|Home | About | Journals | Submit | Contact Us | Français|
Organs are made of the organized assembly of different cell types that contribute to the architecture necessary for functional differentiation. In those with exocrine function, such as the breast, cell–cell and cell–extracellular matrix (ECM) interactions establish mechanistic constraints and a complex biochemical signaling network essential for differentiation and homeostasis of the glandular epithelium. Such knowledge has been elegantly acquired for the mammary gland by placing epithelial cells under three-dimensional (3D) culture conditions.
Three-dimensional cell culture aims at recapitulating normal and pathological tissue architectures, hence providing physiologically relevant models to study normal development and disease. The specific architecture of the breast epithelium consists of glandular structures (acini) connected to a branched ductal system. A single layer of basoapically polarized luminal cells delineates ductal or acinar lumena at the apical pole. Luminal cells make contact with myoepithelial cells and, in certain areas at the basal pole, also with basement membrane (BM) components. In this chapter, we describe how this exquisite organization as well as stages of disorganization pertaining to cancer progression can be reproduced in 3D cultures. Advantages and limitations of different culture settings are discussed. Technical designs for induction of phenotypic modulations, biochemical analyses, and state-of-the-art imaging are presented. We also explain how signaling is regulated differently in 3D cultures compared to traditional two-dimensional (2D) cultures. We believe that using 3D cultures is an indispensable method to unravel the intricacies of human mammary functions and would best serve the fight against breast cancer.
The mammary gland is composed of a series of branched ducts that connect the functional glandular units (acini) to the nipple (see Fig. 1). As early as 1840, Sir Astley Paston Cooper published his observation of a branched organization of the mammary gland with distinct lobes or ductal systems, which each opens at the nipple (1). The existence of multiple ductal systems, the number of which varies between individuals, is now well established (2, 3).
The branching ducts, as well as acini located at the extremities of each ductal system, are composed of two cell layers: an inner layer of secretory luminal epithelial cells, with apical microvilli, surrounded by contractile myoepithelial cells. The luminal cells of acini are arranged radially with tight junctions between cells located at the narrow width of the cells, near the central lumen (see Fig. 1). The primary function of the acinus is milk secretion. Milk flow into the ducts is powered by the contractions of the myoepithelial cells.
The breast epithelium has a unique behavior compared to other tissues in the human body since it continues to develop after birth. It undergoes extended remodeling with cycles of branching, acini formation, and dissolution of epithelial structures during puberty, pregnancy, lactation, and involution. Moreover, there are less extensive and repeated modifications during menstrual cycles (4). The BM, a specialized form of ECM linking epithelial and connective tissues, and the adjacent stroma that traps an abundance of soluble factors constitute the microenvironment of the epithelium. Myoepithelial cells and luminal cells are in contact with the BM (5, 6), the composition of which also undergoes modifications according to the physiological status of the mammary gland (7, 8). The BM is constituted of a polymeric network of collagen IV and laminins, notably laminins-111 and -332. The laminins and collagen IV are interconnected by nidogen and perlecan (9). Many of the BM components are involved in crucial signaling events that regulate tissue-specificity and function. The framework for such signaling was proposed already in the early 1980s (10), and the first proof in the mammary gland was provided for the expression of the milk protein β-casein which is controlled by laminin-111/β1 integrin signaling (11). The BM is also a repository for growth factors and cytokines that upon binding to their receptors trigger specific intracellular signals (12, 13).
A key feature of all luminal epithelia is the basoapical polarity axis. Transmembrane integrins at the basal side of cells serve as anchorage points and receptors for BM components. They trigger intracellular signaling and participate in the perception of the cells’ microenvironment. They cooperate with growth factor receptors to control essential cellular processes such as survival, proliferation, and differentiation (14–16). Among the cell–BM contacts, basal polarity is specifically determined by the interaction between laminin-332 and α6/β4 integrin dimers that form hemidesmosomes (15). Lateral cell–cell contacts are mediated by apical tight junctions, adherens junctions, and in some instances desmosomes (17). The location of tight junctions, the uppermost apical cell–cell adhesion complex, is paramount as it permits to separate cell membrane components and receptors between the apical and basolateral cell membranes and thus, strictly defines apical polarity. The tight seal generated by tight junctions prevents milk leakage in-between cells during lactation. The apical junctional complex formed by tight and adherens junctions also organizes the cytoskeleton and associated signaling pathways, which ultimately impinges on nuclear functions. Thus, the basoapical polarity axis permits unidirectional secretion of milk components in the lumen, as well as structured integration of hormonal and mechanical signals exerted by the microenvironment.
Characterizing the mechanisms underlying normal cell behavior in the context of an organized ductal system is critical to understanding which alterations are necessary for breast cancer to progress. This is particularly important for prevention research related to breast cancer that aims at reducing the burden of this important public health concern.
Tissue architecture (i.e., the organized arrangement of cells into specific multicellular structures) has been shown to be critical for the maintenance of functional differentiation and cell survival (18). It comes as no surprise that alterations in tissue architecture are needed to permit tumor formation and that tissue and cellular organization is commonly used by pathologists to precisely diagnose breast cancer. It has been proposed that the loss of apical polarity is a critical event necessary for tumor development (19). Indeed, 3D culture models of breast acini show that only cells with disrupted apical polarity can be pushed into the cell cycle (19). This hypothesis is supported by the observations reported by several laboratories that tight junction proteins can influence cell proliferation and act as tumor suppressors (17). The impact of basal polarity on the maintenance of apical polarity is not clearly determined. Our recent results suggest that collagen IV and hemidesmosomes both influence the integrity of tight junction organization (20), indicating that microenvironmental alterations might be sufficient to perturb apical polarity. The steps that follow initial apical polarity alterations and ultimately lead to the multilayering of epithelial cells characteristic of preinvasive neoplastic stages (hyperplasia and carcinoma in situ) remain to be uncovered. Among the internal cellular changes associated with apical polarity loss is the relocation of the cell nucleus away from the basal side, as observed in cells of preinvasive neoplastic stages (21).
Breast cancer progression toward invasive stages is accompanied by the breakage of the BM (22) which allows cells to invade the underlying ECM and move into the surrounding tissue. Invasive breast tumors in vivo have been characterized by altered expression and localization of BM proteins (23) and receptors, indicating that both the organization of the microenvironment and receptor–ligand interactions are profoundly altered (24, 25). Changes in intracellular organization have also been observed during tumor progression. Notably the nucleus of invasive tumor cells displays striking changes in the distribution of splicing factor speckle components, heterochromatin and euchromatin domains, and the nuclear mitotic apparatus (NuMA) protein compared to phenotypically normal acinar cells in culture (refs. 19, 26, 27 and unpublished results from the Lelièvre laboratory). Interestingly, the distribution patterns of the chromatin-associated NuMA protein observed in 3D cultures that mimic phenotypically normal and cancerous tissues have been observed also in vivo using archival biopsy samples (Lelièvre and Knowles, unpublished data), suggesting that architectural changes observed in 3D cultures are good predictors of what could be seen in real tissues.
The pioneering work of Mintz and Illmesee (28) with teratocarcinomas brought the paradigm shifting concept that tissue architecture can override genetic and genomic changes (29). The subsequent demonstration that it is possible to induce acinus-like structures from cancer cells with profoundly altered genomes by simply modulating signaling pathways (see Subheading 3.3) has unambiguously revealed the critical role played by epithelial architecture in controlling cell fate (30).
Monolayer cultures (2D cultures) have been very useful models for gene discovery and early work on viral transformation but are a far cry from physiologically relevant models. Flattened cell morphologies and the spatial plane of cell–cell contacts obtained in 2D cultures are strikingly different from those observed in tissues. Cell shape is known to influence cell behavior including growth and nutrient uptake (31, 32) and gene expression (29) (see Subheading 1.7), which may partially explain the decreased expression of tissue-specific genes often observed in 2D cultures.
Organ cultures are physiologically relevant model systems but are often technically challenging and typically short-lived due to necrosis in tissue explants. Culturing cells in 3D is a more flexible alternative to organ culture. One important goal is to enable the formation of tissue structures that have precise geometrical and functional signatures. This can be achieved by providing cells with proper mechanical and chemical signals from both specific types of architectural components of the ECM and soluble molecules. Serum varies in composition and contains high levels of growth factors and hormones in unpredictable concentrations. It is known to disrupt the ability of cells to express their tissue-specific functions (33) and its use should be avoided if at all possible.
Although most cells in culture synthesize ECM components, the establishment of 3D structures from single cells usually requires the use of hydrogels (viscoelastic meshworks consisting of two or more components, one of which is water) containing exogenous ECM components that provide the structural and biochemical signaling necessary for the formation of the correct architecture and differentiation status. Originally, 3D cultures were performed in floating collagen gels, as demonstrated with murine cells (34). Using this method, morphological characteristics of differentiation, including basoapical polarity, was maintained in culture. Cells at the surface of the gel formed monolayers and some cells below the gel surface rearranged themselves to form acinus-like structures. Nowadays, Engelbreth-Holm-Swarm (EHS) extracts are commonly used as hydrogels (see Subheading 1.5). While a large number of human breast cancer cell lines can be cultured in 3D to mimic tumor development (35), the recapitulation of phenotypically normal acinar phenotypes is more challenging and usually requires the presence of BM components in the exogenous hydrogel (20, 36). A few non-neoplastic breast epithelial cell lines have been used in 3D cultures including MCF-10A (37) and HMT-3522S1 (38) cells (for a comparison of the MCF-10A and S1 models please refer to ref. 20). This chapter mainly focuses on the 3D culture of the HMT-3522 series derived from S1 cells that provides a well-studied model of breast cancer progression. The use of primary cells as complements to immortalized cell lines is also discussed.
The HMT-3522 progression series is derived from a benign mammary fibrocystic lesion (38). HMT-3522 cells became spontaneously immortalized in culture, giving rise to the non-neoplatic S1 cell line. In the presence of the appropriate substratum (i.e., BM components and a specific mechanical environment) (20), S1 cells differentiate into basoapically polarized acinus-like structures of approximately 30 μm in diameter containing 25–35 cells (36) (see Fig. 2). Basoapical polarity is a critical feature of normal breast epithelia. The polarity axis is evidenced by the presence of basal BM components (laminin-332 and collagen IV) deposited by the cells, basally localized α6/β4 integrins—the bona fide laminin receptor dimer in breast epithelia—lateral cell–cell adherens junctions (with Ecadherin and β-catenin used as markers), and lateroapical tight junctions (with ZO-1 and ZO-2 core plaque proteins used as markers). The presence of a tiny lumen (often less than the size of a single cell) in S1 acini further indicates close resemblance to acini in the resting mammary gland (20). S1 cells express the luminal marker cytokeratin 18 (30). They have a number of genetic alterations (39) probably linked to their immortalized cell line status. A sub-population of S1 cells carries a mutation in p53 (His to Asp at codon 179). This mutation may confer a slight growth advantage, since its frequency in the cell population progressively increases with passage numbers (40). This phenomenon requires that the use of the S1 cells be restricted to passages below 60, to avoid the drift of the cell population. Acini formed by HMT-3522S1 cells are typically composed of one layer of luminal epithelial cells. In contrast, acini in the mammary gland are constituted by a layer of luminal cells surrounded by contractile myoepithelial cells (see Subheading 1.1 and Fig. 1). Therefore, in 3D monoculture models, luminal cell–ECM contacts are prevalent, whereas in mammary glands in vivo, luminal cells are mostly in contact with myoepithelial cells and make only punctual contacts with the BM (5, 6). Myoepithelial cells are largely responsible for making BM components (41). Surprisingly, although S1 cells are luminal in their behavior and main characteristics, the basal portion of these cells displays myoepithelial characteristics (e.g., presence of vinculin). We believe that these myoepithelial characteristics permit the formation of the appropriate endogenous BM necessary for acinar differentiation.
Malignant transformation of the S1 cells was achieved in vitro by altering the composition of the culture medium (39, 42). The HMT-3522S2 subline, growing independently of EGF and bearing characteristics of preinvasive carcinoma (43–45), was isolated. These cells are actually sensitized to EGF and would not thrive in its presence. They also require coating the flasks with collagen I for 2D culture. S2 structures in 3D culture are heterogeneous in size. Homogeneous S3 sublines (HMT-3522S3-A, S3-B, and S3-C) have been derived from S2 3D cultures by selecting for colony size (45). S3 cell lines display progressive loss of polarity, genomic anomalies, and gene expression changes characteristic of preinvasive to invasive transition (45).
S2 cells that had reached passage 238 were found to produce tumors in nude mice. The T4-2 subline was obtained from a tumor after two rounds of in vitro–in vivo mouse passage. T4-2 cells are highly tumorigenic and have a triple-negative phenotype (i.e., no expression of estrogen, progesterone, and ErbB2 receptors) (45). In 3D culture, T4-2 cells develop structures reminiscent of invasive breast tumors (19, 30). The nodules reach approximately 200 μm in diameter after 10 days in culture (see Fig. 2). Thereafter, the size of the nodules remains relatively stable due to a balance between cell division and apoptosis.
Compared to widespread models of aggressive and metastatic cancers that give little insights into the early events of cancer progression, the HMT-3522 cancer progression series offers unique opportunities to study early phases of tumorigenesis. The series allows for direct comparisons between nonmalignant acini models (S1), preinvasive nodules (S2), and invasive tumors (T4-2). There exists other interesting breast cancer progression series. Non-neoplastic MCF-10A parent cell lines were reported to form acinus-like structures devoid of apical polarity as their acini usually lack apical tight junctions (20, 46, 47). These cells can undergo malignant transformation upon exogenous expression of chimeric ErbB2 receptors (48). This receptor consists of the ErbB2 intracellular domain fused to the synthetic ligand-binding domain of the FK506-binding protein (FKBP) and to the extracellular and transmembrane domains of the p75 nerve growth factor receptor. Homodimerization of chimeric p75-ErbB2 can be induced with a FKBP ligand, leading to the activation of ErbB2 signaling without interfering with endogenous ErbB2 receptors. Activation of p75-ErbB2 in MCF-10A acini produced in the presence of EHS extracts induces cell proliferation and multilayering. The resulting structures retain epithelial properties and are not invasive; thus, they may represent a model of early stage mammary cancer in vitro (48). MCF-10A cells have also been transformed using T24 Ha-ras, giving rise to the MCF10-AneoT cell line (49). The MCF10-AT cell lines, derived from xenograft-passaged MCF10-AneoT, represent premalignant stages. However, a subset of MCF10-AT xenographs developed into preinvasive and invasive carcinomas in immunodeficient mice. One such tumor was isolated and cultured in vitro, leading to the malignant MCF10CA1 cell line (50, 51).
One of the major challenges of 3D culture is to reproduce a microenvironment close enough to the in vivo situation to permit proper differentiation into phenotypically normal tissues, and to accurately mimic different tumor stages. During the development of 3D models of specific tissues, it is important to test for compliance with the architecture and physiology of the in vivo counterparts. An ECM-like hydrogel substratum and an appropriate medium need to be carefully chosen, in order to solve the problem under study using conditions relevant to the organ in vivo. Ideally, well-characterized hydrogels with defined ECM components should be used. However, with the exception of collagen I, pure ECM components are very costly and difficult to formulate as hydrogels. As an alternative, an ECM mixture isolated from Engelbreth-Holm-Swarm mouse sarcoma cells (52) provides an acceptable BM approximation in terms of components and organization, especially when studying noninvasive breast tissue structures and tumors. Commercially available EHS extracts contain laminin-111, type IV collagen, proteoglycans, and entactin. The composition of commercially available EHS extracts is not fully elucidated and varies between lots; this implies that lots need to be tested for a particular application (see Note 1). Engineered ECM-like hydrogel substrata are being developed as alternatives to EHS extracts, but, to our knowledge and to this date, there is no report that they allow non-neoplastic breast epithelial cells to recapitulate acinar differentiation.
In the mammary glands, acini are connected to ducts (see Fig. 1); whereas human acinar models in 3D cultures are isolated sphere-like entities that rarely branch out. The 3D culture protocols aim at keeping the growth conditions as constant as possible using chemically defined cell culture media and carefully tested lots of EHS-derived hydrogels. Despite these efforts, variations inherent to the biological origin of the additives used in the cell culture media and components of the hydrogel are difficult to control, as is the heterogeneity of primary cultures or even established cell lines. Moreover, the physiological changes found in vivo (e.g., estrous cycles and functional changes associated with puberty, pregnancy and menopause) are yet to be readily explored using 3D culture models.
Ultimately, results obtained using 3D models need to be validated in vivo. This is a challenge when working with human cells. Available in vivo alternatives are murine models bearing human cells (e.g., cells cultured in cleared fat pad) and human tissues from biopsies and reduction mammoplasties. Experiments with xeno-graft models can be influenced by the animals’ physiology and the context into which the cells are injected, and human biological samples provide only frozen moments in time. Nevertheless, there have been several compelling examples showing that what was discovered in 3D cultures indeed illuminated important phenomena in vivo (see Subheading 1.7).
Primary cells are nonimmortalized cells obtained directly from tissues. Breast tumor cells can be obtained from breast cancer patients undergoing surgical treatment whereas non-neoplastic mammary cells are typically derived from reduction mammoplasty or from milk. Primary cells have proven to be useful to study breast phenotypes (53). They are heterogeneous and hence truly represent their tissue of origin. Primary cells do not carry genetic (polyploidy, mutations) and phenotypic (e.g., rapid growth) alterations linked to immortalization. However, their use in culture is limited due to the small number of divisions achievable in vitro, which hampers long-term studies. Variability within primary cell populations also represents a challenge since it may reduce experimental reproducibility. Finally, compared to immortalized cell lines, the access to primary cells is limited.
This section is a nonexhaustive description of the use of 3D human breast epithelial cultures in a number of research areas (54). Due to space limitations, we have focused only on a few studies presenting compelling and original demonstrations of the usefulness of 3D cultures to study different aspects of breast tissue biology.
Three-dimensional tissue models have provided unique information pertaining to cell signaling, notably that there exist crosstalks between signaling pathways that are only established under certain architectural conditions (i.e., when an acinus is formed). The crosstalk between integrin and growth factor receptors signaling in the glandular epithelium is now well established (16). Originally, it was shown that epidermal growth factor receptor (EGFR) and β1 integrin signal transduction pathways were coupled in breast acini obtained in 3D culture, but not in 2D cultures of the same breast epithelial cell line (55). The predominance of tissue architecture in establishing specific signaling networks was confirmed by the restoration of the crosstalk in cancer cells induced to form basally polarized acinus-like structures. This example illustrates the large impact of tissue architecture on signaling from the microenvironment (29, 56).
Three-dimensional cultures are being used extensively also to study how basoapical polarity is established, maintained, and compromised in mammary epithelia. In particular, laminin-111 was shown to be essential for proper polarization. Luminal epithelial cells cultured on collagen I substratum adopted an inverted polarity; but proper polarity could be rescued by laminin-111 (23). Blocking ECM signaling through β4 integrins prevented the phenotypic reversion of the malignant T4-2 HMT3522 cells (30, 57), and blocking the same signaling pathway in S1 acini compromised apical polarity (20), suggesting that the establishment of basal polarity is needed to maintain apical tight junctions. The underlying mechanisms remain to be discovered.
Intracellular organization in 3D cultures often closely resembles the organization in vivo. In particular, the nuclear structure of acini produced in 3D culture and that observed on sections of acini in resting mammary glands are strikingly similar. This observation is illustrated by the distributions of the nuclear mitotic apparatus (NuMA) protein (ref. 58, Knowles and Lelièvre, unpublished), and certain markers of higher order chromatin organization (59). A tremendous advantage of 3D culture models over fixed human tissues is that they can be used for functional experiments with physiological relevance. Studies conducted with 3D acini and 3D tumor models have revealed that nuclear organization actively modulates cell and tissue phenotypes (60). A clear demonstration of the impact of nuclear organization on cell behavior in epithelial cells came from studies performed with mammary acini in 3D culture (26). Alterations induced in the distribution of the nuclear protein NuMA using function blocking antibodies, expression of dominant negative truncated forms of the protein, and siRNAs were shown to impair acinar differentiation, alter BM integrity and lead to proliferation and cell death (19, 26, 59). This “dynamic reciprocity” between NuMA and the cell phenotype was not observed in 2D cultures. More particularly, the work performed in 3D culture demonstrated that the protein NuMA was influencing the higher order organization of chromatin (59) normally achieved upon acinar differentiation (19). Recurring findings that changes in cell shape during differentiation are accompanied by the remodeling of nuclear organization and alterations in gene expression profiles (60) strengthen the importance of studying biological processes in tissue contexts.
Specific recognition of the influence of tissue architecture on tumor development has come from the demonstration that cancer cell phenotype can be reverted by selectively modulating cell communication with the microenvironment (18). Relatively simple experiments had already indicated that the arrangement of tumor cells could impact their behavior. Indeed tumor cells were found to change their sensitivity to cancer chemotherapeutic drugs when placed on soft agarose (61). The behavior displayed by cells organized into a tumor was in fact similar to that of xenografts, indicating that the formation of a tumor nodule (instead of a flat monolayer of cells) mimicked the in vivo situation concerning the response to treatments. Later on, it was demonstrated that in breast cancer cells placed in 3D culture, drug sensitivity was influenced by basal polarity, and notably that hemidesmosome-directed signaling was conferring resistance to treatments aimed at killing cancer cells (57). These findings have important implications for research on DNA repair (Rizki, Jasin, and Bissell, unpublished) and apoptosis, and for the development of chemotherapeutics. They clearly indicate that 2D cultures, although easily amenable to high-content assays, are poor predictors of the effect of drugs in vivo.
Finally, 3D cultures in hydrogels have revealed the importance of mechanical stimuli for tissue differentiation. Manipulation of the mammary gland's microenvironment stiffness (using glutaraldehyde fixation of collagen I gels or by mixing collagen I to laminin-111 gels) was found to affect β-casein expression (24). Interestingly, increased ECM stiffness characteristic of malignant tumors resulted in increased cell growth and altered organization of non-neoplastic mammary epithelial cells (62).
In summary, virtually all cellular processes studied so far appear to be influenced by the architectural and microenvironmental contexts. Hence, 3D culture models provide invaluable tools to elucidate fundamental biological questions under physiologically relevant conditions.
Unless indicated otherwise, additives are dissolved in Milli-Q water, filter-sterilized (0.22 μm), aliquoted, fast-frozen in liquid nitrogen and stored at –80°C.
Culture cell lines from the HMT-3522 breast cancer progression series (39) in the absence of serum in chemically defined H14 medium (38, 63) (see Note 2 and Subheading 3.1, step 1). The same culture medium is used for nonmalignant S1, preinvasive S2/S3 cells, and malignant T4-2 cells, except that EGF is omitted from S2, S3, and T4-2 cultures. HMT-3522 cell lines are propagated in 75 cm2 cell culture flasks. For S2, S3, and T4-2 cultures, PureCol®-coated flasks are used (see Subheading 2.2, step 4). Rigorous attention should be paid to seeding density to avoid changes in phenotypes (see Note 3).
Splitting cultures of HMT-3522S1, S2, and T4-2 cells:
Cells are cultured in 3D in the presence of the hydrogel. Several techniques have been developed (refs. 20, 64, see Fig. 3) for this type of culture. Irrespective of the culture method used, normal and malignant phenotypes obtained in 3D culture need to be validated (see Note 4). The embedded method was initially used for HMT-3522 cultures (36). In this setting, individual cells are included within EHS-derived gels covered by culture medium. The technique offers the advantage of providing a homogeneous microenvironment to the cells. It is useful to study the invasive phenotype of tumors rather than the invasive capabilities of individual cells measured in Boyden chambers (65). However, the thick gel layer renders a number of applications challenging, in particular direct fluorescence imaging. Embedded cultures can be cryo fixed and sectioned for immunostaining experiments (see Subheading 3.6).
The drip method is an alternative to the embedded method which offers greater flexibility in the experimental design, especially for high-resolution imaging applications.
The drip method has also been used for 3D culture of other non-neoplastic (e.g., MCF-10A) and malignant breast cell lines including MCF-7, MDA-MB-231, BT20, and BT-474 (35).
A new technique has been developed for HMT-3522S1 acini culture (ref. 20, see Fig. 3). This technique, referred to as “high throughput” (HTP), can be used to obtain S1 acini with basoapical polarity and morphological characteristics indistinguishable from acini in “drip” or “embedded” cultures. The HTP technique circumvents the need to coat cell culture surfaces with EHS-derived gel, hence reducing handling time and permitting the production of large quantities of acini for high-content screening methods and biochemical analyses. The absence of a gel coat on the culture surface greatly reduces background signals in live cell and immunostaining experiments.
The phenotype of malignant cells can be “reverted” to mimic phenotypically normal acinus-like spheroids or partial acinar differentiation (see Fig. 2). This approach is particularly useful to determine if a difference in phenotype or behavior between non-neoplastic and neoplastic cells results from differences in tissue architecture, cell cycle status, or the genetic background. Moreover, models of non-neoplastic multicellular structures like the acini can be manipulated, notably to affect the basoapical polarity status. Established protocols for tumor cell reversion and for the manipulation of non-malignant phenotypes in 3D culture are referenced in Table 1. A powerful method is to add function-blocking antibodies targeting integrins to the culture medium starting the day of cell plating. 1 mL of cell suspension is first incubated for 30 min with 15 μg/mL antibodies at 37°C; then cells are plated in 3D culture in the presence of 15 μg/mL antibodies. This technique has been used to obtain reverted T4-2 structures resembling S1 acini but lacking apical polarity.
The same approach was applied to block the establishment of basoapical polarity (57) in S1 cultures. It is also possible to treat with antibodies or small molecules after the establishment of non-malignant acinar structures and tumor-like colonies, hence mimicking clinically relevant situations to assess specifically the effect of potential anticancer drugs on their tumor targets and/or the side effects on normal tissue (66). An example of the application of this method is illustrated by the suppression of β4-integrin signaling in S1 acini (20). Preformed acini are released from the drip culture with dispase (see Subheading 3.5) and incubated with function blocking antibodies (30 min at 37°C), before replating in the presence of antibodies using the drip method (see Subheading 3.2.2).
Cells in 3D culture usually have a morphology that does not seem amenable to the use of classical transfection reagents. Moreover, non-neoplastic breast epithelial cells usually become quiescent after a few days in the presence of EHS-derived or collagen 1 gels. To circumvent these limitations, cells can be transiently transfected in 2D culture, just before plating them under 3D culture conditions. This approach can be applied to silence the expression of genes coding for proteins with relatively slow turnover using RNA interference (59). Stable transfection of cells in 2D culture prior to use in 3D culture is also possible (57, 59), but clonal selection should be avoided since some of the clonal populations might lack/lose differentiation capability. Other possibilities include viral infection followed by selection of stably transduced cells in 2D culture prior to placing them in 3D cultures (67). If the transgene should only be expressed following differentiation in 3D culture, conditional expression should be used (68).
Function-blocking antibodies or peptides can be introduced inside cells in 2D and 3D cultures to disrupt protein functions or to manipulate signaling pathways. This method requires plasma membrane permeabilization (58, 59):
Although technically challenging, microinjection can be applied to 3D tissue-like structures (20) and could be used for nucleic acid delivery.
Classical biochemical analyses of RNAs and proteins are possible using 3D cultures. Although some of these experiments can be performed by extracting the cellular material directly from 3D cultures containing the hydrogel, it is often better to first isolate the tissue structures from the exogenous substratum, typically by solubilization of the ECM using proteases (collagenase for collagen I gels or dispase for EHS-derived gel). Cell suspensions are then processed using standard protocols. Below is a protocol used for the isolation of HMT-3522S1 acini and T4-2 tumor nodules:
In contrast to trypsin/EDTA treatments, dispase (or collagenase) treatment does not affect cell–cell contacts. This is important since disruption of cell–cell junctions with EDTA would result in profound changes in cellular organization and gene expression that may influence the outcome of the experiments. The recently developed HTP culture for HMT-3522S1 acini (ref. 20, see Subheading 3.2.3) is an alternative for larger-scale production of acinus-like structures. With the HTP culture, dispase treatments with shorter incubation time (10 min) can be used to release the acini.
Immunostaining can be performed on tissue structures obtained with different 3D culture techniques (see Subheading 3.2). Embedded cultures are too thick for direct immunolabeling and imaging. Therefore, cryosections have to be prepared (20, 30):
We thank Dr. Kurt Hodges for providing micrographs from tissue sections used in Fig. 1. Support was provided by the National Institutes of Health (R01CA112017 and R03CA112613) and the Susan G. Komen Breast Cancer Foundation (BCTR-0707641) to SAL; the US Department of Energy, Office of Biological and Environmental Research, via a Distinguished Fellow Award and Low Dose Radiation Program (DE-AC02-05CH1123), the National Institutes of Health (R37CA064786, U54CA126552, R01CA057621, U54CA112970, U01CA143233, U54CA143836—Bay Area Physical Sciences–Oncology Center, University of California, Berkeley, California), and the US Department of Defense (W81XWH0810736) to MJB; postdoctoral fellowships from the Novartis Foundation and the Swiss National Science Foundation (PBNEA–116967) to PAV.
1Matrigel™ is only good for 2 years when stored at –80°C. Due to variability inherent to biological materials, new gel lots need to be tested in order to ensure reproducibility between experiments using different lots. A typical test procedure consists of culturing side-by-side S1 cells using new as well as currently used gel lots. After 10 days in culture, morphological characteristics (diameter and smoothness of the multicellular structures) are evaluated and basoapical polarity and cell proliferation are scored after immunostaining for specific markers (see Note 4). T4-2 tumor phenotypes also need to be validated when changing the gel lot.
2A major reason for not using serum for the culture of HMT-3522 cells is to have a defined medium to monitor tissue phenotypes and signaling. It is also desirable to prevent exposure to concentrations of growth factors and cytokines that the cells do not generally experience in vivo.
3In order to avoid phenotype drifts (i.e., selection of cells with more or less aggressive behavior or less differentiation capabilities), cells are only used within defined passage numbers (passages 52–60 for S1 cells; T4-2 cells can be split usually ten times without seeing a shift in aggressiveness compared to the first passage used in 3D culture). In addition, S1 cell cultures are split no sooner than day 8 and no later than day 12. These cells need time to deposit BM components on the cell culture surface. Passages performed before day 8 would select for rapidly adhering cells that do not differentiate well. Waiting longer than 12 days may lead to difficulties in detaching cells with trypsin. T4-2 cell cultures are split when 70–80% confluence is reached (typically after 4–5 days in culture) as waiting longer would select for cells with less aggressive phenotype in 3D culture. Finally, the culture medium is replaced no earlier than 48 h after plating the cells in order to avoid selection for fast-adhering (more aggressive) S1 cells and to allow cells to benefit from the production of paracrine and autocrine factors during the initial phase of the culture. In 2D culture, HMT-3522S1 cells organize into patches (islands) delineated by a layer of cells that display a morphology typical of epithelial cells, with nuclei basally located against the inside of the island and cytoplasm extending toward the outside of the island (see Fig. 2). The proliferation rate of S1 and T4-2 cells in seed cultures needs to be monitored since aberrant growth rates often reflect changes in cellular metabolism that have profound consequences on the phenotypes obtained in 3D cultures.
4Quality control is essential to confirm that tissue structures obtained in 3D culture indeed mimic physiologically relevant tissue structures in vivo. The first step in the validation process is the morphological comparison between structures obtained in 3D culture and their counterpart in vivo. Under optimal conditions, S1 cells in 3D culture should differentiate into smooth spherical structures of ~30 μm in diameter, whereas T4-2 cells should produce irregular structures with a broader size distribution (50–200 μm range). Differentiated mammary acini are quiescent (i.e., cells have exited the cell cycle) and basoapically polarized (see Figs. 1 and and2).2). In a typical S1 acini culture, close to 100% of the acini display correctly distributed basal markers and approximately 75% of the acini display correctly distributed apical markers under optimal culture conditions. The percentage of apically polarized acini may drop to 50% or 60% from one experiment to another since apical polarity seems very sensitive to microenvironmental conditions (20). In contrast, T4-2 tumor-like structures contain proliferating cells and, although certain basal and apical polarity markers are expressed, these makers typically show highly disorganized distributions. The percentage of quiescent cells can be monitored by Ki67 immunostaining. Basal polarity markers include type IV collagen, laminin-332, and α6 and β4 integrins. Tight junction protein ZO-1 is used as a robust apical polarity marker.
5In our experience, cells are much more resilient to alterations in their environment when cultured in 2D compared to 3D. For example, subtle variations in the cell culture medium—such as altered water quality or variations in additive concentrations—may have no visible effect on cells cultured in 2D while profoundly altering cell behavior (differentiation, proliferation, survival) in 3D culture. Moreover, treatments with certain drugs or dyes that may be well tolerated by cells in 2D cultures might drastically change phenotypes in 3D culture and even lead to cell death.