Matrix Composition Modulates Vitamin D3's Effects on 3D Collagen Fiber Organization by MCF10A Cells

Abstract

Vitamin D3 (vitD3) has been implicated in various cellular functions affecting multiple tissue types. Epidemiological and laboratory studies suggest that vitD3 may be effective as a preventive or therapeutic option for breast cancer. However, randomized clinical trials have yet to confirm these suggestions. Breast neoplasias can arise from developmental alterations; based on this evidence, we seek to understand vitD3's role in normal breast development, particularly its role in epithelial morphogenetic processes such as ductal elongation, branching, and alveolar formation. These processes require extensive changes in the extracellular microenvironment, such as collagen fiber organization, and are largely influenced by hormones. Here, we build upon our past work to shed light on calcitriol's effects on collagen fiber organization by breast epithelial cells, and how such effects are modulated by extracellular matrix composition. We embedded MCF10A normal human breast epithelial cells in two different matrices—collagen type I and collagen type I + 10% Matrigel; treatment with calcitriol resulted in flatter epithelial structures. Next, using two-photon microscopy, we examined changes in collagen fiber organization and corresponding changes in epithelial structures. Applying a novel three-dimensional (3D) image analysis method, we show that increasing doses of calcitriol result in denser collagen fiber bundles in the localized area surrounding the epithelial structures, and that these bundles are aligned in a more parallel direction to epithelial structures when exposed to the highest vitD3 dose. Changed patterns in fiber organization may explain the flattening of epithelial structures; in turn, changes in biophysical forces in the matrix abutting these structures may be responsible for changes in the referred patterns. Addition of 10% Matrigel dampened the effects of calcitriol on both epithelial morphogenesis and fiber organization. Overall, we report novel functions of calcitriol in the breast epithelium and add to the growing body of evidence documenting how hormones affect biophysical processes.

Impact statement

In this study, we report novel functions of calcitriol in the breast epithelium and use a novel quantitative metric to parse the effects of calcitriol on collagen fiber organization that cannot be detected through conventional histological procedures. Despite the large body of literature on vitamin D3 (vitD3) and calcitriol's effects on cellular functions across tissue types, little is known about how they affect collagen fiber organization, an early critical step for breast epithelial development. This work provides further evidence that hormones affect morphogenesis by means of biophysical forces, with implications for a comprehensive view on vitD3's effects in breast development and neoplasia.

Introduction

Vitamin D3 (vitD3) and its active metabolite calcitriol (1,25-dihydroxyvitamin D3) are thought to act as a steroid hormone by affecting a host of cellular functions in multiple tissue types. While vitD3 has traditionally been studied in the context of bone development, more recently, vitD3's role in breast cancer has been explored as a potential therapeutic and/or preventive agent. However, randomized clinical trials have not provided conclusive evidence in its favor, despite promising results from preclinical models.¹ This discrepancy can be attributed to the underlying premises framing these studies, as well as technical limitations. Regarding the latter, epidemiology is considered, first, a blunt tool with which to find statistically significant differences between the experimental and the control arms, and second, the preclinical studies may be based on models that are far removed from the clinical situation.

Regarding theoretical issues on carcinogenesis, the current hegemonic tenets of the somatic mutation theory are cell-centered and rely on DNA mutations as the causal event. Conversely, the Tissue Organization Field Theory (TOFT), which posits that cancer is a tissue-based disease arising from aberrant stromal–epithelial interactions, frames carcinogenesis as development gone awry and therefore argues for a developmental origin for breast cancer.^2–4 Altogether, evidence from both epidemiological and laboratory studies support this latter assertion. For example, women exposed to the pesticide dichlorodiphenyltrichloroethane (DDT) in utero and during puberty had significantly higher risk of breast cancer later on in life.^5–8 Similarly, laboratory studies on endocrine disrupting chemicals (EDCs) revealed that rats exposed perinatally to bisphenol-A exhibited higher incidences of mammary carcinomas compared to unexposed controls^9–12; perinatal exposure in mice resulted in altered mammary gland development and preneoplastic lesions.¹³ Taken together, exposure to estrogenic EDCs during morphogenesis alters mammary gland development and increases the propensity to neoplastic development. This outcome justifies investigating the role of another hormone, namely, vitD3’s, which modulates estrogenic effects, to better understand its potential role as a therapeutic and/or preventive role against breast cancer.¹⁴

Several stages of mammary gland development in rodents, such as puberty, pregnancy, and involution, provide informative models comparable to those of human breast development.¹⁵ At these stages, the mouse mammary epithelium manifests major morphological changes: for instance, during puberty, the rudimentary branching structure expands to fill out the fat pad, and throughout pregnancy, the branched ducts generate grape-like clusters called alveoli, that is, the structures that produce milk. These features are largely influenced by the main mammotropic hormones estrogen, progesterone, and prolactin.¹⁶ These structures regress postweaning, leaving behind a ductal tree histologically similar to that of a prepregnancy state, but functionally different.¹⁷ These changes are due to a complex interplay among a host of factors present in the stroma and the epithelium.

Welsh and colleagues established that in mice lacking vitamin D receptor (VDR), mammary glands show accelerated development and a florid phenotype during puberty and pregnancy.^18,19 Conversely, in mice lacking CYP24A1 (the key calcitriol degradation enzyme) in the mammary epithelium, the mammary glands exhibited stunted development during puberty.²⁰ It should be noted that in both cases, when there is no VDR (VDR KO) or when vitD is not degraded (CYP24A1 KO), the resulting phenotypes are observed at the tissue level of biological organization, that is, changes are observed in tissue morphology and not at the cellular level since when observed through immunostaining, the proliferation index of mammary epithelial cells remains unaffected.^19,20

In the rodent mammary gland stroma, fibrillar collagen (collagen type I) is the most prevalent structural protein; individual mammary acini and ducts are frequently circumscribed with 10–100 μm or larger bands of organized collagen-rich fibrillar stroma.²¹ The “dynamic reciprocity” between epithelial cells and structural proteins in the stroma are critical for development and disease.²² Changes in collagen fiber organization and the extracellular matrix (ECM) are required for the morphogenetic changes in the mammary epithelium to take place^21,23–30; similar findings have been associated with malignancy in breast tumors.^21,30–34 The significance of the ECM, contained within the stroma, in neoplastic development is better understood under the tenets of TOFT, which posits that neoplasms arise due to alterations in stroma–epithelial interactions.^35,36

Changes in fiber organization and the ECM are also related to the physical forces that act on the tissue during morphogenesis.^{21,27,37–39} Previous work reported decreased expression levels of nonepithelial cadherins, α6 and β4 integrins, and α-smooth muscle actin in human breast cancer cell lines treated with calcitriol in two-dimensional (2D) culture conditions.⁴⁰ In a separate study, calcitriol treatment increased expression of F-actin and E-cadherin and decreased vimentin expression in MCF7 and MDA-MB-231 human breast cancer cells.⁴¹ In addition, VDR KO mice showed high levels of α-SMA and P-cadherin in the mammary gland⁴⁰; calcitriol also decreased tenascin C expression in mouse and human mammary epithelial cells in 2D culture conditions.⁴² However, evidence regarding the effects of calcitriol on the biomechanics of morphogenesis or on fiber organization remains scarce. While there is no data on collagen fiber density in mammary glands from VDR KO mice, calcitriol treatment reduced fibrosis in the liver while VDR KO mice spontaneously developed hepatic fibrosis.^18,43 In addition, in a model of pancreatitis associated with pancreatic adenocarcinoma, calcitriol reduced fibrosis by acting through stromal cells.⁴⁴

The use of traditional 2D cultures to answer questions about morphogenesis has proved to be inadequate when considering that development occurs in a complex three-dimensional (3D) environment of biological organization. Moreover, in vivo conditions do not allow experimenters to exercise a tight control over the multiple factors involved in the changes of matrix composition, fiber organization, and other constraints determining the organization of the mammary epithelial compartment. Therefore, we considered preferable to investigate calcitriol's effects in a 3D environment that recapitulates as closely as possible in vivo conditions and allows for a controlled manipulation to tease out specific effects in a quantitative manner. To this purpose, our laboratory has developed 3D culture models to study breast morphogenesis in vitro that recapitulate ductal and acinar structures as observed in vivo.^26,39,45,46 We have also used a novel quantitative metric of 3D collagen fiber organization (3D variance) to characterize subtle remodeling changes in a hormone sensitive 3D breast tissue model.^28,47 Using such models, we have shown that calcitriol restrains estrogen's effects on mammary epithelial morphogenesis, and also that calcitriol exerts estrogen-independent effects that are conveyed through changes in collagen fiber organization.¹⁴ Here, we build upon our previous results and provide a quantitative and more comprehensive view of calcitriol's estrogen-independent effects on collagen fiber organization during epithelial morphogenesis and identify previously undescribed effects of calcitriol in the human breast epithelium.

Methods

Cell maintenance and culture reagents

Hydrocortisone, cholera toxin, insulin, phosphate-buffered saline (PBS, 10 × ), and calcitriol were purchased from Sigma-Aldrich (St. Louis, MO). DMEM/F12, L-Glutamine and Trypsin were purchased from Life Technologies (Carlsbad, CA). Equine serum was purchased from Thermo Fisher (Waltham, MA). Epidermal growth factor (EGF), growth factor reduced Matrigel, and rat tail type I collagen were purchased from Corning (Tewksbury, MA). Calcitriol was resuspended in ethanol to make 10⁻³ and 10⁻⁴ M stocks, respectively, and stored under nitrogen at −20°C. MCF10A cells used in the study were purchased from American Type Culture Collection (Manassas, VA) and maintained in our laboratory in DMEM/F12 with phenol red, 5% equine serum, 20 ng/mL EGF, 0.5 μg/mL hydrocortisone, 0.1 μg/mL cholera toxin, and 10 μg/mL insulin. Experiments performed using MCF10A cells used the same medium. In all cases, cells were kept in a humidified environment at 37°C with 6% CO₂.

3D cultures

3D gels were formulated with a final collagen concentration of 1 mg/mL as described previously²⁶; for mixed gels, 10% Matrigel was added into the collagen solution before self-assembly. Cells were suspended in the gel solution at a density of 75,000 cells per gel, and 1.5 mL of mixture (per well) was poured into 12-well plates. The collagen mixture was allowed to self-assemble into hydrogels for 30 min at 37°C, and 1.5 mL of media was added to each well. Gels were detached as previously described.²⁷ Cultures were maintained for 2 weeks after which imaging and morphological assessments were performed. Calcitriol was administered every 3 days when cell culture media was replenished. After 2 weeks, gels were harvested and processed as described previously.²⁶ For morphological assessments, gels were fixed with 10% phosphate-buffered formalin overnight, and either embedded in paraffin for histological analyses or whole mounted and stained with carmine alum. Gel diameter was measured postfixation using a ruler and before further processing.

Whole mount analysis

Whole-mounted gels stained with carmine alum were visualized with a Zeiss Axioskop 2 microscope. For morphometric analyses, images were captured using a Zeiss LSM 800 confocal microscope. In brief, ∼1 mm² area of the gel periphery was imaged to a depth of ∼100 μm. Complete images were reconstructed in ImageJ by stitching tiles and compiling z-stacks together; resulting images were analyzed using Software for Automated Analysis (SAMA).⁴⁸ Statistical analyses of morphometric parameters were performed using GraphPad Prism software.

Immunocytochemistry

For F-actin staining, gels were harvested at the 2-week timepoint and fixed with freshly prepared 2% paraformaldehyde in PBS and permeabilized with 0.2% Triton X-100 in PBS for 30 min. Following PBS washes, gels were incubated with rhodamine-labeled phalloidin (1:100 dilution in PBS; Cytoskeleton, Inc., CO) for 1 h at room temperature, protected from light. Gels were then washed with PBS and counter-stained with DAPI (2.5 μg/mL; Sigma-Aldrich), mounted with antifade mounting medium (0.02% n-propyl gallate in 80% glycerol) and visualized with Zeiss LSM 800 confocal microscope. For Filamin-A staining, 5 μm thick formalin-fixed paraffin embedded gel sections were rehydrated in xylene, serial ethanol dilutions (100%, 95%, and 70%), and DI water. Antigen retrieval was achieved by microwave treatment in 10 mM, pH 6.0 citrate buffer; sections were then blocked in normal goat serum in 1.5% milk for 1 h at room temperature before incubating with mouse antihuman Filamin A antibody (PM6/317 clone; Thermo Fisher) overnight at 4°C. Biotinylated secondary antibody was applied to sections for 1 h in a humidified chamber at room temperature the following day. Slides were washed with PBS, treated with streptavidin conjugated horseradish peroxidase, and positive detection was accomplished using DAB (Sigma-Aldrich, MO). Samples were counterstained with Harris' hematoxylin, dehydrated, mounted with a permanent mounting medium, and imaged using Zeiss Axioskop 2 microscope.

Two-photon microscopy

Second harmonic generation (SHG) images of collagen fibers and two-photon excitation fluorescence (TPEF) images of epithelial structures in 3D gels were obtained at 920 nm excitation using a Leica TCS SP8 confocal microscope equipped with a tunable (680–1300 nm) fs laser (Insight DS+, Spectra Physics, CA). Non-descanned, low-noise Leica hybrid detectors collected signals in the 460 ± 20 nm (SHG) and 525 ± 25 nm (TPEF) ranges; a water-immersion 40 × objective (NA 1.10, 650 μm working distance) was used for imaging. Images with a size of 512 × 512 pixels (pixel size 284 × 284 nm) were acquired at an interval of 1 μm along the z-axis. Images were reconstructed in 3D using ImageJ software.

Collagen fiber organization analysis

Collagen fiber density and 3D directional variance were assessed based on TPEF/SHG images of MCF10A gels, both collagen-only and mixed gels. The 3D directional variance of collagen fibers within a volume varies between 0 and 1, with these extreme values representing fibers that are totally aligned and disorganized, respectively. The localized 3D variance with the SHG 3D stacks was calculated for voxels, 7 × 7 × 5 pixels (3.88 × 3.88 × 5 μm in size), using an algorithm which has been described in detail elsewhere (related MATLAB code is available here—http://ase.tufts.edu/biomedical/research/Georgakoudi/researchNonDestructive.asp).⁴⁷ Next, to define the borders of the epithelial structures, we relied on the TPEF images, which were processed using a combination of a Sobel edge detection and Otsu's thresholding filters. The regions surrounding the cells were segmented in two different ways to characterize different aspects of 3D collagen organization, as shown in Supplementary Figure S1A-(1) we defined concentric disks extending 10–20 μm each, up to a distance of 63 μm from the border, following the selection of a disk region extending 3 (or 6) um from the border of an epithelial structure (isolated), and (2) we defined shells of increasing widths (from 3 to 63 μm) from the epithelial border (cumulative). For analysis, we selected 12 acini per treatment group (0, 10, 25, 50 nM calcitriol) across 3 biological replicates, and in two different matrix conditions (collagen, collagen +10% Matrigel).

Statistical analyses

Statistical analyses were performed using GraphPad Prism, SPSS, and MATLAB. One-way analysis of variance followed by Tukey's post hoc test was used to determine differences in the number of elongated structures in MCF10A mixed gels and gel diameters, collagen fiber density, and 3D directional variance across treatment groups. Kruskal-Wallis test with post hoc Dunn's multiple comparison test was performed on morphometric data obtained from MCF10A mixed gels using SAMA; Mann–Whitney U-test was used for comparisons between collagen only and mixed gels. For all statistical tests, results were considered significant at p < 0.05.

Results

Matrix composition alters organization of MCF10A cells in 3D culture

Previous studies revealed that cells embedded in mixed matrices (collagen type I plus Matrigel) organize differently than those containing collagen only.^25,39,45 As shown in Figure 1A and B, we observed that addition of 10% Matrigel to rat tail type I collagen (mixed gel) resulted in less spherical and more elongated MCF10A epithelial structures compared to structures growing in only-collagen (col) containing 3D cultures. We also observed that the elongated structures in the mixed gels (col +10% Matrigel) were smaller and thinner compared to col gels; they showed a cord-like morphology, whereas the elongated structures in col gels exhibited tubular/ductal morphology.

FIG. 1.

Matrix composition alters MCF10 epithelial organization and response to calcitriol. (A) MCF10A cells form less spherical and more elongated structures in collagen +10% Matrigel (col+M; mixed) matrices compared to collagen only (col) matrices, as analyzed using SAMA, *p < 0.05, Mann–Whitney U-test; error bars = SEM. (B) Representative cross-section z-stack images of MCF10A acinar structures observed in col (top) and mixed/col+M (bottom) gels; scale bar = 20 μm. (C) (Left) Representative images of carmine alum stained whole mounts of col+M MCF10A gels showing elongated structures (arrowheads, magnified representatives in insets; scale bar = 100 μm); (right) quantification of elongated structures showing a decrease in number in response to calcitriol treatment (n = 8), *p < 0.05 compared to 0 nM, one-way ANOVA, Tukey's post hoc; error bars = SEM. (D) Morphometric analyses of MCF10A structures in mixed (col+M) gels treated with calcitriol showing increased flatness, decreased sphericity, and reduced volume in a dose-dependent manner, *p < 0.05 compared to 0 nM, **p < 0.05 compared to 50 nM, Kruskal-Wallis; error bars = min to max. Analyses were performed after 2 weeks of culture. ANOVA, analysis of variance; SAMA, Software for Automated Analysis; SEM, standard error of the mean. Color images are available online.

Calcitriol affects MCF10A organization in mixed gels

MCF10A cells treated with different doses of calcitriol in mixed gels showed both similarities and differences when compared to the outcomes in col gels. In the col gels, the number of elongated MCF10A structures changed in a dose-dependent manner, with the highest numbers observed at 10 nM calcitriol with decreasing numbers at higher doses.¹⁴ In the mixed gel model, we observed that treatment with calcitriol also resulted in a reduction in the number of elongated structures at all doses tested; however, there was no statistically significant differences between the doses used (Fig. 1C). Analysis of the structures using confocal images of whole mounted gels and of our in-house software SAMA showed that calcitriol treatment resulted in flatter structures than those observed in col gels, but, again, without significant differences between the doses (Fig. 1D).⁴⁸ Calcitriol treatment also resulted in a decrease in sphericity and volume of the structures (Fig. 1D), but in a dose-dependent manner—the 50 nM dose group showed significantly less sphericity and volume compared to the 10 and 25 nM groups.

Calcitriol affects contractile properties of the mixed gels in dose-dependent manner

Similar to col gels that exhibited increased contraction upon addition of calcitriol, mixed gels also showed significant contraction as evidenced by the reduction in gel diameter (Fig. 2A).¹⁴ This reduction occurred in a dose-dependent manner—gels treated with the 25 and 50 nM doses had a greater reduction in diameter compared to those in the 10 nM dose group. Considering that contraction of these gels is directly related to the organization of collagen fibers by MCF10A cells, we performed second harmonic imaging microscopy to visualize collagen fibers in the mixed gels (Fig. 2B).^14,25 We noticed that the 25 nM calcitriol concentration showed the most intense collagen signal in this matrix, which could explain the significant difference in gel diameter between the 10 and 25 nM groups.

FIG. 2.

Calcitriol affects contractile properties of mixed (col+M) gels in a dose-dependent manner. (A) MCF10A mixed gels show increased contraction in response to calcitriol treatment, (n = 8, *p < 0.05 compared to 0 nM, ^#p < 0.05 compared to 50 nM, one-way ANOVA); error bars = SEM. (B) Representative SHG images showing collagen fibers in green and MCF10A acini in red. (C) Visualization of actin filament organization in MCF10A acini in mixed matrix using rhodamine-labeled phalloidin. (D) Immunostaining for Filamin A showing dark positive stain along cellular borders (arrowheads). Scale bars = 50 μm. Analyses were performed after 2 weeks of culture. SHG, second harmonic generation. Color images are available online.

The cellular cytoskeleton exists in a functional and physical continuum with the external environment; thus, any change in the physical environment affects the organization of the filamentous actin network in the cell. Concomitantly, visualization of F-actin network using rhodamine-labeled phalloidin showed more pronounced actin staining in the periphery of the epithelial structures (Fig. 2C). Filamin A, which serves as a receptor for integrins that cells use to manipulate collagen fibers, also connects to the actin cytoskeleton; immunostaining for Filamin A revealed a pronounced staining at the periphery of the epithelial structures (Fig. 2D). Taken together, the data collected provide a basis for the extrapolation that calcitriol may alter the physical forces experienced by MCF10A cells in the mixed gel matrix and their consequent reactions.

Effects of calcitriol on collagen fiber organization in col-only gels

Effects of MCF10A cells on collagen I fiber organization have been reported a few hours postseeding.²⁵ In addition, hormones act on cells to change how collagen fibers are organized^26,28; we previously reported a qualitative assessment of calcitriol's effects on collagen fiber organization in col gels.¹⁴ Here, we performed a quantitative analysis of fiber density and alignment in calcitriol-treated col-only gels. Since the majority of the epithelial structure population consisted of acini, we chose to perform our analysis on such structures. We utilized a previously described model that measures the density of fibers and 3D directional variance within each voxel of the image.^28,47 Using this model, we measured the two parameters in the perimeter of the structures (0–3 μm radius) and in close proximity to the structures (3–23 μm). In both cases, calcitriol treatment significantly increased fiber density in a dose-dependent manner. Increasing calcitriol concentrations led to lower directional variance, indicating more aligned fibers in the perimeter of the structures (0–3 μm) (Fig. 3); however, in the proximity of the structures (3–23 μm), the opposite was true: namely 25 and 50 nM groups showed higher variance, that is, more disorganized fibers (Fig. 3).

FIG. 3.

Effects of calcitriol on collagen fiber density and alignment in col only gels. (Top) Calcitriol treatment results in increased fiber density at both the perimeter (0–3 μm) and local (3–23 μm) regions of MCF10A acini (n = 8, *p < 0.05 compared to 0 nM, one-way ANOVA). (Bottom) Calcitriol decreases fiber variance, that is, increases fiber alignment in perimeter (0–3 μm) region but increases fiber variance (decreased alignment) in the local region (3–23 μm) (n = 8, *p < 0.05 compared to 0 nM, one-way ANOVA). Error bars = SEM. Analyses were performed after 2 weeks of culture.

To discern whether the fiber organization patterns were also observed at larger distances away from the examined epithelial structures, we measured 3D directional variance and fiber density up to a radius of 63 μm from each structure. The two different masking approaches (isolated vs. cumulative disks) were applied to parse any location-dependent differences in fiber density and orientation (Supplementary Fig. S1A). In both cases, fiber density decreased with distance, but the overall differences between the treatment groups persisted, with higher calcitriol doses exhibiting increased density (Supplementary Fig. S1B). For 3D directional variance, the isolated approach showed that variance increased sharply between 3 and 23 μm for all treatment groups, before plateauing out up to 63 μm (Supplementary Fig. S1C). While the variance was lower in higher calcitriol doses initially (0–3 μm), the differences between treatment groups did not persist in the isolated masks. In the cumulative approach, we saw that the variance steadily increased over the distance; the treatment groups showed a difference in the mean variance—the higher the calcitriol dose, the larger the mean variance (Supplementary Fig. S1C).

Effects of calcitriol on collagen fiber organization in mixed gels

Previous work from our laboratory has shown that addition of small amounts of Matrigel (as low as 5%) changed some gel properties, impairing the cells' ability to interact with collagen fibers and formed a coating around acinar structures.²⁵ We therefore investigated the interaction between changes in matrix composition and additions of calcitriol at the fiber organization level. Acini from mixed gels were imaged using TPEF, and fibers were imaged using second harmonic imaging microscopy; fiber organization was assessed using the same algorithm as above. However, due to an increased thickness of the collagen ring around the acini in mixed gels, the radii of analysis were changed for these gels—for the perimeter of the structure, a radius of 6 μm was used and for regions proximal to the structures, a 6–23 μm radius was used; distal regions (>53 μm) away from the structures were also analyzed. Calcitriol treatment in this model did not show a significant difference in fiber density in the perimeter (0–6 μm) of the structures, but we did observe an increasing trend in fiber density with increasing calcitriol concentrations in the local proximity (6–23 μm); this effect was statistically significant at the 50 nM dose (Fig. 4A). At the distal (>53 μm) regions, an overall increase in fiber density was observed that did not change with calcitriol concentrations (Fig. 4A). For 3D directional variance, we observed that at 10 nM calcitriol dose, there was a statistically significant increase in variance in both the proximal (6–23 μm) and distal (>53 μm) regions; fiber variance was also increased in those regions in the 25 and 50 nM doses, but the increase was not statistically significant (Fig. 4B). When measured over distance with isolated and cumulative masks as done for col gels, we observed a similar phenotype; namely fiber density decreased with distance while fiber variance increased with distance in both the isolated and cumulative approaches (Supplementary Fig. S2). While the effect of calcitriol doses on fiber density in mixed gels showed similar trends to those in col gels (Supplementary Fig. S1B), the impact in mixed gels appeared smaller, especially in the isolated mask approach.

FIG. 4.

Effects of calcitriol on collagen fiber density and alignment in mixed (col+M) gels. (A) Calcitriol increases fiber density only at 50 nM in the local region (6–23 μm) and shows an increased density trend in the far region (>53 μm) of MCF10A acini. (B) Calcitriol increases fiber variance, that is, decreases alignment, at both local (6–23 μm) and far regions (>53 μm) of MCF10A acini. *p < 0.05, compared to 0 nM, one-way ANOVA. Error bars = SEM. Analyses were performed after 2 weeks of culture.

Effects of calcitriol on fiber orientation

To get a more detailed view of the orientation of collagen fibers, we analyzed the distribution of the polar (φ) and azimuth (θ) angles, which are used to depict an orientation in 3D space, within 23 μm of the epithelial structures. Mean distributions for these angles from each treatment group are shown in Figure 5.⁴⁷ As shown in Figure 5A, calcitriol treatment in col gels has a nonmonotonic effect on the distribution of the polar angle—the 10 nM dose shows a lower and less sharp peak at 90° and a wider spread compared to 25 and 50 nM doses and the control (0 nM). With increasing doses, calcitriol resulted in an increase in the % of voxels (peak) at ∼90°; 50 nM has the highest peak and the narrowest spread compared to the other groups. The azimuth angle distribution also changed between the calcitriol doses and the control (0 nM)—in the control, there was a sudden sharp peak ∼65°; the spread at 10 and 25 nM doses were similar, but at 50 nM, there was a shift of the peak of the spread to 65°.

FIG. 5.

Effects of calcitriol on spatial orientation of collagen fibers in local regions of MCF10A acini. Mean azimuth (θ, green) and polar (φ, blue) orientation distributions of collagen fibers of representative 2D frame from each image stack analyzed within 23 μm from the edge of the epithelial structure in (A) collagen only (col) and (B) mixed (col+M) gels. Differences between polar (φ) angle distributions are compared by calculating full width half maxima of distribution curves in both (C) col and (D) mixed (col+M) gels (*p < 0.05 compared to 0 nM, one-way ANOVA, n = 12 per treatment group). Differences between azimuth (θ) angle distributions in (E) col and (F) mixed (col+M) gels are compared by calculating the 0–90:90–180° area under the curve ratios (*p < 0.05 compared to 0 nM, one-way ANOVA, n = 12 per treatment group). Error bars = SEM. Analyses were performed after 2 weeks of culture. 2D, two-dimensional. Color images are available online.

In the mixed gel, the distribution of polar angle was affected similarly, but the effect was not as pronounced as in the col gels (Fig. 5B). We observed a decrease in the peak height (% of voxels) at ∼90° in calcitriol doses; at 50 nM dose, the distribution was broader at ∼90° compared to the sharp single peak noticed in 0, 10, and 25 nM groups.

To compare the differences in polar (φ) angle distributions, we calculated the full-width half maxima for each treatment group in both col (Fig. 5C) and col+M (Fig. 5D) gels. Although the differences were not statistically significant, we observed a decreasing trend with increasing calcitriol dose in col gels; in mixed gels, a decrease was only observed at the 50 nM dose. These observations support our previous observations regarding the spread of the distribution at high calcitriol doses in both matrices. Similarly, we compared the azimuth (θ) angle distributions in both gel types by calculating the ratio between the areas under the curve for 0–90 and 90–180° (0–90:90–180). In col gels (Fig. 5E), we saw a significant decrease in the ratio in the 50 nM dose, indicating the skewness of the distribution toward the left. In mixed gels (Fig. 5F), no apparent differences were observed.

The sharp peak of the polar angle at ∼90° denoted that the majority of fibers in the region were aligned in a primarily parallel direction to the optical sections. The broader distributions observed for both angles denoted a more random alignment. When compared to the data presented in Figures 4 and 5B, we observed congruence between the distribution of orientation angles and the corresponding variance for each treatment group. For example, the 10 nM treatment group had a higher variance/more random alignment in the mixed gel (Fig. 4B), which correlated with the shorter polar angle peak at ∼90° (Fig. 5B). The lack of significant differences in fiber density and variance observed in calcitriol-treated mixed gels compared to the control was in concordance with the angle distribution data; in contrast, the angle distribution patterns in col gels (control and treated) were consistent with the dramatic differences observed in these gels.

Discussion

Matrix composition is a key determinant of breast epithelial morphogenesis. Understanding of the role of its many components requires a careful manipulation of the known and suspected protagonists that dynamically affect the phenotypes observed along the development of this gland. By reducing the number of variables affecting the in situ phenotypes of the breast, in culture experiments help the observer to more narrowly focus attention on the role of a more manageable number variables. For instance, addition of Matrigel to collagen type I matrices changes the type and frequency of structures formed.^25,39,45 In this study, we presented evidence showing how matrix composition influences calcitriol's effects on breast epithelial morphogenesis. In accordance with what we reported previously, we observed that MCF10A cells mostly formed acini and some elongated structures in a rat tail col type I matrix and the mixed matrix in 3D cultures; however, the elongated structures in the mixed gels differed morphologically from those in the col gels.^14,27 The effects of calcitriol were markedly different in the two matrices—in the col matrix, we observed a non-monotonic response with regard to the frequency of elongated structures, but in the mixed gels, we noticed that calcitriol resulted in a decreased number of elongated structures without any significant difference among the doses tested (Fig. 1C).¹⁴ We did observe similar phenotypes between the two matrices upon calcitriol treatment, although the extent of the effects differed—we observed contraction of the gels and flattening and reduction of the size of acini with the highest effect at the 50 nM dose. Addition of Matrigel resulted in gels with larger diameters overall and the extent of contraction was less obvious compared to col gels.¹⁴

The current study adds to our growing understanding of how hormones influence the ability of cells to organize collagen fibers and, reciprocally, how collagen fibers influence epithelial organization. Previously, we highlighted the necessity for collagen fibers to become organized as an early step for breast epithelial morphogenesis.^25,26,49 We have also shown that the mammotropic hormones (estrogen, progesterone, prolactin) affect the cells' ability to organize fibers differentially in concordance with the structures they form under the influence of these hormones.^26,28 Here, we show that the effects of calcitriol on fiber organization vary depending on matrix composition, subsequently resulting in the morphological differences observed in MCF10A structures. We utilized second harmonic imaging to visualize collagen fibers in both gel types and used a 3D variance metric to analyze fiber orientation given its high sensitivity compared to 2D and other existing metrics.²⁸

Effects on collagen fiber density

Similar to our previous work on hormone-sensitive breast cancer cells, here we examined calcitriol's effects on fiber density, variance and orientation angle distribution in an estrogen-receptor negative normal breast cell line.²⁸ The density measurements showed that calcitriol treatment of MCF10A cells in col I gels increased fiber density in the perimeter (0–3 μm radius) and local proximity (3–23 μm) of the epithelial structures. Fiber density decreased with distance; however, the rate of decline varied among treatments (Supplementary Fig. S1B). Treatment with increasing calcitriol concentrations resulted in a decrease in the rate of decline. Using the isolated mask approach, we found that untreated controls had the steepest decline until 33 μm (away from the structure); 10, 25, and 50 nM treatments showed smaller decreases in the rate of decline in a dose-dependent manner. We noted that in the respective treatment groups, the fibers were the least dense at 33 μm away from the MCF10A structures. Past this point, the density steadily increased—the rate of increase was also higher in the untreated group. In the cumulative approach, the graph line for 50 nM group was almost flat when density was measured against distance. Data from whole mounts and gel contraction measurements published previously can explain this observation—epithelial structures are overcrowded in higher calcitriol concentrations, and picrosirius staining showed a more homogeneous distribution of bundles of fibers in 25 and 50 nM treatment groups.¹⁴

In the mixed gels, we observed a similar pattern in the decline of density and the increase in variance over distance in both the isolated and cumulative approaches (Supplementary Fig. S2). However, the differences between the treatment groups and control in the rates of change were less pronounced when compared to those in the col gels (Supplementary Fig. S1). Previous data from our laboratory showed that Matrigel, at concentrations as low as 5% changed the fibrillar collagen matrix to a globular one; increased concentrations (at 50%) made the globular matrix more compact.²⁵ We also observed that Matrigel at 5% formed a localized coating in areas surrounding acini and protrusions from cells were observed only in areas that were not masked by Matrigel. Thus, the impaired ability of cells to interact with the collagen fiber network in the mixed gels may explain the differences in epithelial morphology we observe here. We confirmed this in the mixed gels—at 50 nM calcitriol treatment, we observed the highest density in the proximity (6–23 μm) of the structures (Fig. 4A). Corroborating the idea that presence of Matrigel impairs cells' ability to interact with collagen fibers, we only noticed a statistically significant increase in fiber density at the 50 nM calcitriol in the mixed gels compared to the significant increase observed in 10, 25, and 50 nM doses in the col gels (Fig. 3A). This may also explain the expanded ring we observed around acini in the mixed gels during SHG imaging, which is why we utilized a larger radius for analysis compared to the col gels.

Increased collagen fiber density has been attributed to both normal development and malignancy, further cementing the need to view carcinogenesis as development gone awry. Ductal extension in the mouse mammary gland during puberty, under the influence of estrogen, require denser bundles of collagen I fibers along the flank of the nascent duct where cell proliferation is inhibited.²⁴ Our current findings that collagen fiber density is increased in the localities of epithelial structures, paired with our earlier findings that calcitriol reduces cell yield in MCF10A 3D cultures, add to the understanding of a concerted series of actions that result in a directional growth of a developing mammary epithelium.¹⁴

Effects on fiber organization as measured using 3D variance

Hormones act on cells to change collagen fiber organization.^26,50 Here, we noticed that in addition to estrogen, progesterone, and prolactin, calcitriol directly affects how cells organize collagen fibers. We used 3D variance as a measure for organization—a lower variance indicates a higher alignment of fibers. Compared to the density measurements, the variance data showed a more nuanced effect of calcitriol. Calcitriol appeared to lower variance (higher alignment) in the perimeter of the structure while increasing variance (lower alignment) in the local proximity of the MCF10A structures in col gels (Fig. 4B). The isolated disk approach showed that the increase in variance plateaued at 33 μm away from the structures; this is also the inflection point for the density-distance curves (Supplementary Fig. S1B). However, the variance did not change past 33 μm up to 63 μm. The variance in the untreated group posed a contrasting view because in this group, the variance was higher than in calcitriol-treated groups in the perimeter but was lower in the 3–23 μm range. From the variance data, it can be inferred that calcitriol-treated cells organize fibers in a more aligned manner in the perimeter of the structure while increasing overall density of fibers.

In the mixed gels, we were unable to analyze fiber variance in the perimeter of the structure, possibly due to the coating of Matrigel, as explained above. The overall trend in increase in variance both in the proximal and distal regions in the mixed gels treated with calcitriol (Fig. 5B) may explain the reduction in the number of elongated structures compared to control; elongation requires parallel alignment of fibers, that is, reduced variance, in the local region of the structures.²⁵ The lack of statistical significance in some of the observed trends, in both fiber density and variance, described above may be explained by the variations observed within the images used for analyses. Collagen heterogeneity, and in 50 nM calcitriol doses, the presence of more than one acinus in the field of view due to increased contraction of the gels, may have contributed to such variations.

During development, aligned collagen fibers radiate out from the cap cells of the terminal end buds (TEBs) before invasion; these aligned fibers allow for cell migration and advancement of TEBs.^24,51 Similarly, radially organized collagen fibers along tumor boundary allow for tumor cells to migrate.³² Thus, fiber alignment is important for both normal and neoplastic development. Our current findings indicate that calcitriol facilitates local organization of fibers (lower variance in the proximal areas), but not distant ones (higher variance in distal regions), thus presenting additional constraints on a developing epithelial structure.

Effects on fiber orientation in 3D

The 3D variance metric used for our analysis depends on measurements of two angles, polar and azimuth, used to describe orientation in 3D space. As discussed in the Results section and in previous work, we observed that acinar structures in untreated MCF10A gels (both col and mixed) take on a flatter, and more oval shape with increasing calcitriol concentrations, similar to the flattened shapes observed in T47D collagen gels treated with estrogen+promegestone (E2+Prmg).^14,26,28 Liu et al. attributed the heterogeneity observed in overall fiber alignment in E2+Prmg treated T47D gels to the changes observed in the polar angle rather than the azimuthal angle. In our analysis, we observed pronounced changes in the polar angle distribution at the proximal regions of acini in col gels (Fig. 5A); the pattern of azimuthal angle distribution shows only slight differences between the treatment groups. In the mixed gels, the changes in the angle distribution are insignificant (Fig. 5B)—this may be due to the impaired ability of the cells to interact with collagen fibers in the mixed gels. This also may explain the lack of a pronounced effect of calcitriol treatment on epithelial morphology in mixed gels. However, further investigation is required to elucidate the processes underlying this phenomenon.

Conclusion

Collagen fiber density and organization in the mammary gland epithelium varies spatially and temporally. More importantly, as explained above, the fiber alignment at the perimeter and local level are most probably the major contributors. Our data reflect only certain aspects of how calcitriol might affect fiber organization by acting on epithelial cells at a given time. No literature on the effect of vitamin D and/or calcitriol on collagen fiber organization is available—gross histological assessment of mammary glands in VDR KO mice at puberty did not show any differences in overall connective tissue content but these researchers did not investigate collagen fiber organization.¹⁸ As we have shown above, however, calcitriol affects fiber orientation and density, especially in a localized manner, which, in turn, affects epithelial morphogenesis. Calcitriol's effects on fiber organization are further modulated by matrix composition, possibly by altering the ability of cells to adhere to and manipulate collagen fibers in their vicinity. Our current findings, along with our previously published work generates a more coherent, but still incomplete, understanding of how each hormone can differentially affect collagen fiber organization by acting on the epithelial cells during mammary gland development; it necessarily warrants more detailed investigations in the future. Altogether, these findings pave the way to a more refined understanding of how vitD3 mitigates breast cancer risk within the context of carcinogenesis as development gone awry.

Disclosure Statement

No competing financial interests exist.

Funding Information

This work was supported by the National Institute of Environmental Health Sciences grant ES030045 (AMS), and the NIH Research Infrastructure grant S10 OD021624. The content does not necessarily represent the official views of the funding agencies.

Supplementary Material

Supplementary Figure S1

Supplementary Figure S2