3D bioprinted mammary organoids and tumoroids in human mammary derived ECM hydrogels

Abstract

The extracellular matrix (ECM) of tissues is an important mediator of cell function. Moreover, understanding cellular dynamics within their specific tissue context is also important for developmental biology, cancer research, and regenerative medicine. However, robust in vitro models that incorporate tissue-specific microenvironments are lacking. Here we describe a novel mammary-specific culture protocol that combines a self-gelling hydrogel comprised solely of ECM from decellularized rat or human breast tissue with the use of our previously described 3D bioprinting platform. We initially demonstrate that undigested and decellularized mammary tissue can support mammary epithelial and tumor cell growth. We then describe a methodology for generating mammary ECM extracts that can spontaneously gel to form hydrogels. These ECM hydrogels retain unique structural and signaling profiles that elicit differential responses when normal mammary and breast cancer cells are cultured within them. Using our bioprinter, we establish that we can generate large organoids/tumoroids in the all mammary-derived hydrogel. These findings demonstrate that our system allows for growth of organoids/tumoroids in a tissue-specific matrix with unique properties, thus providing a suitable platform for ECM and epithelial/cancer cell studies.

1. Introduction

Gaining a better understanding of the microenvironmental signals that dictate tissue homeostasis is critical to achieve a complete understanding of cell fate and cancer progression. In particular, the adipose tissue compartment found within the breast is comprised of a complex mixture of cells, structural extracellular matrix (ECM), and tethered signaling molecules [1–5]. The protein signature of ECMs from different epithelial tissues can vary greatly. Notably, those of the mammary gland encompass a unique microenvironment that governs tissue identity [5]. This characteristic is a critically important element of maintaining a normalized tissue, where breakdowns in microenvironmental cues are pivotal in promoting tumorigenic events [6–12]. Furthermore, we have recently demonstrated that derivatives of normal mammary ECM are capable of redirecting non-mammary cells co-injected into the cleared mouse mammary gland [8]. Importantly, this was shown to only occur with mammary ECM, not omental adipose, lung ECM, or Matrigel. These data further support an established finding that the regenerating mouse mammary gland is capable of redirecting to a mammary fate: neuronal stem cells [13], testicular cells [14], bone marrow[15] and cancer cells [10, 11, 16]. Thus, in aggregate, these data establish that the normal breast microenvironment is sufficient to initiate and/or maintain a normal mammary cell fate even in the presence of epigenetic variations or cellular mutational events.

Deciphering the critical elements of these observations is difficult due to the intrinsic complexities of the cleared mouse mammary fat pad model, including the fact that human cells will not proliferate on their own[17]. In order to overcome this, in vitro 3D cultures have been employed using derivatives of rat mammary gland ECM, collagen I from rat tails, and basement membrane from mouse Engelbreth-Holm-Swarm sarcoma (EHS) tumors (Matrigel/Geltrex) [18, 19]. The procedure developed by O’Brien et. al to isolate rat mammary ECM has established that tissue-specific components are capable of eliciting more biomimetic reactions in in vitro culture systems [19]. However, the methodologies developed in these studies depend on isolation of only the acid soluble components of mouse mammary tissue. Thus, these mouse derivative ECMs require alternative structural elements (Matrigel/Geltrex, rat tail collagen etc.) in order to polymerize into hydrogels. Further studies have also used whole decellularized human abdominal/breast tissue or cross-linked mouse mammary ECM in an effort to overcome these limitations [20–22]. Yet, as we and others have demonstrated, the human breast microenvironment is a unique adipose-laden tissue with spatial and temporal dynamics found nowhere else in the body. Furthermore, as cell are placed into polymerizing gels without any predetermined orientation random and unpredictable cellular behaviors are the typical result. Thus, control of cell placement and orientation in 3D culture systems is essential for reproducibility. To achieve this, our group has recently described the adaptation of an accessible bioprinting system with the capacity to direct formation of large mammary organoids in hydrogels [23, 24]. Combining this technology with better biomimetic substrates holds promise for improving current 3D culture systems.

To this end, here we describe the combination of our 3D bioprinter with a novel all mammary ECM hydrogel. We use rat- and human-derived breast tissue as a source for ECM. We demonstrate the use of decellularized whole human breast tissue for in vitro culture and present an extraction protocol capable of deriving a self-gelling in vitro hydrogel system. Data collected revealed that these ECM scaffolds contain growth factor patterns unique to the mammary gland. We show that this ECM system allows two common cancer cell lines (MCF-7 and MDA-MB-468) and a mammary epithelial cell line (MCF-12A) to proliferate uninhibited, similar to the commercially available Rat tail collagen and Geltrex. Further, when MCF-12As were 3D bioprinted into human mammary ECM, they showed self-organizational potential to form directionally oriented large ductal branching structures different than those we generated in rat tail collagen and Matrigel/Geltrex[25]. These differences in morphological growth were associated with differences in ki67 and cytokeratin 5 expression in bioprinted MCF12a organoids. We additionally demonstrate that culture in human breast ECM results in significant changes in protein and gene expression profiles of normal mammary and breast cancer cells as compared to traditional Geltrex culture, highlighting the unique signaling components of mammary ECM.

2. Materials and Methods

2.1. Cell culture

MCF-7 and MDA-MB-468 cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) containing Glutamax (Gibco, Gaithersburg, MD) complete media, supplemented with 10% fetal bovine serum (Life Technologies), and 1% Antibiotic-Antimycotic (100×) (ABAM; Life Technologies). MCF-12A cells were cultured in DMEM/F12 with Glutamax (Gibco), supplemented with 5% horse serum (Gibco), 20 ng/mL hEGF (Gibco), 10 μg/mL bovine insulin (Sigma Aldrich), 500 ng/mL hydrocortisone (Sigma Aldrich), and 1% ABAM. All cells were maintained in a 5.0% CO₂ incubator at 37.0°C. Media was replenished every 2–3 days and cells were passaged upon reaching 80% confluency. For 2-D culture, cells were plated at 3.0*10⁴ cells/cm². Cells were passaged using TrypLE™ Express Enzyme (Thermo Fisher) according to the manufacturer’s suggested protocol.

2.2. Extracellular matrix production

Rat mammary tissue was donated by Dr. Roy Ogle’s laboratory here at ODU and obtained immediately post-mortem from female Sprague Dawley rats. Handling and housing of rats was approved by the Old Dominion University Animal Care and Use Committee (IACUC). Human mammary tissue was obtained from patients undergoing breast reduction surgery. These studies were approved as exempt human subjects’ research by the ODU Institutional Review Board for Human Subjects Research (IRB), as tissue was collected as discarded medical waste. Human samples were first dissected and cleared by a pathologist to be free of all tumorigenic growths. Mammary tissue was stored at −80°C until processing. All wash/rinse steps were performed at room temperature in a temperature controlled orbital shaker (MaxQ 4000; Thermo Scientific). Tissue was cut into ~ 1–2 cm³ pieces, placed into large, sterile Erlenmeyer flasks and washed 3 times in PBS supplemented with 2% ABAM for 10–15 minutes for each wash. We then decellularized the tissue pieces with 2% v/v N-lauroylsarcosine sodium salt solution (Sigma Aldrich; Cat. No. 61717) and 2% ABAM in PBS for 96 hours under agitation on an orbital shaker, changing the solution every 8 hours. Decellularized tissue was then vigorously rinsed in sterile-filtered water for the duration of an hour, changing the water every 5 minutes. To remove lipid deposits, samples were soaked in 100% isopropanol for 6–8 hours under gentle agitation, replacing the isopropanol every 2 hours, and then gently agitated in 100% acetone for 12 hours. Samples were transferred to a sterile Erlenmeyer flask and vigorously washed with sterile-filtered water for 2 hours, replacing the water every 5–10 minutes.

Tissue was mechanically digested on ice in sterile water with an electronic homogenizer (Pro200; PRO Scientific) with MULTI-GEN 7XL homogenizers (PRO Scientific) until the majority of the sample was processed. Remaining large pieces of tough tissue were manually cut into fine pieces with sterile surgical scissors and a scalpel. The homogenized tissue slurry was poured into a pre-weighed 10 cm² non-treated petri dishes and placed in a lyophilizer (FreeZone Triad Cascade Benchtop Freeze Dryer; Labconco; Kansas City, MO) for 48 hours to remove liquid content. Post-lyophilization, the plates were weighed to acquire mass of dried samples obtained. The samples underwent a second mechanical disruption with liquid nitrogen and a mortar to increase surface area for enzymatic digestion. Samples were transferred to sterile 50 mL conical tubes with 10 mg/mL pepsin digest solution (Sigma Aldrich) containing sterile 0.5 M glacial acetic acid. The pepsin solution was added to each sample at a concentration of 10 mg pepsin per 100 mg of sample. Samples were allowed to digest for 48–96 hour period depending on sample size. After digestion, samples were diluted with 0.5 M glacial acetic acid to ~ 6 – 7 mg/ml and filtered through either a 40 μm (rat) or 100 μm (human) filter and stored at 4°C and used within 24–48 hours.

2.3. Preparation of huMECM, rtMECM, rtCOL, and Geltrex hydrogels

Human and rat-derived hydrogels were formed using ice-cold reagents and ice-cold 15 ml conical tubes, while maintaining sterility. For complete gelation of huMECM and rtMECM, reagents were added in the following order to create a master-mix for each hydrogel type: 1 part PBS, 2 parts 1X DMEM, 4 parts huMECM (human mammary ECM) or rtMECM (rat mammary ECM), ensuring to mix the ECM thoroughly by pipetting up and down. Finally, using a chilled, large-bore, 1000 μl pipet tip, 3 parts 0.8M NaOH was added, quickly mixing the components by pipetting up and down. This formulation yielded hydrogels with concentrations of ~ 2.6 mg/ml. After plating complete hydrogel precursor solution onto cell-culture treated plates, the solutions were incubated at 37°C in a humidified incubator overnight. Upon achievement of full polymerization, hydrogels were gently washed with 1X DMEM for 5 minutes twice. After washing, pH registered between 7.2 and 7.4.

Rat-tail collagen (rtCOL) (Corning; Cat. No. 354249, Corning, NY) hydrogels were made according to the manufacturer’s protocol. Briefly, for one 250 μl gel, we used the following mixture: 80.6 μl 1X PBS, 80.6 μl DMEM, 86.6 μl 4.33 mg/ml rat tail collagen (rtCOL), and 2.2 μl 1 N NaOH per collagen gel. This formula was scaled up appropriate for experimental necessity. A master mix was made for the number of collagen gels needed plus one and 250 μl of the master mix was plated for each well. Plates were placed in incubator for 20 – 30 minutes. Collagen gel pH registered at 7.2. Geltrex™ LDEV-Free, hESC-Qualified, Reduced Growth Factor Basement Membrane Matrix (Life Technologies; Lot: 1937101) was thawed in the refrigerator overnight on ice, and pipetted into each well (undiluted 16 mg/ml) and allowed to polymerize overnight at 37°C.

2.4. Extracellular matrix characterization

Oil Red O (ORO) histological stain was used to qualitatively evaluate the lipid content from tissue samples collected at several points throughout the decellularization process (unprocessed tissue, decellularized tissue, and delipidized tissue). Sample size was adjusted to ~ 25 mg for each sample. Each sample was placed in a 24-well plate (1 sample per well). The ORO stock solution (3 mg/ml) was mixed using ORO powder into isopropanol and filtered. The ORO working solution was mixed using 3 parts ORO stock and 2 parts ddH₂O (total volume dependent upon sample number). The working solution was incubated for 10 minutes at room temperature and filtered through a 0.22 μm filter. Samples were incubated for 5 minutes in 100% isopropanol. Isopropanol was removed and samples were incubated for 5 minutes in ORO solution. Following incubation, excess ORO was removed, and samples were washed with 100% isopropanol until the liquid came off clear. Lipid staining was visualized under a light microscope.

The DC Assay kit (Bio-Rad) was used following the provided protocol to determine total protein concentration of the huMECM and rtMECM. Sample ECM was diluted with 0.5 M glacial acetic acid. Mixtures were incubated for 15 minutes at room temperature before being analyzed on the Nanodrop™ 2000 at an absorbance of 750 nm.

Mini SDS-PAGE pre-cast gels (7.5%) (Bio-Rad) were used to resolve ECM structural protein content. For protein electrophoresis, mammary-derived ECM was diluted to 2.5 μg/μl. A total of 20 μg or 40 μg of each rtMECM, huMECM, Geltrex, and VitroCol® (Advanced Biomatrix; San Diego, CA) substrate was loaded into each well. Spectra™ Multicolor High Range Protein Ladder (Thermo Scientific) was used to character sample size. Gels were resolved at 150 V for approximately 1 hour and then washed 3 times for 5 minutes each in ddiH₂O. Gels were then counterstained with coomassie G-250 on a rocker for approximately 1 hour. After staining, gels were rinsed 5 times for 5 minutes in ddiH₂O and then allowed to wash overnight. Bands were visualized using a MYECL Imager (Thermo Scientific).

Molecular component characterization of 41 targets was conducted on each sample using Growth Factor Antibody Array (ABCAM; Cat. No. ab134002) according to the manufacturer’s protocol. Overall protein concentration was adjusted to 0.2 μg/μl in 1.0 ml in 1X blocking buffer (provided in the kit) for each sample (huMECM, rtMECM, rtCOL, and Geltrex). Wash and incubations were performed under gentle rocking following the kit protocol. Images of growth factor arrays were acquired using a MYECL Imager.

2.5. Rheological measurements

The Young’s modulus of each hydrogel was determined using the unconfined compression test, where only lateral deformation of the hydrogels was induced. The force (F) required to compress the gels was recorded to deduce the stress (σ [kPa]) versus strain ratio. The strain was defined by the amount of deformation (ΔL) divided by the initial height (L) of the uncompressed hydrogel. The yield strength was determined by the stress over the strain, and by equation:

$$E=\frac{\sigma }{\raisebox{1ex}{$\Delta L$}\!\left/ \!\raisebox{-1ex}{$L$}\right.}$$

2.6. 3D bioprinting of cells

Cells were delivered directly into preformed hydrogels using our accessible 3D bioprinter described previously [23]. Briefly, the cells were counted via hemocytometer and cell density was adjusted to 6.0 × 10⁵ cells/ml in 1X DMEM. A portion of this cell-laden media was transferred to a pulled glass needle. The needle was attached to the operating arm of the printer and the plate of hydrogels was fixed to the printing bed. The printer operated on a toolpath generated by a custom matlab program to penetrate the surface of each hydrogel and deliver cells to the middle portion of each gel. Cells were cultured for up to 14 days post-print and fluorescent images were captured at 2 and 14 day time points using a Zeiss Axio Observer Z1 microscope (Carl Zeiss AG, Jena, Germany). Samples were fixed in 10% neutral buffered formalin on day 14 for 1 hour for histological analysis.

2.7. Manual injections into whole decellularized tissue

Samples of decellularized mammary tissue were collected at the post-decellularized time point of the decellularization process. Cells were counted with a hemocytometer and cell density was adjusted to 1.0 × 10⁶ cells/ml. A 1 cc syringe (Becton, Dickinson and Company; East Rutherford, NJ) was filled with 3.3 × 10⁵ cells in 330μl of media and fitted with a 22-gauge needle for each cell type and tissue type. Injection were made in a five point pattern for each sample. All 3.3 × 10⁵ cells were delivered to each sample. Injections were returned to a 37°C humidified incubator and cultured for 14 days with media changes every 48 hours.

Following culturing, samples were collected, grouped according to cell line/tissue type, and fixed in 10% neutral buffered formalin for 1 hour. Fixed samples were then dehydrated cleared, paraffin embedded, and processed into 5μm sections for slide imaging.

2.8. Histology and immunofluorescent staining

Hydrogels and decellularized tissues were fixed in 10% neutral buffered formalin for 1 hr at room temperature, dehydrated, paraffin embedded and cut into 5μm sections. For immunostaining, sections were deparrafinized, rehydrated and subjected to heat mediated antigen retrieval in pH 9.0 Tris-EDTA buffer containing 0.05% v/v Tween 20 for 20 mins using a steam cooker. The following antibodies were used: rabbit polyclonal anti-ki67 (3.5μg/ml; Abcam; ab15580) and rabbit monoclonal (EP1601Y) anti-cytokeratin 5 (1.22μg/ml; Abcam; ab52635), and polyclonal goat anti-rabbit alexafluor 488 conjugated secondary (1μg/ml; ThermoFisher; A32731. Sections were all counterstained with DAPI. All images were taken on a Zeiss Axio Observer Z1 microscope and analyzed using the Zeiss Zen software. Cell counts were performed manually and fluorescence intensity was measured using the Zen software on a minimum of 6 random locations per field of view across a minimum of 4 sections of 3 separate samples. For Laminin 111/112 measurements, hydrogels were formed in a chamber slide, and stained with a rabbit polyclonal anti laminin 111/112 antibody (10μm/ml; Abcam; ab7463) and a goat alexafluor 488 conjugated secondary as described above. Fluorescence intensity was performed as described above across 3 independent samples.

2.9. Differential gene expression

Cells printed into hydrogels were collected at days 3 and 7 of culture. RNA was extracted with TRIzol™ (Thermo Scientific) using the provided protocol for tissue. RNA purity and concentration was determined using a Nanodrop 2000, with 260/280 ratio of every sample falling between 1.8 and 1.9. The RT² Profiler™ PCR Array for Human Signal Transduction (Qiagen, Cat. No. PAHS-014ZC) was used to identify differentially expressed gene expression of cell injections in each hydrogel. Polymerase chain reaction was performed on a StepOnePlus™ Real-Time PCR System (Life Technologies). Differential expression was conducted using the 2^−ΔΔCt method using the 5 housekeeping genes provided in the assay as endogenous controls. Each experimental variation was processed independently in triplicate.

2.10. Statistical analysis

Each independent experiment was performed with 3 or more biological repeat samples (n ≥ 3). Data are presented as a mean (± standard deviation) or median (± interquartile range). Oneway ANOVA analysis with Tukey multiple comparison test was determined for statistical significance of laminin levels and mean fluorescent intensity of cytokeratin 5 (Prism 7 software). For comparison of total growth factor content differences amongst the four hydrogels, a repeated measures one-way ANOVA with a Tukey multiple comparisons post-hoc was performed. Differences in growth factor content from decellularized tissue to pepsin digested tissue was compared by a repeated measures T-test. A standard T-test was performed to compare duct length of organoids formed in rtCOL and huMECM. For comparison of gene expression changes between geltrex and huMECM were compared by a paired T-test (pairing each individual gene). Differences in frequencies of organoid morphology and ki67 positivity were compared by a chi-square analysis.

3. Results

3.1. Culture of breast cancer cells within decellularized breast ECM

As in vitro model systems that accurately recapitulate native human tissue environments hold clear advantages, we initially aimed to generate a minimally manipulated, intact whole tissue human and rat mammary culture system. To do this, we used a previously reported decellularization system to remove all cellular components but leave intact the structural and tethered signaling molecules in situ [26]. Thus, we took whole tissue samples from the breasts of patients’ who were undergoing either prophylactic breast mastectomy or elective breast reduction. Alongside this human tissue, we also obtained rat mammary glands from nulliparous Sprague Dawley rats, as this tissue strain has been extensively studied as an isolated mammary extract [8, 19, 27]. We then processed all samples into 3 mm × 3 mm cubes of whole adipose tissue, avoiding large areas of connective tissue. All samples were then placed into decellularization solution for 48 hours, washed to remove all residual detergent and analyzed for cellular, lipid, and DNA content. At this stage of decellularization, we discovered there was still substantial levels of lipids remaining in our samples (Fig. 1A), but undetectable levels of intact cells (Fig. 1B). We then began treatment with both isopropanol and acetone to remove the remainder of the lipids, which successfully reduced them to undetectable levels (Fig. 1A). Furthermore, our DNA content dropped to undetectable levels as determined by picogreen detection (data not shown). Using these decellularized rat and human samples we then injected into multiple sites a total of ~ 3.3 × 10⁵ cells with a 22-gauge syringe. Following 14 days of culture, the tissues were then fixed, sectioned and analyzed. The MDA-MB-468, MCF-7 and MCF-12A all grew throughout both the rat and human tissues (Fig. 1C). The basal MDA-MB-468s grew extensively within the existing framework of the luminal areas of the tissue sections, with occasional clusters (Fig. 1C middle). The luminal A MCF-7 grew into large clusters reminiscent of tumors, while not reacting to the nascent structures (Fig. 1C right). The non-transformed MCF-12A cell line intriguingly grew as multilayered clusters lining the nascent cavities left by either epithelial or endothelial structures (Fig. 1C left).

Figure 1 – Decellularized and Delipidized human and rat mammary tissue support growth of human breast epithelial and cancer cells.

A) Whole mammary tissue from rat and humans was cut into uniform pieces and subjected to decellularization and delipidization which reduced lipid content to undetectable levels demonstrated by Oil Red O staining. Imaged under a stereo microscope at 5×. Scale bars represent 1.5mm B) Histological analysis demonstrated that whole tissues treated with both decellularization and delipidization were free from all cells via both H&E staining and DAPI. Scale bars represent 200 μm. C) MCF-12A, MDA-MB-468, and MCF-7 cells were injected into decellularized/delipidized rat and human mammary tissues, which resulted in growths throughout. MCF-7 and MB-MDA-468 tended to grow without apparent order, whereas MCF-12A lined previously formed luminal structures (black arrows). Scale bars represent 200 μm.

3.2. Extraction and characterization of self-gelling ECM from breast tissue

While growing cells within 3D decellularized tissue constructs provides a means of generating more biomimetic in vitro culture substrates, there are a number of culture-based drawbacks. Due to the density of the tissue it is difficult to track cells during experiments, and post processing for DNA/RNA and cellular proteins is hindered by the surrounding tissues ECM. Furthermore, cell placement is difficult to control due to the generally amorphous and soft features of adipose tissue. In order to overcome these limitations and develop a system compatible with our 3D bioprinter, we aimed to generate a self-gelling 3D extract from our tissues. Previous breast ECM extractions have relied on a protocol that eliminated the urea-insoluble fraction [19]. While these methods preserve many of the signaling molecules present in the tissue, the larger structural components (collagen I, collagen IV, etc.) are at a concentration which prevents spontaneous gelation. To maintain a total ECM extract, we lightly digested our tissues in a pepsin solution, thus preserving a recapitulation of each tissue that was as accurate as possible. Upon completion of the ECM extraction protocol, our initial extracts resulted in a concentration of ~15 – 30 mg/ml solution, which we then adjusted to ~ 6 – 7 mg/ml with 0.5 M glacial acetic acid to easier manipulate. Upon completion of our gelation protocol with 0.8 M NaOH, the gels would set within an hour into slightly opaque hydrogels (Fig. 2A). We then did a comparison of the gelation capability of our extracted ECM by generating gels from concentrations 10.4–0.7 mg/ml (Fig. 2B), which revealed successful gelation down to 1.3 mg/ml. We then standardized our rtMECM and huMECM gels to 2.6 mg/ml and analyzed their compressive mechanical properties using a MTS Criterion Model 42 (MTS; Eden Prairie, MN). Using the unconfined compressive test we found that rtMECM and huMECM maintained a slightly varied difference in yield strength, with rtMECM (2.6 mg/ml) at 39 ± 9 and huMECM (2.6mg/ml) at 42 ± 6 kPa. These mechanical properties are similar to undiluted Geltrex (16.6 mg/ml) hydrogels with 30 ± 2 kPa and rtCOL gels (1.5mg/ml) at 10 ± 4 kPa. We next compared the overall protein contents by performing an SDS-PAGE analysis with coomassie staining comparing our huMECM, rtMECM, rtCOL, Geltrex, and human collagen (huCOL) (Fig. 2C). Importantly, this revealed that our total ECM extraction maintains the structural collagens at a somewhat lower, but similar level to rtCOL. Likely this feature is what allows for the huMECM and rtMECM to polymerize into firm hydrogels when neutralized (Fig. 2B).

Figure 2 – Enzymatically digested whole mammary tissue retains intact structural ECM components and spontaneously gels.

A) Representative images of human and rat mammary tissue as it is decellularized, delipidized, homogenized, lyophilized and enzymatically digested. Once neutralized the whole tissue extracts would spontaneously gel at 37°C. B) Image showing successful polymerization of huMECM (top) and rtMECM (bottom) at various concentrations C) Coomassie stained denaturing gel comparing rat tail collagen (rtCOL), human derived collagen (VitroCol), Geltrex, rat mammary ECM (rtMECM), and human mammary ECM (huMECM).

Beyond the structural components of the ECM, the signaling molecules present in each tissue form a signature that helps establish cell identity [8, 28]. To determine if our extraction protocol maintained native signaling molecules, we compared the growth factor content of each of huMECM to rtMECM, rtCOL, and Geltrex. All four ECM substrates varied significantly in overall growth factor content (p<0.001; Fig 3, Supp Fig 1). Furthermore, we found that processing samples with pepsin did not significantly alter growth factor content from that present in decelluarized samples (p=0.34; mean difference = 0.006896). We found that both rtMECM and huMECM retained significant amounts of growth factors, and that they differed from each other in quantity but not content (Fig 3). Conversely, rtCOL had significantly higher concentration of all the signaling molecules (Fig. 3). Geltrex, which is growth factor reduced, had the least and in some cases undetectable levels of the various growth factors. These results confirm that we could generate a self-gelling hydrogel from mammary ECM that retained intact growth factors.

Figure 3 – Comparison of growth factor concentrations between huMECM, rtMECM,Geltrex, and rtCOL.

Relative concentration of growth factors in huMECM, rtMECM, Geltrex, and rtCOL were evaluated using spot densitometry. Resulting measured concentrations of growth factors are displayed as a heat-map relative to the levels measured in growth factor reduced Geltrex. After the completed decellularization/ECM extraction process, both huMECM and rtMECM show overall similar, but individually different levels of the various growth factors. rtCOL contained the highest, and most unique pattern of each growth factor, distinct from both mammary ECM products. Differences between substrates were significantly different as measured by a repeated measures ANOVA (p<0.001).

3.3. Cellular growth of 3D bioprinted cells into mammary derived ECM extracts

We next examined whether our bioprinting system could be combined with our novel hydrogels for in vitro analysis of mammary epithelial and tumor cell growth. Variability of stiffness has been well established in the literature to have impacts on the activity of cells [29–31]. As our goal here was to determine the viability of our unique ECM hydrogel to support mammary cell growth we chose to create similar stiffness to the commonly used Geltrex/Matrigel with our 2.6 mg/ml gels to control for this variable. We made 250 μl gels of huMECM and rtMECM in 48-well plates and used previously established protocols with our 3D bioprinter (Supp. Fig. 2A) [24]. We printed lines of 50 (± 3) MCF-12A, MCF-7 or MDA-MB-468 cells at 500 μm intervals using a CNC driven toolpath (Supp. Fig. 2B). We cultured these cells over a 14-day period with media changes every other day and tracked their growth. At days 2 and 14, the MCF-12A grew into large organoids in all four matrices (Fig. 4A). Similar to our previous findings [24], the MCF-12A formed a contiguous luminal structure in each matrix exceeding 3 mm in total length. MCF-7 also grew well in all of our substrates; however, there were morphological variations within the rtMECM with tight clusters, and the Geltrex tended to favor MCF-7 growth in multiple large grape-like clusters within each print site (Fig. 4B). The MDA-MB-468s had the largest morphological differences between the various substrates, including an overall lack of general growth in the huMECM (Fig. 4C). The rtMECM seemed to elicit a bridging reaction, which was different from the clustering morphology that both the Geltrex and rtCOL elicited.

Figure 4 – 3D bioprinted growth of epithelial organoids and tumoroids in rat and human mammary ECM hydrogels.

Cellular encapsulation into select hydrogels was performed using our 3D bioprinter. Each cell-type injection contained 50 (±3) cells, spaced at 500 μm intervals. A) MCF-12A cells were printed in linear pattern within Geltrex, rtCOL, rtMECM, and huMECM hydrogels and cultured for 14 days. Cells grew into large organoids in all substrates, typically exceeding 3 mm in length. B) MCF-7 cells were injected in a linear pattern within Geltrex, rtCOL, rtMECM, and huMECM hydrogels and cultured for 14 days. Cells grew in grape like clusters in Geltrex, in all other substrates cells retained a single spherical shape which all grew in size by day 14. C) MB-MDA-468 cells were printed in linear pattern within Geltrex, rtCOL, rtMECM, and huMECM hydrogels and cultured for 14 days. Cells grew in small clusters in Geltrex, but grew in single tumoroids in rtCOL, and rtMECM. In huMECM, the cells failed to flourish with sporadic small surviving cells. Scale bars 200μm.

As the ultimate goal of these studies is to develop human biomimetic further evaluate morphological differences in the growth, we compared hematoxylin and eosin (H&E) stained histological sections of organoids/tumoroids grown in huMECM vs those formed in the industry gold standard substrates Geltrex and rtCOL after 14 days. MCF-7 cells produced solid structures that lacked lumens in huMECM, rtCOL, and Geltrex, consistent with their tumorigenic phenotype. However, in huMECM, MCF-7s did produce tumoroids with more irregular borders, as opposed to the nearly perfectly cylindrical borders seen in rtCOL and Geltrex (Fig. 5). In addition, we noted significant differences in the morphology of MCF12a organoids and MCF7 tumoroids between the three substrates (Fig 5). This was most striking in MCF12a cells, which formed nearly exclusively large duct like luminal structures (Fig 5a and b). Conversely, in Geltrex, MCF12a cells only produced clusters of alveolar-like structures (Fig 5a and c). In rtCOL a mixture of small ducts and other structures that could not be easily classified as duct-like or alveolar like (Fig 5a and c). Particularly striking, in huMECM, we found several large ductal structures extending over 2mm in length (Fig 5b). While we have previously identified large duct-like structures in rtCOL [25], these structures are rare and never exceeded ~1mm in length. To quantitate these morphological differences, we scored resulting organoids as either duct-like (luminal structures with a length at least three times its width), alveolar-like (luminal structures with a length no more than twice the width) and other structures that could not fit in either category (Fig 5c). 78% of organoids formed in huMECM were ductal, compared to 42% of organoids in rtCOL (p=0.0001) and 0% in geltrex (p=0.00002). Furthermore, we found the mean length of ductal structures formed in huMECM (1030μm ± 707) were significantly longer than those formed in rtCOL (339μm ± 144).

Figure 5 – huMECM hyrdrogels facilitate large duct-like luminal structure formation of bioprinted MCF-12As.

A) Histological sections stained with H&E for MCF-7 and MCF-12A encapsulated in Geltrex, rtCOL, and huMECM after 14 days. MCF-12A cells formed luminal structures in all 3, with duct-like structures appearing in the collagen and huMECM and clusters of aveolar-like structures in Geltrex. MCF-7 cells formed solid tumoroids in all 3 substrates, with irregular borders only observed in huMECM. Scale bars = 100μm. B) Tiled image of a H&E section of a large ductal network formed in huMECM that exceeded 2mm in length. C) Quantitation of morphological differences in MCF12a organoids grown in huMECM. Left panel shows frequencies of ductal vs alveolar/other structures in each of the substrates. huECM generated significantly more ductal organoids than geltrex (p=0.00002) or rtCOL (p=0.0001). Right panel shows ductal structures formed in huMECM were significantly longer than those observed in rtCOL (*p<0.05). Images were acquired with a Zeiss Axio Ovserver.Z1 microscope.

To further evaluate the difference in MCF12a growth in huMECM compared to the industry standard geltrex and rtCOl hydrogels, we evaluated expression of proliferation marker ki67 and cytokeratins (Fig 6). Compared to geltrex and rtCOL, significantly more MCF12a cells expressed ki67 in huMECM than in the other two substrates (p<0.00001) consistent with greater proliferation (Fig 6a and 6b). Furthermore, evaluation of cytokeratin expression revealed a significant increase in the expression level of the basal cytokeratin 5 based on immunofluorescence intensity in MCF12a cells grown in huMECM verse either geltrex or rtCOL (Fig 6a and 6c). These differences are likely associated with the differences in organoid morphologies noted above. No differences in luminal cytokeratin 8 were noted (not shown).

Figure 6 – huMECM alters ki67 and cytokeratin 5 expression in MCF12a bioprinted organoids.

A) Representative images of immunofluroscent staining of proliferation marker ki67 and cytokeratin 5 (green) in organoids within geltrex, rtCOL, and huMECM. Nucleir are counterstained with DAPI (blue). Scale bars = 50μm. B) Quantitation of ki67 staining demonstrates that huMECM has statistically significant higher proportion of ki67+ cells (p=0.0001 and p=0.0088 versus geltrex and rtCOL, respectively). C) Quantitation of relative expression of cytokeratin 5 demonstrates significantly higher amounts of expression in MCF12a cells bioprinted into huMECM versus either rtCOL or geltrex (***p<0.001).

Taken together, these results demonstrate that cells have unique morphological and cellular properties when grown in huMECM. The growth responses exhibited by the epithelial MCF-12A, luminal MCF-7 and the basal MDA-MB-468 demonstrate the importance of using the correct substrate, as any morphological variations could indicate an important underlying influence fromthe ECM. This influence could easily drive reactions that are not tissue appropriate when studying various cancer related subjects such as the dynamics of cellular response to drug treatment [32, 33]. To the best of our knowledge, this is the first evidence that large epithelial organoids and mammary tumoroids can be grown in a hydrogel comprised solely of mammary ECM.

3.4. Variations in ECM activated signaling pathways

To determine if differences in ECM constituents resulted in measurable molecular changes in the cells, we compared gene expression profiles of MCF-12A, MCF-7, and MDA-MB-468 cells 3D bioprinted into huMECM and Geltrex. We chose Geltrex as a control as this is one of the most common methods of 3D culture for breast cancer research. We chose to investigate the retained signal transduction capabilities of huMECM because of the lack of readily available human models for breast cancer, as our ultimate goal is to develop biomimetic models for the human breast microenvironment. Changes in signal transduction were monitored by the Signal Transduction Pathway Finder RT² Profiler Array (Qiagen), which monitors expression of genes known to be up/down-regulated by specific signal transductions pathways. Differences were seen in all pathways monitored (Fig 7a) and the effect of huMECM and Geltrex on transcriptional activity of signal transduction associated genes was significantly different (p < 0.0001). This resulted in poor correlation between gene expression levels among cells grown in each of the hydrogels (Fig 7b; slope = 0.3174; n=3). We found a subset of genes were altered by at least 10-fold across all cell lines (Supplemental Tables 1 and 2). These genes were associated with Hedgehog, NF-κB (nuclear factor kappa-light-chain-enhancer of activated B cells), Notch, oxidative stress, PPAR-γ (peroxisome proliferator-activated receptor gamma), TGF-β (transforming growth factor beta), and WNT (wingless/integrated) signaling pathways. Most notably, 4 of 9 tested genes associated with Notch signaling that were tested were up/downregulated by 100-fold or more across all 3 cell lines. Three of these genes are positive effectors of Notch signaling and were down-regulated. The remaining gene, inhibitor of DNA binding 1, dominant negative helix-loop-helix protein (ID1), is an inhibitor of Notch signaling and was up-regulated. The changes in both global signaling, as well as NOTCH signaling specifically is not surprising because of the differences in growth factor content between the substrates (Fig 3) and the fact that Geltrex is a concentrated basement membrane ECM. Geltrex thus contains a high concentration of laminin 111, which is a known effector of several signal transduction pathways including NOTCH, [34–37]. To further explore this, we evaluated laminin 111/211 content of our hydrogel precursor solutions via antibody-based fluorescence quantification. We found Geltrex contained significantly higher levels of laminin 111/211 than any of the other hydrogels tested (Fig. 7c). No statistically significant difference was seen between rtMECM, huMECM, or rtCOL. These data combined indicate that the huMECM has a novel structural and growth factor composition that elicits unique responses in mammary epithelial and cancer cells. Therefore, the combined use of self-gelling tissue-specific ECM and our bioprinting system offers a novel approach to studying cell/tissue-specific ECM interactions in vitro.

Figure 7 – huMECM elicits altered molecular signaling compared to Geltrex.

A) Heat map demonstrating differences in median gene expression of genes associated with each of the displayed signal transduction pathways. B) Comparison of gene expression of MCF-12A, MCF-7, and MDA-MB-468 cells 3D bioprinted into huMECM or Geltrex hydrogels. Each point represents a single gene in a single cell line measured by qRT-PCR pathway finder array plotted for expression in the two substrates. Repeated measures T-test revealed significant differences in overall expression and regression analysis found poor correlation (slope = 0.314) between the two substrates indicative of differential pathway activation. C) Mean fluorescent intensity of immunostaining comparison of Laminin 111/211 levels in all hydrogels used demonstrates significantly higher levels in Geltrex compared to the other substrates.

4. Discussion

The primary components of model development have always attempted to recreate physiological-normal, human body orchestration, wherein discrete cellular and ECM constituents establish unique microenvironments that are tissue-to-tissue specific. The most successful and accessible models to date came from the development of standard 2D culture systems. The culture of many cell types in these systems depends solely on the presence of an activated polystyrene surface, sufficient nutrients, and growth signals to illicit functional cellular responses. However, the study of convoluted phenomenon, such as cancer progression, demonstrates variable and unreliable reactions. This is reinforced by many studies that demonstrate substantial changes of cellular behavior when they are transitioned into a 3D in vitro or in vivo environment [38–40]. The two most common substrates used in these 3D models are Matrigel/Geltrex from Engelbreth-Holm-Swarm mouse sarcoma cells and Collagen I extracted from rat tails. These two ubiquitous systems appear to illicit more biomimetic responses; however, we were surprised in our studies to find that the rat-tail collagen had an overabundance of essentially every signaling molecule of interest. This makes cellular reactions within these 3D substrates a difficult element to interpret due to the likely over stimulation of key signaling pathways. These overrepresented levels were in stark contrast to the levels seen in the growth-factor reduced Geltrex, which is produced in an effort to control for batch to batch variability, and for undesired stimulations. While this feature of growth factor reduced Geltrex does allow for some confidence when distinguishing signaling based reactions in 3D culture, the structural basement membrane elements of Geltrex (and Matrigel) stand alone as a confounding variable. Although cells in vivo may come into contact with laminins from basement membranes in high frequency, there is never a context in which a cell will grow within a 3 mm thick, three-dimensional version if it. Not only is the 3D element of Matrigel/Geltrex an oddity, so too is the reaction cells have when interacting with the structural molecules of basement membrane. This was particularly notable when we examined the laminin 111/211 signaling between our ECMs and Geltrex, where notch signaling was vastly upregulated when cells were encapsulated within the Geltrex hydrogels. Thus, as we anticipated, the rat and human mammary ECMs had median concentrations of all of the targeted signaling molecules, as well as exhibited low levels of basement membrane components. Moreover, even though there were overall comparable levels of all of the growth factors between rat and human mammary ECM, there was a clear signature pattern that emerged which could easily distinguish the two. This tissue specific signature clearly imparts microenvironmental cues that would elicit different cellular responses. This was clearly demonstrated by the statistically significant morphological and pathway activation differences we witnessed between the various substrates. Furthermore, we and others have demonstrated previously the power of the nulliparous rat mammary ECM to illicit a normalizing, or redirecting, response from cancer or stem cells [8, 41, 42]. Thus, while these variations from rat to human are clearly unique, the overall mammary based signaling can still likely can be achieved in either of their use. This was also clearly highlighted by our previous publication describing in vivo response from only rat mammary adipose ECM, and not from omental adipose sources [8].

In our studies we also examined the concept of using a whole, decellularized mammary gland as a culture substrate for breast epithelial cells and breast cancer cells. Previous reports had demonstrated that human abdominal, and more recently, breast tissue could be used for this purpose [20, 22]. Thus, we hypothesized that substituting breast adipose would clearly be advantageous for studying these mammary cell types due to our previous findings. In these experiments all of our cell types grew in a manner consistent with what we expected with the tumor cells growing into isolated masses, some very large, mimicking in vivo studies using the same cells. Interestingly, the epithelial cells seemed to gravitate toward cavities within the decellularized tissue, in some instances lining what were previous luminal structures. While these findings were in line with our expectations, this system faced several obstacles. Due to the inherent amorphous nature of adipose tissue, it was difficult to get a completely homogenous shape. While we could easily control for this with categorical size/shape separation, we sought to move into using these tissues as a more readily accessible 3D ECM similar to Geltrex or rat tail collagen. An advantage of the decellularized 3D mammary ECM is that it presents both the structural and corresponding signaling cues to cells when placed within their confines. A draw back to this system is that the deconstructed and extracted ECM, when allowed to form back into structural lattices, may not have the precise original order necessary to be fully biomimetic. While, this is a definitive drawback, the advantages of these hydrogels are vast, allowing for accessible in situ visualization, straightforward extraction, and easy in situ manipulation. Furthermore, as discussed earlier, our ECM extracts have demonstrated pronounced in vivo reactions indicating their intrinsic functional nature [8]. Finally, another inherent feature of each of the ECMs used here is lot-to-lot variability. In our studies we controlled for this by batching normal breast adipose tissue and using it throughout our experiments in order to control for any variations. However, the lot-to-lot variability may provide a unique opportunity to elucidate patient-specific phenotypes, providing real world data. Thus, these batch-specific variations in human samples may provide the grounds for patient-specific models, furthering the understanding of breast cancer through direct cancerous to non-cancerous breast comparisons. Regardless, the combination of our bioprinter with an all mammary self-gelling hydrogel provides a novel 3D culture system for the biomimetic study of cell/microenvironmental interactions.

5. Conclusion

The development of truly biomimetic in vitro systems to study complex diseases such as cancer will improve the reliability of laboratory discoveries leading to quicker bedside cures. To this end, here we demonstrate the use of a self-gelling human mammary derived ECM 3D culture system and an accessible 3D bioprinter platform to generate large organoids/tumoroids. The hydrogels retain novel structural and signaling components that elicit unique morphological, cellular, and molecular responses in the cells, providing a novel methodology for studying cell and native ECM interactions. We have established this system to study both fundamental elements of cancer initiation and cancer cell reaction to various treatments, and future studies will enable us to uncover the mechanistic nature of our previously reported redirecting experiments performed in vivo.

Supplementary Material

1

Supplemental Figure 1 – Growth factor concentrations between huMECM, rtMECM, Geltrex, and rtCOL. Bar graphs demonstrating the relative concentrations of each of the measured growth factors in huMECM, rMECM, growth factor reduced geltrex, and rtCOL.

Supplemental Figure 2 – Modified 3D bioprinter enables accurate cell printing. An image of our 3D bioprinter A) with modifications to allow accurate and repeatable placement of cells in various tissue culture ware. B) An example of a CNC driven toolpath for printing cells into a linear pattern into a 24 - well dish.

Supplemental Table 1 – Total genes from each pathway showing 10/100 fold change vs Geltrex.

Supplemental Table 2 – List of genes showing 10 fold+ up or down-regulation across all three cell lines. Relative fold change in cells grown in huMECM verses Geltrex is shown. Note all positive effectors of NOTCH signaling are down-regulated, while negative regulator ID1 is significantly up-regulated.