ECM: Extracellular Matrix; TDLU: Terminal Ductal Lobular Unit; CAF: Cancer-Associated Fibroblast

The prospect of precision or personalized medicine has revived the entire field of predictive drug testing originally pursued in clonogenic assays and soft agar colony forming assays some four decades ago [1]. In particular, the human breast and its derived tumors, while relatively trivial in composition, have proven enormously difficult to establish and interpret ex vivo. Early attempts relying on complex media formulations including addition of conditioned media or modifications of elaborate serum-free media originally designed to support growth of normal breast epithelium revealed that even low numbers of residual normal tissue in primary breast tumors have a growth advantage over cancer cells in culture [2-4]. Moreover, an exact identification of cancer cells in primary culture of a mixed breast tumor has proven to be quite challenging [3]. In a collaborative effort together with Dr. Mina J. Bissell, Lawrence Berkeley National Laboratory, CA, we originally took advantage of the differentiating role of the extracellular matrix (ECM) component laminin-1, which is the main constituent of commercially available Matrigel, to rapidly distinguish non-malignant cells from cancer cells in primary culture [5]. Whereas primary normal luminal cells in this assay formed polarized acinus-like structures and were growth arrested within twelve days, and basal or myoepithelial cells formed larger multilayered colonies still surrounded by a basement membrane, cancer cells like those from established cell lines stayed as irregular unpolarized colonies with some potential to grow [5,6]. We could not passage these cultures though, since cancer cells did not sustain the enzyme treatment necessary to dissolve the Matrigel. Therefore, the assay has served its purpose primarily as a very reliable and fast tool in dissecting the molecular background of the individual steps in breast tumor progression and reversion [7]. It has, nevertheless, also paved the way for a next generation of three-dimensional cell-based assays now commonly referred to as organoid assays [8]. The ambition still is to examine drugs for prediction purposes, and this often requires passaging of cells, since primary breast cancer biopsies commonly yield small cell numbers.

Hence, culture conditions preferentially supporting growth of breast cancer cells are highly sought-after. In 2018, a protocol published by the Dr. Hans Clevers lab, Hubrecht Institute, Utrecht, NL, described the successful cultivation of primary breast tumors as organoids in Matrigel using a special breast cancer organoid medium [9]. Hoping to be able to use their protocol to culture primary breast cancers ourselves, we meticulously reproduced the method. To validate if breast cancer-derived organoids cultured in this breast cancer organoid medium are indeed malignant or normal-like, we applied our previously described approach to distinguish between non-malignant and malignant breast cancer organoids [5]. Accordingly, we gauged for two key characteristics of cancer cells: 1) Loss of polarization and organized growth, and 2) capability to grow without deposition of a basement membrane. We succeeded in reproducing the heterogeneity of the primary breast cancer-derived organoids [9,10]. Thus, many organoids were organized in a bilayered manner typical for normal epithelium with correct polarization and expressed proteins required for attachment to a basement membrane, i. e. integrin β4 [10]. As expected, however, we also observed unorganized organoids that lacked polarization and did not express adherence proteins at the cell-ECM junction reminiscent of the tumor cell mass in the parental tumors. However, these -by our definition -cancer cells did not sustain passaging as we have previously experienced. Cells that survived passaging were submitted to genomic characterization and compared with breast tumor biopsies of origin. As in the original publication, we found significant differences between the two, confirming that passaged organoid cultures are submitted to drifting [9,10]. Hence, we concluded that these new improvements for cultivating primary breast cancer like previous protocols for that purpose are best applied to primary culture, as is the case with normal-derived cells [3,6]. However, there is reason to keep on trying not least in light of the latest detailed protocol published emphasizing on the use of specific homemade ingredients and more than one medium recipe [11].

These findings demonstrate how important a correct microenvironment is for breast cancer cells to grow, as well as for the normal epithelium to maintain its homeostasis. Identifying the interactions between the microenvironment and the normal epithelium as well as the changes that occur during tumorigenesis will help not only to understand how breast cancer arises, but also to define suitable culture conditions for human normal breast cells as well as for primary breast cancer. The impact of the microenvironment in development, morphology, homeostasis and disease has already been studied for more than a century (reviewed in [8]). In the literature, several examples can be found where the microenvironment plays a crucial part for the experimental design in breast research. In addition to the ECM, the cellular compartment of the breast stoma is known to influence the cultivation of the breast epithelium. For example, co-injection of human fibroblasts is required for growth of human breast epithelial cells upon inoculation in murine fat pads [12]. Furthermore, use of fibroblasts feeder cells has proved essential for preventing phenotypic drifting of pure cultures of human breast myoepithelial cells [13]. Subtleties of stromal composition may have hitherto underappreciated significance. Thus, we recently discovered that the human breast stroma contains two functionally distinct phenotypes of fibroblasts, which can be maintained and immortalized in culture; one surrounding the mammary ducts, the other surrounding the terminal ductal lobular units (TDLUs) [14]. This may be important for successful cultivation of normal human breast epithelial cells that differ in their anatomical location. While mature luminal epithelial cells are likely to reside in TDLUs, current evidence suggests that luminal progenitor cells reside in ducts [15,16]. Hence, signaling from the distinct types of fibroblasts might be involved in maintaining this local separation of mature and progenitor cells or vice versa. A similar supportive role of normal fibroblasts towards normal breast epithelial cells may apply also to cancer-associated fibroblasts (CAFs) in breast cancer [17]. Consequently, incorporation of hitherto underappreciated cues from CAFs may further development of assays for primary breast cancer. In 2018, two publications independently described different types of mammary CAFs [18,19]. Evidence suggests that these distinct CAF populations have different origins as well as diverse cancer-promoting as well as inhibitory functions [18,19]. More research is required to characterize the function of human breast CAFs in detail and to investigate their potential for the cultivation of primary breast cancer.

Taken together, the current status is that still cultivation of primary human breast cancer is possible only to a very limited extent that is far from ideal for drug testing or personalized medicine. Fortunately, an increasing interest in the microenvironment as contributor to all aspects of cell biology has evolved over the past 20 years (reviewed in [8]). We are hopeful that unravelling the interactions between the stroma and the epithelium under physiological as well as under pathological conditions will provide a major contribution to better culture models of the normal and malignant human breast. However, as was discussed already in relation to the original cancer culture assay using soft agar [20] translation into the clinic with the purpose of predictive drug testing probably requires an assay that truly recapitulates tumor heterogeneity, which is a challenge still to be met.

The authors declare no conflict of interest.

The Novo Nordisk Foundation Center for Stem Cell Biology is supported by Novo Nordisk Foundation grant number NNF17CC0027852. NG is covered by grant NNF18CC0033666, Copenhagen BioScience PhD Programme, a Novo Nordisk Foundation initiative. This work was also supported in part by the Novo Nordisk Fonden and Danish Research Council Grant 10-092798 (to DanStem), and the Kirsten and Freddy Johansens Fond (to OWP).

We thank Dr. Lone Rønnov-Jessen, Department of Biology, University of Copenhagen, DK, Dr. René Villadsen, and Dr. Jiyoung Kim, both Department of Cellular and Molecular Medicine, University of Copenhagen, DK, for critical reading of the manuscript.