Dual TGFβ/BMP Pathway Inhibition Enables Expansion and Characterization of Multiple Epithelial Cell Types of the Normal and Cancerous Breast

This article is featured in Highlights of This Issue, p. 1415

The breast tissue is composed of the epithelial and stromal cells cushioned by layers of adipose cells (1). The ability to culture and propagate various types of cells that form the breast hierarchy, which include the stem-progenitor-mature epithelial cells and stromal cells, is of vital importance for all functional studies related to in vitro characterization of different cell types, epithelial–stromal interaction, elucidation of molecular mechanisms of normal cell differentiation, cancer initiation, and progression. Several protocols have been developed that support the propagation of breast epithelial cells, which, in large part, are biased toward outgrowth of basal cells (summarized in ref. 2). Several of these culture protocols utilize irradiated mouse embryonic fibroblast or human foreskin fibroblasts as feeder cells to maintain the pluripotent state of stem cells (3, 4). Normal breast stromal cells have morphologic features and characteristics of fibroblasts. It is also difficult to distinguish mesenchymal and adipose stem cells from fibroblasts without profiling for cell surface markers. Therefore, an ideal system should allow growth of as many cell types of a tissue as possible and should utilize alternatives to feeder layers such as extracellular matrix proteins and small molecules that enable the maintenance of adult stem cells and their lineage commitment properties. In this respect, vitamin C and inhibitors of Rho-associated coil–coil containing protein kinase (ROCK) have been shown to be effective in stem cell reprogramming and in preventing matrix detachment–induced apoptosis of stem cells, respectively (3, 5).

A recent study reported that plating adult epithelial cells from lungs on tissue culture dishes precoated with laminin-5-rich–conditioned media from the rat bladder cancer cell line 804G and in media containing dual inhibitors of SMAD/BMP pathways permits propagation of diverse epithelial cells (6). We had previously shown that maintaining primary breast epithelial cells in low glucose–containing media and limiting the use of DMSO as a solvent for small molecules enable long-term culture of breast epithelial cells from a core-needle biopsy on a feeder layer and these cultured cells maintain stem-progenitor–differentiated cell hierarchy (7). The goals of this study were to determine whether combining these two methods would permit growth of multiple cell types of the breast without the need for feeder layer and whether the technique can be extended to other cancer types as well as biopsies from metastatic sites. Because there is growing evidence for heterogeneity between primary tumor and metastasis within an individual patient, despite sharing large number of driver mutations (8), there is an urgent need to develop methods that enable characterization of metastatic cells in vitro for therapeutic decision-making. We show that the method described here is effective in propagating primary epithelial cells that include but is not limited to (i) normal breast, (ii) carriers of BRCA1, BRCA2, PALB2 mutations, or breast cancer–protective alleles, (iii) primary breast cancer, (iv) pleural effusions of patients with breast cancer, (v) ascites fluids from patients with ovarian cancer, and (vi) liver metastasis of breast cancer and pancreatic cancer of neuroendocrine origin. The majority of cells cultured with this protocol maintained the epithelial phenotype and could be cultured for approximately 10–12 passages and provided >5 ×10⁶ cells for cryopreservation, functional studies, and/or to immortalize/transform in case of primary nontransformed cells. Nonepithelial cells could be easily separated from epithelial cells by flow cytometry and cultured for further evaluation of the microenvironment. In addition, tumor cells formed tumors in NSG mice thus allowing simultaneous characterization of patient-derived primary and metastatic tumor cells both in vitro and in vivo.

Fresh or cryopreserved, deidentified tissue samples were obtained from the Indiana University (IU) Simon Cancer Center Tissue Bank or Komen Tissue Bank (KTB) at IU School of Medicine, Indianapolis, after informed consent from the donors. All experiments were carried out in accordance with the approved guidelines of the Indiana University Institutional Review Board. The 804G cells were a generous gift from Dr. Rajagopal (Harvard Medical School, Boston, MA). The 804G cells were grown in RPMI media (10-0400CV, Corning) with 10% FBS and 1% penicillin and streptomycin. Conditioned medium (CM) from the 804G cells was harvested once the plates were confluent. No more than 2–3 batches of CM were collected from one culture. The CM was filtered through 22-μm filters, divided into aliquots in 50 mL tubes and frozen at −20°C for future use. A day before the tissues were processed, 60-mm culture dishes were prepared by adding 5 mL of the thawed 804G CM followed by overnight incubation at 37°C. Just prior to use, the CM was aspirated and the dish was washed with PBS once. This CM is enriched for laminin-related components (9).

Supplementary Table S1 provides details of tissues used in this study. Solid tissues dissociation protocol along with primary cell culture media has been described previously (7). The primary cell media was supplemented with 1 μmol/L of BMP inhibitor DMH-1 (#4126, Tocris), 1 μmol/L of TGFβ inhibitor A-83-01 (#2939, Tocris), 5 μmol/L of ROCK inhibitor Y-27632, and 4 μL/mL of adenine (6 mg/mL stock, A9001, Sigma-Aldrich). Pleural effusion or ascites fluids were centrifuged at 1,000 rpm for 5 minutes to collect cell pellet and the pellet was washed twice in PBS before plating on 804G CM–coated plates. Medium was changed the next day to remove cellular debris and floating cells. The cells could be cultured for 10–12 passages using this method.

Adherent cells were collected by trypsinization, stained using antibodies CD49f-APC (FAB13501A), CD140b-FITC (FAB1263F; R&D Systems), PROCR (CD201)-PE (130-105-256), EpCAM−PE (130-091-253), EpCAM−APC (130-091-254; Miltenyi Biotec Inc.), CD271-APC (345108; BioLegend), CD44-APC (559942), CD24-PE (555428), CD73-PE (561014), CD90-APC (559869), CD166-PE (559263), JAM-1-PE (552556), MUC-1-FITC (559774; BD Pharmingen), CD10-PE (340920; BD Biosciences), and CD117-FITC (11-1178-42; eBioscience), and were acquired using a BD LSR II flow cytometer. Data were analyzed using CellQuest or FlowJo software. Forward and side scatter were used to ensure that only live cells were considered in the analysis. Gating was done using appropriate FITC (555573), PE (555749), and APC (555576; BD Pharmingen) isotype control antibodies and only a representative isotype control for two fluorescent markers are shown.

Mammosphere assay in 6-well plate at a density of 5,000 cells/mL has been described previously (10). Phase contrast images were taken at day 5. Mammospheres were collected, washed, and trypsinized to obtain single-cell suspensions. Cells were stained with indicated antibodies and subjected to flow cytometry.

We used the KTB200, KTB201, and KTB202 primary cells to characterize the growth kinetics and in vitro survival of primary cells in culture. Growth kinetics was evaluated from successive cell counts. A total of 2.5 × 10⁵ cells per cell type were seeded on 60-mm dishes at day zero and incubated under standard conditions. Cells were harvested every 2 days by trypsinization, counted using a hemocytometer, and 2.5 × 10⁵ cells were reseeded for the next culture.

Floating collagen scaffold was prepared on the basis of a protocol from Linnermann and colleagues, with some modifications as detailed in Kumar and colleagues (2, 11). Matrigel basement membrane (354234, Corning) was slowly thawed on ice. Eight-chambered coverglass system (155409, Lab-Tek II) was coated with 40 μL of Matrigel per well and spread evenly to cover the bottom surface. The chamber was incubated for 30 minutes at 37 °C for the gel to solidify. A total of 400 μL of overlay media comprising primary cell media as described above with 2% Matrigel and 6,000 cells per well were added. The medium was changed every 3–4 days and the cells were cultured to 10 to 12 days. Immunofluorescence staining of 3D structures has been described previously (2).

Total RNA was isolated using RNeasy Kit from Qiagen and 1 μg of RNA was used to synthesize cDNA with Bio-RAD iScript cDNA Synthesis Kit. qRT-PCR was performed using Taqman universal PCR mix and predesigned Taqman assay primers with best coverage from Applied Biosystems. The following assays were used in our study: ACTB (Hs01060665_g1), ESR1 (Hs01046816_m1), GATA3 (Hs00231122_m1) FOXA1 (Hs04187555_m1), and FOXC1 (Hs00559473_s1). Applied Biosystems StepOneplus real time PCR system was used for PCR and fluorescence detection and the StepOne software was used for analysis.

Cells were cultured in glass-bottomed Microwell Dish (P35G-0-14-C, MatTek Corporation) and stained with KRT14 (ab7800, Abcam), KRT19 (ab52625, Abcam), and KRT17 (ab53707, Abcam) primary antibody according to protocol described previously (2). Cells were stored in PBS in dark and images were taken within 48 hours of staining with Olympus FV1000 MPE inverted confocal microscope.

The Indiana University Animal Care and Use Committee approved the use of animals in this study and all procedures were performed as per NIH guidelines. Primary cells created from pleural effusions of breast cancer or triple-negative breast cancers (TNBC; 5 × 10⁵ cells) with 50% Basement Membrane Matrix (BME) Type 3 (100 μL volume) were implanted into the mammary fat pad of 5- to 6-week-old female NSG (NOD/SCID/IL2Rgnull) mice. A 60-day slow-release estradiol (0.72 mg) pellet was implanted at the time of mammary fat pad injection, irrespective of the injected breast cancer subtype. Tumor growth was measured weekly and tumor volume was calculated as described previously (12). After 2–5 months, lungs and primary tumors were collected and processed for hematoxylin and eosin (H&E), ERα, PR, GATA3, and FOXA1 staining.

H&E, ERα, PR, GATA3, and FOXA1 immunostaining was performed at the CLIA-certified Indiana University Health Pathology Laboratory and the whole-slide digital imaging system of Aperio (ScanScope CS) was used for imaging. The following antibodies were used: ER (clone EP1, Dako IR 084), PR (#RB-9017-P, NeoMarkers), FOXA1 (sc-6553, Santa Cruz Biotechnology), and GATA3 (sc-268, Santa Cruz Biotechnology).

Statistical analyses were performed using GraphPad Prism software program (version 6.0). P values below 0.05 were considered statistically significant.

We have developed a method for propagating the primary cells from normal, cancerous, and metastatic tissue samples with appropriate modifications to growth media. Cells propagated from more than 30 primary breast tissues (>6 healthy donors, two BRCA1 mutants, three BRCA2 mutants with one right and left breasts of the same donor, three PALB2 high-risk, two protective alleles, tumors with adjacent normal tissues from the same patients, several TNBCs, three breast cancer pleural effusion fluids, four breast cancer metastasis from different organs, one liver metastasis of pancreatic cancer–neuroendocrine subtype, and three ascites fluids of patients with ovarian cancer) were analyzed (Supplementary Table S1). These extensive analyses recognized enormous phenotypic heterogeneity in the normal breast and tumors.

We subjected primary breast cells to phenotypic analyses using established and recently discovered markers that define the normal and cancer stem cells (CSC), pluripotent somatic stem cells, progenitor, differentiated, luminal, basal/myoepithelial cells, and estrogen receptor alpha (ERα)-positive subpopulation of cells (2, 13, 14). Similar to results obtained with cells propagated using the previous epithelial reprogramming assay (7), this new culturing method enabled us to document enormous phenotypic heterogeneity in the normal breast. The expanded cell populations are described as CD49f⁺/EpCAM⁻ stem/basal–enriched cells, CD49f⁺/EpCAM⁺ luminal progenitor cells, CD49f⁻/EpCAM⁺ mature luminal cells, CD44⁺/CD24⁻ stem/basal cells, CD44⁺/CD24⁺ stem/luminal progenitor cells, CD44⁻/CD24⁺ differentiated cells, CD201⁺/EpCAM⁻ multipotent stem cells, CD271⁺ rare basal cells in luminal breast cancer (relevant in antiestrogen resistance models), CD73⁺/EpCAM⁺/CD90⁻ rare endogenous pluripotent somatic stem cells, CD73⁺/CD90⁺/EpCAM⁻ potential mesenchymal stem cells, CD10⁺/EpCAM⁻ basal/myoepithelial cells, ALCAM (CD166)⁺ cells that are enriched for ERα, and JAM-A⁺ cells, which are enriched for cancer stem cell phenotype (15–17). Flow cytometry characterization of breast epithelial cells from six healthy women (KTB200-205) showed remarkable variability in stem/basal, progenitor, and mature/differentiated cell subpopulations (Fig. 1A; Supplementary Fig. S1). All samples were from premenopausal women (20–55 age range, 23–41 BMI range, Supplementary Table S1). Using a combination of CD44 and EpCAM, we were able to document the presence of luminal (CD44⁻/EpCAM⁺ or CD44^low/EpCAM⁺) and basal (CD44⁺/EpCAM^low) cells (18). Proportion of cells that express the cell adhesion molecule MUC1 also varied between samples (Supplementary Fig. S1). MUC1 is expressed predominantly in CD24⁺ luminal cells and these cells often display higher self-renewal capacity and are metastatic (19, 20). With respect to rare pluripotent somatic stem cells (13), our technique expanded this population, as a significant portion of CD73⁺/CD90⁻ cells was EpCAM⁺ (Fig. 1B). Cells, propagated using this assay, are free of fibroblasts as they lacked the expression of CD140b, which identifies fibroblasts (Supplementary Fig. S1; ref. 21). We also did not find cells expressing CD117, a putative marker of luminal progenitor cells (Supplementary Fig. S1; ref. 22). These results suggest interindividual heterogeneity among the normal breasts, particularly in number of breast epithelial cells–expressing markers that identify cells as stem, basal, or luminal cells, similar to our previous report (7). However, as we indicated in our previous report, because of limited sample size, the observed heterogeneity cannot be attributed to differences in age, BMI, or time of sample collection (luteal or follicular phase).

We performed serial dilution cell proliferation assay to determine the number of passages these cells can be maintained. Proliferation kinetics varied from samples to samples and declined by around passage 7 (Supplementary Fig. S1). However, by the time cells reached passage 7, couple of million cells could be easily obtained for immortalization and functional studies.

To evaluate the suitability of the assay for growing cells from high-risk women, we created primary cells from prophylactic mastectomy or biopsies of women carrying germline mutations in BRCA1 (two cases), BRCA2 (three cases, one with tissues from both breasts), or PALB2 (three cases). Staining with multiple antibody combinations did not show any patterns that are distinct from the cells cultured from healthy breasts (Fig. 2A; Supplementary Fig. S2). It is also remarkable that interindividual differences in staining patterns were noted among high-risk patients analyzed in the study (Supplementary Fig. S2). For example, CD271/EpCAM, CD10/EpCAM, and MUC1/EpCAM staining patterns of two BRCA2 mutants and three PALB2 mutants were different (Supplementary Fig. S2). Thus, it is unlikely that the breasts of high-risk women contain phenotypically distinct population of cells compared with breasts of healthy women. In general, growing epithelial cells from women with PALB2 and CHEK2 mutations was more difficult than others because of outgrowth of cells with fibroblastic appearance or epithelial cells undergoing rapid epithelial-to-mesenchymal transition (EMT).

Cells with BRCA1 mutations, similar to cells from healthy women, formed acini on collagen gel and these acinis were composed of KRT5/6 and KRT17⁺ cells (Supplementary Fig. S3). We compared the staining pattern of cells from both the breasts of a BRCA2 carrier to ensure that observed staining pattern is not due to collection of biopsy from specific regions of the breast. Randomly collected breast tissues from both the breasts showed similar cell surface marker profiles (Fig. 2B; Supplementary Fig. S3). Immunofluorescence with basal cell marker KRT14 and luminal cell marker KRT19 showed that the majority of cells from high-risk women are hybrid cells that expressed both KRT14 and KRT19 (Fig. 2C), which is a suggested characteristic of cells with luminal progenitor phenotype (23).

We had previously demonstrated that immortalized cell lines derived from healthy breast self-renew as well as undergo differentiation when cultured under mammosphere growth conditions (2). To determine whether cells from high-risk women differ in their capacity to undergo self-renewal and differentiation, we subjected cells from BRCA1 and BRCA2 carriers to mammosphere assay followed by flow cytometry (Fig. 2D and E). Phase contrast images of these mammospheres showed similarity between samples. However, as noted previously with cells from healthy breasts (2), interindividual variability in differentiation capacity was noted with cells from high-risk women. For example, the proportion of cells with differentiated luminal phenotype (CD49f⁻/EpCAM⁺ or CD44⁻/EpCAM⁺ cells) increased under mammosphere condition compared with 2D growth condition but the degree of increase varied between samples (Fig. 2A and E). Staining patterns of mammospheres with additional markers are shown in Supplementary Fig. S3.

Although epidemiologic and genome-wide association studies have identified various risk, predisposition, and protective alleles including new 65 risk loci for breast cancer (24, 25), functional analyses of these risk loci have been difficult because of limited cell resources. To determine whether culture system developed here is useful in this regard, we focused on the recently described breast cancer–protective variant (rs140068132) in Latinas on chromosome 6q25 located 5′ of Estrogen Receptor 1 (ESR1) gene, which is limited to Indigenous American Ancestry (26). We examined DNA from self-reported 43 Hispanic women from whom cryopreserved normal breast tissues were available at the Komen Tissue Bank (KTB). Among these women, three carried the protective allele (Fig. 3A) but viable cells could be generated from two samples. These cells were less proliferative than cells generated from healthy or high-risk women and only limited number of experiments could be performed. Cell surface marker profiles showed clear enrichment of cells with differentiated/luminal characteristics (Fig. 3B; Supplementary Fig. S4A and S4B). For example, unlike in cases of healthy or high-risk women, the majority of cells from women with protective alleles were CD49f⁻/EpCAM⁺, CD271⁻/EpCAM⁺, and CD201⁻/EpCAM⁺. In addition, these cells had higher proportion of cells that were KRT14⁻/KRT19⁺ compared with cells from noncarriers of protective alleles or high-risk individuals (Fig. 3C), and were KRT17⁺ (Supplementary Fig. S4B). Furthermore, cells from one protective allele carrier, which we were able to grow for additional analyses, expressed higher levels of luminal cell markers ESR1, FOXA1, and GATA3 compared with cells from normal (KTB51) and high-risk individuals (Fig. 3D). In mammosphere assays, cells with protective alleles generated very small spheres compared with cells generated from two Hispanic women not carrying the protective allele (Fig. 3E). Thus, the assay established here can be used to characterize cells from women carrying distinct risk loci.

Because gene expression in normal tissues shows interindividual variations due to SNP in the gene regulatory regions (27), we had recently proposed that normal cells from the same patient need to be used to determine cancer-specific gene expression changes (28). Although the use of cells from normal tissues adjacent to tumor is far from perfect because of cancer-induced field defects on adjoining cells (29, 30), it is still better to use those cells or from the unaffected contralateral breast than normal cells from unrelated individuals as controls to identify cancer-specific gene expression changes. Toward achieving this goal, we established culture conditions to grow tumor and tumor adjacent–normal from the same patient. Representative flow cytometry profiles with various markers are shown in Fig. 4A; Supplementary Fig. S4C and S4D. Both tumor and the adjacent normal showed similar marker profiles except that tumors had a minor CD49f^high/EpCAM⁺ population. Cells derived from not all tumors that were characterized showed similar profiles as samples from TNBCs showed a distinct profile than the sample described above (Fig. 4B; Supplementary Fig. S4D). Tumor cells were KRT14⁺/KRT19⁺ suggesting their luminal progenitor characteristics, similar to flow cytometry stain pattern (Fig. 4C). We have created a bank of tumor cells from at least 15 patients with few cases of paired tumor and tumor adjacent–normal, which could be utilized for functional studies.

Because organoid culture system is gaining traction for various types of studies including screening for anticancer therapies (31), we determined whether tumor cells form organoid-like structures in 3D. On Matrigel, tumor cells formed spheres, whereas on collagen gels, cells formed elongated structures that resembled breast ducts (Fig. 5A). Immunofluorescence of collagen gel structures with KRT14, KRT17, and KRT19 showed these ducts are KRT17⁺ and are lined by KRT14⁺/KRT19⁺ double-positive cells (Fig. 5B). Most breast cancers are believed to originate from luminal progenitor cells, which are typically KRT14⁺/KRT19⁺ (23).

Although it has proven extremely difficult to cultivate breast epithelial cells that express ERα, there have been few recent advances in culture methods that enabled propagation of ERα-positive cells (2, 14). Similarly, patient-derived tumor xenograft (PDX) models heavily favor ERα-negative breast cancers relative to ERα⁺ PDXs (32). In addition, very rarely, tumor cells and PDXs have been created from the same tumors. Because we had >90% success rate in generating cells from tumors, we next investigated whether cells derived from breast tumors or pleural effusions can create tumors in NSG mice. Detailed characteristics of tumor cells derived from pleural effusions developed from two ERα⁺ tumors are described in Fig. 6A; Supplementary Fig. S5. Flow cytometry patterns of cells grown from two pleural effusions were different with pleural effusion #2 containing higher proportion of EpCAM⁺ cells. Cells from both pleural effusions contained significant levels of ALDEFLUOR⁺ cancer stem cells (33). Western blotting analysis showed expression of ERα in effusion #2, although very low levels of ERα were detectable in the effusion #1 (Fig. 6B). Cells derived from effusion #2 generated tumors in the mammary fat pad of NSG mice by two months postimplantation (Fig. 6C), which expressed all three markers of luminal breast cancer—ERα, FOXA1, and GATA3 but lacked progesterone receptor (PR), indicating luminal B characteristics (Fig. 6D). A lung metastatic nodule, which expressed GATA3, was also detected (Fig. 6E). We have generated tumors from primary TNBC cells, which could be reimplanted into additional animals similar to PDX models (Fig. 6F). Whole-genome sequencing of these TNBC cells showed mutation patterns observed in breast tumor samples (data not shown). Thus, the method described here provides resources of patient tumor–derived primary cells and tumors from these cells for functional studies.

Metastasis is the main cause of cancer-related death and several studies have shown distinct differences between primary tumors and metastasis (34, 35). We had previously shown that epithelial reprogramming assay could be used to grow and perform genomics of liver metastasis of various cancers (36). However, because mouse embryonic feeder layer cells were needed to grow cancer cells, sorting of cancer cells by flow cytometry was required before sequencing, and bioinformatics analyses of genomic data required additional steps to ensure that sequence reads are not from residual contaminating mouse DNA. To determine whether metastatic cells grown under reprogramming growth conditions can adapt to the current method, we transferred cells from a liver metastasis of a breast cancer grown initially under reprogramming assay to the new method. Indeed, cells readily adapted to the new method and a significant number of cells could be obtained for phenotypic characterization (Fig. 7A; Supplementary Fig. S5). The majority of cells were CD49f⁺/EpCAM⁺, Jam-A⁺/EpCAM⁺, and CD271⁺/EpCAM⁺.

We extended this method to propagate cells from other cancer types. Cells from ascites fluid of patients with ovarian cancer (OV351-MT to OV353-MT) and liver metastasis of pancreatic cancer–neuroendocrine subtype were generated and characterized with various cell surface markers. Cells from ascites fluid of patients with ovarian cancer displayed the phenotypic heterogeneity and exhibited predominant CD49f⁺/EpCAM⁻, CD44⁺/CD24-, CD44⁺/EpCAM⁻, CD73⁺/CD90⁺, CD201⁺/EpCAM⁻, CD166⁺/EpCAM, and CD10⁺/EpCAM⁻ phenotypes (Fig. 7B; Supplementary Fig. S6). However, it is likely that few of the cell types that do not express EpCAM are noncancer stromal cells. Nonetheless, the method permitted propagation of EpCAM⁺ cells, which can be sorted by flow cytometry for functional studies.

The cells from liver metastasis of pancreatic cancer–neuroendocrine subtype contained predominantly CD49f⁺/EpCAM⁺, CD44⁺/CD24^low, CD44⁺/EpCAM⁺, CD73⁺/CD90⁻, CD201^low/EpCAM⁺, CD271⁻/EpCAM⁺, CD10⁻/EpCAM⁺, CD166⁺/EpCAM⁺, CD117⁻/EpCAM⁺, MUC-1⁻/EpCAM⁺, JAM-1⁺/EpCAM⁺, and CD140b⁻/EpCAM⁺ subpopulations (Fig. 7C; Supplementary Fig. S6). A fraction of these cells were CXCR4-positive, which is expressed only in metastatic cancer stem cells of pancreatic cancer and is required for metastasis (37). Taken together, detailed characterization of various cells from different sources documents that our propagating method is suitable for growing primary cells, establishing heterogeneity of cancer cells in tumors, and provides a replenishable source of live primary cells for functional studies.

Functional modeling of breast epithelial hierarchy and stromal–epithelial cell interactions have been difficult due to inability to obtain sufficient stem-progenitor-mature epithelial cells and stromal cells. Establishment of such techniques for modeling of the normal breast in vitro is critical if progress made in identifying various risk and protective alleles of cancers to have an impact in clinic. Similarly, further advancement is needed to translate knowledge gained through single-cell sequencing technologies, which are currently descriptive cataloguing of different cell types in primary and metastatic tumors without allowing functional evaluation of different cell types (35). In this study, we report in vitro propagation method for primary cells from different sources including breast, ovary, and metastasis. These cells are suitable for “omics” studies as TNBC cells from one of the patients have already been sequenced and found to have mutations enriched in breast cancers and additional samples are currently being sequenced (data not shown). Most importantly, sequence information can be obtained when primary tissue is very limited. We expect these tumor-derived cells to be also compatible for proteomics and metabolomics studies with and without prior exposure to chemotherapeutics. Other advantages of the method are that it is relatively inexpensive, adaptable in any labs, and cancer cells can be grown from cryopreserved tissues. We have found that cryopreserving tissues with ROCK inhibitor yields higher number of live cells (7). Although continuous passage leads to drift in proportion of different cell types, the original cell types remain after passage. Samples with disproportionately higher number of fibroblast-like cells compared with epithelial-like cells within a week of initiating culture, which can be identified through phase contrast microscopy, show a higher tendency of drift to nonepithelial (EpCAM⁻) cells with continuous passage. Differential trypsinization (two minutes at room temperature) or flow cytometry sorting can be used to enrich epithelial cells in culture. The only caveat of this modification is that cancer cells that have undergone EMT would get excluded from functional studies.

Single-cell studies are gaining popularity for better understanding of heterogeneity of cell types in an adult organ. Fifteen and three epithelial cell types have been described in the mouse mammary gland and the human breast, respectively (38, 39). Because the assay described here allows propagation of cells from a core biopsy and cells form distinct structures on a 3D matrix, functional studies of cells from different clusters can now be performed with and without addition of stromal cells. Additional advantage of this system is the availability of cells that carry breast cancer risk alleles. Depending on the type of risk alleles, these cells can be easily immortalized with human telomerase. Although we could easily immortalize cells with BRCA1 or BRCA2 mutations, generating cell lines from PALB2 and CHEK2 mutations carriers as well as those with protective alleles has so far proved difficult. Phenotypic analyses using approximately 10 cell surface marker combinations did not reveal unique features of cells obtained from high-risk individuals, which is bit surprising considering a previous study, which demonstrated enrichment of cells with luminal progenitor properties in the breast of BRCA1-mutant carriers (17). However, we were able to document enhanced differentiation properties of cells obtained from women who carry breast cancer–protective alleles, which clearly indicates the role of proper differentiation in protecting against breast cancer.

We had previously demonstrated that short-term culturing followed by sequencing of cultured tumor cells could detect functionally important mutations that are otherwise undetectable upon direct sequencing of tumors (36). Culturing for short duration as such did not introduce mutations. Culturing method used at that time required propagation of tumor cells on a mouse fibroblast feeder layer and purifying tumor cells by flow cytometry before sequencing. The new method described here eliminates those requirements. Despite TNBCs displaying higher rate of mutations than other breast cancer subtypes (40) and data from many samples already available, very few are detected across multiple TNBCs and very rarely the impact of these mutations using the same tumor material could be studied. In general, cancers believed to contain 2–8 “driver” mutations and 30–60 protein-coding changes in passenger genes that alter cellular functions (41). We hope that the assay described here would not only enhance detection of additional mutations but also evaluation of presumed driver and passenger mutations in future. Furthermore, new mutations detected from these cell lines can be incorporated for screening of distant metastasis using ctDNA technology (42). In this respect, tumor cells generated from our culturing method form unique structures under 3D growth conditions and develop tumors in NSG mice, similar to a direct PDX model. Availability of millions of cancer cells from patients would permit further refinement of these models including characterizing growth properties of specific subclones with or without coculture with stromal cells, in vitro high-throughput drug screening on an individual patient basis and further validation of the effective drugs in xenograft models. Sufficient cells can be obtained within a month of initiating culture for rapid in vitro screening for drug sensitivity. In this respect, a recent study that utilized single-cell genomics and TNBC samples from patients receiving neoadjuvant therapy concluded that drug-resistant genotypes are preexisting in tumors and adaptively selected during neoadjuvant therapy (43). Relatively inexpensive way to generate cells from TNBCs described here would enable identification and characterization of such preexisting drug-resistant clones. While breast cancer research community has placed significant emphasis on finding cure for TNBCs, similar degree of attention is also needed for ERα-positive breast cancer because of higher number of patients with this disease type and its late recurrence (44). Recent studies have identified set of genes that predict early and late recurrence (45). However, developing in vitro models with patient-derived cells to functionally characterize genes in the signature has proven to be a challenge. Our assay has overcome this challenge as the cultured cells derived from recurrent tumors maintain ERα and one of our future goals is to establish cultures from metastasis that carry mutations in ERα, which are frequently observed in metastasis but not in primary tumors (46). We do acknowledge that, similar to PDX model (47), clonal selection of different tumor cells during culturing is one of the limitations of this assay. Nonetheless, simplicity of the assay allows propagation of cells from multiple cancers and serves as a resource for validation of pan-cancer genomic signatures including DNA methylation and chromatin accessibility identified through bioinformatics analyses (48–50).

No potential conflicts of interest were disclosed.

Conception and design: K.D. Miller, H. Nakshatri

Development of methodology: M. Prasad, B. Kumar, M. Anjanappa, G. Sandusky, H. Nakshatri

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Prasad, B. Kumar, M. Anjanappa, G. Sandusky, A.M. Storniolo, H. Nakshatri

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Prasad, B. Kumar, M. Anjanappa, G. Sandusky, H. Nakshatri

Writing, review, and/or revision of the manuscript: M. Prasad, B. Kumar, K.D. Miller, H. Nakshatri

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P. Bhat-Nakshatri, K.D. Miller, H. Nakshatri

Study supervision: H. Nakshatri

We thank Dr. Rajagopal from Harvard Medical School for 804G cell line. We thank the members of IU Simon Cancer Center tissue procurement facility, flow cytometry core, clinical trial office, IU School of Medicine IHC core, Komen Tissue Bank for their service, and a countless number of women who donated their breast tissue for research purposes as well as volunteers who facilitated tissue collection. Department of Defense DOD-W81XWH-15-1-0707, Susan G. Komen for the Cure SAC110025, 100 Voices of Hope, and R03CA195250-01A1 (to H. Nakshatri) funded this study. Susan G. Komen for the Cure, Breast Cancer Research Foundation, Vera Bradley Foundation for Breast Cancer Research, and the Catherine Peachey Fund provided funding support to the Komen Tissue Bank.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.