Design of hydrogels with features of human breast tissue
Because we were interested in engineering a 3D scaffold that could stimulate the growth of human breast tissues, we explored hydrogel formulations that contained protein and glycosaminoglycan components found in the ECM of human breast tissue. We focused our efforts on ECM hydrogels with defined components and evaluated various hydrogel formulations by assessing their ability to support the growth of primary human breast tissue fragments. The seeded tissue fragments contained 50–100 cells per fragment and were harvested by dissociating breast tissues from patient reduction mammoplasties (Additional file 2: Figure S1).
Through heuristic optimization, we established a novel hydrogel formulation that supported the growth of human breast tissues (Fig. 1a). These hydrogels had several features that were important for supporting breast tissue growth: (1) They were fabricated with collagen, fibronectin, and laminin, three ECM proteins present in human breast tissue in vivo [25]; (2) they incorporated hyaluronic acid, a glycosaminoglycan polysaccharide present in many human tissues, including the breast; (3) they were loaded with three growth factors (insulin, epidermal growth factor, and hydrocortisone) that support the growth and differentiation of mammary epithelial cells [21, 26, 27]; and (4) they were cultured in suspension [6, 28, 29].
To assess their physical properties, we compared the swelling ratio and elasticity (Young’s modulus) of the ECM hydrogels with those of gels consisting only of collagen. The ECM hydrogels exhibited a significantly higher swelling ratio than collagen gels (306.94 ± 6.29 vs 290.10 ± 0.81, p < 0.01). This difference was likely due to inclusion in the ECM hydrogels of hyaluronans, which bind and retain water. Atomic force microscopy (AFM) indicated that the ECM hydrogels had significantly increased elasticity relative to collagen gels (Young’s modulus of 256.7 ± 20.0 Pa vs. 559.2 ± 204.0 Pa, p < 0.05) (Additional file 2: Figure S2). One potential explanation for this difference is that the additional ECM components in the hydrogels could be partially disrupting the efficiency of collagen polymerization, thereby resulting in a more elastic hydrogel that more closely approximated the elasticity of breast tissue in vivo [30]. These findings indicated that the ECM hydrogels were more elastic with increased water content relative to collagen-only gels.
ECM hydrogels support the growth of complex breast tissues
When seeded into the ECM hydrogels, primary mammary epithelial cell clusters isolated from reduction mammoplasties rapidly grew into complex breast tissues with a seeding efficiency of approximately 33 % (Fig. 1b, c). The majority of breast tissues that expanded in the hydrogels had complex ductal and lobular morphologies that closely resembled the epithelial structures present in the human breast (67 %) (Fig. 1b). The expanded breast tissues exhibited similar morphologies across all seven of the patient samples that were assessed. In contrast, and consistent with prior findings [31], there was minimal or no outgrowth when primary mammary cells were seeded either into collagen-only gels or into basement membrane with or without additional ECM components (Fig. 1b and c; Additional file 2: Figure S3a). The rare outgrowths that formed in collagen-only gels were either thin ducts or spheres, whereas the outgrowths in basement membrane were spheres with some ruffling at the edges.
The efficiency of tissue formation was much lower when single primary epithelial cells were seeded into the hydrogels (0.16 %), when compared with the efficiency observed with primary cell clusters (33 %). Moreover, only 4.5 % of the organoids derived from single cells exhibited the complex ductal and lobular morphologies that were exhibited by the majority of tissues grown from primary cell clusters. Significantly, the few single-cell–derived organoids with complex morphologies contained only cytokeratin 14 (CK14)-positive basal cells and did not contain CK8/18-positive luminal cells (Additional file 2: Figure S4). On an absolute scale, 0.0075 % of single cells gave rise to tissues with complex ductal and lobular morphologies, whereas 26 % of primary cell clusters gave rise to tissues with complex ductal and lobular morphologies (Additional file 2: Figure S4). The structures that formed from single cells were primarily thin ducts (83.6 %) and, less frequently, simple lobules (11.9 %). These findings indicated that single cells can form topologically complex structures at a low frequency, but the resulting structures contained only one of the two major cell lineages present in the mammary gland in vivo and in the tissues grown from primary cell clusters. In light of these observations, we focused subsequent experiments on growing tissues from primary cell clusters.
Tissues exhibit morphological response to hormones
We next assessed if the human mammary tissues grown in hydrogels responded to steroid, pituitary, and lactogenic hormones, which are known to stimulate the development of mammary epithelial tissue in vivo. Staining the hydrogel tissues for hormone receptors revealed that approximately 5 % of the cells expressed each of the estrogen and progesterone receptors (ER and PR, respectively) (Additional file 2: Figure S5). Treatment of the hydrogels with estrogen and progesterone stimulated the mammary tissues to hollow, resulting in the formation of ducts and lobules with evident lumens (Fig. 1d, Additional file 2: Figure S5). This suggested that estrogen and progesterone were promoting further maturation of the mammary tissues.
Extracts of the pituitary gland contain factors important for mammary development, including growth hormone, fibroblast growth factors, and follicle-stimulating hormone [32, 33]. Consistent with this, addition of pituitary extracts to the ECM hydrogels caused a significant increase in both secondary and tertiary ductal branching of the expanded breast tissues (Fig. 1e). Moreover, addition of both pituitary extract and prolactin further stimulated lobular expansion with a fourfold increase in lobular volume accompanied by the formation of large lipid droplets that were visible upon hematoxylin and eosin staining (Fig. 1e and f, Additional file 2: Figure S5).
Kinetics of tissue growth and maturation in hydrogels
To examine the kinetics with which these tissues matured, we captured bright-field images over a span of 8 days, beginning at the earliest time point at which we observed ductal outgrowths (day 4) (Fig. 2a and Additional file 2: Figure S6). These primary ductal outgrowths gave rise to secondary and tertiary ducts over the next week, either through bifurcation of elongating ducts or through side branches that sprouted from ducts. After 8–12 days of tissue growth, there was a rapid increase in the number and size of lobules (Fig. 2a and b).
The primary cells seeded into hydrogels were initially disorganized clusters with intermixed basal (CK14+) and luminal (CK8/18+) cells (Fig. 2c). However, by 7 days, the cells had self-organized into an outer CK14+ basal layer with some CK8/18+ luminal cells in the interior of the expanding tissues (Fig. 2c, center, Additional file 2: Figure S7A, Additional file 3: Movie S1). At this early time point, the majority of newly initiated ducts were small and exclusively composed of CK14+ basal cells. However, as the organoids expand and mature, CK8/18+ luminal cells can be seen lining their interior (Fig. 2c, right, Additional file 2: Figure S7B). In all patients, at least 60 % of the mature organoids contained distinct luminal and basal layers.
By 21 days, there was clear evidence of tissue maturation, with the lobule interiors staining strongly for both the luminal lineage marker GATA3 and the luminal differentiation marker MUC1 (Fig. 2d). At this time, some of the lobule interiors also showed evidence of cavitation (Fig. 2d). Fully mature structures expanded to sizes of up to 3 mm in diameter (Additional file 2: Figure S8) and remained viable for at least 8 weeks in culture in the same hydrogel. During this time, the developing and expanding tissues radically remodeled and condensed the hydrogels in which they were cultured, with evidence of this condensation up to 2 mm away (Fig. 2a and Additional file 2: Figure S6, S9). After 3–6 weeks, the organoids fully expanded to the size of the condensed pad and were unable to grow further. However, these structures could be removed from the hydrogels by enzymatic digestion and reseeded into new hydrogels, which support their continued growth (Additional file 2: Figure S3B).
Prior studies of the morphogenesis of mouse mammary organoids have indicated that the process of ductal initiation and elongation involves a dynamic reorganization of cells within 3D cultures [9]. To assess if this was also occurring in our primary human organoids, we stably labeled the primary cell clusters with fluorescent proteins before seeding them into the hydrogel scaffolds. Because the fluorescent proteins were delivered by lentivirus at a low multiplicity of infection, it was possible to assess the contributions of individual clones and their progeny to the formed mammary tissues. Using this approach, we found that the progeny of individual clones were dispersed throughout the tissue structures rather than being localized to clonal patches (Fig. 2e). This suggested that cells underwent dynamic rearrangements as they proliferated to grow tissues. Time-lapse movies also showed dynamic rearrangements. Mass cell migrations could be seen in the organoid cores, along ducts, and also within terminal ductal lobular units (TDLUs) (Additional file 4: Movie S2, Additional file 5: Movie S3, and Additional file 6: Movie S4).
Mammary stem cell behavior in ductal initiation and maturation
To identify putative mammary stem cells (MaSCs), we performed immunofluorescence staining for the transcription factors SLUG and SOX9, which, when coexpressed, mark MaSCs in the murine mammary gland [34]. SLUG+/SOX9+ cells were rarely seen within the core and ducts of organoids, but they made up roughly half of the cells in the TDLUs. These TDLUs were typically five to eight cells thick, and the layer of cells in direct contact with the ECM (termed the cap region) was most enriched for the dual-positive cells, with roughly two-thirds of cells coexpressing SLUG and SOX9 (Fig. 3a, e). In both ducts and lobules, the dual-positive cells were enriched in the cap region of the expanding outgrowth, in direct contact with the ECM, suggesting that this contact could be involved in maintaining stem cells in an undifferentiated state (Fig. 3b, e).
To assess the topological properties of these organoids, we rendered a surface model from 3D confocal microscopic images and used a Dimension Elite 3D printer (Stratasys, Eden Prairie, MN, USA) to fabricate a high-resolution 1500X scale physical model of an organoid stained for filamentous actin (Fig. 3c). Examination of this physical model revealed that the outgrowths containing the highest fraction of SLUG+/SOX9+ cells were also the shortest. When side-branches started to form, nearly all of the cells were dual-positive, but as the ducts elongated, there was a gradual decrease in the fraction of dual-positive cells (Fig. 3d). This suggested that side branches were initiated by the proliferation of SLUG+/SOX9+ cells, which subsequently differentiated to give rise to interior cells concurrently with ductal elongation.
SLUG+/SOX9+ leader cells direct ductal elongation
Examination of the printed 3D model also revealed the presence of small tips at the leading edges of elongating ducts. Confocal microscopy showed that these tips contained one or two leader cells that were polarized in the direction of ductal elongation. The leader cells stained positively for filamentous actin and protruded from the structures in the direction of ductal elongation (Fig. 4a and b). The leader cells expressed basal cytokeratins (Fig. 4b) and coexpressed SLUG and SOX9 (Fig. 4a). While the majority of outgrowths contained one leader cell, occasionally outgrowths contained multiple leader cells in different orientations (Fig. 4c).
Time-lapse microscopy provided additional insights into the relationship between these leader cells and ductal elongation. Ductal elongation was always preceded by a transient extension of leader cells that physically engaged with and deformed the ECM (Fig. 4d and e, Additional file 4: Movie S2, Additional file 7: Movie S5). At times, the force of this interaction between leader cells and the matrix caused them to break away from the ducts and become isolated in the matrix (Additional file 4: Movie S2). The direction in which the leader cells extended was always the direction of the next wave of ductal elongation. When the direction in which the leader cells emanated was different from the previous direction of elongation, the ducts reoriented in the new direction specified by the leader cells before the next wave of elongation (Fig. 4f, Additional file 8: Movie S6). This ductal reorientation appeared to be induced by the collective rotation of cells in the lobule, which occurred before ductal elongation (Additional file 8: Movie S6). After the ducts reoriented, they elongated for a period of time, after which the elongation ceased. After ductal elongation ceased, new leader cells emanated from the ductal tips to initiate the next cycle of elongation.
A prior study of murine mammary organogenesis indicated that ductal elongation is driven not by leader cells but rather through the collective expansion and migration of luminal cells [9]. Because this prior study was conducted using Matrigel, the discrepancy with our observations could be due either to differences in the types of 3D scaffolds used or to differences in biology between mouse and human mammary cells. When mammary tissue fragments from C57BL/6 J mice were seeded into our ECM hydrogels, they grew and ruffled but did not exhibit any leader cell activity (Additional file 2: Figure S3). This finding suggested that leader cells may play a role specifically in the morphogenesis of human mammary tissue and not that of mice.