Development of cancer therapeutics is an ongoing effort by researchers in the academy and pharmaceutical industry. To evaluate optimal dose of therapeutics, conventional two-dimensional (2D) cell cultures are utilized prior to testing on animal cancer models. However, 2D culture models do not mimic the complexity of the tumor microenvironment (tumor stroma). The interactions between the cells and their microenvironment govern various processes, such as cell differentiation, proliferation and gene expressions in regulation of tumor initiation and progression1. While animal experiments are necessary prior to any clinical trials, there is a large gap in the knowledge obtained between 2D in vitro and in vivo models to completely understand the therapeutic efficiency2. Data from 2D models rarely predicts magnitudes of therapeutic efficiency in vivo. One of the explanations for these discrepancies is the fact that in vivo cells are arranged in three-dimensional (3D) structures and not attached to planar surfaces. In vitro 3D cultures provide an additional step that can bridge the gap between conventional 2D culture and animal models3. It was shown that in vitro 3D cultures enable a better understanding of the molecular and cellular mechanisms, which are more relevant to animal and human studies, thus facilitating the development and screening of new drugs2,4. This affects several aspects related not only to cell-cell interactions, but also to biophysical parameters such as transport of nutrients and therapeutics to different cell populations.

One of the main requirements for a representative 3D in vitro tumor system is the presence of a scaffold that can support cancer cells, allow for nutrient, gas and signal exchanges among cells and mimic extracellular matrix (ECM) conditions. Current scaffolds used are either made from synthetic polymers, such as polyethylene glycol, which is not an appropriate material for cellular recognition, or naturally-derived polymers, such as collagen, which often poses difficulty to produce a controlled matrix5. Biodegradable scaffolds have also been tested, but cells may display slow growth and delayed formation of cell-cell interactions, causing a misrepresentation of the in vivo environment. Additionally, commercially-available Matrigel™, is commonly used for 3D culture, which is a reconstituted basement membrane from the mouse Englebreth-Holm-Swarm tumor6. Matrigel's animal-derived origins, however, bring concern misrepresenting human tumors and potentially affect experimental results.

In order to accurately mimic the in vivo environment, 3D in vitro models without scaffolds have been produced, such as the spheroid model. The spheroid model is a popular approach, especially with breast cancer stem cells, in which cells form heterogeneous aggregates with each other and do not attach to an external surface for support. This model has shown to provide more relevant data than the same cells in the 2D configuration due to the natural formation of cell-cell interactions and the production of tumor-like hypoxia and necrotic regions7. The spheroid model, however, does not take into account the presence of and influence from an important tumor component: the stroma.

The breast tumor stroma consists of fibroblasts, adipocytes, endothelial cells and inflammatory cells with many different enzymes and growth factors, which makes up to 80% of a tumor8,9. Thus the addition of these other cells in an in vitro model significantly changes cell-cell contacts and signals within tumors10. Moreover, the heterogeneous tumor environment affects cell proliferation rates, produces irregular regions of acidity and hypoxia and influence malignant cell transformations, impacting the sensitivity of tumor to therapeutics11.

In this study, we investigate a novel in vitro model to mimic heterogeneous breast tumors without the use of a scaffold while allowing for homotypic and heterotypic cell-cell interactions. Breast cancer cells were co-cultured with fibroblasts and then magnetically levitated. It was shown that the conventional culturing conditions using the magnetic levitation system can result in the formation of 3D structures, closely resembling in vivo tissues (i.e. adipocytes12, vascular smooth muscle cells13). In this paper, 3D heterogenous tumor models for breast cancer were designed to effectively model the influences of the tumor microenvironment on drug efficiency. First, the formation of the 3D in vitro breast tumor using various breast cancer and fibroblast cell lines were physically characterized (tumor size and density). Then, the model was phenotypically compared to the in vivo tumor and examined for the presence of ECM proteins. Lastly, the effect of the presence of a tumor stroma on drug transport and efficiency in the 3D in vitro tumor model were assessed.


Fig. 1A schematically represents the step-by-step methodology of formation of 3D in vitro breast tumors using a co-culture of breast cancer and fibroblast cells. Breast cancer and fibroblast cells pre-incubated with Nanoshuttles™ were co-cultured at different ratios and magnetically levitated. In less than 24 h, 3D structures containing a mixture of the breast cancer and fibroblast cells were formed. The diameter and optical density of 3D in vitro breast tumors were dependent upon the number of cells initially seeded (Fig. 1B, C). It has been previously demonstrated that 3D in vitro cultures can be grown for up to 12 weeks using the magnetic levitation system6. This 3D in vitro breast tumor model with 50:50 ratio of breast cancer cells to fibroblasts was also grown for up to ten days with replenishment of cell media. Further, histological analysis has shown that the 3D in vitro breast tumor model grown for ten days resembles tumors grown in vivo (Fig. 1D). We found that the combination of 70% Fibroblast and 30% breast cancer cell worked the best in the SCID mice model to mimic the condition in human breast cancer which has more stroma than usual breast cancer models in mice (based on clinical samples, data not shown in this manuscript). A few ratios of tumor cells to fibroblasts were injected and, although 293T cells were the fastest dividing cells in in vitro system, allowing them to occupy higher volumes after being co-cultured with breast cancer cells, their in vivo growth was less prominent when seeded in 50:50 ratio. Thus, we decided to compare the above two ratios between in vivo and in vitro experiments. Both in vitro and in vivo samples showed disorganized cell structures, which is typical to tumor cells and both featured similar composition of stromal and cancer cells. Similar structures can also be observed in previously published studies in vivo with MDA-MB-231 cells14.

Figure 1
figure 1

(A) Schematic representation of the development of 3D in vitro breast tumor model from a dispersed mixture of fibroblasts (in blue) and breast cancer cells (in green) after the addition of nanoshuttles. The magnet placed on top aids the attraction of the nanoshuttle internalized cells to form a 3D tumor mass composed of fibroblasts and breast cancer cells and can be grown for several days. (B,C) Plot of the diameter and optical density of 1 day grown 3D in vitro breast tumors in 96-well plates, which was measured by GelCount ®. (D) H&E staining of 3D in vitro tumor models at 50% breast cancer and 50% fibroblast ratio grown in 6 well plates for 10 days comparing to 14 day old in vivo mouse breast tumor, Scale bar = 20 μm.

The 3D in vitro breast tumor model was compared to the conventional 2D in vitro co-cultures with various breast and fibroblast cell lines. Fibroblasts and breast cancer cells were fluorescently labeled with Tracer DiL and Tracer DiO dyes, respectively. Fig. 2 displays the fluorescent images of different cell mixtures grown for 3 days in 2D and 3D configurations. 2D in vitro co-cultures of breast cancer and fibroblast cells displayed a dispersed state of growth and attachment, forming a monolayer. 3D in vitro model resulted in the formation of tumors with heterogeneous cell distribution and in vivo tissue like structure and morphology.

Figure 2
figure 2

Comparison of 2D co-culture with the 3D in vitro breast tumor model composed with different ratios and types of breast cancer cells (in green) to different types of fibroblasts (in red) grown for 3 days in cell culture conditions (37 C, 5% CO2).

Blue signal is from DAPI, staining the nucleus. Images were taken with 10 × objective magnification, scale bar = 100 μm.

Heterogeneous breast tumors were grown with different ratios of fibroblasts and breast cancer cells using the magnetic levitation system. It is clearly demonstrated that fibroblasts encapsulate breast cancer cells in the 3D in vitro system. There is a clear rearrangement of the cells in 3D in vitro culture showing a tumor tissue-like organization. This organization is dependent upon specific characteristics of its cellular components. Similarly to in vivo situation, the ratio of fibroblasts (red) and cancer cells (green) depended heavily on their nature. It can be seen from Figures 2 and 3 that different cancer cell lines and different fibroblasts vary in their growth patterns and organization, producing heterogeneous in vitro tumors. This heterogeneity cannot be seen when the cells were grown in 2D. Moreover, growth rate plays an important role in the formation of 3D tumors in vitro. As the fastest growing cell line, samples with 293T cell lines (Fig. 2) were dominated by the fibroblast in most of the combinations tested. Other samples with fibroblasts with slower proliferation rate such as Hs578bst and CAF (Figures 2 and 3) seemed to be growing in sync with the cancer cells and maintained similar ratio to the seeded cells. Primary cancer breast tumor associated fibroblasts, CAF, formed tight spheres with breast cancer cells. Interestingly, all samples with fibroblast cell lines displayed the accumulation of fibroblasts in the periphery region of the spheres, while with the primary CAF cells, the localization of fibroblasts is less pronounced in the periphery. Based on the self-organization of the cells in the produced 3D tumors, in a majority of cases, the concentration of fibroblasts at the tumor edge is substantially higher than in the core, which corresponds to the fibrotic capsule phenomenon observed in vivo (Fig. 4A). Additionally, the structure with CAF cells were less organized and formed rough edges of the spheres, showing the slow growing rate of CAF might inhibit the formation of relevant model for in vitro spheres. CAF, isolated from tumor lesion of cancer patients and able to secrete relevant factors for tumor survival and growth, is a suitable cell line for modelling fibroblasts in co-culture tumor model. However, the characteristics of slow growth and low availability may limit its utilization for rapid high-throughput assays. The results showed the versatility of the method to obtain various tumor types depending on the cell ratio and the types of cells used in the model, opening the possibility to study heterogeneous tumor types in high throughput system.

Figure 3
figure 3

3D in vitro tumors grown with primary fibroblasts.

3D in vitro breast tumor model with different ratios and types of breast cancer cells (green) and primary fibroblasts (red) co-cultured for 3 days using magnetic levitation system. All cells were counterstained with DAPI (blue) for nucleus. (A) Images taken by confocal microscopy of the spheres. (B–D) Fluorescent images of the spheres taken after cryo-sectioning (4 micron slices) (B) MDA-MB-231 and Hs578bst (C) MDA-MB-231 and HPF (D) MDA-MB-231 and CAF. Images were taken with 10 × objective magnification, scale bar = 100 μm.

Figure 4
figure 4

Characterization of in vitro 3D co-cultures: (A) Fluorescent images comparing the phenotype between 3D in vitro co-culture grown in 6 well plates and in vivo tumors composed of breast cancer cells (green signal) and fibroblasts (red signal) after 7 days growth. Blue signal is from DAPI, staining the nucleus. Yellow arrows indicate the fibrotic capsule formed in both 3D in vitro and in vivo tumors. Scale bar = 100 μm. (B) Collagen, fibronectin and vimentin (in blue) immuno-fluorescent whole mount staining on 7 days old 3D in vitro mono- and co- cultures for breast cancer (in green) to fibroblast cells (in red) grown in 24 well plate, overlayed on brightfield image of tumor, Scale bar = 100 μm (C) Fibronectin concentration detected in 7 days grown 2D and 3D mono- and co-cultures of breast cancer and fibroblasts cells. F = fibroblasts (293T) and B = breast cancer cells (SUM159).

The observed fibrotic capsule in 3D in vitro tumors expressed different levels of common ECM proteins, such as collagen, vimentin and laminin (Fig. 4B). Moreover, cells in 3D produced higher concentrations of fibronectin than cells in 2D form (Fig. 4C), providing evidence that co-culture of breast cancer and fibroblasts cells can produce an ECM matrix without a scaffold.

Evidently the density of a tumor can also be controlled by the number of cells initially seeded in to the system. Optical density, or the amount of light that can pass through an object, was used as a parameter to measure tumor density. The darker the tumor appeared in bright field images, a denser tumor can be assumed. Low density tumors were formed by seeding 100,000 cells and high density tumors were formed by seeding 300,000 cells grown for 1 day. To test the penetration of molecules in tumors with different densities, TRITC-tagged dextran (70 kDa) was administered and the change in the TRITC intensity was observed over time. As expected, penetration of dextran through low density tumors was 35% greater than high density tumors after 2 h (Fig. 5A). Similarly, penetration was studied on tumors grown with low (30% fibroblasts/70% breast cancer cells) and high (70% fibroblasts/30% breast cancer cells) fibroblast content. Penetration through tumors with less fibroblast cells was around 10% greater than tumors with high fibroblast content (Fig. 5B).

Figure 5
figure 5

Penetration of TRITC-tagged dextran: (A) 3D in vitro tumors at low and high density tumors (normalized to highest measured TRITC intensity from dextran penetration after 2 hours in low density tumors); (B) 3D in vitro tumors with low (30% fibroblasts) and high (70% fibroblasts) stromal content (normalized to TRITC-tagged dextran penetration profile in 3D in vitro tumor composed of 100% breast cancer cells (0% fibroblasts).

To further examine the 3D in vitro tumor model as an in vivo representation, the penetration and effects of clinically-used anticancer drugs, doxorubicin and Doxil®, was assessed (Fig. 6A). Fluorescent images displayed differences in fluorescent intensity and location of the two formulations within the tumor. Low molecular weight (579 Da) doxorubicin, exhibited higher penetration into the 3D in vitro tumor cultures when compared to the 90 nm liposomal formulation of doxorubicin (Doxil®).

Figure 6
figure 6

Distribution and therapeutic efficacy of doxorubicin and Doxil® on 3D in vitro tumors: (A) Fluorescent images of 3D in vitro tumors composed of mono- or co-culture of fibroblast (in red) and breast cancer cells (in green) –inset images, comparing 72 h treatment with doxorubicin and Doxil ®, blue – nucleus and orange – fluorescent emission from doxorubicin, Scale bar = 100 μm, (B) Viability assay treated with either doxorubicin or Doxil ® (100 nM) for 72 h, comparing on three different in vitro systems: (1) 2D in vitro (grown for 1 day), (2) 3D in vitro (grown for 1 day) and (3) 3D Matrigel™ (grown for 7 days) * = statistically significant difference to 2D in vitro with the same treatment, n = 4, p < 0.05.

F = fibroblast (293T) and B = breast cancer cells (SUM159).

Cell viability after doxorubicin/Doxil® treatment was measured using the WST-1 assay (Fig. 6B). For comparison, 2D co-cultures and 3D Matrigel™ mono- and co- cultures were treated for 72 h with the same drugs. Cell viability was normalized to control (cells treated with media alone). Doxorubicin and Doxil® significantly affected the viability in 2D systems when compared to the 3D systems. In both the Matrigel™ and the magnetic levitation system for 3D models, co-cultured tumors treated with doxorubicin exhibited 10% lower overall cell viability when compared to tumors treated with Doxil®.

Tumor area and optical density were measured for monitoring growth of 3D in vitro breast tumors in response to doxorubicin treatment (Fig. 7A, B and C). The area and density for 3D in vitro tumors composed of mono- and co-culture of fibroblasts and breast cancer cells increased during the 7 days growth period. On day 7, doxorubicin (100 nM) was added to the 3D in vitro mono- and co- cultures. For 3D in vitro culture of fibroblasts, the area decreased by 87% for up to 5 days after doxorubicin treatment (from day 7 to day 12). The density, however, did not change after doxorubicin treatment. For the 3D in vitro co-culture of breast cancer and fibroblast cells, the area decreased by around 80% and the density decreased by 45% 5 days after doxorubicin treatment. The decrease in tumor size and density after doxorubicin treatment is consistent with in vivo findings15. For the 3D in vitro tumor composed of only breast cancer cells, although the area and density changed over time, doxorubicin treatment did not cause additional changes in the tumor area or optical densities.

Figure 7
figure 7

Effect of doxorubicin treatment on 3D in vitro tumors: (A) Tumor area measurements and (B) Optical density measurements for 3D in vitro tumor composed of mono- and co-cultures of breast cancer and fibroblast cells. Arrows designate the start of doxorubicin (100 nM) treatment to 3D in vitro tumors. F = fibroblast and B = breast cancer cells.


One of the major barriers to the development of efficient cancer therapeutics is the availability of experimental models that can accurately characterize various forms of cancer16. While animal models have been used extensively, these studies are time-consuming and may not be representative of the human tumors17. Thus there is a need to develop in vitro models that can better represent the human tumors, while still keeping the study times to a minimum. Many three dimensional models have been developed to study cell behaviors such as cell-cell as well as cell-matrix interactions17. Recently, the tumor-stroma ratio has been found to be a prognostic factor for cancer with higher stromal component contributing to poor prognosis and increased risk of relapse18,19,20. Thus, it is important to incorporate the stromal components in in vitro breast tumor models. Fibroblasts is one of the prominent players in tumor stroma and responsible for secretion of a collagenous ECM and producing different enzymes, inhibitors, growth factors and structural proteoglycans17,21. Most breast tumors are characterized by loss of epithelial cell polarity leading to a disorganized cellular architecture with randomly interspersed tumor cells and fibroblasts. In this study, we used the magnetic levitation system to mimic this disordered tumor formation in vitro by co-culturing breast cancer and fibroblast cells. Additionally, it serves as a versatile tool to form heterogeneous types of 3D in vitro breast tumor at various sizes, densities and compositions by controlling the number and type of cells. Thus, this 3D in vitro tumor system proves to be an important model to produce breast tumors with different physical characteristics to better understand tumor biology and drug efficiency.

The rate of tumor growth is fast when using the magnetic levitation system compared to the more commonly used 3D cultures formed with Matrigel™ and the tumor forms intrinsic ECM and does not rely on the extrinsic ECM components supplemented to the culture. A tumor structure formed by the magnetic levitation system was observed within 24 h, whereas, the formation of in vitro tumor with the same type of cells using Matrigel™ were observed to be slow and delayed and was only comparable after 7 days growth. It has been observed that 3D in vitro cultures made with Matrigel™ and other scaffolds take a long time to accurately mimic the in vivo tumor phenotypically22,23. Additionally, the cluster size in these slow growing tumor models are generally not large – they are at most in μm2 dimensions23,24,25. This is important for recapitulating necrotic and/or hypoxic areas in tumors, which occurs only when the tumors are much larger than these dimensions. In the magnetic levitation system, 3D in vitro tumors were observed to be in the size range of mm2, especially when grown in the 6-well configuration and so can be used to characterize hypoxic regions in the tumor.

Another study has also demonstrated similar formation of large co-cultured tumors of breast cancer in short time scales. However, this model requires the use of an external scaffold to accurately represent the tumor ECM26. While using external scaffolds to mimic the ECM is feasible and representative for a majority of cell cultures, it may not always accurately characterize the breast tumor. For example, a common scaffold, Matrigel™, is animal derived and may contain endogenous growth factors and signals that do not represent the human tumor environment. These might have independent effect on the tumors and can affect drug transport and efficiency studies27. More relevant and naturally formed ECM components (collagen, laminin and fibronectin) provide an advantage over other systems28. In the magnetic levitation system, tumor and fibroblast cells are allowed to interact with each other and naturally form a complex matrix, thus mimicking a more relevant tumor environment without externally-added or other species-derived components.

The presence of ECM in tumors affects drug efficiency due to the molecule's inability to penetrate through the complex matrix and reach the targeted cancer cells11. It has been previously shown that in vivo tumors with rich collagen networks inhibited penetration of high molecular weight drugs when compared to low molecular weight drugs29. Similarly, lower drug penetration was observed in densely packed in vivo tumors than in low density tumors15. The 3D in vitro breast tumor made by magnetic levitation system was able to mimic similar physical barriers (stromal composition and density) that are observed in vivo. Due to the impaired drug penetration, the cell viability in 3D in vitro tumors were not drastically reduced in response to both doxorubicin and Doxil® when compared to 2D cultures, in which they showed to cause significant cell death. Similar results of doxorubicin penetrating more through 2D mono-cultures than through 3D in vitro mono-culture for liver cancer were reported30. The study attributed the difference in penetration on the presence of ECM. Further, the fluorescent signal from doxorubicin (a small molecule) in the 3D in vitro breast tumors was higher than from Doxil ® (the liposomal form of doxorubicin), confirming that small molecules were able to penetrate better than large drug structures through the 3D in vitro breast tumor. Correspondingly, due to the high penetration of doxorubicin in this present study, cell viability of 3D in vitro culture treated with doxorubicin was lower than those treated with Doxil ®. Comparable results were observed in 3D tumors grown using Matrigel™ system, signifying that either the production or addition of ECM plays an important role in accurately representing drug efficiency under in vitro conditions.

In clinical practice, efficiency of tumor therapy is monitored by the tumor size and density over time. The 3D in vitro breast tumor model was used to monitor similar physical characteristics of the tumor by measuring the changes in tumor area and optical density after treatment with doxorubicin for breast tumors in vivo15. While there is evidence that cell viability for mono-culture of breast cancer cells in the 3D in vitro tumor model is affected following doxorubicin treatment, it is not reflected in tumor area and optical density. It has been shown that certain tumor cell mono-cultures tend to migrate closer to each other, decreasing in size, when using the magnetic levitation system31. The observed decrease in area and increase in density validates that the monoculture of breast cancer cells in the 3D in vitro configuration is dominated primarily by migration and therefore, physical changes of the tumor from doxorubicin is difficult to detect using the GelCount™. The use of GelCount™, however, enabled monitoring of growth and physical changes for the most relevant tumor model – the 3D in vitro co-culture of breast cancer and fibroblast cells.

In conclusion, taking into account the tumor environmental and physical factors that may impair drug transport are essential for understanding overall drug efficiency. This piece of knowledge is conventionally discovered during in vivo animal experiment. To understand drug transportation mechanisms prior to the use of in vivo models provides researchers an immense awareness and prediction on drug efficiency. Therefore, the 3D in vitro breast tumor model allows researchers to re-create or modify their drug designs prior to initiating expensive, time-consuming animal experiments. The results from this study suggest that the proposed 3D in vitro breast tumor is advantageous due to the ability – (1) to form large-sized breast tumor models within 24 h, (2) to control tumor compositions and densities, (3) to accurately mimic the in vivo tumor microenvironment and (4) to test drug efficiency in a more representative model for in vivo tumors.



Dulbecco's Modified Eagle Medium (DMEM), fetal bovine serum (FBS) and penicillin-streptomycin (P/S) were purchased from Gibco® (Invitrogen Corporation, Carlsbad, CA). SUM159 cell line was purchased from Asterand, Inc (Detroit, MI) and MDA-MB-231, Hs785bst, Hs371.t and 293T were purchased from ATCC (American Tissue Culture Collection, Manassas, VA). Vybrant™ Cell-Labeling (Molecular Probes) was used to fluorescently label the cells. Breast cancer cells were labeled with Tracer DiO (Em: 501 nm – Green Signal) and fibroblast cells were labeled with Tracer DiD (Em: 665 nm – Red signal) or DIL (EM 565 nM). Nanoshuttles™ and the Bio-Assembler System™ were purchased from n3D Biosciences, Inc. (Houston, TX). Doxil® was a kind gift from Prof. Y. and doxorubucin and Dextran-TRITC (70 kDa) were purchased from Sigma Aldrich (St. Louis, MO). Matrigel™ (Basement Membrane Matrix, Growth Factor Reduced (GFR), Phenol Red-Free, LDEV-free) was purchased from BD Biosciences (Franklin Lakes, NJ) and WST1 Solution was purchased from Roche Diagnostics Corporation (Indianapolis, IN). Mammocult media was obtained from Stem cell technologies.

Cell culture

Human Pulmonary Fibroblasts (HPF) SUM159, MDA-MB-231, 293T, Hs785bst and Hs371.t cell lines were cultured in DMEM with 10% FBS and 1% penicillin and streptomycin at 37°C in 5% CO2 conditions for both 2D and 3D cultures. CAF were patient derived tumor associated fibroblast cells, obtained by mechanical shearing and separation of patient tumors into stromal and epithelial components. The cells were maintained in 50% Mammocult media and 50% DMEM with 10%FBS.

Growth of 3D in vitro tumor model

The Bio-Assembler™ System (in 6-, 24- and 96- well configuration) from n3D Biosciences, Inc (Houston, TX) was used to construct the 3D in vitro breast tumor model. The assembly of 3D in vitro cultures was performed as previously reported32. In short, Nanoshuttles™(NS) were added to cells at a ratio of 1 μL of NS per 10,000 cells and incubated at 37°C overnight. Cells for 2D and 3D cultures were fluorescently labeled with either Tracer DiO or Tracer DiL (5 μL dye per 1 ml of media) for approximately 30 minutes at 37°C. The cells for 3D cultures were then detached and co-cultured in the Bio-Assembler™ system. For example, when using a 24-well ultralow-attachment plate (Corning, Inc. Tewksbury, MA), a total of 100,000 cells were added in each well at different ratios of breast cancer to fibroblast cells in a total volume of 350 μL of fibroblast medium. Immediately afterwards, a special 24 well lid insert was placed on top of the plate, followed by the magnetic driver which sits inside the insert and then the 24 well plate on top. The plate was gently shaken to agitate cells and placed in the incubator. After 4 hours, the 24 well plates were briefly observed with bright field microscopy using a 2.5 × objective to determine if the structures were forming cohesive levitating structures. In brevity, most cells aggregate within a few hours and continue to become denser as more time passes. Fibroblast cells tend for form more dense spheres whereas epithelial-like cells tend to form more sheet like structures and remain flat. After 48 hours, structures from each condition were magnetically removed using a custom magnetic pen (Teflon cap and magnetic cylinder insert), then transferred to cryomolds and were frozen in OCT, cut into 4 μm sections and processed for histological imaging. Histological cut and Haematoxylin and Eosin staining was performed by Baylor College of Medicine and Houston Methodist Pathology Cores. 2D co-cultures were cultured on slides and then imaged. The slides were fixed in 4% paraformaldehyde (PFA) for fluorescent imaging. Imaging was performed using the NIKON® Eclipse fluorescent upright microscope at 10 × magnification.

Whole mount staining for 3D cultures

All immunohistochemistry steps were carried out within a 96 well ultra-low attachment plate while using a custom magnetic plate to hold down 3D cultures during washing and aspiration steps. After being fixed with 4% PFA, all structures were washed with PBS and then permeabilized with 0.2% Triton-X100 for 10 minutes on an orbital shaker. The structures were then blocked with 10% donkey serum in PBS for 15 minutes and incubated with either mouse anti-collagen type I (Millipore), mouse anti-vimentin (Calbiochem), or rabbit anti-laminin (Abcam) at a concentration of 1:100 in PBS with 1% bovine serum albumin overnight at room temperature (RT) while being shaken. The samples were washed and then incubated at RT with either donkey anti-mouse 555 or donkey anti-rabbit 555 (AlexaFluor 555; Invitrogen) at a concentration of 1:400 in PBS for 1 h. Samples were washed again and then counterstained with using 4′,6-diamidino-2-phenylindole (DAPI; KPL) at 1:1000 in PBS. Samples were washed 4 times and kept in PBS until imaged. Imaging was performed using the NIKON® Eclipse fluorescent upright microscope at 10 × magnification.

In vivo tumor sample

3–4 weeks old SCID-beige female mice (Harlan) were used as the animal model to grow breast tumors. Fluorescently labeled cells at a ratio of 70% fibroblasts (293T) to 30% breast cancer cells (SUM159) were injected into the mammary fat pad. After 7 days, tumors were collected, flash frozen in OCT and cut in 4 μm sections for fluorescent imaging. All animal experiments were performed in accordance to and approved by the Houston Methodist Research Institute, Institutional Animal Care and Use Committee (# AUP-1112-0050).


The QuantiMatrix Human Fibronectin ELISA kit (EMD Millipore, Inc., Billerica, MA) with minimum detection limit 10–20 ng/ml fibronectin was used to measure the concentration of fibronectin in 2D and 3D whole cell lysates in in vitro cultures. The standards and samples were prepared as per manufacturer's protocol. Samples were used fresh and were diluted at 1:10 ratio before the analysis.

Dextran penetration study

The amount of dextran, a model therapeutic agent, penetrating the cultures was studied by measuring the change in TRITC fluorescent intensity over time (0–120 min). 3D in vitro tumors formed with different cell numbers and compositions were incubated with TRITC-tagged dextran (70 kDa) for upto 2 h. Time lapse images were taken using the NIKON® Inverted microscope. The change in fluorescence intensity in the 3D cultures was analyzed using the NIS Elements software.

Treatment with anti-cancer drugs

Fluorescently-labeled 3D in vitro mono- and co- cultures of breast cancer and fibroblast cells were incubated with doxorubicin or Doxil® (at 100 nM doxorubicin concentration) for 72 h frozen in OCT, sectioned and fluorescently imaged for doxorubicin (Ex: 480 nm/Em: 580 nm). WST1 assay for cell viability was used to examine 2D and 3D in vitro systems after 72 h of treatment with either doxorubicin or Doxil®. Briefly, cells were incubated with 10% WST1 solution for 2 h at 37°C and the absorbance at 450 nm was measured. Three dimensional cultures formed using the standard Matrigel™ were compared33. To obtain the comparable volume of the tumor spheres, the cells embedded in Matrigel™ had to be in culture for 7 days. Student T-test was performed on triplicates with 95% confidence at p < 0.05.

Monitoring of 3D in vitro tumor model

The 3D in vitro tumors were imaged using the GelCount™ instrument and tumor parameters such as diameter, optical density and cross-sectional area were measured. The growth of the tumors was monitored by changes in area and optical density for 7 days. Then, the in vitro tumors were treated with doxorubicin (100 nM) the changes in area and optical density were compared to non-treated 3D in vitro tumors. The Gelcount™ instrument uses visible (i.e full spectrum) light to generate images and form the optical density values. ImageJ Analysis was used to quantify GelCount™ images.