Breast fibroblasts modulate epithelial cell proliferation in three-dimensional in vitro co-culture

MCF10A cells (American Type Culture Collection, Manassas, VA, USA) and MCF10AT cells (Karmanos Cancer Institute, Detroit, MI, USA) were cultivated in DMEM/Ham's F-12 (Cambrex, Walkersville, MD, USA) supplemented with 0.1 μg/ml cholera toxin (Calbiochem, San Diego, CA, USA), 10 μg/ml insulin (Sigma, St Louis, MO, USA), 0.5 μg/ml hydrocortisone (Sigma), 0.02 μg/ml epidermal growth factor (Upstate Biotechnology, Lake Placid, NY, USA) and 5% horse serum (Invitrogen, Carlsbad, CA, USA). Subconfluent cultures (80–90% confluence) were utilized in experiments.

Fibroblasts were derived from mammary reduction specimens (NAF) and from primary breast cancers (CAF). The tissues were remnants of diagnostic surgical specimens and were obtained from The University of Alabama at Birmingham Tissue Procurement Facility after Institutional Review Board approval. H&E-stained, frozen histologic sections were prepared from each tissue sample to confirm benignity or malignancy. The tissue samples from the breast reduction specimens consisted predominantly of adipose tissue, but interspersed fibrous areas were selected for fibroblast isolation. The tissue was minced and digested for 18–24 hours at 37°C in DMEM (Vitacell, Manassas, VA, USA) with 10% fetal bovine serum (Invitrogen) supplemented with 100 U/ml streptomycin, 100 μg/ml penicillin, 2.5 μg/ml Fungizone (GibcoBRL, Life Technologies, Grand Island, NY, USA), 150 U/ml hyaluronidase (Sigma) and 200 U/ml collagenase type III (GibcoBRL). The digested tissue was centrifuged at 100 relative centrifugal force and plated in T25 tissue culture flasks with DMEM and 10% fetal bovine serum. Differential trypsinization was applied during subculturing to select for the growth of fibroblasts [26].

Early passages (below passage 9) of all fibroblasts were subjected to immunocytochemical evaluation with anti-vimentin (mouse IgG₁, clone V9; Neomarkers, Fremont, CA, USA), anti-epithelial membrane antigen (mouse IgG_2a, clone ZCE113; Zymed, San Francisco, CA, USA), and anti-cytokeratin (CK) 5/CK 8 (mouse IgG₁, clone C-50; Neomarkers) as confirmation of their stromal origin (i.e. strong vimentin expression, and absence of epithelial membrane antigen and CK 5/CK 8). Epithelial membrane antigen and CK 8 are expressed primarily in luminal breast epithelial cells, whereas CK 5 is found in myoepithelial cells [27]. For immunocytochemical evaluation, fibroblasts were grown on glass coverslips, fixed in 70% ethanol and were permeabilized with acetone. Fibroblasts were incubated in 3% hydrogen peroxide followed by 3% goat serum at room temperature. Anti-vimentin (0.5 μg/ml), anti-epithelial membrane antigen (1 μg/ml), or anti-CK 5/CK 8 (0.3 μg/ml) were applied for 1 hour at room temperature. Secondary detection was accomplished with a streptavidin/horseradish peroxidase secondary detection system (Signet Laboratories, Dedham, MA, USA) and diaminobenzidine (BioGenex, San Ramon, CA, USA). Harris hematoxylin was used as a counterstain.

The fibroblasts were routinely maintained in DMEM and 10% fetal bovine serum. Subconfluent cultures (70–90% confluence) of lower passages (below passage 9) were utilized for the experiments described. Only early-passage fibroblast cultures are used in experiments to more closely simulate their in vivo phenotype. It has been shown that early-passage (below passage 9) colonic primary fibroblasts maintain many of their in vivo characteristics, including expression of alpha-smooth muscle actin, collagen IV and laminin 1 [28]. Multiple NAF and CAF cultures were utilized because of potential variation among fibroblasts from different individuals and different breast cancers.

The effect of NAF and CAF on MCF10A breast epithelial cells and MCF10AT breast epithelial cells was studied using a 3D in vitro model. In co-cultures, epithelial cells and fibroblasts were mixed in Human Endothelial–SFM Basal Growth Media (Invitrogen) supplemented with 10 ng/ml epidermal growth factor and 20 ng/ml basic fibroblast growth factor (Invitrogen). This was followed by dispersal on 100 μl basement membrane material (Growth Factor Reduced Matrigel^®; BD Biosciences, Bedford, MA, USA) in each well of eight-well chamber slides (Lab-Tek^® Chamber Slide™ System; Nalge Nunc International, Naperville, IL, USA).

The ratio of epithelial cells to fibroblasts (E:F) ranged from 2:1 to 1:3 by varying the number of fibroblasts while keeping the number of epithelial cells constant (100,000 cells/well). Controls consisted of 3D monocultures of MCF10A cells and MCF10AT cells (100,000 cells/well). All cultures were incubated in a 37°C, 5% CO₂ humidified incubator for 14 days with supplementation of fresh Human Endothelial-SFM Basal Growth Media supplemented with 10 ng/ml epidermal growth factor and 20 ng/ml basic fibroblast growth factor at 4-day intervals.

Morphologic development was observed by phase contrast microscopy. 5-bromo-2'-deoxyuridine (BrdU) (0.2 mg/ml; Calbiochem) was applied to all cultures for 24 hours. The cultures were removed from the chamber slides, were embedded in HistoGel specimen processing gel (Richard-Allan, Kalamazoo, MI, USA), were fixed in 10% neutral-buffered formalin, and were embedded in paraffin. Histologic sections were prepared for H&E staining and immunocytochemistry.

Distinction of epithelial cells from fibroblasts in the 3D cultures was accomplished by examination of the cell morphology and location within the culture. To ensure that only epithelial cells were counted, however, immunostaining with anti-BrdU (mouse IgG₁, clone Bu20a; DAKO, Carpinteria, CA, USA) was followed by immunostaining with anti-CK 5/CK 8, expressed only in MCF10A cells and MCF10AT cells in 3D cultures. The staining entailed pretreatment with low-temperature antigen retrieval (i.e. incubation in 0.01 M citric acid monohydrate, pH 6.0, for 2 hours in an 80°C water bath), followed by sequential incubation in 1 N HCl, 3% hydrogen peroxide, 1% goat serum, and anti-BrdU (3 μg/ml). Secondary detection was as previously described. This was followed by incubation with anti-CK 5/CK 8 (4 μg/ml) and secondary detection using a streptavidin/alkaline phosphatase reagent (Signet Laboratories) and New Fuchsin (BioGenex) as the chromogen. The slides were lightly counterstained with Harris hematoxylin.

The resulting dual coloration of anti-BrdU (brown nucleus) and anti-CK 5/CK 8 (pink cytoplasm) enabled identification of proliferating epithelial cells. A BrdU-labeling index in the epithelial cells was determined by calculating the percentage of epithelial cells with nuclear staining for anti-BrdU in complete cross-sections of the 3D cultures. A minimum of 500 epithelial cells was counted. Negative controls consisted of histologic sections of each 3D culture processed without the addition of primary antibodies.

Cells were gently removed from 3D cultures after treatment with Dispase (BD Biosciences Discovery Labware, Bedford, MA, USA), followed by 5–10 mM EDTA. Recovered cells were washed with cold PBS. To distinguish co-cultured epithelial cells from fibroblasts, allophycocyanin-conjugated anti-EpCAM (mouse IgG₁, clone EBA-1; BD Biosciences Immunocytometry Systems, San Jose, CA, USA) was used to label MCF10A cells and MCF10AT cells. The cells were permeabilized and fixed (BD Cytofix/Cytoperm and Perm/Wash kit; BD Biosciences Pharmingen, San Diego, CA, USA). Prior to staining with FITC-conjugated anti-BrdU (mouse IgG₁, clone B44; BD Biosciences Immunocytometry Systems), cells were treated with DNase I (Roche Diagnostics GmbH, Penzberg, Germany). Samples were analyzed on a BD FACS Calibur™ flow cytometer (BD Biosciences). The percentage of BrdU-labeled epithelial cells (positive for anti-EpCAM and anti-BrdU) was calculated from the total acquired events.

Cells were gently removed from Matrigel^® in 3D co-cultures of MCF10AT cells (100,000 cells/well) with fibroblasts (50,000 cells/well) and monocultures of MCF10AT cells (100,000 cells/well) or fibroblasts (50,000 cells/well) using dispase, as previously described. The released cells were washed with cold PBS to remove residual Matrigel^®. Cell lysates were prepared from 3D cultures and monolayer cultures with NEB lysis buffer (Cell Signaling Technology, Beverly, MA, USA) containing 20 mM Tris–HCl (pH 7.5), 150 mM NaCl, 1 mM Na₂EDTA, 1 mM EGTA, 1% Triton, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM Na₃VO₄, 1 μg/ml leupeptin, 10 mM sodium fluoride, a Complete Mini Protease Inhibitor Cocktail tablet (Roche, Indianapolis, IN, USA) and 1 mM phenylmethylsulfonylfluoride, and were cleared by centrifugation. All lysates were concentrated with Millipore Microcon^® Centrifugal Filter Devices YM-3 (Millipore Corporation, Bedford, MA, USA). The protein concentration of lysates was measured with the BCA Protein Assay Kit (Pierce, Rockford, IL, USA).

The protein levels of IGF II in cell lysates were measured by ELISA for human IGF II (DSL-10-9100 Active™ IGF-II ELISA; Diagnostic System Laboratories, Webster, TX, USA) as per the manufacturer's instructions. The kit includes a modified version of the standard acid–ethanol extraction prior to ELISA. The ELISA for IGF II shows no detectable cross-reactivity with IGF I, insulin, IGF binding protein 1, and IGF binding protein 3. The minimum detection limit of the assay is 0.25 ng/ml. The amount of IGF II present in lysates from cell lines and 3D cultures was normalized to the total protein concentration.

RNA isolation (Trizol reagent; GibcoBRL) was followed by RNA clean-up with RNeasy columns (Qiagen, Valencia, CA, USA). Spectrophotometric ratios of A₂₆₀ to A₂₈₀ were greater than 1.8. The forward, reverse and probe oligonucleotides were synthesized and purified by HPLC (Applied Biosystems, Foster City, CA, USA) after complete evaluation of the IGF II and ribosomal S9 gene sequences (GenBank database) (IGF II forward, 5'-GTCGATGCTGGTGCTTCTCA-3' and reverse, 5'-GGGCGGTAAGCAGCAATG-3' ; probe, 5'-6FAMCTTCTTGGCCTTCGCCTCGTGCTTAMRA-3' ; ribosomal S9 forward, 5'-ATCCGCCAGCGCCATA-3' and reverse, 5'-TCAATGTGCTTCTGGGAATCC-3' ; probe, 5'-6FAMAGCAGGTGGTGAACATCCCGTCCTTTAMRA-3') using the Primer Express software (Applied Biosystems).

Fluorescent signal data were collected by the ABI Prism 7700 Sequence Detection System (Applied Biosystems). The log-linear phase of amplification was monitored to obtain the threshold cycle (defined as the fractional cycle number at which the amount of the amplified target reaches a fixed threshold) values for each RNA sample. Ribosomal S9 was used as the internal reference and was selected because it exhibits minimal variability in tissues of different origins [29]. The comparative threshold cycle method was employed to determine IGF II expression levels in each sample relative to a calibrator, in this case MCF10AT cells [29, 30]. Each sample was run in triplicate.

The immunocytochemical characterization of fibroblasts used in the described experiments confirmed their stromal nature. Immunostaining for vimentin was strongly positive and staining for epithelial membrane antigen and CK 5/CK 8 was negative.

In 3D monocultures, both MCF10A cells and MCF10AT cells initially form a lattice-like network of duct-like structures. After several days, the lattice-like network is replaced by a predominance of rounded epithelial cell groups (spheroids) (Fig. 1a,1b,1c,1d). MCF10A cells form small, rounded spheroids (Figs 1b and 2a). MCF10AT cells aggregate into larger solid groups or sheets with extensive squamous metaplasia (Figs 1d and 2c). The formation of larger three-dimensional structures and the abnormal differentiation (i.e. squamous metaplasia) of MCF10AT cells supports their greater degree of transformation.

In 3D co-cultures with NAF or with CAF, the epithelial cells and fibroblasts form large rounded structures (Fig. 1e). In histologic sections of co-cultures with NAF or with CAF, MCF10A cells and MCF10AT cells form spheroids or sheets within Matrigel^®, as seen in monocultures, and these epithelial groups surround an aggregate of fibroblasts (Fig. 2b,2d). These 3D co-cultures resemble a terminal duct-lobular unit in the normal breast or in proliferative breast disease/DCIS (Fig. 3a,3b,3c), in that epithelial cells are arranged in groups (similar to an acinus or a terminal duct) surrounded by a laminin-rich basement membrane with fibroblasts located outside the basement membrane and separating epithelial groups. The 3D in vitro model differs from normal breast or in situ breast disease in vivo, however, in that the number of fibroblasts centrally located and between epithelial cell groups is greater than that typically seen in vivo. Additionally, the fibroblasts in the in vitro model are surrounded by an ECM rich in laminin and collagen IV [31], whereas in vivo collagen I would typically predominate [32].

In 3D co-cultures, both NAF and CAF markedly suppressed squamous metaplasia of MCF10AT cells as observed on H&E-stained sections of monocultures, thereby normalizing the morphology to a glandular phenotype (Fig. 2c,2d). No obvious differences between NAF and CAF in morphology, growth pattern or adhesion to substrate, in either a monolayer or the 3D culture, were observed.

In replicate co-cultures of MCF10A cells with three different NAF and CAF grown in an E:F of 2:1, both types of fibroblasts significantly reduced proliferation of MCF10A cells. The mean BrdU-labeling index of MCF10A cells, when measured by immunocytochemistry, was decreased by 47% in co-culture with NAF (n = 19, P = 0.009) and by 39% in co-culture with CAF (n = 19, P = 0.024) relative to the MCF10A monoculture (Table 1 and Fig. 4). The BrdU-labeling index of MCF10AT cells was reduced by 49% in the presence of NAF (n = 20, P = 0.013), relative to the MCF10AT monoculture, whereas co-culture with CAF failed to significantly lower the MCF10AT BrdU-labeling index (n = 22, P = 0.935) (Table 2 and Fig. 4). The effect of NAF versus CAF on the rate of proliferation of MCF10AT cells was significantly different (P < 0.001). The effect was further confirmed by repeating the co-cultures to measure the BrdU-labeling index by flow cytometry, rather than by immunocytochemistry (Fig. 5).

There was variability among NAF cultures and among CAF cultures in their ability to suppress proliferation of MCF10A cells and MCF10AT cells (Tables 1 and 2) in this 3D culture system, potentially reflecting heterogeneity among the individuals from which the fibroblasts were derived. Because of this variability, detection of a significant difference in the function of NAF and CAF required many replicates and multiple fibroblast cultures derived from different individuals.

In a prior report, CAF was found to promote, rather than inhibit, the growth of MCF10A cells in a similar 3D co-culture system [22]. One of several possible explanations for this discrepancy between the prior result and the present result is a difference in E:F. Shekhar and colleagues used an E:F of 1:1 rather than the E:F of 2:1 we initially used [22]. The number of fibroblasts has been shown to have an effect on the response of epithelial cells [7, 9, 14]. We therefore repeated the 3D co-cultures of MCF10A cells using NAF-2 and CAF-1 with increasing numbers of fibroblasts (i.e. a decreasing E:F) (Fig. 6). BrdU labeling was assessed by immunocytochemistry of histologic sections of 3D cultures.

As previously, NAF-2 at an E:F of 2:1 suppressed proliferation of MCF10A cells. However, with increasing numbers of NAF-2, this suppression effect was gradually weakened (P = 0.043). Although we found no significant difference in the suppressive effect of NAF-2 in an E:F of 2:1 versus an E:F of 1:1 or of 1:2, there was a significantly greater rate of proliferation of MCF10A cells with NAF-2 in an E:F of 1:3 compared with in an E:F of 2:1 (P = 0.028). More importantly, CAF-1 at an E:F of 1:1 did not significantly suppress proliferation, whereas our original ratio of 2:1 did (P = 0.025). CAF-1 at an E:F of 1:2 also conferred a higher rate of proliferation of MCF10A cells than the E:F of 2:1, but this did not reach statistical significance (P = 0.054). At an E:F of 1:3, however, CAF-1 caused a decrease in proliferation of MCF10A cells and enhanced cell death, as assessed by microscopic morphology. At an E:F of 1:3, the total number of viable MCF10A cells was reduced in co-culture with both NAF-2 and CAF-1; however, this reduction was more marked with CAF-1.

Neither NAF nor CAF had a significant effect on the rate of apoptosis of MCF10A cells or MCF10AT cells when grown at an E:F of 2:1 after 2 weeks of co-culture, as assessed by TUNEL assay (Fig. 7).

As an initial attempt to identify differences between NAF and CAF that explain our observed results, expression of IGF II in NAF and in CAF was assessed. A higher level of expression of IGF II in CAF than in NAF may provide an explanation for the higher rate of proliferation of MCF10AT cells allowed by CAF in comparison with NAF.

ELISA performed on cell lysates of NAF and CAF cultures demonstrated variability in expression of IGF II among cultures, but no significant difference was observed in the mean IGF II quantity between NAF and CAF in monolayer cultures (Table 3) or in 3D monocultures (Table 4). Although in monolayer cultures more CAF than NAF had IGF II levels > 10 ng/μg protein, the mean levels of IGF II were similar (NAF versus CAF, P > 0.05) (Table 3). This comparison included five different NAF cultures and six different CAF cultures. ELISA was also performed on NAF-conditioned media and CAF-conditioned media but, despite the concentration of samples, the levels were too low for reliable quantification by ELISA (data not shown).

Additionally, because IGF II mRNA was previously reported to be expressed at a higher level in CAF than in NAF, IGF II mRNA was assessed by quantitative real-time PCR in monolayer cultures of NAF and CAF. The relative expression levels for each fibroblast culture are provided in Fig. 8. Although more CAF cultures than NAF cultures expressed IGF II mRNA at relatively high levels, the mean relative expression level of IGF II mRNA for NAF (4.2, n = 5) and for CAF (7.7, n = 5) did not differ significantly (P = 0.390).

Co-culture of MCF10A cells or MCF10AT cells may enhance expression of IGF II in fibroblasts, as has been reported in fibroblasts co-cultured with MCF-7 cells [33]. Furthermore, expression of IGF II may be enhanced to a greater degree in CAF versus NAF, thus possibly explaining the greater rate of proliferation of MCF10AT cells in co-culture with CAF than with NAF. 3D culture of NAF and CAF also could alter expression of IGF II in comparison of the monolayer culture. To investigate these possibilities, IGF II levels were assessed by ELISA in 3D monocultures of the same three NAF cultures and three CAF cultures used in previous co-cultures, in 3D monocultures of MCF10AT cells, and in 3D co-cultures of MCF10AT cells with NAF and CAF in an E:F of 2:1, identical to those co-cultures assessed for BrdU labeling previously (Table 4). In the 3D cultures, fibroblasts expressed IGF II at a significantly higher level than the epithelial cells (P < 0.01), and there was no difference in IGF II between NAF and CAF (P > 0.05). In co-cultures, the overall expression of IGF II was lower than in fibroblast monocultures. This latter result is expected because of the addition of a relatively large number of MCF10AT cells (E:F of 2:1) that express IGF II at a lower level than fibroblasts. Furthermore, there was no difference in mean IGF II levels in co-cultures with NAF versus CAF. The IGF II levels in 3D cultures were generally higher than in monolayer cultures, and this was particularly true when comparing 3D monocultures of fibroblasts versus two-dimensional monocultures of fibroblasts. The potential reasons for this include an effect of 3D growth, or the presence of Growth Factor Reduced Matrigel^® on the expression of IGF II, and/or deviations between the ELISA runs.

When the protein levels of IGF II in 3D monocultures of fibroblasts or in 3D co-cultures were correlated with the rate of proliferation of MCF10AT cells in co-culture with the matching NAF or CAF, no significant correlation was observed (r = 0.030 or r = 0.258, respectively; P > 0.05, Pearson Product Moment Correlation).