Tumor size is strongly correlated with breast cancer metastasis and patient survival. Increased tumor size contributes to hypoxic and metabolic gradients in the solid tumor and to an aggressive tumor phenotype. Thus, it is important to develop three-dimensional (3D) breast tumor models that recapitulate size-induced microenvironmental changes and, consequently, natural tumor progression in real time without the use of artificial culture conditions or gene manipulations. Here, we developed size-controlled multicellular aggregates (“microtumors”) of subtype-specific breast cancer cells by using non-adhesive polyethylene glycol dimethacrylate hydrogel microwells of defined sizes (150–600 μm). These 3D microtumor models faithfully represent size-induced microenvironmental changes, such as hypoxic gradients, cellular heterogeneity, and spatial distribution of necrotic/proliferating cells. These microtumors acquire hallmarks of tumor progression in the same cell lines within 6 days. Of note, large microtumors of hormone receptor–positive cells exhibited an aggressive phenotype characterized by collective cell migration and upregulation of mesenchymal markers at mRNA and protein level, which was not observed in small microtumors. Interestingly, triple-negative breast cancer (TNBC) cell lines did not show size-dependent upregulation of mesenchymal markers. In conclusion, size-controlled microtumor models successfully recapitulated clinically observed positive association between tumor size and aggressive phenotype in hormone receptor–positive breast cancer while maintaining clinically proven poor correlation of tumor size with aggressive phenotype in TNBC. Such clinically relevant 3D models generated under controlled experimental conditions can serve as precise preclinical models to study mechanisms involved in breast tumor progression as well as antitumor drug effects as a function of tumor progression. Cancer Res; 76(13); 3732–43. ©2016 AACR.

Malignant tumor phenotype and secondary metastasis are the major causes of breast cancer–associated deaths. Recent statistical data suggested that survival of breast cancer patients decreased from 84% to 24% as the tumor spread from local to distant lymph nodes (1). Thus, understanding metastatic tumor progression is critical to design effective breast cancer treatment strategies.

Mechanisms of metastatic breast tumor progression are studied widely in two-dimensional (2D) cultures by exposing them to external stimulus like hypoxia and in xenograft models by manipulating various signaling pathways and molecular targets (2, 3), which do not recapitulate natural events of breast tumor progression (4, 5). Cells in 2D monolayers lack microenvironmental context, cell–cell contacts in 3D, integrated cell signaling and cell–matrix interaction, making them physiologically different from the cells in solid tumors in vivo (4, 5). These cells in vivo experience metabolic stress due to oxygen and nutritional gradients and physiological heterogeneity (6). Such cellular heterogeneity plays an important role in determining drug diffusion, efficacy, and development of drug resistance (7, 8). Several studies utilize different cell lines to represent noninvasive and invasive phenotypes or manipulate gene/protein expression in the same cell lines by gene knock-in or knock-out (3, 9). These studies have improved understanding of epithelial–mesenchymal transition (EMT) and various signaling pathways involved in breast tumor metastasis (10–13). However, given the complexity of cellular signaling networks, gene/protein manipulations may cause simultaneous activation/inhibition of alternative pathways, hampering systematic understanding of processes involved in metastatic progression.

Xenograft mouse models and patient derived xenografts resemble solid breast tumors more closely than 2D monolayers (14). However, differences in pathophysiology and etiology of tumor development in animal models, immune system and lack of human breast stromal component limits applicability of these findings to human breast tumor progression (3, 14). The breast tumor biopsies and tissue microarrays can be useful to study metastatic and aggressive tumor phenotypes (15). Although promising, the unavailability of biologically matched early- and advanced-stage tumor biopsies limits mechanistic understanding of size-dependent metastatic progression in clinical samples. The 3D tumor models generated using human cells under precisely controlled conditions have the ability to mimic in vivo microenvironments and pathophysiology of human disease and can fill the gap between currently used 2D monolayers and animal models.

Many 3D breast tumor models generated using scaffolds (16), microfluidic system (17), and cocultures (18) have significantly improved our understanding of molecular mechanisms involved in metastasis. However, preclinical 3D models recapitulating tumor growth/size-induced early to malignant progression remain to be developed. Such tumor models are of great importance as breast tumor size is positively associated with lymph node positivity, metastasis, and patient survival depending on the hormone receptor status (19–21). In this study, we developed size-controlled 3D breast tumor models to recapitulate “tumor size–induced” changes in the microenvironment and their effect on metastatic progression in vitro.

We hypothesized that the microtumor size–associated microenvironmental changes positively contribute to the aggressive phenotype, which is characterized here by migratory behavior and upregulation of mesenchymal markers at mRNA and protein levels. To test this hypothesis, we developed first of its kind in vitro human tumor progression model using microfabrication to generate uniform and defined size microtumors of various breast cancer cell lines (MCF7, T47D, BT474, MDA-MB-231, and HCC1187). We could link breast tumor size to aggressive and migratory behavior by precisely modulating just the microtumor size and without any exposure to artificial hypoxic chambers or chemical stimuli (22). Our results demonstrated that large size (600 μm) microtumors of noninvasive estrogen/progesterone receptor–positive (ER+/PR+) breast tumor cells (MCF7, T47D) upregulate mesenchymal markers and “migrate out” from the hydrogel microwells without losing their epithelial phenotype. The hypoxic core (increased expression of Hif-1α) and necrotic cells in large microtumors matched the spatial cellular distribution observed in solid tumors. Size-dependent upregulation of mesenchymal markers was also observed in ER+/PR+/HER2+ BT474 cell lines. Interestingly, 3D microtumors of triple-negative MDA-MB-231 (MM-231) and HCC1187 cells did not exhibit size-dependent upregulation of mesenchymal marker expression reproducing clinically observed poor correlation between size and aggressive phenotype in triple-negative breast cancer (TNBC) and, thus, further validating clinical relevance of our 3D model.

Breast cancer cell lines were purchased from ATCC between the years 2002 to 2012. Characterization of cell lines was done by the University of Arizona Genetic Core facility using their well-established method. Briefly, genomic DNA from all the cell lines were amplified using Promega Power Plex16HS PCR kit, and PCR products were separated by capillary electrophoresis by an AB 3730 DNA Analyzer. Electropherograms were analyzed using Soft Genetics, Gene Marker Software Version 1.85. Alleles were matched for STR Profile recorded with DMSZ (http://www.dsmz.de/fp/cgi-bin/help.html) and ANSI (http://webstore.ansi.org/RecordDetail.aspx?sku=ANSI%2FATCC+ASN-0002-2011#.UMuAP6yRPTo) of the same cell line using 80% match as threshold. Details of cell culture, chemicals and reagents can be found in Supplementary Materials and Methods.

3D culture and microtumor fabrication

Uniform-size 3D microtumors of subtype specific breast cancer cell lines were fabricated using polyethylene glycol dimethacrylate (PEGDMA) hydrogel microwell arrays as described earlier (Fig. 1A; Supplementary Materials and Methods; refs. 23, 24). Although the same cell seeding density (0.5 × 106 cells/device) was used in different size microwell devices (150–600 μm), cell number/microtumor was dependent on the microwell diameter. To confirm that acquisition of aggressive phenotype is truly tumor size dependent and not due to differences in initial cell number/microtumor (cell density), we generated density-matched microtumors by seeding cells proportionate to the total microwell volumes in each device (see Supplementary Materials and Methods; Supplementary Table S1). These microtumors were referred to as “density-matched” hereafter. Density-matched microtumors were harvested and EMT marker expression was measured by qRT-PCR analysis.

Figure 1.

Fabrication of size-controlled microtumors of MCF7. A, schematic representing microtumor fabrication. B, microtumors in hydrogel microwells. C, calcein-AM (green, live)– and ethidium homodimer (red, dead)–stained microtumors. D, day 6 harvested microtumors. E, growth kinetics of microtumors estimated by measuring diameter. F, number of cells/microtumor. Data, mean ± SD (n ≥ 50; *, P < 0.05, compared with day 1). G, i, spatial cellular distribution, Ki-67–positive cells (green) on periphery and PI-positive cells (red) in the core of 600-μm microtumors. G, ii, Ru-dpp–stained microtumors showing higher fluorescence in the core of 600-μm microtumors due to limited oxygen diffusion. H, Western blot of Hif-1α and Vegf. I, densitometry data of Hif-1α and Vegf in 150- and 600-μm microtumors. Data, mean ± SEM (n = 6; #, P < 0.05; ###, P < 0.001 with respect to 150-μm microtumors). Scale bar, 500 μm.

Figure 1.

Fabrication of size-controlled microtumors of MCF7. A, schematic representing microtumor fabrication. B, microtumors in hydrogel microwells. C, calcein-AM (green, live)– and ethidium homodimer (red, dead)–stained microtumors. D, day 6 harvested microtumors. E, growth kinetics of microtumors estimated by measuring diameter. F, number of cells/microtumor. Data, mean ± SD (n ≥ 50; *, P < 0.05, compared with day 1). G, i, spatial cellular distribution, Ki-67–positive cells (green) on periphery and PI-positive cells (red) in the core of 600-μm microtumors. G, ii, Ru-dpp–stained microtumors showing higher fluorescence in the core of 600-μm microtumors due to limited oxygen diffusion. H, Western blot of Hif-1α and Vegf. I, densitometry data of Hif-1α and Vegf in 150- and 600-μm microtumors. Data, mean ± SEM (n = 6; #, P < 0.05; ###, P < 0.001 with respect to 150-μm microtumors). Scale bar, 500 μm.

Close modal

Intratumoral oxygen availability

Oxygen bioavailability in 3D microtumors was assessed by oxygen-sensitive ruthenium-tris(4,7-diphenyl-1,10-phenanthroline)dichloride dye (Ru-dpp; Supplementary Materials and Methods; ref. 25). Fluorescence intensity of Ru-dpp (absorbance λmax 455 nm, luminescence λmax 613 nm) is inversely dependent on oxygen concentration and is greatly reduced in the presence of oxygen. Fluorescent images of Ru-dpp–stained microtumors were acquired on a confocal microscope (Olympus Fluoview) using a 543-nm He-Ne laser to excite the dye and 604 LP emission filters.

qRT-PCR and Western blotting

RNA isolation, protein extraction followed by qRT-PCR and Western blotting were performed as described in Supplementary Materials and Methods. Primer sequences are given in Table 1.

Table 1.

List of primers used

NameSequence
1. β-actin F 5′-ACC TTCTAC AAT GAG CTG CG-3′ 
 β-actin R 5′CCT GGA TAG CAA CGT ACA TGG-3′ 
2. Fibronectin F 5′-CCG CCG AAT GTA GGA CAA GA-3′ 
 Fibronectin R 5′-TGC CAA CAG GAT GAC ATG AAA-3′ 
3. Vimentin F 5′-CAA CCT GGC CGA GGA CAT-3′ 
 Vimentin R 5′-ACG CAT TGT CAA CAT CCT GTC T-3′ 
4. Snail F 5′-TGC AGG ACT CTA ATC CAA GTT TAC C-3′ 
 Snail R 5′-GTG GGA TGG CTG CCA GC-3′ 
5. Slug F 5′-AGC ATT TCA ACG CCT CCA-3′ 
 Slug R 5′-GGA TCT CTG GTT GTG GTA TGA C-3′ 
6. Twist1 F 5′-TGT CCG CGT CCC ACT AGC-3′ 
 Twist1 R 5′-TGT CCA TTT TCT CCT TCT CTG GA-3′ 
7. N-cadherin F 5′-GAC GGT TCG CCA TCC AGA C-3′ 
 N-cadherin R 5′-TCG ATT GGT TTG ACC ACG G-3′ 
8. E-cadherin F 5′-GAA CAG CAC GTC CAC AGC CCT-3′ 
 E-cadherin R 5′-GCA GAA GTGTCC CTG TTC CAG-3′ 
NameSequence
1. β-actin F 5′-ACC TTCTAC AAT GAG CTG CG-3′ 
 β-actin R 5′CCT GGA TAG CAA CGT ACA TGG-3′ 
2. Fibronectin F 5′-CCG CCG AAT GTA GGA CAA GA-3′ 
 Fibronectin R 5′-TGC CAA CAG GAT GAC ATG AAA-3′ 
3. Vimentin F 5′-CAA CCT GGC CGA GGA CAT-3′ 
 Vimentin R 5′-ACG CAT TGT CAA CAT CCT GTC T-3′ 
4. Snail F 5′-TGC AGG ACT CTA ATC CAA GTT TAC C-3′ 
 Snail R 5′-GTG GGA TGG CTG CCA GC-3′ 
5. Slug F 5′-AGC ATT TCA ACG CCT CCA-3′ 
 Slug R 5′-GGA TCT CTG GTT GTG GTA TGA C-3′ 
6. Twist1 F 5′-TGT CCG CGT CCC ACT AGC-3′ 
 Twist1 R 5′-TGT CCA TTT TCT CCT TCT CTG GA-3′ 
7. N-cadherin F 5′-GAC GGT TCG CCA TCC AGA C-3′ 
 N-cadherin R 5′-TCG ATT GGT TTG ACC ACG G-3′ 
8. E-cadherin F 5′-GAA CAG CAC GTC CAC AGC CCT-3′ 
 E-cadherin R 5′-GCA GAA GTGTCC CTG TTC CAG-3′ 

Immunostaining and confocal microscopy

Day 6 harvested microtumors were immunostained for Vim, Snail, E-cadherin (E-cad), and Ki-67 (Supplementary Materials and Methods).

Plasticity of mesenchymal phenotype

To investigate the plasticity of size-dependent upregulation of mesenchymal markers, a size reversal experiment was performed. Briefly, 150- and 600-μm MCF7 microtumors were harvested on day 6. Subsequently, 150-μm microtumors were disaggregated using trypsin and reseeded onto 600-μm microwells (referred to as size reversal of 150; SR-150). Similarly, 600-μm microtumors were disaggregated and reseeded onto 150-μm microwells (size reversal of 600; SR-600). Both SR-150 and SR-600 microtumors were harvested on day 6, and EMT marker expression was compared with parent 150- and 600-μm microtumors by qRT-PCR.

4-Hydroxytamoxifen treatment

Advanced-stage tumors often acquire drug resistance by EMT-mediated genetic changes (26). Hence, the drug response was studied by measuring cell growth in 150 and 600-μm microtumors after 50 μmol/L 4-hydroxytamoxifen (4-OHT) treatment (Supplementary Materials and Methods).

siRNA transfection

To investigate the role of hypoxia in the acquisition of aggressive phenotype, Hif-1α was knocked down in MCF7 microtumors by Hif-1α siRNA and the Hif-1α inhibitor methyl 3-[[2-[4-(2-adamantyl)phenoxy] acetyl]amino]-4-hydroxybenzoate (Supplementary Materials and Methods). EMT marker status in both the groups was determined by qRT-PCR.

Statistical analysis

The results were analyzed by two-way analysis of variance (ANOVA) followed by a Tukey multiple comparison test (Graph Pad Prism, V6.0). P values less than 0.05 are considered statistically significant.

Microfabricated hydrogel microarrays generate microtumors of controlled sizes

It is widely accepted that tumor growth, progression, and drug response are controlled by the microenvironment. The tumor microenvironment consists of non-cellular components (pH, hypoxia/necrosis, metabolic stress, etc.), extracellular matrix (ECM), and cellular components (tumor and stromal cells). Tumor spheroids remain the best characterized and widely used 3D models to study various aspects in cancer biology and drug resistance; however, techniques such as non-adhesive surfaces, spinner flasks, NASA rotary system, hanging drop method used to fabricate spheroids result in a wide range of sizes and shapes (27, 28). Heterogeneity in spheroid sizes can affect noncellular factors in the tumor microenvironment, including nutrient/oxygen gradients, hypoxia and metabolic stress, which can further affect tumor biology and response to drug. To circumvent this challenge, we have microfabricated non-adhesive PEGDMA hydrogel microarrays with hundreds of defined diameter microwells using photolithography (29) that allows generation of uniform, defined size microtumors of various cell lines such as cervical, breast, and head and neck cancer (Fig. 1A; ref. 24). Microwell diameter, cell seeding density, and cell lines determine the size of generated microtumors (24). Here, we have used this microfabricated hydrogel platform to develop and validate 3D breast tumor progression model. We hypothesized that controlled variation in microtumor sizes will positively contribute to aggressive behavior through changes in the noncellular factors in the tumor microenvironment.

Using PEGDMA microwells of 150-, 300-, 450-, and 600-μm diameters, we generated uniform sized 3D microtumors of noninvasive MCF7 cells (Fig. 1B and C). The microtumor size was precisely controlled by microwells as shown by the photomicrographs of harvested microtumors (Fig. 1D). After harvesting on day 6, 150-μm microtumors appeared tightly packed, while 600-μm ones were loosely packed. Growth of the microtumors in the microwells was determined by measuring the diameter and number of cells/microtumor on days 1 and 6 of culture. The microtumor diameter and number of cells/microtumor increased with microwell size as well as time in culture (day 1 to day 6). Albeit the rate of proliferation was much slower than that is generally observed for 2D monolayer culture (approximate doubling time of 24 h; ref. 11). For instance, the diameter of 150-μm microtumors increased from 82 ± 33 to 128 ± 16 μm on day 6 and cell number/microtumor increased from 550 ± 130 to 816 ± 144; while diameter of 600-μm microtumors increased from 394 ± 46 to 550 ± 57 μm and cell number/microtumor increased from 1,751 ± 386 to 3,690 ± 117 (Fig. 1E and F). Thus, cell-doubling time was approximately 6 days for 3D microtumors compared with 24 hours reported for MCF7 2D monolayers (11).

Microtumors mimic size-dependent microenvironmental changes

Aggressive cell proliferation in solid tumors creates cellular heterogeneity in response to hypoxia and metabolic stress. The effect of microtumor size (150 and 600 μm) on intratumoral microenvironment was evaluated by staining with propidium iodide (PI; dead cells marker) and Ki-67 (proliferation marker). Consistent with previous observations in solid tumors (6, 28), large (600 μm) microtumors had more PI-positive cells in the center, indicating dead cells, possibly due to diffusion limitations for oxygen and nutrients compared with small (150-μm) microtumors (Fig. 1G, i). Moreover, proliferating Ki-67–positive cells were restricted to the outer layers in large microtumors while they were observed throughout the small microtumors. Limited oxygen availability inside large 600-μm microtumors was determined by Ru-dpp staining followed by confocal imaging. Increased fluorescence in the core of large (600-μm) microtumors compared with 150-μm ones (Fig. 1G, ii) suggested lack of oxygen availability (25) with increase in microtumor size. Existence of hypoxia in 3D microtumors was estimated by measuring protein levels of Hif- 1α and vascular endothelial growth factor (Vegf), a downstream effector of Hif-1α in 2D monolayers, 150- and 600- μm MCF7 microtumors (Fig. 1H). Densitometry showed more than 3-fold increase in Hif-1α and significant upregulation of Vegf expression in 600-μm microtumors compared with 150-μm ones (Fig. 1I). Small and large microtumors of ER+/PR+ T47D cells showed similar results for Ki-67/PI staining as well as Hif-1α and Vegf expression (Supplementary Fig. S1C–S1E). As tumors grow in size, they experience metabolic stress and fulfill their increasing energy demand by upregulating glucose transporters such as Glut1 (6). To evaluate this metabolic shift, we determined Glut1 expression in protein lysates of 2D, 150- and 600-μm T47D microtumors (Supplementary Fig. S1F). Densitometry analysis showed more than 3-fold increase in Glut1 expression in 600-μm microtumors compared with 150-μm ones (Supplementary Fig. S1G). Cumulatively, these results suggest that size-controlled 3D breast microtumors mimic the microenvironmental complexity and spatial cellular distribution observed in solid tumors in vivo (28).

Snail-negative MCF7 microtumors acquire size-dependent migratory and mesenchymal phenotype independent of cell seeding density

MCF7 and T47D cell lines are widely used as models for noninvasive, Snail-negative breast cancer. Interestingly, 600-μm microtumors of MCF7 (Fig. 2A) and T47D (Supplementary Figs. S1A and S2) displayed migratory behavior and consistently “migrated out” of the wells by day 6 (indicated by arrows). However, their extent of migration was heterogeneous on the same device (Supplementary Fig. S1B). The observed migratory behavior may also stem from the differences in the cell density (cell number/microtumor) in 150- and 600-μm microtumors (Fig. 1F). To rule out the effect of different cell density present in different size microtumors on migratory behavior, we matched the cell number/microtumor/unit volume of microtumors for all sizes (referred to as “density-matched” microtumors) by seeding cells proportionate to the total microwell volumes in each device as described in Supplementary Materials and Methods and Table 1. As shown in Fig. 2B, 600-μm microtumors generated using the same initial seeding density (0.5 × 106 cells/device) displayed approximately 4-fold higher cell number/microtumor than 150-μm microtumors after 1 day of culture. On the other hand, density-matched 600-μm microtumors generated using 12.5-fold higher cell density (6.25 × 106 cells/device) exhibited approximately 11.7-fold higher cell number/microtumor (Fig. 2B). Interestingly, we observed similar migratory behavior in the density-matched 600-μm MCF7 microtumors on day 6 (black arrows, Fig. 2A, last column), demonstrating density-independent migration in large microtumors. Migratory behavior of large microtumors resembled “collective or cohesive migration” observed in clinically aggressive breast tumors (30). We then measured the transcriptional status of various mesenchymal and epithelial markers including VIM, FIB, SNAIL, SLUG, E-CAD, and N-CAD in MCF7 microtumors of all sizes generated using same initial seeding density and matched seeding density. SNAIL and SLUG (mesenchymal mRNA) were increased 6- to 14-fold in 600-μm microtumors compared with 150-μm ones in Snail-negative MCF7 cells (Fig. 2C, same seeding density). Similar increase in mesenchymal marker mRNA expression was observed in density-matched microtumors of all sizes (Fig. 2D). These data indicate that upregulation of mesenchymal marker expression in noninvasive MCF7 microtumors was truly size-dependent and not density-dependent. Interestingly, microtumors of all sizes did not show change in epithelial markers E-CAD and N-CAD at the transcriptional level (Fig. 2C and D).

Figure 2.

Snail-negative MCF7 microtumors acquire size-dependent migratory and mesenchymal phenotype independent of cell seeding density. MCF7 cells were seeded in 150- to 600-μm microwells at two different densities. In one set, all devices were seeded with 0.5 × 106 cells/device. In density-matched set, 150-, 300-, 450-, and 600-μm devices were seeded with 0.5, 1.25, 3.5, and 6.25 × 106 cells/device. A, photomicrographs of 150 μm, 600 μm, and density-matched 600-μm microtumors on day 1 (top) and day 6 (bottom). Scale bar, 500 μm. B, number of cells/microtumor in 150 μm, 600 μm, and density-matched 600-μm microtumors on day 1 (N > 50 microtumors/experiment). Number of cells/microtumor in density-matched 600-μm microtumors was similar to the theoretically calculated seeding densities based on total microwell volumes (approximately 12 times total microwell volume of 150-μm devices). Data, mean ± SD (n = 3). C and D, mRNA expression of EMT markers in microtumors generated using 0.5 × 106 cells/device (C) and density-matched microtumors (D). Density-matched microtumors exhibited size-dependent but density-independent upregulation of mesenchymal marker expression. Data, mean ± SEM in C and D (n = 6; #, P < 0.05; ##, P < 0.005; ###, P < 0.001 with respect to 150-μm microtumors).

Figure 2.

Snail-negative MCF7 microtumors acquire size-dependent migratory and mesenchymal phenotype independent of cell seeding density. MCF7 cells were seeded in 150- to 600-μm microwells at two different densities. In one set, all devices were seeded with 0.5 × 106 cells/device. In density-matched set, 150-, 300-, 450-, and 600-μm devices were seeded with 0.5, 1.25, 3.5, and 6.25 × 106 cells/device. A, photomicrographs of 150 μm, 600 μm, and density-matched 600-μm microtumors on day 1 (top) and day 6 (bottom). Scale bar, 500 μm. B, number of cells/microtumor in 150 μm, 600 μm, and density-matched 600-μm microtumors on day 1 (N > 50 microtumors/experiment). Number of cells/microtumor in density-matched 600-μm microtumors was similar to the theoretically calculated seeding densities based on total microwell volumes (approximately 12 times total microwell volume of 150-μm devices). Data, mean ± SD (n = 3). C and D, mRNA expression of EMT markers in microtumors generated using 0.5 × 106 cells/device (C) and density-matched microtumors (D). Density-matched microtumors exhibited size-dependent but density-independent upregulation of mesenchymal marker expression. Data, mean ± SEM in C and D (n = 6; #, P < 0.05; ##, P < 0.005; ###, P < 0.001 with respect to 150-μm microtumors).

Close modal

TNBC microtumors show poor association between size and mesenchymal marker upregulation

We further investigated if acquisition of aggressive phenotype (“migrating out” behavior and upregulation of mesenchymal marker expression) is also observed in other molecular subtypes of breast cancer. Hence, we fabricated 150- and 600-μm microtumors of different subtype-specific breast cancer cell lines, such as ER+/PR+ (T47D), ER+/PR+/HER2+ (BT474) and triple negative (MM-231 and HCC1187). Although all cell lines formed uniform-sized microtumors (Fig. 3A), T47D and BT474 microtumors were more compact than triple-negative MM-231 and HCC1187 microtumors, which could be attributed to their invasive phenotypes. Indeed, it was recently shown that distinct colony morphologies in 3D is measure of tumor cell invasiveness (31). For example, MCF7, T47D, and BT474 cells formed round colonies with disorganized nuclei and filled colony centers (labeled as “mass” colonies) whereas MDA-MB-231 cells formed “stellate” colonies due to their invasive phenotypes (31). EMT marker expression in large T47D and BT474 microtumors showed statistically significant upregulation of mesenchymal markers compared with small microtumors (Fig. 3B). For example, large microtumors exhibited more than 3-fold increase in VIM, 4- to 7-fold increase in SNAIL, and 2- to 5-fold increase in TWIST mRNA expression in both BT474 and T47D. In contrast, triple-negative MM-231 and HCC1187 microtumors did not show significant differences (P > 0.05) in mesenchymal marker (VIM, FIB, SNAIL and TWIST) expression in small and large microtumors, suggesting poor correlation between tumor size and aggressive phenotype. Consistent with MCF7 data, we did not observe statistically significant differences in E-CAD mRNA levels between 150- and 600-μm microtumors of any cell line except T47D (Fig. 3B). These data collectively indicated positive correlation between tumor size and aggressive phenotype in size-controlled hormone receptor–positive breast cancer 3D microtumor models, but not in TNBC.

Figure 3.

Characterization of subtype-specific (T47D, BT474, MM231, and HCC1187) breast cancer microtumors. A, photomicrographs of subtype-specific microtumors. Scale bar, 500 μm. B, EMT marker expression in different 150- and 600-μm microtumors. The expression levels were normalized with 2D (dotted lines). Size-dependent upregulation of mesenchymal markers was observed in T47D and BT474, while MM231 and HCC1187 did not show any change. Data presented as mean ± SEM (n = 6; #, P < 0.05; ##, P < 0.005; ###, P < 0.001 with respect to 150-μm microtumors).

Figure 3.

Characterization of subtype-specific (T47D, BT474, MM231, and HCC1187) breast cancer microtumors. A, photomicrographs of subtype-specific microtumors. Scale bar, 500 μm. B, EMT marker expression in different 150- and 600-μm microtumors. The expression levels were normalized with 2D (dotted lines). Size-dependent upregulation of mesenchymal markers was observed in T47D and BT474, while MM231 and HCC1187 did not show any change. Data presented as mean ± SEM (n = 6; #, P < 0.05; ##, P < 0.005; ###, P < 0.001 with respect to 150-μm microtumors).

Close modal

Size-dependent upregulation of mesenchymal markers at translational level

The effect of microtumor size on upregulation of mesenchymal markers was further confirmed by determining protein levels of Vim, Snail, and E-cad in MCF7 and T47D microtumors (Fig. 4A). Compared with small microtumors, large microtumors displayed significant upregulation of Vim (more than 8- and 2-fold in MCF7 and T47D) and Snail (more than 2-fold in MCF7 and T47D; Fig. 4B). Surprisingly, we also observed approximately 10-fold increase in E-cad protein levels in both 150- and 600-μm microtumors compared with 2D monolayers, although there was no significant difference between 150- and 600-μm 3D microtumors. Overexpression of E-cad protein in 3D compared with 2D has been previously reported in ovarian cancer (32) and inflammatory breast cancer (33). As reported for conventional EMT (10, 11, 30), “cadherin switch” from E-cad to N-cadherin (N-cad) was not observed in large microtumors (Supplementary Fig. S3). Taken together, these data showing upregulation of mesenchymal markers at transcriptional and translational level without loss of E-Cad/gain of N-Cad indicate that the observed migratory behavior of large microtumors may not be due to the conventional EMT, instead may represent “cohesive or collective migration” that requires E-cad (30).

Figure 4.

Size-dependent upregulation of mesenchymal markers at the translational level. A, Western blots of Vim, Snail, and E-Cad. B, densitometry analysis of Vim, Snail, and E-cad in MCF7 and T47D microtumors. The protein levels were normalized with 2D (dotted lines; data presented as mean ± SEM (n = 6; #, P < 0.05; ##, P < 0.005, with respect to 150-μm microtumors). C, immunostaining showing in situ expression of Vim, Snail, and E-Cad in 150- and 600-μm MCF7 microtumors. Large microtumors showed more number of Vim and Snail-positive cells, with no change in E-cad protein. (n = 10–15 microtumors/group; scale bar, 100 μm).

Figure 4.

Size-dependent upregulation of mesenchymal markers at the translational level. A, Western blots of Vim, Snail, and E-Cad. B, densitometry analysis of Vim, Snail, and E-cad in MCF7 and T47D microtumors. The protein levels were normalized with 2D (dotted lines; data presented as mean ± SEM (n = 6; #, P < 0.05; ##, P < 0.005, with respect to 150-μm microtumors). C, immunostaining showing in situ expression of Vim, Snail, and E-Cad in 150- and 600-μm MCF7 microtumors. Large microtumors showed more number of Vim and Snail-positive cells, with no change in E-cad protein. (n = 10–15 microtumors/group; scale bar, 100 μm).

Close modal

Spatial localization of Vim, Snail, and E-cad in small and large MCF7 microtumors was investigated by immunostaining (Fig. 4C). As expected, Vim- and Snail was absent in small microtumors of Snail-negative MCF7, while large microtumors of the same MCF7 cell line exhibited presence of Vim and Snail-positive cells. Most importantly, heterogeneous spatial distribution of Vim- and Snail-positive cells was evident from immunofluorescence images. For instance, only cells at the outer periphery of 600-μm microtumors were stained for Vim (Fig. 4C, first column, last row and Supplementary Video 1). Consistent with mRNA and protein expression data, E-cad was not lost in any of the cells in 150- or 600-μm size microtumors (Fig. 4C, last column, last row). This was evident in both 600-μm MCF7 and T47D microtumors. Immunofluorescence images of the migrating microtumors also showed uniform E-cad staining irrespective of whether the microtumor was inside the microwell or migrated out of the well (Supplementary Video 2 and 3).

Size-induced mesenchymal behavior in 3D breast microtumors is not reversible

To investigate the plasticity of size-induced mesenchymal behavior, a size reversal experiment was performed in MCF7 microtumors (Fig. 5A, i–ii). Parent 150- and 600-μm microtumors were harvested on day 6 and disaggregated using trypsin. Subsequently, cells disaggregated from parent 150-μm microtumors were reseeded into 600-μm microwells (labeled as “SR-150”) while those disaggregated from parent 600-μm tumors were reseeded into 150-μm microwells (SR-600) and cultured for additional 6 days. The MCF7 cells of SR-150 acquired aggressive phenotype evidenced by migratory behavior (Fig. 5B, black arrows in SR-150). Interestingly, these SR-150 microtumors were loosely packed compared with tightly packed parent 150-μm microtumors possibly due to size-induced hypoxic microenvironment experienced by the large SR-150 microtumors. Despite of their small (150 μm) size, SR-600 microtumors also acquired “migratory” behavior similar to parent 600-μm ones. Such migratory behavior was never observed in the parent 150-μm size microtumors. Light microscopy images suggest that migratory phenotype acquired by parent 600-μm microtumors cannot be reversed even if they are cultured to form small microtumors, which do not exhibit significant hypoxia (Fig. 1Gii) or Hif-1α upregulation (Fig. 1H). These results are further confirmed by measuring mRNA expression of mesenchymal markers VIM, FIB, SNAIL, and SLUG in parent 150- and 600-μm as well as SR-150 and SR-600 microtumors. All data were compared with parent 150-μm microtumors as a control. Consistent with the results in Fig. 2C and D, parent 600-μm microtumors displayed significant upregulation of mesenchymal markers compared with parent 150-μm microtumors (Fig. 5C). More importantly, SR-150 exhibited 16-, 6-, 2-, and 4-fold higher mRNA expression of VIM, FIB, SNAIL and SLUG, respectively, than parent 150-μm microtumors (Fig. 5C). Thus, acquisition of an aggressive phenotype (migration as well as mesenchymal marker upregulation) in size reversal of 150 μm into 600-μm microtumors (SR-150) further support an important role of tumor size in acquisition of an aggressive phenotype in large microtumors. Interestingly, VIM, FIB, SLUG, and SNAIL were further upregulated in SR-600 compared with both, parent 150- and 600-μm microtumors (Fig. 5C), implying that size reversal of 600 μm into 150-μm microtumors is not reversible. This means that even after disaggregation into single cells, parent 600-μm microtumors maintained their aggressive phenotypes, which was further upregulated after culture for additional 6 days.

Figure 5.

Size-induced mesenchymal phenotype is not reversible. A, schematic representation of the size reversal experiment for 150 to SR-150 (i) and 600 to SR-600 (ii). B, photomicrographs of 150 and SR-150 microtumors on day 6. Unlike parent 150, SR-150 “migrated out” on day 6 (black arrows). Similar “migrating out” was observed in both parent 600 and SR-600 microtumors, suggesting that migratory phenotype of 600-μm microtumors is not reversible even after reversing its size to small 150-μm microtumors. Scale bar, 100 μm. C, mRNA expression of VIM, FIB, SNAIL, and SLUG in 150-μm, 600-μm, SR-150, and SR-600 microtumors. Fold change was calculated using parent 150-μm microtumors as a control. Data, mean ± SEM (n = 4; #, P < 0.05; ###, P < 0.001 with respect to parent 150-μm microtumors).

Figure 5.

Size-induced mesenchymal phenotype is not reversible. A, schematic representation of the size reversal experiment for 150 to SR-150 (i) and 600 to SR-600 (ii). B, photomicrographs of 150 and SR-150 microtumors on day 6. Unlike parent 150, SR-150 “migrated out” on day 6 (black arrows). Similar “migrating out” was observed in both parent 600 and SR-600 microtumors, suggesting that migratory phenotype of 600-μm microtumors is not reversible even after reversing its size to small 150-μm microtumors. Scale bar, 100 μm. C, mRNA expression of VIM, FIB, SNAIL, and SLUG in 150-μm, 600-μm, SR-150, and SR-600 microtumors. Fold change was calculated using parent 150-μm microtumors as a control. Data, mean ± SEM (n = 4; #, P < 0.05; ###, P < 0.001 with respect to parent 150-μm microtumors).

Close modal

EMT also contributes to drug resistance in advanced tumors (26). To understand the effect of microtumor size-induced EMT on drug response, 150- and 600-μm microtumors were treated with 4-OHT, a selective ER modulator. Percentage growth data for untreated and 4-OHT treated microtumors (Supplementary Fig. S4) suggested that 4-OHT effectively inhibited growth of 150-μm microtumors to a much greater extent (>85%) than in 600-μm microtumors (about 40%) compared with their respective untreated controls. These data further indicate acquisition of drug-resistant phenotype in large microtumors.

Hif-1α knockdown only partially downregulates mesenchymal markers

Simultaneous upregulation of Hif-1α and mesenchymal markers led us to investigate if the observed size-induced aggressive phenotype is due to hypoxia in 600-μm microtumors. Hif-1α was knocked down in 600-μm MCF7 microtumors either by siRNA treatment or by chemical inhibition and, subsequently, mRNA expression of EMT markers was investigated. Both siRNA knockdown and chemical inhibition downregulated Hif-1α protein expression (Fig. 6A). MCF7 cells formed microtumors in spite of suppressed Hif-1α (Fig. 6B). The siRNA and chemical inhibition-mediated Hif-1α knockdown in 600-μm microtumors significantly downregulated mesenchymal markers at the mRNA level compared with those treated with scrambled siRNA or untreated control (Fig. 6C and D). It is noteworthy that both Hif-1α inhibition methods were unable to completely abolish the migratory behavior (Fig. 6B) and upregulation of mesenchymal markers in large microtumors, implying that mechanisms other than hypoxia may be involved in acquisition of aggressive phenotype observed in large microtumors.

Figure 6.

Hif-1α knockdown partially downregulates mesenchymal markers. Hypoxia-mediated regulation of mesenchymal phenotype was evaluated by inhibiting Hif-1α with Hif-1α siRNA and Hif-1α inhibitor. A, Western blots of Hif-1α in untreated (control/600), siRNA-transfected (siRNA/600), and Hif-1α inhibitor–treated 600-μm microtumors. B, photomicrographs of 600-μm MCF7 microtumors treated with scrambled and Hif-1α SiRNA. In both groups, 600-μm MCF7 microtumors displayed migratory behavior on day 6 (black arrows), suggesting Hif-1α knockdown could not completely abolish migratory behavior. Scale bar, 500 μm. C, qRT-PCR analysis demonstrated downregulation of VIM, FIB, SNAIL, and SLUG in Hif-1α siRNA/600 compared with control/600 and scrambled siRNA-transfected 600-μm microtumors. D, Hif-1α inhibitor–treated microtumors also showed reduced expression of FIB, SNAIL, and SLUG compared with untreated 600-μm microtumors. Data are represented as mean ± SEM (n = 4; *, P < 0.05; **, P < 0.005; ***, P < 0.001 with respect to control/600; #, P < 0.05; ###, P < 0.001 with respect to scrambled siRNA/600).

Figure 6.

Hif-1α knockdown partially downregulates mesenchymal markers. Hypoxia-mediated regulation of mesenchymal phenotype was evaluated by inhibiting Hif-1α with Hif-1α siRNA and Hif-1α inhibitor. A, Western blots of Hif-1α in untreated (control/600), siRNA-transfected (siRNA/600), and Hif-1α inhibitor–treated 600-μm microtumors. B, photomicrographs of 600-μm MCF7 microtumors treated with scrambled and Hif-1α SiRNA. In both groups, 600-μm MCF7 microtumors displayed migratory behavior on day 6 (black arrows), suggesting Hif-1α knockdown could not completely abolish migratory behavior. Scale bar, 500 μm. C, qRT-PCR analysis demonstrated downregulation of VIM, FIB, SNAIL, and SLUG in Hif-1α siRNA/600 compared with control/600 and scrambled siRNA-transfected 600-μm microtumors. D, Hif-1α inhibitor–treated microtumors also showed reduced expression of FIB, SNAIL, and SLUG compared with untreated 600-μm microtumors. Data are represented as mean ± SEM (n = 4; *, P < 0.05; **, P < 0.005; ***, P < 0.001 with respect to control/600; #, P < 0.05; ###, P < 0.001 with respect to scrambled siRNA/600).

Close modal

Tumor size is an important prognostic determinant for tumor staging, metastasis, nodal involvement, and survival in hormone receptor–positive (luminal) breast carcinoma patients (19, 20, 34). In contrast, the probability of metastasis and patient survival in TNBC is independent of tumor size (21, 35). Increase in tumor size leads to microenvironmental changes in vivo due to diffusional limitations, physiological hypoxia, acidosis, and metabolic stress that can induce alternative adaptive mechanisms for tumor cell survival and, eventually, metastasis and drug resistance (6). Although several clinical studies have shown the correlation between tumor size and metastatic aggressive phenotype in luminal breast cancer (19, 20, 34), tumor heterogeneity and lack of physiologically relevant models hamper mechanistic understanding of malignant tumor progression as a function of tumor size.

Given the clinical relevance of breast tumor size in disease progression and its important role in controlling tumor microenvironment, precise control over the size of multicellular 3D aggregates (microtumors) is important for a mechanistic understanding of tumor progression as a function of tumor size. Current in vitro methods are unable to generate uniform size aggregates with a reproducible phenotype (27, 28). Our 3D model can precisely regulate microtumor size, providing a better control over biological parameters such as diffusion of oxygen, nutrients, and metabolites and recapitulate size-dependent malignant progression observed in the luminal subtype for the first time. Use of defined size PEGDMA microwells allow generation of hundreds of uniformly sized microtumors of ER+/PR+ (MCF7, T47D), ER+/PR+/HER2+ (BT474) and triple-negative (MM-231, HCC1187) cells within 6 days. Triple-negative cells formed loosely packed microtumors compared with luminal MCF7, T47D, and BT474 consistent with the literature (31, 36) and can be linked to their aggressive and metastatic phenotype (31).

Size-controlled microtumors showed PI-positive cells in the central core and proliferating cells at the periphery, mimicking spatial distribution of cells in solid tumors as well as characteristics of previously reported multicellular aggregates (27, 28). Additionally, increased number of Ru-dpp– and PI-positive cells in 600-μm microtumors compared with 150-μm ones demonstrated size-dependent diffusion limitation of oxygen and nutrients leading to hypoxia and cell death in the center of large microtumors. We also observed increased expression of glucose transporter Glut1, suggesting existence of metabolic stress in large microtumors.

EMT is one of the hallmarks of aggressive and metastatic phenotypes in the advanced breast tumors (10–13). It is characterized by the loss of membranous E-cad and gain of N-cad (so-called “cadherin-switch”) along with upregulation of mesenchymal transcription factors SNAIL, SLUG, and TWIST (10, 13, 30). However, recent reports also suggest the existence of partial EMT in advanced breast cancer, where cells acquire mesenchymal characteristics while retaining well-differentiated epithelial characteristics (37–39). For example, expression of TWIST1, an EMT transcription factor, induced rapid dissemination of cytokeratin-positive epithelial cells without loss of membrane-localized E-cad and β-catenin (38). Consistent with these studies, we observed significant upregulation of mesenchymal markers (VIM, FIB, SNAIL, SLUG, and TWIST) with no change in E-CAD and collective migratory behavior in large microtumors of noninvasive MCF7 and T47D cells, implying acquisition of aggressive malignant phenotype. Thus, these 3D models recapitulate clinically observed positive correlation between tumor size and aggressive phenotype in luminal breast cancer. Surprisingly, we did not observe size-dependent changes in mesenchymal markers in triple-negative MM-231 and HCC1187 cells. Indeed, clinically tumor size is a weak prognostic marker for determining node positivity, metastatic risk, and survival in TNBC (21). Thus, size-controlled 3D microtumor models recapitulate clinically observed correlation between tumor size and aggressive phenotype in different breast cancer subtypes, further validating our 3D model.

Drug resistance is another characteristic of advanced tumors. Treatment with 4-OHT inhibited growth of large microtumors to a lesser extent than smaller ones, suggesting that large microtumors acquired drug resistant phenotype. Cumulatively, these results demonstrate metabolically distinct phenotypes of 150- and 600-μm microtumors made of the same cell line.

Unlike conventional EMT, the upregulation of mesenchymal marker expression in large microtumors was not accompanied by loss of E-cad (10, 13, 30). On the contrary, results showed 10-fold increased expression of E-cad in both, small and large microtumors compared with 2D monolayers. Indeed, similar overexpression of E-cad in inflammatory breast cancer is reported to contribute to disease aggressiveness and decreased patient survival (33). In contrast to Snail and Vim-negative small microtumors, we observed spatial heterogeneity of Vim-expressing cells in large microtumors. Cells in the outer periphery of large microtumors showed expression of mesenchymal protein, Vim while maintaining Vim- and Snail-negative phenotype of parental MCF7 in the central core. This is the first 3D in vitro model that has been able to recapitulate spatial intratumoral cellular heterogeneity in vitro. We believe that migratory behavior in 600-μm microtumor represents “collective cell migration” and not “single-cell migration”. Single-cell migration is characterized by the presence of single cells in the adjacent tissue, while in collective cell migration, cancerous tissue pushes forward as a whole (30). Tumor cells can migrate using any of these mechanisms depending upon cell types and microenvironmental context. Gain of peripheral expression of Vim along with maintenance of E-cad expression in large microtumors supports the observed collective migration where Vim-positive peripheral cells acquire mesenchymal phenotype pulling along the neighboring cells due to intact cell–cell contact.

Physiological hypoxia induces aggressiveness and metastatic phenotype in normal breast epithelial cells (22). Hif-1α is a master regulator of various signaling pathways and known to play an important role in tumor progression and angiogenesis in solid tumors (40). Thus, increased expression of Hif-1α and Vegf along with mesenchymal markers upregulation in large microtumors suggested possible role of Hif-1α in mesenchymal behavior of larger microtumors. However, only partial downregulation of FIB, SNAIL, and SLUG by siRNA knockdown and chemical inhibition of Hif-1α demonstrated that size-induced hypoxia might be one of the factors responsible for mesenchymal behavior observed in large microtumors.

Cell seeding density may affect 3D microtumor formation, spatial cellular distribution, and metastatic potential of cells within multicellular aggregates by modulating cellular metabolism, growth kinetics, and/or epigenetic adaptations (41). However, we confirmed that mesenchymal marker upregulation was size-dependent and not affected by cell density. This was further confirmed by size-reversal experiment wherein re-growing small microtumors into larger ones resulted in the mesenchymal marker upregulation, but re-growing larger ones into smaller sizes could not reverse mesenchymal marker expression to their basal levels, implying irreversible acquisition of mesenchymal phenotype, truly dependent upon microtumor size. As tumors grow in size, physical pressure and solid stress within the tumor and surrounding extracellular milieu (42) can also contribute to the EMT. It is demonstrated that TGF-β–induced EMT attenuates stiffness and stiffening response to force and increases invasion in normal murine mammary gland epithelial cells (43). Indeed, we also observed that large microtumors became less compact than the smaller ones on day 6, and this phenomenon was accompanied by acquisition of mesenchymal markers in larger ones. Thus, tumor size–induced changes in the physical pressure may be another factor in inducing the collective migration in large microtumors; however, further studies are needed to confirm this.

In summary, we report a number of important findings pertaining to breast cancer progression. First, the microfabrication technique can generate uniform yet defined size 3D microtumors in 6 days in a high-throughput manner, clearly an advantage over currently used laborious methods. Second, modulation of microtumor size using size-controlled microwells generated hypoxia, metabolic stress, intratumoral cellular heterogeneity and aggressive malignant as well as drug-resistant phenotype in large microtumors using the same non-invasive cell lines (MCF7/T47D), mimicking in vivo tumor growth-induced progression. This was achieved by creating controlled tumor microenvironments without exposure to external stimuli such as cultures in the hypoxic chambers or use of different cells lines representing non-invasive and invasive phenotypes. The model successfully recapitulated clinically observed positive association of tumor size with aggressive phenotype in luminal breast microtumors while also maintaining poor association in the case of TNBC. Large microtumors also recapitulated intratumoral spatial heterogeneity evidenced by Vim-positive cells only in the periphery, making this breast tumor model physiologically more relevant. However, the study also has some limitations. The observed migratory phenomenon in large microtumors could be cumulative results of various factors including hypoxia, oxidative, and metabolic stress that needs further investigation to dissect the mechanism of growth-induced breast tumor progression (44). The hypothesis of collective migration wherein peripheral Vim-expressing cells orchestrate the “migrating out” of the entire microtumor needs to be investigated further. More importantly, the exact role of different cell populations contributing to intratumoral heterogeneity and malignant progression in large microtumors needs detailed investigation. Nonetheless, clinically relevant microengineered 3D models generated within 6 days under controlled experimental conditions will serve as precise preclinical models to study mechanisms involved in breast tumor progression as well as antitumor drug effects as a function of tumor progression.

No potential conflicts of interest were disclosed.

Conception and design: S. Oesterreich, S. Sant

Development of methodology: M. Singh, S. Sant

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Singh, S. Mukundan, M. Jaramillo, S. Sant

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Singh, M. Jaramillo, S. Oesterreich, S. Sant

Writing, review, and/or revision of the manuscript: M. Singh, S. Oesterreich, S. Sant

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Singh, S. Sant

Study supervision: S. Sant

The authors thank Dr. Wen Xie, Director, Center for Pharmacogenetics, University of Pittsburgh School of Pharmacy, for core facilities access. They thank Drs. Paul Johnston and Vinayak Sant, Department of Pharmaceutical Sciences, for critical reading and insightful suggestions to improve the manuscript.

This work is supported by NIH funding (EB018575) to S. Sant and S. Oesterreich, the start-up funds from the Department of Pharmaceutical Sciences, School of Pharmacy, University of Pittsburgh awarded to S. Sant, and the Breast Cancer Research Fund awarded to S. Oesterreich.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Howlader
N NA
,
Krapcho
M
,
Garshell
J
,
Miller
D
,
Altekruse
SF
,
Kosary
CL
, et al
SEER cancer statistics review, 1975–2010, National Cancer Institute Bethesda
.
Bethesda, MD
:
National Cancer Institute
; 
2013
.
2.
Menezes
ME
,
Das
SK
,
Emdad
L
,
Windle
JJ
,
Wang
XY
,
Sarkar
D
, et al
Genetically engineered mice as experimental tools to dissect the critical events in breast cancer
.
Adv Cancer Res
2014
;
121
:
331
82
.
3.
Vargo-Gogola
T
,
Rosen
JM
. 
Modelling breast cancer: one size does not fit all
.
Nat Rev Cancer
2007
;
7
:
659
72
.
4.
Weigelt
B
,
Ghajar
CM
,
Bissell
MJ
. 
The need for complex 3D culture models to unravel novel pathways and identify accurate biomarkers in breast cancer
.
Adv Drug Deliv Rev
2014
;
69–70
:
42
51
.
5.
Alemany-Ribes
M
,
Semino
CE
. 
Bioengineering 3D environments for cancer models
.
Adv Drug Deliv Rev
2014
;
79–80
:
40
9
.
6.
Rundqvist
H
,
Johnson
RS
. 
Tumour oxygenation: implications for breast cancer prognosis
.
J Intern Med
2013
;
274
:
105
12
.
7.
Kim
JW
,
Ho
WJ
,
Wu
BM
. 
The role of the 3D environment in hypoxia-induced drug and apoptosis resistance
.
Anticancer Res
2011
;
31
:
3237
45
.
8.
Videira
M
,
Reis
RL
,
Brito
MA
. 
Deconstructing breast cancer cell biology and the mechanisms of multidrug resistance
.
Biochim Biophys Acta
2014
;
1846
:
312
25
.
9.
Gozgit
JM
,
Pentecost
BT
,
Marconi
SA
,
Otis
CN
,
Wu
C
,
Arcaro
KF
. 
Use of an aggressive MCF-7 cell line variant, TMX2-28, to study cell invasion in breast cancer
.
Mol Cancer Res
2006
;
4
:
905
13
.
10.
Takebe
N
,
Warren
RQ
,
Ivy
SP
. 
Breast cancer growth and metastasis: interplay between cancer stem cells, embryonic signaling pathways and epithelial-to-mesenchymal transition
.
Breast Cancer Res
2011
;
13
:
211
.
11.
Sutherland
RL
,
Hall
RE
,
Taylor
IW
. 
Cell proliferation kinetics of MCF-7 human mammary carcinoma cells in culture and effects of tamoxifen on exponentially growing and plateau-phase cells
.
Cancer Res
1983
;
43
:
3998
4006
.
12.
Scully
OJ
,
Bay
B-H
,
Yip
G
,
Yu
Y
. 
Breast cancer metastasis
.
Cancer Genomics Proteomics
2012
;
9
:
311
20
.
13.
Kalluri
R
,
Weinberg
RA
. 
The basics of epithelial-mesenchymal transition
.
J Clin Invest
2009
;
119
:
1420
28
.
14.
Chakrabarti
R
,
Kang
Y
. 
Transplantable mouse tumor models of breast cancer metastasis
.
Methods Mol Biol
2015
;
1267
:
367
80
.
15.
Eom
KY
,
Jang
MH
,
Park
SY
,
Kang
EY
,
Kim
SW
,
Kim
JH
, et al
The expression of Carbonic Anhydrase (CA) IX/XII and lymph node metastasis in early breast cancer
.
Cancer Res Treat
2016
;
48
:
125
32
.
16.
Pathi
SP
,
Lin
DD
,
Dorvee
JR
,
Estroff
LA
,
Fischbach
C
. 
Hydroxyapatite nanoparticle-containing scaffolds for the study of breast cancer bone metastasis
.
Biomaterials
2011
;
32
:
5112
22
.
17.
Bersini
S
,
Jeon
JS
,
Dubini
G
,
Arrigoni
C
,
Chung
S
,
Charest
JL
, et al
A microfluidic 3D in vitro model for specificity of breast cancer metastasis to bone
.
Biomaterials
2014
;
35
:
2454
61
.
18.
Benton
G
,
DeGray
G
,
Kleinman
HK
,
George
J
,
Arnaoutova
I
. 
In vitro microtumors provide a physiologically predictive tool for breast cancer therapeutic screening
.
PLoS One
2015
;
10
:
e0123312
.
19.
Michaelson
JS
,
Silverstein
M
,
Wyatt
J
,
Weber
G
,
Moore
R
,
Halpern
E
, et al
Predicting the survival of patients with breast carcinoma using tumor size
.
Cancer
2002
;
95
:
713
23
.
20.
Foulkes
WD
,
Reis-Filho
JS
,
Narod
SA
. 
Tumor size and survival in breast cancer—a reappraisal
.
Nat Rev Clin Oncol
2010
;
7
:
348
53
.
21.
Lai
HW
,
Kuo
SJ
,
Chen
LS
,
Chi
CW
,
Chen
ST
,
Chang
TW
, et al
Prognostic significance of triple negative breast cancer at tumor size 1 cm and smaller
.
Eur J Surg Oncol
2011
;
37
:
18
24
.
22.
Vaapil
M
,
Helczynska
K
,
Villadsen
R
,
Petersen
OW
,
Johansson
E
,
Beckman
S
, et al
Hypoxic conditions induce a cancer-like phenotype in human breast epithelial cells
.
PLoS One
2012
;
7
:
e46543
.
23.
Kwon
CH
,
Wheeldon
I
,
Kachouie
NN
,
Lee
SH
,
Bae
H
,
Sant
S
, et al
Drug-eluting microarrays for cell-based screening of chemical-induced apoptosis
.
Anal Chem
2011
;
83
:
4118
25
.
24.
Singh
M
,
Close
DA
,
Mukundan
S
,
Johnston
PA
,
Sant
S
. 
Production of uniform 3D microtumors in hydrogel microwell arrays for measurement of viability, morphology, and signaling pathway activation
.
Assay Drug Dev Technol
2015
;
13
:
570
83
.
25.
Victor
VM
,
Nunez
C
,
D'Ocon
P
,
Taylor
CT
,
Esplugues
JV
,
Moncada
S
. 
Regulation of oxygen distribution in tissues by endothelial nitric oxide
.
Circ Res
2009
;
104
:
1178
83
.
26.
Huang
J
,
Li
H
,
Ren
G
. 
Epithelial-mesenchymal transition and drug resistance in breast cancer (Review)
.
Int J Oncol
2015
;
47
:
840
8
.
27.
Lin
R-Z
,
Chang
H-Y
. 
Recent advances in three-dimensional multicellular spheroid culture for biomedical research
.
Biotechnol J
2008
;
3
:
1172
84
.
28.
Hirschhaeuser
F
,
Menne
H
,
Dittfeld
C
,
West
J
,
Mueller-Klieser
W
,
Kunz-Schughart
LA
. 
Multicellular tumor spheroids: an underestimated tool is catching up again
.
J Biotechnol
2010
;
148
:
3
15
.
29.
Bae
H
,
Chu
H
,
Edalat
F
,
Cha
JM
,
Sant
S
,
Kashyap
A
, et al
Development of functional biomaterials with micro- and nanoscale technologies for tissue engineering and drug delivery applications
.
J Tissue Eng Regen Med
2014
;
8
:
1
14
.
30.
Cheung
KJ
,
Ewald
AJ
. 
Illuminating breast cancer invasion: diverse roles for cell-cell interactions
.
Curr Opin Cell Biol
2014
;
30
:
99
111
.
31.
Kenny
PA
,
Lee
GY
,
Myers
CA
,
Neve
RM
,
Semeiks
JR
,
Spellman
PT
, et al
The morphologies of breast cancer cell lines in three-dimensional assays correlate with their profiles of gene expression
.
Mol Oncol
2007
;
1
:
84
96
.
32.
Myungjin Lee
J
,
Mhawech-Fauceglia
P
,
Lee
N
,
Cristina Parsanian
L
,
Gail Lin
Y
,
Andrew Gayther
S
, et al
A three-dimensional microenvironment alters protein expression and chemosensitivity of epithelial ovarian cancer cells in vitro
.
Lab Invest
2013
;
93
:
528
42
.
33.
Morales
J
,
Alpaugh
ML
. 
Gain in cellular organization of inflammatory breast cancer: a 3D in vitro model that mimics the in vivo metastasis
.
BMC Cancer
2009
;
9
:
462
.
34.
Verschraegen
C
,
Vinh-Hung
V
,
Cserni
G
,
Gordon
R
,
Royce
ME
,
Vlastos
G
, et al
Modeling the effect of tumor size in early breast cancer
.
Ann Surg
2005
;
241
:
309
18
.
35.
Lee
JH
,
Kim
SH
,
Suh
YJ
,
Shim
BY
,
Kim
HK
. 
Predictors of axillary lymph node metastases (ALNM) in a Korean population with T1-2 breast carcinoma: triple negative breast cancer has a high incidence of ALNM irrespective of the tumor size
.
Cancer Res Treat
2010
;
42
:
30
6
.
36.
Imamura
Y
,
Mukohara
T
,
Shimono
Y
,
Funakoshi
Y
,
Chayahara
N
,
Toyoda
M
, et al
Comparison of 2D- and 3D-culture models as drug-testing platforms in breast cancer
.
Oncol Rep
2015
;
33
:
1837
43
.
37.
Dubois-Marshall
S
,
Thomas
JS
,
Faratian
D
,
Harrison
DJ
,
Katz
E
. 
Two possible mechanisms of epithelial to mesenchymal transition in invasive ductal breast cancer
.
Clin Exp Metastasis
2011
;
28
:
811
8
.
38.
Shamir
ER
,
Pappalardo
E
,
Jorgens
DM
,
Coutinho
K
,
Tsai
W-T
,
Aziz
K
, et al
Twist1-induced dissemination preserves epithelial identity and requires E-cadherin
.
J Cell Biol
2014
;
204
:
839
56
.
39.
Zhang
J
,
Tian
X-J
,
Zhang
H
,
Teng
Y
,
Li
R
,
Bai
F
, et al
TGF-β–induced epithelial-to-mesenchymal transition proceeds through stepwise activation of multiple feedback loops
.
Sci Signal
2014
;
7
:
ra91
ra91
.
40.
Gilkes
DM
,
Semenza
GL
. 
Role of hypoxia-inducible factors in breast cancer metastasis
.
Future Oncol (London, England)
2013
;
9
:
1623
36
.
41.
Griffith
LG
,
Swartz
MA
. 
Capturing complex 3D tissue physiology in vitro
.
Nat Rev Mol Cell Biol
2006
;
7
:
211
24
.
42.
Shieh
AC
. 
Biomechanical forces shape the tumor microenvironment
.
Ann Biomed Eng
2011
;
39
:
1379
89
.
43.
Osborne
LD
,
Li
GZ
,
How
T
,
O'Brien
ET
,
Blobe
GC
,
Superfine
R
, et al
TGF-beta regulates LARG and GEF-H1 during EMT to affect stiffening response to force and cell invasion
.
Mol Biol Cell
2014
;
25
:
3528
40
.
44.
Martinez-Outschoorn
U
,
Sotgia
F
,
Lisanti
MP
. 
Tumor microenvironment and metabolic synergy in breast cancers: critical importance of mitochondrial fuels and function
.
Semin Oncol
2014
;
41
:
195
216
.