n = 40 aggregates). (H) Fluorescent images of collagen fiber alignment with aggregates expressing GFP-HRasC20 treated with blebbistatin or Y-27632 for 8 days. Collagen fibrils were labeled with CNA35-mCherry (magenta). (I) Symmetry ratio of aggregates cultured for 8 days with or without blebbistatin or Y-27632 (n = 88 aggregates). (J) Distribution of collagen fiber orientation surrounding elongated aggregates (n = 22 aggregates) or non-elongated aggregates (n = 13 aggregates). (K) Difference between elongation axis of aggregates and average angle of collagen fibers in non-elongating area or elongating area (n = 57 aggregates). All data are means ± SEM, ns, not significant, *p<0.05, **p<0.01, ***p<0.001. Data in (D, E, G, I) were analyzed by one-way ANOVA Tukey’s multiple comparisons test. Data in (K) were analyzed by unpaired Student’s t-test.

Cell proliferation drives anlagen elongation

Multiple cellular processes have been implicated in epithelial elongation (Andrew and Ewald, 2010; Economou et al., 2013; Keller, 2002). We first examined cell proliferation, given the evident increase in the size of aggregates that elongated. Staining for Ki67, a nonhistone nuclear protein commonly used as a marker of proliferating cells (Soliman and Yussif, 2016), showed clear proliferation in aggregates that had elongated 8 days after transplantation into collagen (Figure 1C and D). These elongating aggregates commonly consisted of elongated extensions as well as a rounded region that marked the original aggregate (Figure 1B). Strikingly, Ki67-positive cells were more frequent in the elongating parts of the aggregates (41.00% ± 4.55% of all cells) compared with the non-elongating parts of the aggregates (18.86% ± 3.05%) (Figure 1C and D). This differential pattern of proliferation was also evident when we analyzed just the cells at the surface of the aggregates, which were in contact with the collagen (Figure 1—figure supplement 2A). Live-cell imaging of cells expressing NLS-mCherry also showed that the proportion of nuclei that divided was higher in the elongating than the non-elongating areas within aggregates (Figure 1E and F). Thus, elongation was associated with locoregional differences in cell proliferation, which was enhanced in those parts of the aggregates that elongated.

More detailed inspection revealed that increased proliferation was first evident in aggregates 2 days after transplantation (Figure 1G) and appeared to precede the onset of symmetry breaking (Figure 1H). To quantitate aggregate elongation, we calculated a symmetry ratio, that is, the ratio of the maximum length (L1) and width (L2) of the aggregates. Completely round aggregates will have a symmetry ratio of 1, whereas for elongated aggregates L1/L2 will be >1 (Figure 2B). The symmetry ratio first increased at day 3 (Figure 1H, Figure 1—figure supplement 2C), approximately 24 hr after an increase in proliferation was detected (Figure 1G). To test if any locoregional differences in proliferation were to be found at these early stages of symmetry breaking, we analyzed aggregates that had broken symmetry (defined experimentally as a symmetry ratio >1.5) after 3 days of culture. We compared the proportion of cells that were Ki67-positive in the elongating regions of aggregates with the proportion to be found in the non-elongating areas of the same aggregates (Figure 1I, Figure 1—figure supplement 2D). Ki67 positivity was twofold greater in the elongating areas (Figure 1I), suggesting that locoregional differences in proliferation were established early in the elongation process.

Importantly, elongation was inhibited when proliferation was blocked using the cell cycle inhibitors mitomycin C and aphidicolin (Figure 1J). In control cultures, the proportion of Ki67-positive cells increased progressively after transplantation into collagen (from 2.28% ± 0.85% at day 1 to 26.63% ± 3.34% at day 8). However, cell proliferation was significantly reduced by mitomycin C (3.88% ± 1.18%, day 8) and aphidicolin (11.54% ± 3.55%, day 8) (Figure 1K). Aggregate elongation was also significantly reduced by both mitotic inhibitors (Figure 1J). Whereas the symmetry ratio increased as control anlage elongated, this was blocked by both mitomycin C and aphidicolin (Figure 1L). Formally, cell proliferation might have been necessary to initiate symmetry breaking and cell elongation or ongoing proliferation might also have served to sustain elongation. To pursue this, we added mitomycin C or aphidicolin after the first signs of symmetry breaking were evident (at 3 days after transplantation). This reduced elongation by approximately threefold (Figure 1M). Therefore, cell proliferation was required for elongation, with the capacity to contribute early in the symmetry-breaking process and also later to sustain elongation.

We then used live-cell imaging of labeled nuclei to evaluate other processes implicated in epithelial elongation. Cells within the elongating regions tended to divide along the axis of elongation (Figure 1—figure supplement 2B), as revealed by comparing the axis of cell division and the principal axis of the region. The angle difference between these axes was smaller in elongating areas compared with non-elongating areas within the same aggregate (Figure 1—figure supplement 2B). This suggested that polarized cell division accompanied enhanced proliferation during aggregate elongation (Gong et al., 2004; Keller, 2006).

Tracking also revealed that cells were motile within elongating aggregates (Figure 1—figure supplement 3A). This was evident in rounded aggregates that had not broken symmetry as well as in aggregates that had elongated (Figure 1—figure supplement 3B). When we examined regional differences within aggregates that had undergone elongation, we found that the speeds of migration were identical in the parts that were elongating compared with the non-elongating regions of the aggregates (Figure 1—figure supplement 3C). However, the straightness of the tracks, which was used as an index of the persistence of migration, was slightly greater in the elongating areas than in the non-elongating areas (Figure 1—figure supplement 3D). Furthermore, cells within regions of elongation appeared to orient better with the axis of the aggregate than did cells found in non-elongating areas. We measured this by comparing the orientation of the tracks with the principal axis of the aggregates (track displacement angle, Figure 1—figure supplement 3E). The track displacement angle was less in the elongating areas than in the non-elongating areas.

Inhibition of proliferation with either mitomycin C or aphidicolin did not affect cell migration speeds in either the rounded or elongated areas (Figure 1—figure supplement 3B). However, blocking proliferation slightly reduced track straightness, implying a reduction in persistence (Figure 1—figure supplement 3F) and increased the track displacement angle (Figure 1—figure supplement 3G). Therefore, the apparently orderly migration of cells was compromised by blocking cell proliferation. Together, these results suggest that proliferation was essential for the elongation process and may also have influenced aspects of directional migration that would be predicted to reinforce the elongation process.

Collagen 1 condenses around epithelial aggregates that break symmetry

To gain further insight into the process of elongation, we sought to identify changes in the ECM during this process. Consistent with the change in matrix environment associated with the transplantation process, re-embedded aggregates lost laminin V expression over time (Figure 2—figure supplement 1A). They expressed fibronectin throughout the experiments without any evident regional differences in the elongating areas of aggregates (Figure 2—figure supplement 1B).

Then we visualized collagen 1 by labeling with the collagen-binding peptide CNA35 (mCherry-CNA35; Krahn et al., 2006), mixed with soluble collagen 1 before the incorporation of MCF10A aggregates (Figure 2A). Collagen 1, in contrast to the basement membrane-like Matrigel, is composed of fibrillar polymers that can guide cell movement and whose organization is influenced by the application of cellular forces (Brownfield et al., 2013; Gjorevski et al., 2015; Piotrowski-Daspit et al., 2017). Aggregate elongation appeared to coincide with change in fibril organization. Whereas fibrils appeared isotropic in acellular gels, collagen condensed around the aggregates that had begun to elongate, forming dense bands that extended away from the cells into the gel. However, it was seldom possible to visualize individual fibrils with the low-magnification, long-working distance lenses that were required to visualize aggregates within the gels.

Therefore, we measured the coherency (or co-orientation) of gels with Orientation J, which calculates the local orientation and isotropy for each pixel in an image based on the structure tensor for that pixel (see Materials and methods for details; Rezakhaniha et al., 2012). We interpret coherency as reflecting collagen bundling and condensation (which we shall call ‘bundling’ for short), such as has been observed elsewhere during elongation (Brownfield et al., 2013; Buchmann et al., 2021; Gjorevski et al., 2015). Then, we compared collagen coherency with aggregate shape, as measured by the symmetry ratio (Figure 2B), after 8 days culture, when elongation was established. Overall, collagen coherency increased with the increase in symmetry ratio (Figure 2C), implying that the degree of collagen bundling increased with aggregate elongation. Furthermore, closer examination around elongating aggregates showed that collagen coherency was greater in regions proximate to the elongating parts of the aggregates compared with the non-elongating parts (Figure 2D). In contrast, aggregates that remained spherical showed no regional differences in collagen organization. Similarly, earlier studies showed that collagen reorganized ahead of the tips of elongating aggregates (Brownfield et al., 2013; Gjorevski et al., 2015). This suggested that collagen became increasingly bundled and condensed where aggregates were elongating.

Increased collagen bundling was also evident at the early stages of elongation. Because aggregates varied in the timing of when they broke symmetry, we examined aggregates by live imaging in the first three days of the assays. We then subdivided these based on their symmetry ratio and compared the coherency of collagen 1 around the elongating areas of aggregates with that around the non-elongating areas (i.e., those that remained rounded, expressed as a fold difference). This showed that collagen 1 coherency increased around the elongating areas as they broke symmetry (Figure 2E). Thus, from the early stages of symmetry breaking collagen 1 bundling increased around the sites where aggregates elongated.

Collagen bundling required cell proliferation, as coherency was reduced by treatment with either mitomycin C or aphidicolin (Figure 2F and G). Collagen bundling was also reduced when we inhibited cellular contractility with the myosin antagonist, blebbistatin, or the Rho kinase (ROCK) inhibitor, Y27632 (Figure 2H, I), consistent with reports that cell contractility can condense collagen (Brownfield et al., 2013; Buchmann et al., 2021; Gjorevski et al., 2015). In contrast, collagen condensation persisted around aggregates treated with the Rac1 GEF inhibitor NSC23766, which did not affect aggregate elongation or the speed with which cells moved within the aggregates (Figure 2—figure supplement 2). This suggested that aggregates exerted forces to reorganize their local collagen environment as they broke symmetry.

Collagen polarity co-aligns with cell aggregates during elongation

We then asked if higher order organization in the collagen 1 gels was altered as aggregates elongated. For this, we used Orientation J to extract the principal axis of orientation in the gel, as a measure of its polarity, around elongating or non-elongating parts of an aggregate. Then we compared these gel orientations with the axis of elongation of the cell aggregate. First, we assembled histograms of gel orientation, setting the principal axis of elongation in the whole aggregate as 0o (Figure 2J). The gels around non-elongating areas of the aggregates showed a broad distribution of orientations relative to the axis of the aggregate (Figure 2J). In contrast, the gel around the elongating regions of aggregates tended to orient with the principal axis of the aggregates. This suggested that the gel preferentially co-aligned with the aggregates around the regions of elongation. This notion was reinforced by comparing the angle differences between the principal axis (polarity) of the gel and of the aggregates. In this analysis, a decrease in the difference in angles between these two axes indicates an increase in their co-alignment. We found that the angle differences between the axes of collagen polarity and aggregate elongation was significantly smaller around the elongating regions than around the non-elongating regions of the aggregates (Figure 2K). Thus, the collagen matrix became polarized and co-aligned with the aggregates around regions of elongation.

Collagen polarization stimulates cell proliferation to induce aggregate elongation

This led us to wonder if polarization of the collagen could itself influence the process of aggregate elongation. To test this, we developed a strategy that applied exogenous stretch to polarize the matrix independently of cell-generated forces. In this protocol, a collagen gel was created within a ring-shaped polydimethylsiloxane (PDMS) frame, then stretched uniaxially by expanding a stretcher inserted into the hole of the ring for 4 hr (Figure 3—figure supplement 1A and B). Second harmonic imaging confirmed that collagen fibers were isotropically distributed in the unstretched gels and became more coherent (bundled) and more polarized immediately after stretching (Figure 3—figure supplement 1C and D), with their principal axes oriented in the direction of stretch (Figure 3—figure supplement 1E). Moreover, polarization of the collagen fibrils could be preserved for at least 7 days after the application of stretch by re-embedding the collagen rings in a larger collagen gel (Figure 3A–C, Figure 3—figure supplement 1B). In contrast, when gels were allowed to float in medium after stretching (Figure 3D, Figure 3—figure supplement 1B), gel coherency (Figure 3E) and polarity (Figure 3F) reverted to that of unstretched gels. Thus, re-embedding allowed us to sustain collagen polarization for a prolonged period after the initial stretching.

Figure 3 with 2 supplements see all
Collagen polarization induces mammary aggregate elongation.

(A) Second harmonic generation (SHG) images of collagen fibers in the gel with or without stretching and incubated for 7 days after re-embedding in gel. Double-headed arrow: stretching axis. (B) Coherency of collagen fiber in the gel with or without stretching (n = 24 positions in multiple gels). (C) Distribution of collagen fiber orientation in the gel with or without stretching. 0o is defined as the axis of stretch (N = 3 independent experiments). (D) SHG images of collagen fiber floated for 7 days with or without after stretching. Double-headed arrow: stretching axis. (E) Coherency of collagen fiber in 7 days floated gel with or without stretching (n = 46 positions in multiple gels). (F) Distribution of collagen fiber orientation in gels that had been allowed to float (7 days) with or without prior stretching (N = 3 independent experiments). (G) Time-lapse images of aggregates embedded in stretched gel. Double-headed arrow: stretching axis. (H) Initiation time of aggregate elongation in the gel with or without stretching (– stretch: n = 25 aggregates, +stretch: n = 71 aggregates). (I) Symmetry ratio of aggregates in the early phase of culture (0–3 days) with or without stretching (–stretch, day 0: n = 12; day 1: n = 12; day 2: n = 24; day 3: n = 24; +stretch, day 0: n = 21; day 1: n = 32; day 2: n = 17; day 3: n = 22). (J) Distribution of elongation axes of aggregates in the gel with or without stretching (–stretch: n = 112 aggregates, +stretch: n = 115 aggregates). (K) Difference between elongating axis of aggregates and average angle of collagen fibers in the gel with or without stretching (–stretch: n = 21 aggregates, +stretch: n = 102 aggregates). (L) Fluorescence image of aggregates cultured for 5 days after gel stretching and co-stained with phalloidin (green) and DAPI (blue) in stretched gel. Collagen fibrils were labeled with CNA35-mCherry (magenta). (M) Coherency of collagen fibers surrounding elongated aggregates in the gel with or without stretching (–stretch: n = 21 aggregates, +stretch: n = 102 aggregates). All data are means ± SEM; ns, not significant, **p<0.01, ***p<0.001. Data in (B, E, H, K, M) were analyzed by unpaired Student’s t-test. Data in (I) were analyzed by one-way ANOVA Tukey’s multiple comparisons test.

We then transplanted spherical cell aggregates from Matrigel into the collagen rings and applied our stretching protocol. Interestingly, the aggregates did not become elongated during the period when stretch was actively applied. The symmetry ratios of aggregates immediately after stretching were identical to those in unstretched aggregates (Figure 3—figure supplement 2A and B). Instead, aggregates broke symmetry many hours after the period of active stretching. However, symmetry breaking occurred earlier when stretched gels were allowed to retain collagen polarization by re-embedding (35.14 ± 3.28 hr, Figure 3G and H) compared with aggregates that were allowed to break symmetry spontaneously in non-stretched gels (68.40 ± 5.88 hr, p<0.001). Moreover, the degree of elongation was also enhanced once it had begun. The symmetry ratios of aggregates embedded in stretch-polarized gels were significantly greater than those seen with native gels (Figure 3I). Therefore, polarization of the collagen gel could stimulate the cell aggregates to break symmetry.

Stretch-polarized aggregates also elongated preferentially along the axis of the exogenous stretch, whereas aggregates in non-stretched gels elongated in a random direction (Figure 3G and J, Figure 3—figure supplement 2C and D, Video 2). The angle difference between collagen fiber polarity and aggregate elongation was significantly smaller in stretch-polarized gels than in unstretched control gels (Figure 3K). Of note, collagen gel reorganization was preserved in the stretch-polarized gels (Figure 3L and M), even in the presence of cellular aggregates, and this was oriented in the direction of the original stretch. This suggested that in this assay the elongating aggregates were following the polarity of the gel itself. In contrast, similar intensity of fibronectin staining was seen in stretch-polarized aggregates compared with controls, suggesting that fibronectin deposition was not stimulated by the stretch stimulation (Figure 3—figure supplement 2E).

Video 2
Time-lapse images of aggregates embedded in stretched gel.

Images were taken every 1 hr.

Importantly, the impact of stretch on the aggregates was disrupted when collagen polarization was reversed. The proportion of aggregates that elongated after stretching was reduced when the gels were floated rather than being re-embedded (Figure 4A) and their length of elongation was reduced (Figure 4B). Of note, cell-based forces could not overcome the external forces acting on the gel as the reversal of collagen polarization occurred even when aggregates were incorporated into the gel (Figure 4C). This implied that collagen polarization may have been responsible for allowing external stretch to promote aggregate elongation.

Collagen polarization must be sustained to stimulate mammary aggregate elongation.

(A) Population of elongated aggregates in the re-embedded or floated gel (N = 3 independent experiments). (B) The length of elongated aggregates in the re-embedded or floated gel (N = 3 independent experiments). (C) Coherency of collagen fibers surrounding the aggregates in gels that were re-embedded after stretch or floated for 7 days after stretching (n = 68 aggregates). (D) Cartoon of aggregate from stretched gel into normal collagen gel. Aggregates were isolated from stretched gel by collagenase and re-embedded in naïve gel. (E) Initiation time for aggregate elongation in non-stretched control gels, stretched gels that had been re-embedded to preserve collagen polarization (embed) and after cells were extracted and transferred into non-stretched gels (transfer) (n = 83 aggregates). All data are means ± SEM; ns, not significant, **p<0.01, ***p<0.001. Data in (A–C) were analyzed by unpaired Student’s t-test. Data in (E) were analyzed by one-way ANOVA Tukey’s multiple comparisons test.

To confirm that these results were due to a critical role of collagen polarization, rather than as-yet-unknown impacts of stretch upon the cells, we used collagenase to extract aggregates from gels immediately after their 4 hr stretch, then transplanted them into unstretched gels where the collagen fibers oriented randomly (Figure 4D). We reasoned that if elongation were due to a cell-intrinsic mechanism that bore the memory of the stretch, then this should be preserved even after cells were transplanted into a naïve gel. As noted earlier, aggregate elongation occurred earlier when stretched gels were re-embedded. However, this effect was lost when cells were removed from the stretched gel immediately after stretching and transplanted into a naïve isotropic gel (Figure 4E). Together, these results indicate that collagen polarization can direct aggregate elongation to accelerate its initiation and orient the direction of symmetry breaking.

Collagen matrix polarization stimulates cell proliferation for anlagen elongation

Since cell proliferation was a major driver of spontaneous aggregate elongation, we then asked if it was altered when the collagen gel was polarized by stretch. Indeed, we found that stretch polarization stimulated cell proliferation (Figure 5A, Figure 5—figure supplement 1). An increase in the proportion of Ki67-positive cells was evident at day 1 after stretch polarization, whereas it did not increase till day 2 in unstretched gels. Moreover, the proportion of Ki67 cells was consistently greater in stretch-polarized gels during the early phase of culture (days 1–3) as well as at the end of our experiments (day 8). Cell proliferation mediated the enhanced elongation in the stretched gels as both the length (Figure 5B) and symmetry ratio (Figure 5C) of stretch-stimulated aggregates were reduced by mitomycin C and aphidicolin.

Figure 5 with 1 supplement see all
Collagen polarization induces cell proliferation for aggregate elongation.

(A) Time course of cell proliferation within aggregates in control gels (–stretch) or after stretching (+stretch). Data were percentage of cells that were Ki67-positive (n = 206 aggregates). (B, C) Length (B) and (C) symmetry ratio of aggregates in stretched gel incubated with mitomycin C or aphidicolin for 8 days (n = 214 aggregates). (D) Fluorescence images of elongated aggregates cultured for 7 days in the re-embedded or floated gel after stretching. Aggregates were co-stained with anti-Ki67 antibody (green), phalloidin (red), and DAPI (blue). Collagen fibers were labeled with mCherry-CNA35 (magenta). (E) Percentage of Ki67 positive cells in the aggregates cultured for 7 days in the re-embedded or floated gel after stretching. (n = 56 aggregates). (F) Percentage of Ki67-positive cells in the elongating area of aggregates in the re-embedded or floated gel (n = 47 aggregates). All data are means ± SEM; ns, not significant, **p<0.01, ***p<0.001. Data in (E, F) were analyzed by unpaired Student’s t-test. Data in (A–C) were analyzed by one-way ANOVA Tukey’s multiple comparisons test.

However, proliferation was not increased if gels were allowed to float, rather than being re-embedded, after stretching. Floating the gels reduced the proportion of proliferating cells in the aggregates overall (Figure 5D and E) as well as those specifically within the area of elongation (Figure 5F), compared with gels that had been re-embedded to preserve collagen polarization. This implied that it was sustained polarization of the gel, rather than simply transfer into the collagen environment, that could stimulate cell proliferation to elongate aggregates.

Polarized collagen promotes cell proliferation via ERK pathway

Extracellular signal-regulated kinase (ERK1/2) is one of the major signals that stimulates cell proliferation in mammary epithelial cells (Moreno-Layseca and Streuli, 2014; Streuli and Akhtar, 2009; Walker and Assoian, 2005). To evaluate its possible role in aggregate elongation, we expressed the ERK/KTR-mClover biosensor, which translocates out of the nucleus when ERK is activated (de la Cova et al., 2017). We confirmed the action of the sensor, which accumulated in nuclei as well as in the cytoplasm when monolayer cultures were treated with the ERK inhibitor FR180204 (50 μM) (Figure 6—figure supplement 1A). Using this sensor in 3D cultures, we scored cells that showed nuclear exclusion of the sensor as showing activation of ERK. We found that the proportion of ERK-activated cells was significantly greater in the elongating regions of aggregates that had spontaneously broken symmetry than in the non-elongating areas (Figure 6A and B). This was especially apparent in cells at the surface of aggregates that were in contact with the collagen gel (Figure 6C). In contrast, although Hippo pathway signaling has been implicated in breaking the symmetry of epithelial organoids (Serra et al., 2019), we did not detect any changes in nuclear Yap1 staining that might indicate that this pathway was being activated in our experiments (Figure 6—figure supplement 1B and C). This suggested that ERK 1/2 signaling might be a candidate for collagen polarization to stimulate cell proliferation.

Figure 6 with 1 supplement see all
Polarized collagen promotes cell proliferation via the ERK pathway.

(A) Fluorescent images of aggregates expressing ERK/KTR-mClover biosensor cultured for 7 days and stained with phalloidin (magenta) and DAPI (blue). Cells that showed nuclear exclusion of the biosensor (yellow arrowheads) were scored as ERK-activated, while cells that showed both nuclear and cytoplasmic localization of the sensor (white arrows) were scored as inactive. (B) Percentage of ERK active cells in elongating areas and non-elongating areas of the aggregates (n = 30 aggregates). (C) Percentage of ERK active cells at the surface in elongating and non-elongating areas of aggregates (n = 30 aggregates). (D) Fluorescent images of aggregates cultured for 7 days treated with FR180207 after gel stretching. Aggregates were co-stained with anti-Ki67 antibody (green), phalloidin (red), and DAPI (blue). Collagen fibers were labeled with mCherry-CNA35 (magenta). (E) Percentage of Ki67-positive cells in aggregates incubated with FR180207 for 7 days after stretching (n = 24 aggregates). (F) Effect of inhibiting ERK on stretch-induced aggregate elongation. Proportion of elongated aggregates in stretched gel incubated with FR180207 for 3 days and 7 days (N = 3 independent experiments). (G) Length and (H) symmetry ratio of elongated aggregates incubated with FR180207 for 7 days (n = 166 aggregates). All data are means ± SEM; ns, not significant, **p<0.01, ***p<0.001. Data in (B, C, E, H) were analyzed by unpaired Student’s t-test.

This notion was supported by finding that FR180204 (50 μM) reduced cell proliferation in response to stretch polarization (Figure 6D and E) and decreased the proportion of aggregates that elongated (Figure 6F). As well, the length (Figure 6G) and asymmetric morphology (Figure 6H) were reduced in those aggregates that did elongate. Thus, ERK1/2 appeared to be critical for collagen polarization to stimulate cell proliferation for aggregate elongation.

Integrins are necessary for polarized collagen to stimulate elongation

Integrins are the major ECM receptors in the mammary gland and other epithelia, and they control diverse aspects of cellular function, including cell proliferation (Miranti and Brugge, 2002; Wozniak et al., 2003). We therefore asked if integrin signaling might have stimulated cell proliferation and aggregate elongation in response to collagen polarization (Figure 7A).

Integrins are necessary for polarized collagen to stimulate elongation.

(A) Schematic of potential integrin-ERK pathway that mediates the effect of collagen polarization on cell proliferation. (B) Immunoblot of integrin α2, β1, and ERK1/2 protein levels in MCF10A cell lysate. (C) Fluorescent images of aggregates cultured for 7 days treated with AIIB2 antibody after gel stretching. Aggregates were co-stained with anti-Ki67 antibody (green), phalloidin (red), and DAPI (blue). Collagen fibers were labeled with mCherry-CNA35 (magenta). (D) Percentage of Ki67-positive cells in the aggregates incubated with AIIB2 antibody for 7 days after stretching the gels (n = 35 aggregates). (E) Proportion of elongated aggregates in stretched gel incubated with AIIB2 antibody for 3 days and 7 days (N = 3 independent experiments). (F) Length and (G) symmetry ratio of elongated aggregates incubated with AIIB2 antibody for 7 days (n = 134 aggregates). (H) Fluorescent images of aggregates expressed with ERK/KTR-mClover and incubated with IgG (control) or AIIB2 antibody for 2 days after gel stretching. Aggregates were co-stained with phalloidin (red) and DAPI (blue). Collagen fibers were labeled with mCherry-CNA35 (magenta). (I) Percentage of ERK active cells in aggregates incubated with AIIB2 antibody for 2 days or 7 days in stretched gel (n = 78 aggregates). All data are means ± SEM; ns, not significant, **p<0.01, ***p<0.001. Data in (D–G, I) were analyzed by unpaired Student’s t-test.

We focused on the β1-integrins that have been implicated in regulating cell proliferation during mammary gland development (Li et al., 2005). This class of integrins contains a number of collagen 1 receptors, including α2β1-integrin (Heino, 2000; Käpylä et al., 2000), which was expressed in our MCF10A cells (Figure 7B). Then we used the inhibitory AIIB2 antibody to block β1-integrin during stretch polarization experiments. mAb AIIB2 (15 μg/ml) was added as aggregates were transplanted into the collagen gel rings and replenished after 7 days. mAb AIIB2 blocked the induction of proliferation in stretch-polarized gels, the proportion of Ki67-positive cells being significantly reduced in AIIB2-treated cultures compared with controls (Figure 7C and D). This was accompanied by inhibition of elongation. After 3 days incubation, 49.19% ± 1.61% aggregates started to elongate in controls, whereas less than 20% of aggregates elongated when β1-integrins were inhibited (AIIB2: 17.86% ± 4.78%) (Figure 7E). Moreover, even when mAb AIIB2-treated aggregates eventually elongated upon prolonged culture (comparing the proportion of elongated cells at 7 days with that at 3 days), these aggregates remained shorter (Figure 7F) and less asymmetric (Figure 7G) than in the absence of AIIB2. Finally, we asked if the ERK response required β1-integrins. Indeed, we found that the ability of stretch polarization to stimulate ERK signaling at early (day 2) and later stages (day 7) was reduced when β1-integrins were blocked with mAb AIIB2 (Figure 7H and I). This implied that a β1-integrin-ERK pathway was responsible for stimulating epithelial proliferation.

Discussion

These results lead us to conclude that a polarized collagen 1 matrix can promote epithelial elongation by stimulating locoregional cell proliferation. We infer this because (1) locoregional cell proliferation was a striking feature of MCF10A elongation in our experiments, whose loss was not compensated for by other morphogenetic processes. (2) The pattern of proliferation within aggregates was critically influenced by the polarized organization of the surrounding collagen 1 matrix. Together, we suggest that these reveal an interplay between collagen polarization and cell proliferation that can guide epithelial elongation (Figure 8).

Model of collagen polarization as a structural memory for epithelial anlage elongation.

(i) Initially isotropic epithelia anlage exert isotropic patterns of force on a non-polarized collagen 1 gel. (ii) Initial anisotropies in force associated with symmetry breaking of the aggregate exert strain on collagen fibrils leading to bundling and polarization. (iii) The polarized collagen matrix provides a structural memory that promotes regional cell proliferation to direct further elongation of the anlage.

Cell proliferation was a distinguishing feature of the elongation process in our MCF10A model. Increased proliferation was observed in elongating aggregates and, indeed, began to increase before elongation was detected. Furthermore, proliferation was greater in the regions of the aggregates that underwent elongation compared to those areas that did not. This suggested that a locoregional increase in cell proliferation might be important for the elongation process, a notion that was supported when elongation was inhibited by blocking proliferation. Elongation was reduced when mitomycin C and aphidicolin were added from the beginning of the assays, consistent with earlier evidence that inhibiting proliferation early in the culture process blocked branching morphogenesis in mouse mammary organoids (Ewald et al., 2008). We also found that elongation was compromised if proliferation was inhibited after the elongation process had begun, a later dependence that was not seen in organoids (Huebner et al., 2016). Differences in cell system may therefore influence the impact of cell proliferation on elongation. Nonetheless, our data strongly suggest that locoregional stimulation of cell proliferation was required for epithelial elongation in our model.

This raised the question of how cell proliferation might be preferentially increased within specific regions of MCF10A aggregates. Several of our observations implicate polarized reorganization of the collagen 1 matrix in this specification process. First, locoregional polarization of the collagen 1 matrix accompanied aggregate elongation when cell aggregates spontaneously broke symmetry. This reorganization was distinguished by increased coherency of the matrix, consistent with an increase in collagen bundling, as well as polarization of the gel so that it became oriented with the axis of the elongating aggregate. Matrix reorganization concentrated where the aggregates were elongating and proliferating, as has been reported earlier (Brownfield et al., 2013; Buchmann et al., 2021; Gjorevski et al., 2015). Second, aggregate elongation was stimulated when the gel was polarized by applying an external stretch. Similarly, collagen 1 orientation has been reported to induce mammary organoid branching (Brownfield et al., 2013). But the stimulatory effect on elongation was lost if the stretched gels were allowed to lose their polarization or if aggregates were transplanted into a naïve, unstretched gel. This implied that the cells were responding to the altered organization of the gel, rather than simply exposure to collagen 1. Consistent with what we had observed when aggregates spontaneously broke symmetry, stretch-induced polarization of gels stimulated cell proliferation, and this was necessary for the accelerated elongation process to occur. This appeared to be mediated by a β1-integrin and ERK-dependent pathway.

What property of the polarized collagen 1 was being recognized by the cells to enhance their proliferation? The polarization of collagen networks can have complex effects on cells. Increasing fibril alignment in acellular collagen gels was reported to increase fiber stiffness and decrease pore size (Riching et al., 2014; Taufalele et al., 2019), even in the absence of externally applied forces. Moreover, in single-cell cultures aligned collagen fibers were associated with larger focal adhesions than non-aligned fibrils (Doyle and Yamada, 2016). One possibility, then, is that integrin signaling was responding to the increased stiffness of the polarized collagen gels. This is consistent with the well-characterized role for integrins to sense changes in matrix stiffness (Giannone and Sheetz, 2006). As well, collagen 1 can bind to, and sequester, a variety of growth factors. Whether this reservoir can be released when cells apply tension to collagen is an interesting question for further consideration (Wipff and Hinz, 2008).

Cell migration is a key driver of branching morphogenesis (Gjorevski et al., 2015; Huebner et al., 2016), and earlier experiments reported that the application of external stretch to collagen-embedded aggregates could promote elongation through cell migration (Brownfield et al., 2013). We, too, observed cell migration in our aggregates and, although their speeds were not different, migrating cells within elongating parts of the aggregates tended to align with the axis of elongation and move slightly more persistently than cells moving within the non-elongating regions of the aggregates. These features of collective migration might be expected to contribute to aggregate elongation.

Interestingly, blocking cell proliferation decreased the apparent persistence of the migrating cells and their alignment with the principal axes of the aggregates. Therefore, rather than collective migration being enhanced to compensate for the decrease in proliferation, cell proliferation appeared to support these elongation-facilitating aspects of cell migration. One possible explanation is that persistence and directional alignment were being guided by the reorganized collagen that, in effect, served to create a 3D micropattern around the aggregates. Studies in 2D systems have shown that cells can orient their patterns of migration when collagen networks condense and form bundles (Mohammed et al., 2020; Wang et al., 2018). As collagen reorganization was reduced by inhibiting proliferation, mitomycin C and aphidicolin might have affected cell migration indirectly through matrix organization. Such micropatterning might also influence other aspects of cell behavior during elongation. Micropatterning experiments using 2D substrata have shown that anisotropic confinement can orient patterns of cell division (Théry et al., 2005), potentially by orienting tensile stresses (Campinho et al., 2013; Legoff et al., 2013), as would also be predicted to accompany stiffer, condensed collagen bundles. Consistent with this, we observed that cells within the elongating areas appeared to orient their divisions with the axis of elongation, an effect that would also be predicted to enhance elongation. These observations reinforce the notion that interplay between cell proliferation and collagen (re)organization is a key contributor to epithelial elongation.

The reductionist model that we used in these experiments allowed us to focus on testing how collagen organization could promote epithelial elongation. How might the results from this system operate in the more complex environment of the tissue? One important factor to consider is the basement membrane, which can separate the epithelial cell compartment from collagen 1 in the stroma. However, earlier studies reported that the basement membrane thins substantially and is remodeled in regions of epithelial elongation, such as the tips of mammary or salivary gland buds (Harunaga et al., 2014; Silberstein and Daniel, 1982; Williams and Daniel, 1983), potentially allowing elongating regions of epithelia to engage with collagen 1. Evaluating how cell proliferation and stromal collagen organization are coordinated with other components of the ECM in physiological models of developing glands will be an important question for future research.

In conclusion, we propose the following working model (Figure 8): regional polarization of the collagen 1 matrix begins in response to anisotropies in force that aggregates exert upon their ECM via their integrin adhesions (Figure 7A). Cell-based forces exerted through integrins can apply strain on collagen fibrils (Brownfield et al., 2013; Buchmann et al., 2021; Gjorevski et al., 2015; Hall et al., 2016) and, consistent with this, collagen reorganization in our experiments was compromised by inhibiting cellular contractility. However, compression of collagen by increasing cell numbers or cell movements could also have contributed (Buchmann et al., 2021). By implication, collagen polarization will be greater around the parts of the aggregate that are generating more force. The consequent locoregional collagen polarization then stimulates further cell proliferation in the adjacent parts of the aggregate to sustain elongation. Biochemical contact with collagen may provide a basal stimulus for cell proliferation, which is further enhanced where collagen becomes polarized, to provide a boundary condition that promotes elongation. It was interesting to note that collagen reorganization was compromised when cell proliferation was blocked from the outset of the assays. Possibly, subtle differences in cell proliferation may have contributed to generating initial anisotropies in force to reorganize the gels. If so, we speculate that the increase in cell proliferation may have contributed to further collagen reorganization, with the capacity to develop a feedback system that might help sustain aggregate elongation. This would allow polarized collagen networks to effectively provide a structural memory of the initial axis of symmetry breaking, that is, a relatively long-lived spatial cue that directs further elongation of the epithelial anlagen.

Materials and methods

Cell culture and lentivirus infection

Request a detailed protocol

MCF10A human mammary epithelial cells (CRL-10317) and HEK293T human embryonic kidney 293 cells (CRL-3216) were obtained from ATCC. Both cells were confirmed mycoplasma negative by PCR. MCF10A cells were cultured in DMEM/F12 medium supplemented with 5% horse serum, 10 μg/ml insulin, 0.5 μg/ml hydrocortisone, 100 ng/ml cholera toxin, 20 ng/ml EGF, 100 units/ml Penicillin and 100 units/ml Streptomycin as previously described (Debnath et al., 2002).

HEK-293T cells were transfected using Lipofectamine 2000 (Invitrogen) for lentiviral expression vector pLL5.0 and third-generation packaging constructs pMDLg/pRRE, RSV-Rev, and pMD.G. Third-generation packaging constructs were kindly provided by Prof. James Bear (UNC Chapel Hill, NC). After transducing MCF10A cells with either pLL5.0-EGFP-HRasC20, pLL5.0-NLS-mCherry, or pLenti-ERK/KTR-mClover lentivirus construct, we isolated highly-expressing cells using fluorescence-activated cell sorting (Influx Cell sorter; Cytopeia).

3D Matrigel and collagen culture

Request a detailed protocol

MCF10A aggregates and acini were cultured on 100% growth factor-reduced Matrigel (FAL354230; Corning) and overlayed with Matrigel including medium in eight-well chamber coverglass (0030742036; Eppendorf) as previously described (Debnath et al., 2002; Debnath et al., 2003). In brief, single isolated MCF10A cells were mixed with an assay medium containing 2% Matrigel, 5 ng/ml EGF, and seeded on a solidified layer of 100% Matrigel at 10,000 cells/well. After seeding cells on Matrigel, medium was changed every 4 days.

In collagen culture, single isolated MCF10A cells were mixed with collagen gel at 20,000 cells/well (FAL354236; Corning). This mixture was neutralized with NaOH and HEPES buffer and collagen concentration was adjusted to 1.0 mg/ml with culture medium on ice. The pH of collagen solution was checked by litmus paper. Collagen solution with cells was seeded in eight-well chamber coverglass and solidified at 37℃ for 30 min.

Transplantation of MCF10A aggregates

Request a detailed protocol

MCF10A aggregates cultured on Matrigel for 10 days were washed with PBS and incubated with cold cell recovery solution (FAL354253; Corning) for 30 min on ice. Aggregates were collected into a tube containing the cell recovery solution, spun down at 1200 rpm, and washed with cold PBS. PBS-washed aggregates were resuspended in MCF10A culture medium and mixed with collagen solution neutralized with NaOH and HEPES buffer. Collagen gel solution with aggregates was solidified at 37℃ for 30 min.

Aggregates embedded in collagen gel were isolated from gel by dissolving the gel with collagenase (C2799; Sigma). Gels were washed with PBS and incubated in 20 μg/ml collagenase in Hanks’ Balanced Salt Solution (H8264; Sigma) at 37℃ for 30 min to dissolve the collagen gel. Aggregates were collected with incubated solution into the tube and mixed with DMEM/F12 medium, which contains 20% horse serum. Solution was spun down at 1200 rpm and pellet was washed with PBS. PBS-washed aggregates were resuspended in culture medium and re-embedded in collagen gel.

Plasmids pLL5.0-NLS-mCherry was constructed in a previous study (Leerberg et al., 2014). pLent-ERK/KTR-mClover was obtained from Addgene (#59150). pLL5.0-EGFP-HRasC20 was generated by insertion of EGFP-HRasC20 fragment that was amplified from pEGFP-F (#6074-1; Clontech) by PCR into pLL5.0 vector. pLL5.0-EGFP (gift from Prof. James Bear; UNC Chapel Hill) was digested with EcoRI and SbfI to remove EGFP, and then EGFP-C20 fragment was inserted and ligated by In-Fusion cloning kit (638910; Clontech).

Antibodies and inhibitors

Request a detailed protocol

Primary antibodies used for immunocytochemistry in this study were rabbit anti-Ki67 (ab15580; Abcam), mouse anti-laminin V (MAB19562; Chemicon), rat anti-E-cadherin (131900; Invitrogen), mouse anti-GM130 (610822; BD), rabbit anti-fibronectin (F3648; Sigma), and rabbit anti-YAP1 (4912; Cell Signaling Technology). F-actin was stained with AlexaFluor 488-, 594-, 647-phalloidin (Invitrogen). Primary antibodies used for immunoblot were rabbit anti-integrin α2 (ab181548; Abcam), mouse anti-integrin β1 (610467; BD Transduction Laboratories), and rabbit anti-p44/42 MAPK (Erk1/2) (9102; Cell Signaling Technology). AIIB2 antibody for blocking integrin β1 was purchased from Developmental Studies Hybridoma Bank and treated at 15 μg/ml in this study.

Proliferation inhibitors mitomycin C (M7949; Sigma) and aphidicolin (178273; Merck) were used at 10 μM and 2 μM, respectively. Myosin II inhibitor blebbistatin (203390; Sigma) and ROCK inhibitor Y-27632 (688000; Sigma) were used at 25 μM and 30 μM, respectively. ERK1/2 inhibitor FR180204 (SC-203945; Santa Cruz) was used at 50 μM. Except for mitomycin C, cell aggregates were treated with the inhibitors after transplantation into collagen gels. Inhibitors were then replenished when medium was changed and left in the collagen culture until cell aggregates were ready to be fixed. For mitomycin C washout experiments, cell aggregates were treated with mitomycin C for 1 hr, then washed three times with PBS, and left to recover in MCF10A culture medium.

Immunocytochemistry and microscopy

Request a detailed protocol

3D cultured cells in gel were fixed with 4% paraformaldehyde in cytoskeleton stabilization buffer (10 mM PIPES at pH 6.8, 100 mM KCl, 300 mM sucrose, 2 mM EGTA, and 2 mM MgCl2) at room temperature for 30 min, followed by a treatment for 30 min with 0.5% TritonX-100 and 10% goat serum in PBS for 1 hr at room temperature. Then they were stained with primary antibodies for overnight at 4℃. Subsequently, cells were washed with PBS and incubated with secondary antibodies with phalloidin and DAPI for 1 hr at room temperature.

Confocal images were acquired with an upright Meta laser scanning confocal microscope (LSM710; Zeiss) equipped with plan-Apochromat 20× 0.8 NA or 40× 1.3 NA objectives (Zeiss) and zen2012 software (Zeiss). The fluorescent images of collagen fibrils probed with mCherry-CNA35 were acquired by inverted microscope (Ti2; Nikon) equipped with Dragonfly spinning disc (Andor) by using plan-Apo 10× 0.45 NA, 20× 0.75 NA, or 40× 0.95 NA dry objectives (Nikon) and Fusion software (Andor).

Fluorescent images of the SHG signal from collagen 1 were collected with an inverted confocal microscope (LSM710; Zeiss) equipped with multiphoton laser by using 860 nm excitation with SHG signal obtained with 690 nm bandpass filter. A 40× 1.3 NA or 63× 1.40 NA plan-apochromat oil objectives (Zeiss) were used to obtain SHG signals.

Fluorescence and phase-contrast live images of elongating aggregates were acquired with an inverted fluorescence microscope (IX81; Olympus) equipped with CCD-camera (Hamamatsu) and an incubation box (Clear State Solutions) maintained at 37℃ and 5% CO2 with gas controller (OkoLab), using plan-Apo 10× 0.4 NA objective (Olympus) and CellSens software (Olympus).

Collagen gel labeling and stretching

Request a detailed protocol

Collagen fibrils were labeled with mCherry-CNA35 by mixing with purified mCherry-CNA35 protein at 2 μM before gelling. Gels were solidified at 37℃ for 30 min.

PDMS gel frame and stretchers for gel stretching were kindly gifted from Dr. James Hudson (QIMR Berghofer, Queensland, Australia). 1.5 mg/ml collagen gels were solidified as a ring-shape in PDMS frame. PDMS stretchers were inserted into the hole of the gel and expanded, and incubated for 4 hr in culture medium. Stretched gels were released and then floated in medium or re-embedded in collagen gel. For re-embedding, stretched gels were put in 1.5 mg/ml collagen gel and solidified at 37℃ for 30 min.

CNA35-mCherry protein purification

Request a detailed protocol

Protein expression vector pET28a-mCherry-CNA35 was obtained from Addgene (#61607). Transformed Escherichia coli (BL21) was cultured in 400 ml LB, and induced protein expression with isopropyl β-D-1-tthiogalactopyranoside for 20 hr at 25℃. Cultured bacteria were spun down, and collected bacteria pellet was sonicated. The lysate was centrifuged, and the resulting supernatant was run through a column filled with N-NTA His-band resin (Millipore). Bound protein was eluted and then dialyzed for overnight in PBS at 4℃. Endotoxin was removed by endotoxin removal columns (88274; Thermo) following with manufacturer’s protocol.

Quantitative analysis of collagen fibril alignment and aggregates elongation

Request a detailed protocol

Fluorescent confocal images of collagen fibrils acquired SHG microcopy or CNA35 probes were used for collagen fibril analysis. We analyzed collagen organization in these images using Orientation J, which calculates the local orientation and isotropy for each pixel in an image based on the structure tensor for that pixel (Rezakhaniha et al., 2012). The structure tensor is evaluated for each pixel of the given image by calculating the spatial partial derivatives by using (a cubic B-spline) interpolation. The local orientation and isotropy for each pixel are computed based on the eigenvalues and eigenvectors of the structure tensor. We characterized three features in the organization of collagen fibrils orientations (Clemons et al., 2018): (i) the coherency of fibrils, defined as co-orientation in the same direction, as a measure of bundling; (ii) isotropy, the distribution of fibril orientation in the field of analysis; and (iii) polarization, defined as the principal axis of fibril orientation in anisotropic gels, relative to a reference axis. 50 × 50 µm or 100 × 100 µm size of ROI were selected from SHG images or CNA35 probed images, respectively, and used for analysis. Elongation axis of aggregates was measured by the angle tool in ImageJ, and then we calculated the angle difference between principal axis of collagen fibrils or stretching axis. The fold difference of gel coherency was measured by dividing the coherency of fibrils in the elongating area by the non-elongating area.

The symmetry ratio of aggregates was measured by dividing the longest length by widest width of aggregates.

Quantitative analysis of proliferating cells and nuclear division angle

Request a detailed protocol

The number of cells co-stained with DAPI and anti-Ki67 antibody was counted by Imaris software (Bitplane). The number of Ki67-positive cells was divided by the total number of nucleus stained with DAPI to calculate percentage of Ki67-positive cells. For live imaging, cells were expressed with NLS-mCherry to count their number. The frequency and orientation of cell division were analyzed from the time-lapse images by ImageJ. We set the elongation axis of aggregate as a reference and measured the angle difference between elongation axis and dividing axis of nucleus.

Analysis of nuclear tracking

Request a detailed protocol

To obtain the migration speed of cells, we tracked individual cell nuclei using the Spot function in the Imaris software (Huebner et al., 2016). The fluorescent images of cells expressing NLS-mCherry were acquired every 10 min for at least 50 hr. To obtain the track displacement angle, we first calculated the displacement angle of nuclei from the displacement of X and Y axis between first and last position, and then measured the angle difference from elongating angle of aggregates. Track straightness was calculated from track displacement by track length. Inhibitors were treated 1 hr before the imaging.

Quantitative analysis of ERK/KTR biosensor

Request a detailed protocol

ERK activity was judged by the location of mClover fluorescent signal in individual cells. Briefly, fluorescent-tagged KTRs translocate between nuclei and cytoplasm depends on kinase activity. When ERK activity is high, KTR-mClover should localize in cytoplasm (de la Cova et al., 2017). Aggregates expressing ERK/KTR-mClover were transplanted into collagen gel and were co-stained with phalloidin and DAPI after fixation. To judge the delocalization of KTR-mClover, we used line intensity scan in single-plane images and then manually scored ERK active cells as those where the fluorescent signal was excluded from the nucleus (compared with cells that showed both nuclear and cytoplasmic localization, which were scored as inactive).

Statistical analysis

Request a detailed protocol

Significance was determined by unpaired Student’s t-test and one-way ANOVA by using GraphPad Prism 8 (GraphPad software).

Data availability

All data generated and analysed in this study are included in the manuscript and supporting files. Source data files have been provided for Figure 1-7.

References

Decision letter

  1. Matthew Kutys
    Reviewing Editor; University of California, San Francisco, United States
  2. Jonathan A Cooper
    Senior Editor; Fred Hutchinson Cancer Research Center, United States
  3. Matthew Kutys
    Reviewer; University of California, San Francisco, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This paper makes use of elegant experiments to explore the role of collagen matrix polarization during symmetry break and elongation of three-dimensional epithelial anlagen. Polarized collagen stimulates regional cell proliferation through β1 integrin and ERK signaling which is necessary to drive symmetry break and elongation. These findings will be of interest to a broad audience of developmental and cell biologists.

Decision letter after peer review:

Thank you for submitting your article "Collagen polarization provides a structural memory for the elongation of epithelial anlage" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Matthew Kutys as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Jonathan Cooper as the Senior Editor.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

In this paper, the authors used reductionist models to investigate the interplay between collagen extracellular matrix and the polarized elongation of mammary epithelial anlage and provide new mechanistic insights into this process, which are potentially of broad interest. However, all three reviewers felt that there are several issues that need to be addressed before this paper can be published, most significantly those summarized below:

1) All reviewers felt the data did not fully support the conclusions in several areas. Most notably, there were issues with the suggested causality of proliferation in driving analage elongation and the instructive role of polarized collagen in directing outgrowth that must be addressed.

2) The authors distinguish their work by focusing on elongation rather than branching events, however the approaches and observations are similar to previous studies of branching 3D mammary explant organoids in aligned collagen matrix. How the conclusions of this work, including the observed morphogenic behavior and underlying control mechanisms, advance our conceptual understanding must be made clear.

3) Demonstrating the extent to which the findings from this work apply to mammary tubulogenesis in vivo is required.

Reviewer #1:

Katsuno-Kambe and colleagues investigate the mechanism by which non-polarized MCF10A spheroids can be induced to extend multicellular aggregates when cultured within 3D type I collagen extracellular matrix hydrogels. Transplanting preformed aggregates into collagen is sufficient to induce asymmetric elongation of the spheroid. Branching and elongation depends on region-specific cell proliferation and multicellular outgrowths are characterized by the local accumulation of collagen matrix after the initial symmetry break. 3D hydrogel stretching and preformed aligned collagen extracellular matrix is sufficient to direct elongated outgrowth, which is dependent on proliferation, β1 integrin and ERK signaling. While the work is quantitative and the imaging is of high quality, chemical and physical factors that affect the long axis bias of the developing mammary epithelium have been intensively studied in vivo and in vitro, and it is not clear that this work offers a significant conceptual advance. The study also would need validation of observations from their engineered in vitro systems with relevant aspects of in vivo conditions.

Methods and conclusions from this work are similar to those previously reported by the Bissell group (Brownfield et al. Current Biology, 2013). In this study, unpolarized primary murine mammary explants were embedded within 3D collagen hydrogels and axially stretched using a deformable PDMS-based device. Collagen stimulated explant branching and outgrowth, which required collagen fiber polarization. The authors do reference this study, but argue its focus is on branching rather than elongation. However, given the nearly identical methods, analyses, and morphogenic conclusions, it is unclear how this present work is distinct. Similarly, it is unclear if the mechanisms derived from the reductionist system here are relevant to mammary development/tubulogenesis in vivo.

1) Recent and previous work from the Bissell and Nelson groups have explored the role of collagen fiber polarization and/or axial alignment during patterned elongation of the developing mammary gland. While the role collagen still remains poorly understood, it will be important to reconcile these studies with the work presented here, namely by demonstrating the relevance of the findings from this work to mammary tubulogenesis in vivo.

2) Methods and conclusions from this work are similar to those reported by the Bissell group (Brownfield et al. Current Biology, 2013). In this study, unpolarized primary murine mammary explants were embedded within 3D collagen hydrogels and axially stretched using a deformable PDMS-based device. Collagen stimulated explant branching and outgrowth, which required collagen fiber polarization. The authors do reference this study but argue its focus is on branching rather than elongation. However, given the nearly identical methods, analyses, and morphogenic conclusions, it is unclear how this present work is distinct.

3) "In contrast, Rac1 inhibitor treatment did not affect aggregate elongation and collagen fibril rearrangement (data not shown)". These data should be shown, as it conflicts with the conclusions of the Brownfield study. It should be noted that the Rac1 inhibitor NSC23766 only blocks a subset of Rac1 interactions with Rac1GEFs.

4) From Figure 2A, is the gel coherency metric capturing fibril alignment, or the localized accumulation of collagen fluorescent intensity? Is collagen accumulation a function of cells pushing/expanding out into the matrix, or from active fiber recruitment to the outgrowth surface?

What determines whether localized collagen intensity leads to branching and outgrowth? In Figure 2A, there are two distinct regions of polarized collagen, but only one contains outgrowth. There is an overall variability in the brightness, contrast, and resolution of the representative collagen images, making it difficult to interpret across experiments. For instance, polarized collagen appears to be present in the blebbistatin treated outgrowths, however it is difficult to discern due to the different magnification and non-uniform hydrogel intensity.

5) Is proliferation truly driving this behavior? In the representative images (Figure 1C, 1H) it appears that all peripheral MCF10A making basal matrix contact are Ki67 positive. It is a bit difficult to tell when proliferation arresting inhibitors were added in the evolution of these processes. Does acute addition of mitomycin halt, or regress, established outgrowths?

6) Similarly, do stretch-induced elongations regress after removal from stretch and transfer to floating culture?

7) AIIB2 implicates β1 integrin, but not α2β1. The authors should check to see if fibronectin deposition is stimulated by stretch, and is fibronectin present at stochastic outgrowths in non-patterned collagen?

8) "This regional patterning would be necessary if proliferation drives elongation, as a global increase in cell number would cause aggregates to expand isotropically." It is important to acknowledge that a critical regulatory feature missing from this model is a developed basement membrane. in vivo, the local degradation and remodeling of basement membrane mechanically permit region specific expansion of developing branched tissues, such as in the embryonic salivary gland (Harunaga et al. Dev. Dyn., 2014).

9) "The decision to branch is recognized to be a critical checkpoint that is controlled by developmental signaling." It is unclear what distinction the authors are trying to make. There is robust literature suggesting cell nonautonomous mechanisms regulating branching events, notably localized extracellular matrix deposition and heterotypic cell-cell interactions.

10) The aspect ratios of the images in Figure 5D appear to be off.

Reviewer #2:

The formation of mammary gland tubules begins as non-polarized cellular aggregates, called anlage, which elongate as solid multicellular cords that eventually form lumens. Katsuno-Kambe et al. investigate the role of the extracellular matrix during the process of elongation, making use of 3D culture models of mammary epithelial MCF10A cells. These cells have previously been shown (e.g Krause et al., Tissue Eng Part C Methods 2008, and others) to form multicellular aggregates that polarize and eventually form lumens when cultured in Matrigel, and when embedded in Collagen I (a major component of the stromal environment) form elongated, solid cords of non-polarized cells. Therefore, to capture the process by which aggregates break symmetry and become elongated, the authors first seeded MCF10A cells in Matrigel until proliferation arrested (resembling anlage), after which aggregates where isolated and embedded into collagen I gels. The authors show increased proliferation of cells in elongating areas of aggregates, and that these elongated aggregates are not formed when proliferation is inhibited. The authors further show that this elongation is accompanied by collagen reorganization into polarized bundles around elongating areas, and by artificially polarizing collagen by application of external stretch show collagen polarization can increase proliferation and guide the direction of elongation. Finally, the authors demonstrate aggregate elongation requires beta1-integrin adhesion and downstream ERK activity.

The strength of this manuscript is the use of a reductionist approach to investigate the role of collagen alignment in the elongation of epithelial anlage, by transplanting MCF10A cell clusters from matrigel into Collagen I gels, and artificially manipulating collagen polarization by stretching these gels. In addition, the exploration of the underlying mechanism of elongation, requiring proliferation and integrin-ERK signaling, is of potential great interest to understanding the role of cues derived from the extracellular matrix in epithelial morphogenesis.

The main weakness of this study is that the data still insufficiently supports several conclusions, in particular on proliferation being a key driver of anlage elongation, and on the reciprocal interaction between proliferation and collagen polarization during this process. The induction of proliferation is only analyzed 8 days after cell clusters have been transferred to Collagen gels, and thus after symmetry break and elongation have already occured. Analyses of proliferation and collagen polarization during the different stages of symmetry break and elongation, and detailed analyses of the distribution of proliferative cells during these stages, are lacking. This is in particular relevant as also in non-elongated areas, cells that are in contact with Collagen appear to be proliferative. Alternative driver mechanisms, for instance the induction of collective migration, are not fully tested and it remains unclear whether proliferation is merely required for, or truly a driver event of aggregate elongation.

The authors investigate a role for integrin signaling and downstream ERK in the induction of proliferation and aggregate elongation in response to collagen polarization. While an integrin β 1 inhibitory antibody and ERK inhibitor indeed prevents elongation, key unanswered questions are whether this pathway is indeed activated by polarized collagen during elongation (i.e. in selective regions of the cell clusters) and whether it is required for the proliferation response itself.

Despite the elegance of the reductionist approach in MCF10A cells in vitro, it remains unclear to what extent these findings generally apply to mammary tubulogenesis under physiological conditions in vivo, which remains largely undiscussed in the manuscript.

Comments for the authors:

1. The authors conclude that cell proliferation is a key driver of epithelial anlage elongation. However, the induction of proliferation is only analyzed 8 days after transfer of cell clusters to Collagen I, after symmetry break and elongation have taken place. And here, analysis of proliferation is limited to cells in the elongated areas versus the remaining cell cluster. It is essential to include analyses of proliferation during the different stages of elongation (i.e. before and upon initial symmetry break and at different time points of elongation), as well as detailed analyses of the distribution of proliferative cells during these stages. This is in particular relevant as from the images (e.g. Figure 1C) it is apparent that also in the non-elongated areas the cells on the outside of the cell cluster (thus in contact with Collagen I) are proliferating (e.g. Figure 1C), yet do not form elongated strands (which also does not support the authors conclusions that "proliferation is selectively increased in regions of aggregates that were elongating"). Finally, analysis of the induction of proliferation should also include a control in which cell clusters are transferred back to matrigel.

2. Alternative driver mechanisms, in particular cell migration, are limited to analysis of migration speed that is similar in different regions of the aggregate. This does not exclude that induction of directional, collective migration could be the main driving event underlying elongation, which remains unexplored. As such, it remains undetermined whether proliferation is merely required for, or truly a driver event of aggregate elongation.

3. The authors mention in the discussion that "the initial step of symmetry-breaking was associated with reorganization of collagen fibrils along the aggregate (…) which became increasingly bundled and aligned as elongation proceeded". However, data showing these changes in collagen during the different steps are not presented in the manuscript. This is in particular important as the authors conclude that collagen polarization induces proliferation, and therefore the authors should include a more detailed analyses of both processes (collagen reorganization and proliferation) over time after cell clusters are transferred to Collagen gels, and how they are associated with each other in individual cell clusters.

In addition, can the authors explain the induction of proliferation in cells in contact with collagen in the non-elongating areas (e.g. Figure 1C, see comment 1), as this suggest contact with collagen is sufficient to increase proliferation and collagen polarization is not required?

4. In line with the above comment, the conclusion that collagen polarization promotes proliferation is largely based on the experiments in which collagen gels were stretched. However, the effect on proliferation is again only assessed following 8 days of culture in these gels, and thus after aggregate elongation. To support their conclusion, proliferation should be analyzed at early time points following stretching. This would also address whether the earlier symmetry break observed after stretching is a consequence of increased/faster induction of proliferation, or involves the ability of already proliferating cells to drive epithelial elongation (which will affect the interpretation of these data).

5. The authors demonstrate that inhibition of proliferation prevents aggregate elongation. It is difficult to interpret from these experiments whether proliferation could truly be a key driver of elongation, as aggregates appear to remain much smaller following mitomycin C and Aphidicolin treatment (Figure 1H). It should be shown whether in the absence of the inhibitors, symmetry break and elongation is induced at this aggregate size (and thus whether the observed effect represents a direct or indirect of the inhibitors on elongation). In addition, proliferation inhibitors should also be given after the symmetry break has taken place, to test whether this prevents protrusions that are already present from elongating further.

6. The authors investigate a role for integrin signaling and downstream ERK in the induction of proliferation and aggregate elongation in response to collagen polarization. While indeed both integrin beta1 and ERK activity are necessary for aggregate elongation, whether this pathway is directly involved in aggregate elongation remains unclear, and analysis of cell proliferation is lacking. The authors should test with available antibodies whether this pathway is specifically induced by polarized collagen (by analyzing different regions of the aggregate during symmetry break and elongation). In addition, effect of inhibitors on the level and distribution of Ki67+ cells should be shown.

7. To control that effects observed in stretched collagen gels are truly mediated by the induced collagen polarization, the authors allowed gels to float in the medium to revert gel isotropy to that of unstretched gels. While collagen polarization is indeed reverted, cells cultured in floating gels will not able be able to generate stresses themselves (e.g. Halliday et al., 1995). From these experiments it can therefore not be concluded whether it is truly the collagen polarization or stresses generated by the cells that are needed for aggregate elongation.

8. It remains unclear to what extent the findings from this manuscript apply to mammary tubulogenesis in vivo, which remains undiscussed in the manuscript. While some experiments may be difficult to verify in vivo (e.g. to monitor collagen alignment), it would be relevant to at least to know whether in vivo mammary epithelial cells indeed come in contact with collagen I during anlage elongation, and preferably also show whether this is accompanied by a specific proliferation response.

Reviewer #3:

Understanding tissue morphogenesis requires an understanding of underlying molecular and mechanical mechanisms. Many tissues form a network of branched tubules and understanding their origins is an important question for the field. Here the authors develop a simple, novel assay for exploring this. By plating MCF-10A mammary epithelia cell cysts in 3D collagen, they find that these cells rapidly reorganize, break symmetry, and send out streams of cells that proliferate as they move outward, forming solid cords. This behavior is particular for collagen and is not see in Matrigel. They then proceed to use this system to explore the causal links between collagen, proliferation, directed outgrowth and integrin signaling. They find that cell proliferation coincides with outgrowth and outgrowth is blocked in proliferation is inhibited. They find that collagen is re-organized as outgrowth occurs, altering fibril organization. They then carried out a set of clever experiments to use mechanical stretch to pre-orient fibrils, and found that aggregates preferentially elongated along the fibril preferred axis. Pre-oriented fibrils also sped elongation and proliferation. Finally, blocking integrin signaling reduced/delayed elongation. I found the new system interesting, and the data intriguing. However, I had two major concerns, about the measurements presented and the causality suggested.

1. Most of the experiments involve measuring collagen fibril orientation and coherency. However, in most of the images I could not see individual fibrils (the exception, Figure 3D, was strikingly different than the rest). I would like to see a clearer description in the main text of how orientation was measured and why this is difficult to see in most images.

2. Bigger picture was the issue of causality. The authors present seemingly a linear view – integrin signaling drives proliferation and outgrowth and this triggers fibril orientation that then feeds back into the process. However, I found causality difficult to discern. While proliferation is necessary for outgrowth, its role seemed to be more likely to be permissive rather than instructive. The observation that cells would follow oriented fibrils was interesting but perhaps not surprising, and I found it difficult to tease out whether elongation triggered polarization of collagen or vice versa – I guess the authors favor a feedback process.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Collagen polarization promotes epithelial elongation by stimulating locoregional cell proliferation." for further consideration by eLife. Your revised article has been reviewed by 2 reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Jonathan Cooper as the Senior Editor.

We have received comments from two of the previous reviewers, both of whom indicate the manuscript is strengthened by the addition of new experiments, text clarifications, and discussion of results in context. The reviewers request that specific representative images be added. Notably, the reviewers each agree that it is important to:

1) Provide clarification on how quantification of ERK biosensor activity was performed. Consider selecting a more representative image or using high magnification insets.

2) Provide clarification on how local regions of proliferation were calculated in 3D aggregates (i.e. did it include all cells, or only cells in contact with collagen). If needed, include a direct analysis of cells in contact with collagen in non-elongating vs elongating regions to support the conclusions.

Reviewer #1:

The authors have taken steps to substantially clarify the core findings from their paper, driving home their central conclusion that regional proliferation stimulated by polarized collagen directs aggregate elongation. Importantly, the authors also distinguish the conclusions of this study from prior work that apply similar methods. They have added additional experiments that address most of my concerns regarding the role of proliferation in this process and altogether the manuscript is stronger.

Where I am still not clear is the relationship between collagen and the stimulation of proliferation. This conclusion is derived from reported regional differences in cell proliferation that coincide with elongation and collagen polarization. However, this quantification could be potentially misleading, as there is also a significant difference in the number of cells with surface area contacting the surrounding collagen in non-elongating regions yet those that do appear to be proliferative (Figure 1C, 1J). Further not all instances of accumulated collagen appear to lead to elongation.

The authors clearly demonstrate (Figure 5) that polarized collagen influences the kinetics of proliferation and elongation. Perhaps this can be explained by a slightly refined model in which collagen is sufficient to stimulate cell proliferation in anlagen and its concurrent polarization/remodeling provides boundary conditions that together support the emergent behavior of elongation?

Reviewer #2:

The authors have addressed my main concerns in their rebuttal. The addition of new analyses strengthens the conclusion of the authors that locoregional proliferation is important for the elongation of multicellular aggregates. The data doesn't completely tease out whether this locoregional proliferation is instructive or merely permissive, but the data are more clearly interpreted and discussed in the revised manuscript.

https://doi.org/10.7554/eLife.67915.sa1

Author response

Essential revisions:

In this paper, the authors used reductionist models to investigate the interplay between collagen extracellular matrix and the polarized elongation of mammary epithelial anlage and provide new mechanistic insights into this process, which are potentially of broad interest. However, all three reviewers felt that there are several issues that need to be addressed before this paper can be published, most significantly those summarized below:

1) All reviewers felt the data did not fully support the conclusions in several areas. Most notably, there were issues with the suggested causality of proliferation in driving analage elongation and the instructive role of polarized collagen in directing outgrowth that must be addressed.

We’ve performed additional experiments that analyse early events in anlagen elongation, especially the relationship between collagen polarization, cell proliferation and elongation. These show that:

a) When symmetry-breaking and elongation occur spontaneously, cell proliferation precedes elongation (by ~ 24 hours) and increases preferentially in those regions of aggregates that elongate.

b) Collagen polarization coincides with aggregate elongation even in these early stages (i.e. within the first 3 days of the assays).

c) Locoregional cell proliferation is stimulated further and accelerated when we induce collagen polarization extrinsically by stretching the gels.

Together, this analysis of early events reinforces the importance that enhanced loco-regional cell proliferation plays in the elongation process.

2) The authors distinguish their work by focusing on elongation rather than branching events, however the approaches and observations are similar to previous studies of branching 3D mammary explant organoids in aligned collagen matrix. How the conclusions of this work, including the observed morphogenic behavior and underlying control mechanisms, advance our conceptual understanding must be made clear.

This is a very good point and we have substantially re-written the manuscript to make the novelty of our findings clearer, especially in relationship to what has been reported before using broadly similar experimental approaches. To summarize, we believe that the contribution of our study is to identify (1) A role for loco-regional cell proliferation as a driver of epithelial anlagen elongation; and (2) A role for local reorganization of the collagen 1 matrix as a stimulator of that cell proliferation.

This contrasts with earlier studies that used similar experimental systems. For example: Brownfield et al. (Current Biology, 2013) showed that stretching collagen gels induces the elongation of embedded epithelial aggregates, but they focused on cell motility and did not test cell proliferation. Gjorevski et al. (Scientific Reports, 2013) showed that aggregates can reorganize collagen by exerting tensile forces, but focused their analysis on the impact for cell migration. Similarly, Buchmann et al. (Nat Comms, 2021) did not observe loco-regional differences in cell proliferation, but these were tested after elongation was well established, not at the early stages of symmetry-breaking as we have now done. Finally, Huebner et al. (Development, 2016) reported a role for ERK in epithelial elongation, but did not test the role of ECM organization. Thus, our findings have identified a new driver of elongation that has not been identified before.

3) Demonstrating the extent to which the findings from this work apply to mammary tubulogenesis in vivo is required.

We have expanded our Discussion to consider how our findings may apply to the more complex environment of a tissue. In particular, earlier studies have reported that the basement membrane is remodelled and thins out substantially at the tips of elongating mammary epithelial buds (and this is also seen in other models, such as the salivary gland). This would be predicted to increase cell exposure to collagen 1 in the stroma, a context that could allow the process that we have identified to operate. It will be important to test our predictions in an in vivo model, but this was not something that was feasible in the time frame of a revision. Respectfully, we suggest that it is a problem that deserves a study of its own.

Reviewer #1:

Katsuno-Kambe and colleagues investigate the mechanism by which non-polarized MCF10A spheroids can be induced to extend multicellular aggregates when cultured within 3D type I collagen extracellular matrix hydrogels. Transplanting preformed aggregates into collagen is sufficient to induce asymmetric elongation of the spheroid. Branching and elongation depends on region-specific cell proliferation and multicellular outgrowths are characterized by the local accumulation of collagen matrix after the initial symmetry break. 3D hydrogel stretching and preformed aligned collagen extracellular matrix is sufficient to direct elongated outgrowth, which is dependent on proliferation, β1 integrin and ERK signaling. While the work is quantitative and the imaging is of high quality, chemical and physical factors that affect the long axis bias of the developing mammary epithelium have been intensively studied in vivo and in vitro, and it is not clear that this work offers a significant conceptual advance. The study also would need validation of observations from their engineered in vitro systems with relevant aspects of in vivo conditions.

Methods and conclusions from this work are similar to those previously reported by the Bissell group (Brownfield et al. Current Biology, 2013). In this study, unpolarized primary murine mammary explants were embedded within 3D collagen hydrogels and axially stretched using a deformable PDMS-based device. Collagen stimulated explant branching and outgrowth, which required collagen fiber polarization. The authors do reference this study, but argue its focus is on branching rather than elongation. However, given the nearly identical methods, analyses, and morphogenic conclusions, it is unclear how this present work is distinct. Similarly, it is unclear if the mechanisms derived from the reductionist system here are relevant to mammary development/tubulogenesis in vivo.

1) Recent and previous work from the Bissell and Nelson groups have explored the role of collagen fiber polarization and/or axial alignment during patterned elongation of the developing mammary gland. While the role collagen still remains poorly understood, it will be important to reconcile these studies with the work presented here, namely by demonstrating the relevance of the findings from this work to mammary tubulogenesis in vivo.

i) Although the Bissell and Nelson groups have used broadly similar experimental systems, they did not identify the important role for cell proliferation, as we have done. In particular, it was striking that proliferation in our experiments was increased preferentially in those regions of the epithelial aggregates that underwent elongation; and this appears to be stimulated by the local re-organization of the collagen 1 matrix. Such loco-regional stimulation of cell proliferation was not reported in earlier studies. For example, Brownfield et al. (Current Biology, 2013; Bissell group) showed that stretching collagen gels induces the elongation of embedded epithelial aggregates, but they focused on how this might occur through cell motility and did not test cell proliferation. As well, Gjorevski et al. (Scientific Reports, 2013; Nelson group) showed that aggregates can reorganize collagen by exerting tensile forces, but also focused their analysis on its impact on cell migration.

Thus, we think that the notion that collagen 1 reorganization can promote elongation by stimulating cell proliferation is a key message of our study that opens new avenues for the field. However, we did not present this with adequate focus in the original manuscript. So, we have thoroughly rewritten the revised manuscript to make this more explicit. We thank all the reviewers for prompting us to do a better job.

ii) We have expanded the Discussion to consider how our findings may be relevant for the more complex environment of tubulogenesis in vivo. Of course, our system is explicitly reductionist and does not capture the complexity of the ECM in vivo, especially in terms of other ECM components as well as a basement membrane. However, it is noteworthy that studies in vivo have reported important roles for collagen 1 during mammary tubulogenesis (e.g. Kelly et al., Differentiation 1995) and that the basement membrane often thins in regions of elongation e.g. Williams and Daniel, Dev Biol 97, 274, 1983; Silberstein and Daniel, Dev Biol 90, 215, 1982), making these more likely to engage with stromal collagen 1. We have included this in the Discussion as ways in which our findings may have impact in vivo. While it would be important to test these concepts in the more complex environment of a tissue, we respectfully felt that this was beyond the feasible scope of a revision.

2) Methods and conclusions from this work are similar to those reported by the Bissell group (Brownfield et al. Current Biology, 2013). In this study, unpolarized primary murine mammary explants were embedded within 3D collagen hydrogels and axially stretched using a deformable PDMS-based device. Collagen stimulated explant branching and outgrowth, which required collagen fiber polarization. The authors do reference this study but argue its focus is on branching rather than elongation. However, given the nearly identical methods, analyses, and morphogenic conclusions, it is unclear how this present work is distinct.

This is a very good point, one that indicates that we had not adequately focused the message of our manuscript. As noted above, although the experimental models may be similar, what is distinct in our findings is the role for loco-regional cell proliferation as a driver of epithelial elongation. In contrast, both Brownfield et al. (2013) and Gjorevski et al. (2013) focused on cell migration as the basis for how reorganizing collagen might affect the morphogenetic process of epithelial elongation. Recently, Buchmann et al. (Nat Comms, 2021) also focused on migration; they reported that proliferation occurred throughout their aggregates, but did not examine early stages of symmetry-breaking as we have done. So, we believe that our findings introduce a new driver for elongation. And we have fully rewritten the manuscript to make this clearer and more explicit.

3) "In contrast, Rac1 inhibitor treatment did not affect aggregate elongation and collagen fibril rearrangement (data not shown)". These data should be shown, as it conflicts with the conclusions of the Brownfield study. It should be noted that the Rac1 inhibitor NSC23766 only blocks a subset of Rac1 interactions with Rac1GEFs.

We have now included this data (Figure 2-Supplemental Figure 2). It should be noted that Brownfield et al. tested the potential role for Rac 1 by overexpressing a constitutively-active transgene. It is possible that the differences in our conclusions derive from our use of an inhibitor (albeit one that does not target Rac1 directly). Our data are also consistent with the findings of Gjorevski et al. (2013) and Buchmann et al. (Nat Comms, 2021), who reported that inhibiting ROCK disrupted the ability of cells to reorganize their surrounding collagen matrix.

4) From Figure 2A, is the gel coherency metric capturing fibril alignment, or the localized accumulation of collagen fluorescent intensity? Is collagen accumulation a function of cells pushing/expanding out into the matrix, or from active fiber recruitment to the outgrowth surface?

What determines whether localized collagen intensity leads to branching and outgrowth? In Figure 2A, there are two distinct regions of polarized collagen, but only one contains outgrowth. There is an overall variability in the brightness, contrast, and resolution of the representative collagen images, making it difficult to interpret across experiments. For instance, polarized collagen appears to be present in the blebbistatin treated outgrowths, however it is difficult to discern due to the different magnification and non-uniform hydrogel intensity.

i) We measured collagen coherency using Orientation J, which calculates the local orientation and isotropy for each pixel in an image based on the structure tensor for that pixel. The structure tensor is evaluated for each pixel of the given image by calculating the spatial partial derivatives by using (a cubic B-spline) interpolation. The local orientation and isotropy for each pixel are computed based on the eigenvalues and eigenvectors of the structure tensor.

We therefore used coherency as a measure of fibril co-alignment, which we interpret to reflect bundling, rather than purely collagen accumulation. It should be noted that we used this approach because our experiments required low-magnification, long working distance lenses to capture aggregates and their surrounding matrix. Therefore, we could not resolve individual collagen fibrils (except for the floating, cell-free gel in Figure 3D, where we could image close to the surface).

ii) At this point, our data don’t allow us to determine the cellular mechanism(s) that are responsible for the collagen reorganization. Given the slow time course of the phenomena, it is entirely possible that multiple mechanisms (including tugging forces and pushing forces) may contribute. With respect, we feel that it is somewhat peripheral to the key point of our study, which is how interplay between cell proliferation and collagen reorganization drives epithelial elongation. It will be important to elucidate this issue, but may be better suited to an independent study.

iii) Unfortunately, we can’t address the question of what may control the choice of branching vs elongation, because our analyses focused purely on the elongation process. In our experience, collagen 1 gels commonly show some variability in density (as reflected in differences in the intensity of labelling), even in the absence of cells. This prompted us to compare organization features of the ECM in regions of interest that were placed either adjacent to the parts of aggregates which showed elongation or those parts that did not.

5) Is proliferation truly driving this behavior? In the representative images (Figure 1C, 1H) it appears that all peripheral MCF10A making basal matrix contact are Ki67 positive. It is a bit difficult to tell when proliferation arresting inhibitors were added in the evolution of these processes. Does acute addition of mitomycin halt, or regress, established outgrowths?

i) In our original experiments, we added mitomycin C or aphidicolin when aggregates were first transplanted into collagen 1 gels, so proliferation inhibitors were present throughout the experiments. To test whether proliferation might be required to sustain elongation once it had begun, for this revision we added the drugs at day 3, when elongation had begun. Even though the aggregates had already broken symmetry, inhibition of proliferation significantly reduced further elongation, consistent with the idea that proliferation is necessary to sustain elongation. In the revised manuscript, we:

a) Clarify the timing for when proliferation inhibitors were added.

b) Include this new data.

ii) To summarize the evidence for a important role of stimulated cell proliferation, we find that:

a) There is a regional difference in cell proliferation that coincides with elongation. Greater proliferation was found in the regions of aggregates that elongate compared with those regions that do not elongate. We measured this by quantitating both Ki-67 staining and also counting cell divisions by live-cell imaging.

b) Stimulation of elongation by stretching gels was also associated with stimulation of cell proliferation in a loco-regional pattern that paralleled the elongation.

c) Both spontaneous and stretch-stimulated elongation were reduced by blocking cell proliferation. Together, we feel that these data argue for a necessary role of proliferation in driving elongation in our experiments. We don’t exclude the possibility that other processes, including polarized cell divisions, contributed to the elongation process. But they clearly did not compensate to sustain elongation when proliferation was blocked.

ii) The relationship between proliferation and cell migration is something that we have analysed in greater detail for this revision (prompted by comments from Reviewer 2). Of note, migratory cells were found throughout elongating aggregates, both in the regions that were elongating and in those that did not elongate. In control cultures, we found no regional difference in migration speed but in the elongating regions cells tended to be a little more persistent in their movement and oriented with the principal axis of the aggregate, which is perhaps to be expected. Nor was there any change in the regional patterns of cell speed when we inhibited proliferation with Mitomycin C or aphidicolin. Surprisingly, these proliferation inhibitors slightly decreased the persistence of movement and decreased the alignment with the principal axis of the aggregates. This suggested that features of coordinated collective migration were compromised when proliferation was inhibited. Therefore, by some means coordination of migration may also mediate the impact of regional cell proliferation on epithelial elongation.

6) Similarly, do stretch-induced elongations regress after removal from stretch and transfer to floating culture?

We tried to do this experiment, but it was technically unsuccessful. This is because the gels shrank when they were transferred into floating culture, making it difficult to visualize the aggregates adequately with our microscope set up.

7) AIIB2 implicates β1 integrin, but not α2β1. The authors should check to see if fibronectin deposition is stimulated by stretch, and is fibronectin present at stochastic outgrowths in non-patterned collagen?

i) Fibronectin: We found that fibronectin stained throughout the interfaces between aggregates and the collagen matrix. This appeared to be uniform in control aggregates that elongated spontaneously, without any evident regional variation in regions that elongated compared with regions that did not elongate (Figure 2—figure supplement 1B). Nor did we see any evident increase in fibronectin staining when cultures were stretched (Figure 3—figure supplement 2).

ii) A good point about AIIB2 implicating β1 integrin generally rather than α2β1 specifically: we’ve corrected our text accordingly.

8) "This regional patterning would be necessary if proliferation drives elongation, as a global increase in cell number would cause aggregates to expand isotropically." It is important to acknowledge that a critical regulatory feature missing from this model is a developed basement membrane. in vivo, the local degradation and remodeling of basement membrane mechanically permit region specific expansion of developing branched tissues, such as in the embryonic salivary gland (Harunaga et al. Dev. Dyn., 2014).

This is a good point and one that also speaks to how findings from this reductionist system might be translated into the complex environment of the native tissue (as discussed in our Response ii to point 1 above). Similarly, local thinning and loss of the basement membrane has been reported at the tips of elongating mammary tubules (e.g. Silberstein and Daniel, Dev Biol, 90, 215, 1982; Williams and Daniel, Dev Biol 97, 274, 1983), suggesting something similar may occur, which locally reduces the impact of the basement membrane, exposing the cellular compartment to collagen in the surrounding matrix.

9) "The decision to branch is recognized to be a critical checkpoint that is controlled by developmental signaling." It is unclear what distinction the authors are trying to make. There is robust literature suggesting cell nonautonomous mechanisms regulating branching events, notably localized extracellular matrix deposition and heterotypic cell-cell interactions.

Here we were trying to make the point that less is known about what regulates elongation, compared to what is known about branching. We’ve modified the text to make this clearer; it now reads:

“Tubulogenesis is a highly regulated phenomenon. The decision to branch is recognized to be a critical checkpoint that is controlled by developmental signals and cell-cell and cell-ECM interactions. The elongation of tubule precursors is also thought to be a regulated process that is controlled by receptor tyrosine kinases and other signaling pathways (Costantini and Kopan, 2010; Gjorevski and Nelson, 2010; Sternlicht et al., 2006).”

10) The aspect ratios of the images in Figure 5D appear to be off.

This reflects the different magnifications that were used to capture the full size of the aggregates (the one on top was much longer than that at the bottom). The scale bars, as printed look the same, but are different, and we omitted to define the scale bars in the legend. That was a source of confusion that we’ve corrected.

Reviewer #2:

The formation of mammary gland tubules begins as non-polarized cellular aggregates, called anlage, which elongate as solid multicellular cords that eventually form lumens. Katsuno-Kambe et al. investigate the role of the extracellular matrix during the process of elongation, making use of 3D culture models of mammary epithelial MCF10A cells. These cells have previously been shown (e.g Krause et al., Tissue Eng Part C Methods 2008, and others) to form multicellular aggregates that polarize and eventually form lumens when cultured in Matrigel, and when embedded in Collagen I (a major component of the stromal environment) form elongated, solid cords of non-polarized cells. Therefore, to capture the process by which aggregates break symmetry and become elongated, the authors first seeded MCF10A cells in Matrigel until proliferation arrested (resembling anlage), after which aggregates where isolated and embedded into collagen I gels. The authors show increased proliferation of cells in elongating areas of aggregates, and that these elongated aggregates are not formed when proliferation is inhibited. The authors further show that this elongation is accompanied by collagen reorganization into polarized bundles around elongating areas, and by artificially polarizing collagen by application of external stretch show collagen polarization can increase proliferation and guide the direction of elongation. Finally, the authors demonstrate aggregate elongation requires beta1-integrin adhesion and downstream ERK activity.

The strength of this manuscript is the use of a reductionist approach to investigate the role of collagen alignment in the elongation of epithelial anlage, by transplanting MCF10A cell clusters from matrigel into Collagen I gels, and artificially manipulating collagen polarization by stretching these gels. In addition, the exploration of the underlying mechanism of elongation, requiring proliferation and integrin-ERK signaling, is of potential great interest to understanding the role of cues derived from the extracellular matrix in epithelial morphogenesis.

The main weakness of this study is that the data still insufficiently supports several conclusions, in particular on proliferation being a key driver of anlage elongation, and on the reciprocal interaction between proliferation and collagen polarization during this process. The induction of proliferation is only analyzed 8 days after cell clusters have been transferred to Collagen gels, and thus after symmetry break and elongation have already occured. Analyses of proliferation and collagen polarization during the different stages of symmetry break and elongation, and detailed analyses of the distribution of proliferative cells during these stages, are lacking. This is in particular relevant as also in non-elongated areas, cells that are in contact with Collagen appear to be proliferative. Alternative driver mechanisms, for instance the induction of collective migration, are not fully tested and it remains unclear whether proliferation is merely required for, or truly a driver event of aggregate elongation.

The authors investigate a role for integrin signaling and downstream ERK in the induction of proliferation and aggregate elongation in response to collagen polarization. While an integrin β 1 inhibitory antibody and ERK inhibitor indeed prevents elongation, key unanswered questions are whether this pathway is indeed activated by polarized collagen during elongation (i.e. in selective regions of the cell clusters) and whether it is required for the proliferation response itself.

Despite the elegance of the reductionist approach in MCF10A cells in vitro, it remains unclear to what extent these findings generally apply to mammary tubulogenesis under physiological conditions in vivo, which remains largely undiscussed in the manuscript.

Comments for the authors:

1. The authors conclude that cell proliferation is a key driver of epithelial anlage elongation. However, the induction of proliferation is only analyzed 8 days after transfer of cell clusters to Collagen I, after symmetry break and elongation have taken place. And here, analysis of proliferation is limited to cells in the elongated areas versus the remaining cell cluster. It is essential to include analyses of proliferation during the different stages of elongation (i.e. before and upon initial symmetry break and at different time points of elongation), as well as detailed analyses of the distribution of proliferative cells during these stages. This is in particular relevant as from the images (e.g. Figure 1C) it is apparent that also in the non-elongated areas the cells on the outside of the cell cluster (thus in contact with Collagen I) are proliferating (e.g. Figure 1C), yet do not form elongated strands (which also does not support the authors conclusions that "proliferation is selectively increased in regions of aggregates that were elongating"). Finally, analysis of the induction of proliferation should also include a control in which cell clusters are transferred back to matrigel.

i) We have performed additional experiments to analyse proliferation early in the elongation process.

a) In order to assess when overall changes in proliferation became evident, we measured the proportion of cells that were Ki67-positive in the first 3 days of culture. An increase in Ki67-positivity was first evident on day 2 which was, interestingly, about a day before elongation was first detected.

b) To examine if there were locoregional differences in proliferation rates in these early aggregates, at day 3 we examined aggregates that had broken symmetry and begun to elongate (which we defined experimentally as a symmetry ratio of >1.5). We compared the proportion of Ki67-positive cells in the elongating regions of aggregates with the proportion found in the non-elongating parts of the aggregates. This showed a significant increase in the proportion of Ki67-positive cells in the elongating regions compared with the non-elongating regions. These new data lead us to conclude that loco-regional differences in cell proliferation are evident early in the elongation process, as well as later (as we had initially demonstrated).

ii) Furthermore, as the reviewer notes, cell proliferation was certainly seen in regions of aggregates that did not elongate. However, proliferation was greater in the elongating areas. So, it is the loco-regional difference in proliferation that seemed to be significant in our experiments. We have endeavoured to clarify this point in the revised manuscript.

iii) We also performed the control experiment where aggregates were extracted from Matrigel, then re-embedded in Matrigel. The data are shown in Figure 1—figure supplement 1C: we found that the aggregates did not elongate and, although Ki67 positive cells were evident, they appeared to be distributed randomly in the aggregates.

2. Alternative driver mechanisms, in particular cell migration, are limited to analysis of migration speed that is similar in different regions of the aggregate. This does not exclude that induction of directional, collective migration could be the main driving event underlying elongation, which remains unexplored. As such, it remains undetermined whether proliferation is merely required for, or truly a driver event of aggregate elongation.

We thank the reviewer for this suggestion and took the opportunity to measure a number of other parameters of cell motility for this revision (Figure 1—figure supplement 2B-H). In addition to cell speed, we also measured the straightness of the tracks (as an index of persistence) and the orientation of tracks relative to the principal axis of the aggregates. For these parameters we compared cells in the elongating regions of aggregates with cells in the non-elongating regions. There was no difference in migration speed between these two areas, but cells in the elongating areas migrated slightly more persistently and tended to orient themselves with the axis of elongation. Such an enhanced orientation of the migrating cells would be predicted to facilitate elongation.

Surprisingly, we found that, although it did not affect the speed of movement, inhibition of proliferation decreased the apparent persistence of movement and also caused cells to orient with less fidelity to the principal axis of the aggregates. Such changes in orderly migration could also have contributed to the defect in elongation. However, the implication is that these features of migration depend in some way upon cell proliferation. Our current data do not allow us to explain why this occurs. One possibility is that reorganization of the collagen matrix, which we show to be influenced by cell proliferation, influences cell movement. Polarized collagen could be regarded as a 3D micropattern that surrounds the elongating regions of the aggregates. Irrespective of the mechanism responsible, we suggest that this new data reinforces the concept that regional stimulation of proliferation is key to epithelial elongation in our system: and it may work both directly (by increasing cell number preferentially in the elongating areas) and also (directly or indirectly) by enhancing features of orderly cell migration that would also promote elongation.

3. The authors mention in the discussion that "the initial step of symmetry-breaking was associated with reorganization of collagen fibrils along the aggregate (…) which became increasingly bundled and aligned as elongation proceeded". However, data showing these changes in collagen during the different steps are not presented in the manuscript. This is in particular important as the authors conclude that collagen polarization induces proliferation, and therefore the authors should include a more detailed analyses of both processes (collagen reorganization and proliferation) over time after cell clusters are transferred to Collagen gels, and how they are associated with each other in individual cell clusters.

In addition, can the authors explain the induction of proliferation in cells in contact with collagen in the non-elongating areas (e.g. Figure 1C, see comment 1), as this suggest contact with collagen is sufficient to increase proliferation and collagen polarization is not required?

i) We have now examined collagen organization earlier in the elongation process (Figure 2E). For this, we analysed aggregates within the first 3 days of the assays. Because aggregates do vary in the time when they broke symmetry we subdivided them based on their symmetry ratio, comparing the coherency of the collagen gel around aggregates that remained rounded (symmetry ratio ~ 1) or around those with progressively greater symmetry ratios. We found that gel coherency increased (which we interpret as reflecting collagen bundling) with the increase in symmetry ratio. This confirmed that collagen reorganizes during the earliest stages of aggregate elongation.

ii) It is important to emphasize that cell proliferation was observed throughout the aggregates, both in regions that failed to elongate as well as those that did elongate. This implies that it is the regional difference is one of degree, not all or nothing. In other words, the key is what may increase proliferation in the areas of aggregates which elongate. Therefore, it becomes noteworthy that the regions that show increased proliferation are those which are adjacent to the more polarized collagen. This was evident in aggregates that broke symmetry spontaneously. Also, when we induced collagen polarization by stretching the gels, proliferation was stimulated preferentially in the elongating areas.

Thus, we conclude that the regional increase in cell proliferation associated with elongation is not being stimulated by contact with collagen per se, but rather by some additional property of the collagen gels that is associated with their reorganization. What that property may be is not something which we can answer at the present time; and we’d respectfully suggest that it may be more suitable to a separate, future project. Instead, we discuss some possibilities – including changes in stiffness of the collagen matrix – in greater depth and explicitness in the revised discussion.

4. In line with the above comment, the conclusion that collagen polarization promotes proliferation is largely based on the experiments in which collagen gels were stretched. However, the effect on proliferation is again only assessed following 8 days of culture in these gels, and thus after aggregate elongation. To support their conclusion, proliferation should be analyzed at early time points following stretching. This would also address whether the earlier symmetry break observed after stretching is a consequence of increased/faster induction of proliferation, or involves the ability of already proliferating cells to drive epithelial elongation (which will affect the interpretation of these data).

This was a good point and we have now done these experiments. We find that enhanced proliferation is detectable even one day after stretching. As shown in Figure 5A, proliferation begins earlier and to a greater degree in stretch-stimulated aggregates. This reinforces our interpretation that stimulation of proliferation is a driver of elongation, and we thank the reviewer for this suggestion.

5. The authors demonstrate that inhibition of proliferation prevents aggregate elongation. It is difficult to interpret from these experiments whether proliferation could truly be a key driver of elongation, as aggregates appear to remain much smaller following mitomycin C and Aphidicolin treatment (Figure 1H). It should be shown whether in the absence of the inhibitors, symmetry break and elongation is induced at this aggregate size (and thus whether the observed effect represents a direct or indirect of the inhibitors on elongation). In addition, proliferation inhibitors should also be given after the symmetry break has taken place, to test whether this prevents protrusions that are already present from elongating further.

a. We have observed small control aggregates (with sizes comparable to those of proliferation-inhibited aggregates) that break symmetry. However, aggregates tended to break symmetry at different times in culture (and therefore at different sizes). So, we have not endeavoured to quantitatively compare the sizes when symmetry broke in controls vs proliferation-inhibited cultures. Accordingly, we include representative images (Author response image 1) , but would be happy to add them to the manuscript if the reviewer and editor consider it to be warranted.

Author response image 1

b. To test the impact of inhibiting proliferation after symmetry had broken, we added drugs at day 3, when symmetry breaking had occurred (Figure 1M). Elongation was significantly inhibited, implying that cell proliferation contributes to sustaining aggregate elongation once symmetry has broken. This observation reinforces our interpretation that proliferation plays a critical role in the elongation process, rather than differences in aggregate size.

6. The authors investigate a role for integrin signaling and downstream ERK in the induction of proliferation and aggregate elongation in response to collagen polarization. While indeed both integrin beta1 and ERK activity are necessary for aggregate elongation, whether this pathway is directly involved in aggregate elongation remains unclear, and analysis of cell proliferation is lacking. The authors should test with available antibodies whether this pathway is specifically induced by polarized collagen (by analyzing different regions of the aggregate during symmetry break and elongation). In addition, effect of inhibitors on the level and distribution of Ki67+ cells should be shown.

We performed a number of additional experiments in response to these suggestions.

i) To determine if ERK was, indeed, a downstream element in an integrin-dependent pathway, we used the ERK location sensor to measure how ERK signaling was affected by the β1-integrin-blocking antibody, AIIB2, when elongation was stimulated by stretching of the gels. We found that the proportion of cells showing ERK activation was significantly reduced by AIIB2, both early (day 2) and late (day 7) in the assays (Figure 7H, I). Thus, we consider that β1-integrin is driving proliferation via ERK.

ii) To test if there were loco-regional differences in ERK activation, we measured the proportion of ERK-activated cells in the elongating and non-elongating areas of aggregates (Figure 6A-C). Consistent with what we had observed for proliferation, a greater proportion of cells showed ERK activation in the elongating areas, than in the non-elongating areas. This loco-regional difference was evident when we measured ERK activation in cells throughout the aggregates (Figure 6B) and when we only analysed those at the surface, which were in contact with the collagen (Figure 6C).

iii) We measured the effect on proliferation of inhibiting ERK with FR180204 (Figure 6D,E) or blocking β1-integrin with AIIB2 (Figure 7D). These new data show that proliferation was reduced by these inhibitors.

7. To control that effects observed in stretched collagen gels are truly mediated by the induced collagen polarization, the authors allowed gels to float in the medium to revert gel isotropy to that of unstretched gels. While collagen polarization is indeed reverted, cells cultured in floating gels will not able be able to generate stresses themselves (e.g. Halliday et al., 1995). From these experiments it can therefore not be concluded whether it is truly the collagen polarization or stresses generated by the cells that are needed for aggregate elongation.

This is a very good point, and one that indicates something which warrants clarification. We think that collagen polarization and stress-generation by cells are closely linked processes. When aggregates spontaneously break symmetry, we’d hypothesize that anisotropies in contractile tension are responsible for initiating collagen polarization. Conversely, the stretched gels that we use to exogenously polarize the gels are likely to apply stresses to the cell. Furthermore, as implied by the reviewer it is possible that cells generate different stresses when they are exposed to polarized compared with unpolarized collagen, and this “internal” stress is what is promoting proliferation.

However, the key point of our paper is that collagen polarization promotes elongation by stimulating cell proliferation in the areas of aggregates that are adjacent to the polarized collagen. The source of the stress that polarizes the collagen is less important, in this perspective, than the fact that polarization promotes proliferation. This was not clear in the original manuscript and we’ve endeavoured to correct this flaw in the revision.

8. It remains unclear to what extent the findings from this manuscript apply to mammary tubulogenesis in vivo, which remains undiscussed in the manuscript. While some experiments may be difficult to verify in vivo (e.g. to monitor collagen alignment), it would be relevant to at least to know whether in vivo mammary epithelial cells indeed come in contact with collagen I during anlage elongation, and preferably also show whether this is accompanied by a specific proliferation response.

Here we can draw guidance from the literature, which has reported that: (a) collagen is a major component of the ECM in morphogenetically-active mammary buds (Keely et al., Differentiation, 59: 1, 1995) and, also, (b) the basement membrane thins substantially at those growing tips, which would facilitate contact between cells and collagen, as reduce the mechanical impact of the basement membrane, compared with the surrounding ECM (Silberstein and Daniel, Dev Biol, 90, 215, 1982; Williams and Daniel, Dev Biol 97, 274, 1983). We’ve included discussion of this point in the final paragraphs of the Discussion, to highlight the physiological implications of our in vitro findings. With due respect, we feel that to extend our current observations into an in vivo model would be beyond the reasonable scope of a revision and better suited to a future study.

Reviewer #3:

Understanding tissue morphogenesis requires an understanding of underlying molecular and mechanical mechanisms. Many tissues form a network of branched tubules and understanding their origins is an important question for the field. Here the authors develop a simple, novel assay for exploring this. By plating MCF-10A mammary epithelia cell cysts in 3D collagen, they find that these cells rapidly reorganize, break symmetry, and send out streams of cells that proliferate as they move outward, forming solid cords. This behavior is particular for collagen and is not see in Matrigel. They then proceed to use this system to explore the causal links between collagen, proliferation, directed outgrowth and integrin signaling. They find that cell proliferation coincides with outgrowth and outgrowth is blocked in proliferation is inhibited. They find that collagen is re-organized as outgrowth occurs, altering fibril organization. They then carried out a set of clever experiments to use mechanical stretch to pre-orient fibrils, and found that aggregates preferentially elongated along the fibril preferred axis. Pre-oriented fibrils also sped elongation and proliferation. Finally, blocking integrin signaling reduced/delayed elongation. I found the new system interesting, and the data intriguing. However, I had two major concerns, about the measurements presented and the causality suggested.

1. Most of the experiments involve measuring collagen fibril orientation and coherency. However, in most of the images I could not see individual fibrils (the exception, Figure 3D, was strikingly different than the rest). I would like to see a clearer description in the main text of how orientation was measured and why this is difficult to see in most images.

We appreciate this point of clarification. Individual collagen fibrils were difficult to image in our experiments because low magnification imaging was necessary, for two reasons: (a) We were endeavouring to capture whole aggregates and their surrounding matrix in 3D videos; and (b) higher magnification lenses did not provide the depth of focus needed to image aggregates deep within the gels. (On this point, the fibrils are evident in Figure 3D because this floating gel allowed us to access the gel surface. In contrast, when gels were embedded – including after they had been stretched – then we had to image deeper.)

Accordingly, we analysed collagen organization using Orientation J, which calculates the local orientation and isotropy for each pixel in an image based on the structure tensor for that pixel. The structure tensor is evaluated for each pixel of the given image by calculating the spatial partial derivatives by using (a cubic B-spline) interpolation. The local orientation and isotropy for each pixel are computed based on the eigenvalues and eigenvectors of the structure tensor.

2. Bigger picture was the issue of causality. The authors present seemingly a linear view – integrin signaling drives proliferation and outgrowth and this triggers fibril orientation that then feeds back into the process. However, I found causality difficult to discern. While proliferation is necessary for outgrowth, its role seemed to be more likely to be permissive rather than instructive. The observation that cells would follow oriented fibrils was interesting but perhaps not surprising, and I found it difficult to tease out whether elongation triggered polarization of collagen or vice versa – I guess the authors favor a feedback process.

In revising the manuscript, we have taken the opportunity to perform additional experiments and also to present the data and our interpretation with greater clarity. In particular, we argue that polarized collagen exerts its impact on elongation by promoting proliferation in the parts of the cell aggregate that are adjacent to the polarized collagen. Our new data show that:

a) When aggregates elongate spontaneously, cell proliferation is detectable ~ 24 hours before elongation. And even in these early stages proliferation is greater in the elongating parts of aggregates, rather than in the non-elongating parts of the aggregates. Since inhibitors of proliferation reduce elongation, these correlations suggest that selective increase of proliferation within elongating areas is necessary for that elongation to occur.

b) Cell proliferation is also stimulated when we induce collagen polarization extrinsically, by stretching the gels (Figure 5). Strikingly, we find that an increase in cell proliferation occurs earlier with stretch-polarization than when aggregates are allowed to break symmetry and polarize spontaneously. And, as seen with spontaneous elongation, proliferation is greater in the elongating areas of the aggregates. We take this as evidence that polarization of collagen can stimulate cell proliferation in a loco-regional fashion.

c) We have performed further analyses to characterize cell migration within aggregates in greater detail. Interestingly, we find that although inhibitors of mitosis did not affect the speed of cell migration, they disrupted the directionality of migration. Cells treated with aphidicolin or Mitomycin C did not orient with the principal axis of the aggregates as well did control cells. This suggests that an effect on directional migration may have contributed to the disruption of epithelial elongation that occurs when proliferation is inhibited.

We propose that these data identify the capacity for polarized collagen to stimulate cell proliferation in the parts of the aggregates proximate to the polarized gel. And this is necessary for elongation to occur effectively. We have endeavoured to make this point more clearly in this revision than we did earlier, because we consider this to be a new insight into how elongation occurs. Earlier studies, using experimentally-similar systems, principally focused on how matrix organization might affect cell locomotion to promote elongation. Migration is unarguably important, but our data suggest that it is not enough. Migration did not compensate when proliferation is inhibited and, indeed, may be supported by proliferation in some way.

Finally, we agree that feedback is likely to be important when aggregates break symmetry and extend spontaneously. We found that collagen polarization was reduced when cell proliferation was blocked, suggesting that polarization may be initially induced by some mechanical change associated with proliferation. We discuss this briefly in the revision. Possibly, local variations in cell proliferation induce anisotropies of force to being to polarize the collagen. However, this is largely speculative at the present time: more detailed temporal and spatial analysis that compares proliferation with elongation and collagen polarization will be needed to answer this question. We respectfully suggest that these experiments would make the subject of a separate study in themselves.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

We have received comments from two of the previous reviewers, both of whom indicate the manuscript is strengthened by the addition of new experiments, text clarifications, and discussion of results in context. The reviewers request that specific representative images be added. Notably, the reviewers each agree that it is important to:

1) Provide clarification on how quantification of ERK biosensor activity was performed. Consider selecting a more representative image or using high magnification insets.

In our quantitation, we used nuclear localization of the ERK biosensor as our measure of ERK activation. This was not made clear in the text and we have now made it more explicit in this revision.

And to assist the reader, we have included higher magnification views of Figure 6a, to identify cells that we have scored as “active” (i.e. with nuclear exclusion) as well as those that we scored as “inactive” (i.e. where there was both nuclear and cytoplasmic staining). This scoring criterion is also explicitly mentioned in the Results and Figure caption.

2) Provide clarification on how local regions of proliferation were calculated in 3D aggregates (i.e. did it include all cells, or only cells in contact with collagen). If needed, include a direct analysis of cells in contact with collagen in non-elongating vs elongating regions to support the conclusions.

For proliferation we measured the proportion of cells which were Ki-67 positive throughout the regions of interest. In this revision we further analysed only the cells that were in direct contact with the collagen. This confirmed the result that we obtained when all cells were analysed – i.e. the proportion of proliferating cells was greater in the elongating regions than in the non-elongating regions. This data is now included in Figure 1—figure supplement 2A, corresponding to Figure 1C – which is where we establish the basic observation that proliferation differs in elongating vs non-elongating regions of aggregates.

Reviewer #1:

The authors have taken steps to substantially clarify the core findings from their paper, driving home their central conclusion that regional proliferation stimulated by polarized collagen directs aggregate elongation. Importantly, the authors also distinguish the conclusions of this study from prior work that apply similar methods. They have added additional experiments that address most of my concerns regarding the role of proliferation in this process and altogether the manuscript is stronger.

Where I am still not clear is the relationship between collagen and the stimulation of proliferation. This conclusion is derived from reported regional differences in cell proliferation that coincide with elongation and collagen polarization. However, this quantification could be potentially misleading, as there is also a significant difference in the number of cells with surface area contacting the surrounding collagen in non-elongating regions yet those that do appear to be proliferative (Figure 1C, 1J). Further not all instances of accumulated collagen appear to lead to elongation.

To pursue this, we further analysed the data set of Figure 1C and J to measure proliferation just in the cells that were in contact with the collagen. We found that the proportion of cells which were Ki-67, was greater in regions of aggregates that were elongating, compared to the non-elongating regions. This confirmed our earlier analysis that encompassed all the cells in the respective regions. It further reinforces our conclusion that the rate of cell proliferation was greater in the elongating regions of aggregates.

This data (which extends the analysis in Figure 1C and 1J) is now included in Figure 1—figure supplement 2A.

The authors clearly demonstrate (Figure 5) that polarized collagen influences the kinetics of proliferation and elongation. Perhaps this can be explained by a slightly refined model in which collagen is sufficient to stimulate cell proliferation in anlagen and its concurrent polarization/remodeling provides boundary conditions that together support the emergent behavior of elongation?

This is a very good idea, which we have happily stolen for the final paragraph of our discussion (lines 542-544):

“The consequent loco-regional collagen polarization then stimulates further cell proliferation in the adjacent parts of the aggregate to sustain elongation. Biochemical contact with collagen may provide a basal stimulus for cell proliferation, which is further enhanced where collagen becomes polarized to provide a boundary condition that promotes elongation.”

https://doi.org/10.7554/eLife.67915.sa2

Article and author information

Author details

  1. Hiroko Katsuno-Kambe

    Division of Cell and Developmental Biology, Institute for Molecular Bioscience, The University of Queensland, Brisbane, Australia
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Validation, Writing – original draft
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0292-3506
  2. Jessica L Teo

    Division of Cell and Developmental Biology, Institute for Molecular Bioscience, The University of Queensland, Brisbane, Australia
    Contribution
    Conceptualization, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Robert J Ju

    Division of Cell and Developmental Biology, Institute for Molecular Bioscience, The University of Queensland, Brisbane, Australia
    Contribution
    Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9850-9803
  4. James Hudson

    QIMR Berghofer Medical Research Institute, Brisbane, Australia
    Contribution
    Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
  5. Samantha J Stehbens

    Division of Cell and Developmental Biology, Institute for Molecular Bioscience, The University of Queensland, Brisbane, Australia
    Contribution
    Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8145-2708
  6. Alpha S Yap

    Division of Cell and Developmental Biology, Institute for Molecular Bioscience, The University of Queensland, Brisbane, Australia
    Contribution
    Project administration, Supervision, Writing – original draft, Writing – review and editing
    For correspondence
    a.yap@uq.edu.au
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1038-8956

Funding

National Health and Medical Research Council (Fellowship 1136592)

  • Alpha S Yap

National Health and Medical Research Council (GNT1123816)

  • Alpha S Yap

National Health and Medical Research Council (1140090)

  • Alpha S Yap

Australian Research Council (DP19010287)

  • Alpha S Yap

Australian Research Council (190102230)

  • Alpha S Yap

Australian Research Council (FT190100516)

  • Samantha J Stehbens

Snow Medical Fellowship

  • James Hudson

Uehara Memorial Foundation (Postdoctoral fellowship)

  • Hiroko Katsuno-Kambe

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Selwin Wu for advice and all our lab colleagues for their feedback and support. The authors were supported by the National Health and Medical Research Council of Australia (Fellowship 1136592 and GNT1123816, 1140090 to ASY), Australian Research Council (DP19010287, 190102230 to AY and FT190100516 to SS), a Snow Medical Fellowship to JH, and a postdoctoral fellowship from The Uehara Memorial Foundation to HK.

Senior Editor

  1. Jonathan A Cooper, Fred Hutchinson Cancer Research Center, United States

Reviewing Editor

  1. Matthew Kutys, University of California, San Francisco, United States

Reviewer

  1. Matthew Kutys, University of California, San Francisco, United States

Version history

  1. Received: February 26, 2021
  2. Preprint posted: February 28, 2021 (view preprint)
  3. Accepted: October 13, 2021
  4. Accepted Manuscript published: October 18, 2021 (version 1)
  5. Version of Record published: October 27, 2021 (version 2)

Copyright

© 2021, Katsuno-Kambe et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,511
    Page views
  • 240
    Downloads
  • 4
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Hiroko Katsuno-Kambe
  2. Jessica L Teo
  3. Robert J Ju
  4. James Hudson
  5. Samantha J Stehbens
  6. Alpha S Yap
(2021)
Collagen polarization promotes epithelial elongation by stimulating locoregional cell proliferation
eLife 10:e67915.
https://doi.org/10.7554/eLife.67915

Further reading

    1. Cell Biology
    2. Developmental Biology
    Simon Schneider, Andjela Kovacevic ... Hubert Schorle
    Research Article

    Cylicins are testis-specific proteins, which are exclusively expressed during spermiogenesis. In mice and humans, two Cylicins, the gonosomal X-linked Cylicin 1 (Cylc1/CYLC1) and the autosomal Cylicin 2 (Cylc2/CYLC2) genes, have been identified. Cylicins are cytoskeletal proteins with an overall positive charge due to lysine-rich repeats. While Cylicins have been localized in the acrosomal region of round spermatids, they resemble a major component of the calyx within the perinuclear theca at the posterior part of mature sperm nuclei. However, the role of Cylicins during spermiogenesis has not yet been investigated. Here, we applied CRISPR/Cas9-mediated gene editing in zygotes to establish Cylc1- and Cylc2-deficient mouse lines as a model to study the function of these proteins. Cylc1 deficiency resulted in male subfertility, whereas Cylc2-/-, Cylc1-/yCylc2+/-, and Cylc1-/yCylc2-/- males were infertile. Phenotypical characterization revealed that loss of Cylicins prevents proper calyx assembly during spermiogenesis. This results in decreased epididymal sperm counts, impaired shedding of excess cytoplasm, and severe structural malformations, ultimately resulting in impaired sperm motility. Furthermore, exome sequencing identified an infertile man with a hemizygous variant in CYLC1 and a heterozygous variant in CYLC2, displaying morphological abnormalities of the sperm including the absence of the acrosome. Thus, our study highlights the relevance and importance of Cylicins for spermiogenic remodeling and male fertility in human and mouse, and provides the basis for further studies on unraveling the complex molecular interactions between perinuclear theca proteins required during spermiogenesis.

    1. Cell Biology
    2. Structural Biology and Molecular Biophysics
    Bronwyn A Lucas, Benjamin A Himes, Nikolaus Grigorieff
    Research Advance

    Previously we showed that 2D template matching (2DTM) can be used to localize macromolecular complexes in images recorded by cryogenic electron microscopy (cryo-EM) with high precision, even in the presence of noise and cellular background (Lucas et al., 2021; Lucas et al., 2022). Here, we show that once localized, these particles may be averaged together to generate high-resolution 3D reconstructions. However, regions included in the template may suffer from template bias, leading to inflated resolution estimates and making the interpretation of high-resolution features unreliable. We evaluate conditions that minimize template bias while retaining the benefits of high-precision localization, and we show that molecular features not present in the template can be reconstructed at high resolution from targets found by 2DTM, extending prior work at low-resolution. Moreover, we present a quantitative metric for template bias to aid the interpretation of 3D reconstructions calculated with particles localized using high-resolution templates and fine angular sampling.