Organoid model of mammographic density displays a higher frequency of aberrant colony formations with radiation exposure

Abstract

Motivation

Aberrant three-dimensional (3D) colony organization of premalignant human mammary epithelial cells (HMECs) is one of the indices of dysplasia. An experiment has been designed where the stiffness of the microenvironment, in 3D culture, has been set at either low or high level of mammographic density (MD) and the organoid models are exposed to 50 cGy X-ray radiation. This study utilizes published bioinformatics tools to quantify the frequency of aberrant colony formations by the combined stressors of stiffness and X-ray exposure. One of the goals is to develop a quantitative assay for evaluating the risk factors associated with women with high MD exposed to X-ray radiation.

Results

Analysis of 3D colony formations indicate that high stiffness, within the range of high MD, and X-ray radiation have an approximately additive effect on increasing the frequency of aberrant colony formations. Since both stiffness and X-ray radiation are DNA-damaging stressors, the additive effect of these stressors is also independently validated by profiling activin A-secreted protein. Secretion of activin A is known to be higher in tissues with a high MD as well as tumor cells. In addition, we show that increased stiffness of the microenvironment also induces phosphorylation of γH2AX-positive foci. The study uses two HMECs derived from a diseased tissue (e.g. MCF10A) and reduction mammoplasty of normal breast tissue (e.g. 184A1) to further demonstrate similar traits in the frequency of aberrant colony organization.

Supplementary information

Supplementary data are available at Bioinformatics online.

1 Introduction

Our goal is to develop a surrogate quantitative model where risks associated with X-ray radiation for women with high mammographic density (MD) can be evaluated. We begin by summarizing the (i) significance of high MD and its relationship with cancer risks; (ii) metrics for quantifying MD and (iii) increased cancer risk as a result of benign breast disease in the presence of high MD. Next, we proceed to the high content screening of the surrogate models of MD, using the 3D cell culture models, with the BioSig3D imaging bioinformatics system. The manuscript concludes with results.

MD varies considerably in the female population and has been shown to be an important breast cancer risk factor. Independent of radiation (breast tissue is radio-sensitive), high mammographic density (HMD) is a known breast cancer risk factor. This association is stronger than almost all other risk factors. Only age and BRCA1/2 mutation status are associated with a larger relative risk of breast cancer compared with breast density (McCormack and dos Santos, 2006). Women with dense breasts have up to a 4- to 6-fold higher risk of breast cancer compared with women with less dense breasts. One prospective cohort study (Vachon et al., 2007) indicated that approximately 26–32% of women in the general population have 50% or higher breast density, where 16–32% of breast cancers may be attributed to this trait (Byrne et al., 1995; Conroy et al., 2012), with an even larger proportion of breast cancers being associated with this trait in premenopausal women (Byrne et al., 1995). A second large-scale prospective cohort study (Kerlikowske et al., 2015) suggested that 47% of women aged 40–74 years had dense breasts. Summarizing the current multiple cohort studies, the data indicate that on average, 40% of women have dense or extremely dense breasts.

Low to high MD is currently documented during mammographic screening and classified using a 1–4 rating based on the Breast Imaging, Reporting & Data System (Boyd et al., 1995). Using this system, a rating of 1 denotes 25% dense tissue (almost entirely fat), a 2 denotes 26–50% dense tissue (scattered fibroglandular densities), 3 denotes 51–75% dense tissue (heterogeneously dense) and a 4 signifies more than 75% dense breast tissue (extremely dense). Women with a score of 3–4 are designated as having dense breasts. MD has become a standard clinical measure due to its strong association with breast cancer risk.

Benign breast disease, particularly atypical hyperplasia, is an established risk factor for breast cancer (Connolly and Schnitt, 1993). One study at the Mayo Clinic followed 9087 women with benign breast disease for 15 years (Hartmann et al., 2005). More than 700 patients in this cohort developed cancer, giving an overall relative risk factor of 1.56 for breast cancer for this group. The relative risks for nonproliferative, proliferative without atypia and proliferative with atypia were 1.27, 1.88 and 4.24, respectively. Another study, with 42 181 women, has shown that women with high breast density and proliferative benign breast disease are at a further increased risk of breast cancer, while women with low breast density are at low risk regardless of their benign diagnosis (Tice et al., 2013). The biologic and/or mechanistic relationships remain unknown, but it appears that proliferative cells grow into aberrant 3D architectures and this may be a rate-limiting step in the development of invasive breast cancer (Bissell and Hines, 2011).

2 High content screening of the 3D colony organization

One of the key components of high MD is increased collagen production that is associated with the increased tissue stiffness (Cheng, 2016, #30260). In a previous manuscript, we (i) collected histology sections from patients from low and high MD and measured the stiffness of the collagen component using atomic force microscopy and (ii) designed and validated matrices corresponding to the stiffness levels of low and high MD (Cheng et al., 2016). We refer to the 3D culture condition of premalignant cell lines, under high stiffness of the microenvironment, as the organoid models of high MD. The 3D culture uses the embedded protocol (Lee et al., 2007) to reduce colony heterogeneity, in matrices that modulate the stiffness of the microenvironment and provides uniform force for every cell in the colony. Regardless, due to cellular plasticity, colony heterogeneity persists, and high content screening is needed. Colony formations are imaged using confocal microscopy and profiled quantitatively with BioSig3D (Bilgin et al., 2013, 2016) imaging bioinformatics system. BioSig3D computes a number of indices from colony organization. Among these indices, colony flatness and surrogate marker of lumen formation are the most relevant to the biologically relevant dysplasia (a stage that precedes development of cancer). In the case of premalignant cells, aberrant colonies are not strictly flat but more ellipsoidal. However, the index is referred to flatness in order to maintain consistency with previous nomenclatures (Bilgin et al., 2016; Cheng et al., 2016). In this study, we use two premalignant cell lines. One is MCF10A, which has been derived from a patient with fibrosis (a benign breast disease). The second one is 184A1, which is also non-malignant and derived from a normal patient having gone through reduction mammoplasty. This line has been derived from normal tissue exposed to Benzo(a)pyrene and transformed immortal line is established (Stampfer et al., 1997).

These two lines were selected to study the differences in colony formation as a function of surrogate models of benign and normal mammary epithelial cells. Finally, it is important to point out that higher MD induces DNA damage in mammary epithelial cells, which has been quantified by increased level of γH2AX and secretion of activin A (DeFilippis et al., 2014). Activin A is a member of the TGFβ superfamily that is also overexpressed as a result of DNA damage (Fordyce et al., 2010). Therefore, increased secretion of activin A, as a result of higher stiffness coupled with the X-ray radiation (e.g. an exogenous DNA damaging), can provide an independent benchmark toward increased frequency of aberrant colony formations. In all experiments, the (i) low and high stiffness of the microenvironment are set at 250 and 2000 Pa, respectively; (ii) radiation exposure is set at 50 cGy, which is equivalent to several CT scans and (iii) colonies are exposed on day 2 of their culture conditions.

3 Results

High content screening of 3D colony formations revealed that colony structures are heterogeneous. Therefore, the consensus cluster (Monti et al., 2003) has been used to identify a taxonomy of cluster formations that would lead to a more sensitive computational assay. BioSig3D performs consensus clustering on a single computed morphometric index with and without injection of noise to assure stability of clusters. The details of sample size and statistical analysis are included in the Supplementary Material section. The results are summarized below and supported by an independent assay quantifying the production of activin A.

3.1 The frequency of colony flatness is increased as a function of combined radiation and the increased stiffness of the microenvironment

The normal pattern of colony morphogenesis is a 3D hollow sphere in about 7–10 days. Colony morphogenesis involves a series of delicate biological processes, where malignant and non-transformed cell lines have distinct phenotypic signatures in 3D (Han et al., 2010; Kenny et al., 2007). For example, loss of lumen formation and colony flatness are strong indicators of aberrant organization and loss of polarity in malignant cell lines and are associated with dysplasia. Aberrant colony organization is also associated with dysplasia, which is an early marker of malignancy. Figure 1 summarizes experimental results for four conditions as a function of two stiffness values of the microenvironment (e.g. 250 and 2000 Pa) and radiation exposure for MCF10A. Colony flatness, a potential marker for dysplasia, was computed for each colony, subtypes of colony formations were computed and the frequency of occurrence of flat colonies for each condition was computed. In the case of colony flatness (e.g. ellipsoid shape), increased stiffness and radiation exposure act additively to increase the frequency of aberrant colony formation. Similar results are observed with 184A1 but not included here.

Fig. 1.

Frequency of aberrant colony organization is increased as a function of radiation and increased stiffness of the microenvironment. (a–d) BioSig3D explores the number of subtypes to conclude that two subtypes are adequate representation of the data. (e–f) Computed subtypes correspond to the phenotypes of round and flat (e.g. elongated) colony organizations. (g) Frequency of each subtype cultured in low stiffness with/without IR. (h) Frequency of each subtype cultured in high stiffness with/without IR. Star indicates a P value < 0.001

3.2 The frequency for the loss of surrogate markers of lumen formation is increased at stiffness value corresponding to high MD and radiation exposure

Lumen formation is an important index for characterizing colony formation; however, it requires three markers (e.g. apical, basal and lateral) for validation, which is not conducive to high content screening. Hence, we proposed computing a surrogate marker of lumen formation in our earlier manuscripts (Bilgin et al., 2013, 2016; Cheng et al., 2016). This surrogate marker quantifies the amount of empty space inside of each colony. The rationale being that malignant colonies do not form lumen (e.g. a hollow space inside an organoid) and are filled with cancer cells. Results are shown in Figure 2, which indicate that increased stiffness (i) is associated with a higher frequency of aberrant lumen formation, and (ii) with low-dose radiation exposure increases the frequency of aberrant lumen formation even further. These observations are shown in comparison with the malignant control cell lines of MCF7 and SKBR3 (e.g. positive controls), which do not form lumens at all.

Fig. 2.

Frequency for the loss of lumen formation is increased as a function of X-ray radiation and high stiffness of the microenvironment: (a–d) BioSig3D examines the number of subtypes for lumen formation with three subtypes being statistically significant (shown in red). These are (i) hollow (e.g. lumen preserving colonies), (ii) partially hollow and (iii) solid colonies. (e) At low stiffness and without radiation, colonies form lumens and at low stiffness and radiation, some colonies do not form lumens. (f) At high stiffness, the frequency of lumen formation is reduced further and at high stiffness and radiation, the frequency of lumen formation is reduced further. (g) Positive controls for MCF7 (subtype of Lumina A) and SKBR3 (ERBB2 positive) show no lumen formation. * and # correspond to P < 0.001 when comparing to none irradiated samples

3.3 Secretion of activin A is increased at the higher stiffness of the microenvironment following X-ray radiation exposure in 3D cultures of MCF10A and 184A1 cells

Association of the DNA damage and repair (DDR) with increased activin A has been shown by mRNA and protein analysis (Fordyce et al., 2010). However, these experiments were performed in 2D cell culture models in the absence of physiologically relevant microenvironments. In the current experiments, secretion of activin A was quantified by ELISA for the two human mammary epithelial cell (HMEC) lines. First, a standard curve was constructed using purified activin A recombinant protein. Next, MCF10A and 184A1 were cultured in 3D in triplicate at different levels of stiffness (250, 2000 and 4000 Pa). Plates were irradiated with 50 cGy on day 2 post seeding. On day 4, the media was collected, and frozen for later analysis, and fresh media was added for normal cell maintenance. On day 9, the media was collected again and frozen. The results, shown in Figure 3, indicate that (i) activin A is overexpressed as a function of the increased stiffness of the microenvironment, (ii) acute radiation exposure and increased microenvironment stiffness have additive effects on the secretion of activin A and (iii) secretion of activin A is persistent and increases as the cultures continue to grow. While previously published research showed that activin A is increased in vHMEC isolated directly from patients with high MD (DeFilippis et al., 2014), our new experiments have shown that increased stiffness of the microenvironment, by itself, can induce this effect on cell lines and that radiation acts separately and additively to increase secretion of activin A.

Fig. 3.

Secretion of activin A is increased at the higher stiffness of the microenvironment following X-ray radiation exposure in 3D cultures of MCF10A and 184A1 cells. (a, c) Activin A secretion from 184A1 and MCF10A w/o radiation on day 4. (b, d) Activin A secretion from 184A1 and MCF10A w/o radiation on day 9. * and ** correspond to P < 0.001 when comparing to 250 Pa

3.4 Expression of γH2AX is increased as a result of increased stiffness of the microenvironment

Both cell lines were cultured in 3D at low and high stiffness of the microenvironment, fixed on day 9, stained and imaged. The positive control for the antibody was UV radiation in monoculture of MCF10A, which is shown in the Supplementary Figure S1. For each condition, 25 colonies were selected randomly, the middle slice of each of the 3D colonies was imaged with a confocal microscope, and foci (e.g. spots) were counted on a cell-by-cell basis for each colony. Quantitative results and representative images are shown in Table 1 and Figure 4, respectively. The main rationale for the immunofluorescence microscopy, as opposed to western, are (i) fewer colonies are needed and (ii) the agarose crystals in the high stiffness gel interferes with western assays.

Table 1.

Number of γH2AX foci is increased in the organoid models of two premalignant cell lines as a result of increased stiffness of the microenvironment

Cell line	MCF10A		184A1
Stiffness	High	Low	High	Low
No. of foci	26.5 ± 6.76	0.5 ± 2.2	13.62 ± 0.5	0.63 ± 0.3

Note: The table shows the number of foci per positive cells, where a positive cell refers to a cell, in a colony, with a least one foci.

Fig. 4.

Immunofluorescent staining shows DNA damage is introduced by high stiffness (HS) microenvironment for both MCF10A and 184A1 cell line, whereas low stiffness (LS) microenvironment triggers DNA damage at minimal level. Scale bar is 10 um

4 Discussion

The organoid models, derived from the two distinct cell lines, reveal that increased stiffness of the microenvironment, within the range of normal MD, and X-ray radiation have an additive effect toward increasing the frequency of aberrant colony formations. Here, an additive is not meant in a purely linear sense, but in the sense, that causal effects are integrated. These observations were made through application of BioSig3D imaging bioinformatics system. An emerging hypothesis is that since increased stiffness of the microenvironment (Cheng et al., 2016) and radiation-induced DNA damage cause aberration in colony formation then these external stressors should operate through the same pathways. In addition, the choice of the cell line also predicated normal (e.g. 184A1) versus a line originating from a benign breast disease tissue (e.g. MCF10A). When normalized for the number of colonies, MCF10A colonies consistently secrete more activin A than 184A1. Similarly, the number of foci per positive cells (e.g. cells having at least one foci), representing the DNA damage, is higher in MCF10A than in the 184A cell lines, which provide further correlative support. Finally, increased expression of activin A coupled with the number of foci per positive cell, because of the higher stiffness of the microenvironment, indicate that MCF10A is more sensitive than 184A1. This can be attributed to the fact that MCF10A was isolated from a patient with benign breast disease, which can potentially place those patients at a higher risk factor.