Introduction
BRCA1 was the first identified breast cancer susceptibility gene and was localized to 17q21 by positional cloning more than 15 years ago [
1].
BRCA1 is mutated in about 2.5% to 5% of all breast cancers, in 45% of inherited breast cancer families, and in up to 80% of breast/ovarian cancer families.
BRCA1 mutation is associated with a high incidence of bilateral disease, and confers an 82% risk for developing breast cancer and an 54% risk for developing ovarian cancer by age 80 years [
2]. Somatic mutations of
BRCA1 have been reported in up to 10% of cases of sporadic ovarian cancer, but they are extremely rare in sporadic breast cancer [
3‐
5]. However, reduced BRCA1 protein expression is detected in high-grade sporadic breast and ovarian tumors, suggesting that epigenetic downregulation of
BRCA1 contributes to their aggressive clinical course [
6‐
8]. The existence of cancer stem cells associated with
BRCA1 mutations or downregulation has not been reported.
In spite of early detection and aggressive surgical and chemotherapeutic approaches, no significant 5-year survival benefits have been achieved in patients harboring
BRCA1 mutations [
6]. During the past several years, cancer stem cells have been subjected to increasing scrutiny as a potential cause of relapse and drug resistance [
9]. Several groups [
10,
11] identified a small subpopulation of highly tumorigenic cells from human breast tumors bearing the CD44
+CD24
-/low lineage phenotype, which have drug-resistant phenotype and the capacity to form tumors after transplantation in nonobese diabetic/severe combined immunodeficient mice. Subsequent enrichment in Sca-1 positive cancer stem cells was shown for mouse mammary tumor models, such as mouse mammary tumor virus (MMTV)-Her2/neu and MMTV-Wnt1 [
12], and Thy1/CD24 expression further defined cancer stem cells in the Wnt1 model [
13]. No studies have yet been conducted to characterize
Brca1-deficient cancer stem cells.
Multiple mouse models with targeted deletion of
Brca1 in the mammary gland generate tumors with low penetrance [
14]. Increased incidence of these tumors is observed in mice harboring two
Brca1Δexon11 genes in a
p53+/- background, with uniform deletion of p53 in these tumors. Lymphomas were also reported in this model [
15]. However, the
Brca1 deficient mouse mammary tumors have variable penetrance and latency, which makes it nearly impossible to use these models to standardize therapies and to study the stem cell population. To overcome these difficulties, we developed and characterized 16 cell lines from five independent
Brca1Δexon11/
p53+/- tumors. We examined these cell lines for specific cell populations using multiple known stem cell markers. Cell populations expressing putative stem cell markers were more resistant to chemotherapeutic agents than were parental cells, and had other characteristics of cancer stem cells, including reconstitution of tumors by as few as 50 to 100 cells.
Materials and methods
Generation of cell lines from Brca1 mouse mammary tumors
Brca1 tumor cell suspensions were prepared as described by Varticovski and coworkers [
16] from
Brca1Δ11p53+/- mammary tumors. Briefly, mice were euthanized with CO
2, and tumors were collected aseptically and mechanically dissociated. Cells were passaged through a 40 μm mesh screen, and were further dissociated by serial passage through a syringe with 18 to 25 gauge needles. Cells were plated at low density for selection of individual clones. Cells were grown at 37°C in 5% carbon dioxide in RPMI 1640 media supplemented with penicillin/streptomycin, glutamine, and fetal bovine serum starting with 2% and progressively increasing to 10%. More than 40 clones were isolated using cloning cylinders and a total of 16 cell lines were developed from five independent primary tumors. Each cell line was passaged weekly and maintained in culture for up to 50 passages.
Characterization of unsorted cells that survive in long-term cultures in the absence of attachment
The ability of cells to grow in the absence of attachment was tested by plating cells on low-binding plates (Fisher Scientific, Hampton, NH, USA). Under these conditions, some cell lines showed surviving floating colonies that formed compact spheroids after 3 to 4 weeks of culture. Spheroid formation frequency for each cell line was tested by plating cells in limiting dilution from 500 to 1 cell/well on 96-well low-binding plates in sextuplicate. The number of colonies was scored weekly, and the spheroids were dispersed into single cells and upstaged to six-well plates after 3 weeks. To obtain single cell suspension, spheroids were collected and washed in phosphate-buffered saline (PBS). Single cell suspension was obtained by incubation in collagenase/dispase and DNAse (Roche, Switzerland), filtering through a 40 μm filter to remove remaining aggregates (Fisher Scientific), and plated directly onto 96-well tissue culture plates for drug testing or passaged onto a new low-binding plate for expansion.
Analysis of cell surface markers
Cell lines grown as monolayer were trypsinized, and cells grown as spheroids were dispersed to obtained single cell populations, as above. Cells were washed in PBS with 1% bovine serum albumin and stained with PE anti-mouse CD24 (BD Pharmingen, San Jose, CA, USA), or APC anti-mouse CD24 (Biolegend, San Diego, CA, USA), FITC anti-mouse CD44 (Southern Biotech, Birmingham, AL, USA), or PE anti-mouse prominin-I CD133 (eBioscience, San Diego, CA, USA). Rat IgG (CHEMICON, Billerica, MA, USA) was used as the isotype control, in accordance with the manufacturer's instructions. Cells were analyzed by flow cytometry using LSR II (BD Biosciences, San Jose, CA, USA). Data were collected using FACSDiVa software (BD Biosciences) from no fewer than 30,000 cells.
Cytotoxicity assay and calculation of combination index
Doxorubicin was obtained from Sigma (St. Louis, MO, USA) and dissolved in PBS as 10 mmol/l stock. Aliquots were frozen and diluted in media immediately before use. Cisplatin and etoposide were obtained from Sigma-Aldrich (St. Louis, MO, USA). 17-DMAG (17-dimethylaminoethylamino-17-demethoxygeldanamycin hydrochloride) was obtained from Invivogen (San Diego, CA, USA). These drugs were dissolved in dimethyl sulfoxide, stored as 10 mmol/l stock aliquots, and diluted in media immediately before use.
Cells were seeded in six replicates in 96-well plates at 10,000 cells per well. Serial dilutions of doxorubicin, cisplatin, etoposide, and 17-DMAG in medium were added to the cells on the following day or as indicated. Dose-response curves to single drugs at 24 hours, 48 hours, and 72 hours after exposure were generated to determine the range of concentrations to be used in combination. For drug interaction studies, drugs were added sequentially (with the 17-DMAG introduced with a 24-hour delay) or simultaneously in a final volume of 100 μl. Cytotoxicity was measured using the Cell Titer 96 Aqueous One Solution Cell Proliferation Assay (Promega, Madison, WI, USA), which is a colorimetric method for determining the number of viable cells based on bioreduction of the tetrazolium compound MTS (3-[4,5-dimethylthiazol-2-yl]-5-[3-carboxymethoxyphenyl]-2-[4-sulfophenyl]-2
H-tetrazolium, inner salt) by metabolically active cells. After exposure to a single drug or after a total of 48 hours in sequential addition experiments, 20 μl MTS reagent was added to each well and the plates were incubated in a humidified 37°C incubator with 5% carbon dixoide for 1 to 4 hours. Absorbance at 490 nm was recorded using a 96-well plate reader. Data were averaged and normalized against the average survivals of untreated samples and analyzed using CalcuSyn (Biosoft, Ferguson, MO, USA), software based on the multiple drug-effect equation of Chou and Talalay [
17].
Tumor growth in vivo
All studies were conducted in an AAALAC (Association for Assessment and Accreditation of Laboratory Animal Care International) accredited facility, in compliance with the US Public Health Service guidelines for the care and use of animals in research. Parental cells and cells sorted for indicated cell surface markers were resuspended in 100 μl RPMI media and injected into the inguinal mouse fat pad (pad #4) of 6- to 8-week-old female nonobese diabetic/severe combined immunodeficient mice. The growth of tumors was monitored using caliper measurements to determine tumor mass. Weights (milligrams) were calculated from measurements (millimeters) of two perpendicular dimensions (length and width) using the formula for a prolate ellipsoid and assuming a specific gravity of 1.0 mg/mm
3 [
18].
RNA extraction and analysis of stem cell-associated genes
Total RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA, USA), in accordance with the manufacturer's instructions, followed by DNAse treatment and RNA clean-up (RNeasy Mini Kit; Qiagen, Valencia, CA, USA). RNA concentration and integrity was determined using the RNA 6000 Nano LabChip Kit (Qiagen) on Agilent 2100 Bioanalyzer.
The mRNA levels of 84 genes associated with stem cell biology were examined using human Stem Cell RT2 profiler arrays (SuperArray Bioscience, Frederick, MD, USA), in accordance with the manufacturer's instructions. RNA (250 to 500 ng) was reverse transcribed using the First Strand Synthesis Kit (Qiagen), and cDNA was subjected to real-time PCR using SYBR green/ROX Master Mix (Qiagen) and PCR cycles were performed on a 7500 Real Time PCR System (Applied Biosystems, Foster City, CA, USA). A dissociation curve was run as a quality control using default melting curve settings. Values obtained for the threshold cycle for each gene were normalized using the average of housekeeping genes amplified on the same array.
Real-time quantitative RT-PCR
Real-time quantitative RT-PCR was used to measure mRNA expression levels of selected mouse ATP-binding cassette (ABC) transporters using LightCycler RNA Master SYBR Green Kit and the LightCycler 480 instrument (Roche Biochemicals, Indianapolis, IN, USA). Specific PCR primer sequences for genes are listed in Additional file
1. All RT-PCR reactions were performed on 200 to 400 ng total RNA with 250 nmol/l specific primers. Negative controls consisting of no-template (water) reaction mixtures were run with all reactions. Melting curves were determined for each primer set following all RT-PCR runs using the LightCycler 480 software. Crossing points for each transcript were determined using the second derivative maximum analysis with the arithmetic baseline adjustment. Crossing point values for each transporter were normalized to the respective crossing point values for reference gene
Pmca4 (plasma membrane calcium ATPase 4) [
19]. Data are presented as fold change in gene expression using the ΔΔCt method.
Immunostaining
Cells growing as spheroids were expanded in low-binding six-well plates, collected, allowed to attach over 2 hours to a multiwell chamber coated with D-poly L-lysine, fixed in 4% paraformaldehyde, and permeabilized with 0.3% Triton X-100 in PBS (15 minutes) and washed in PBS. Cells were blocked in donkey serum for 1 hour, and incubated with anti-Numb, anti-Oct4, anti-Nestin, or anti-CD133 primary antibodies overnight (Abcam, Cambridge, MA, USA). Secondary antibodies conjugated to Alexa Fluor 488 and Alexa Fluor 568 (Molecular Probes, Eugene, OR, USA) were added for 1 hour, washed in PBS, and mounted under a coverslip using Vectashield Mounting Medium (Vector Laboratories, Burlingame, CA, USA). Immunofluorescence was visualized in an Olympus 1X51 fluorescence microscope (Olympus America Inc., Center Valley, PA, USA.
Discussion
Cell lines derived from individual
Brca1 mouse mammary tumors had distinct and non-overlapping populations of putative cancer stem cell markers CD44
+/CD24
- and CD133
+. Only cell lines that contained a significant fraction of cells with these markers formed spheroid structures without preliminary sorting, and expansion of these spheroids
in vitro led to further spontaneous enrichment in cells with stem cell makers. In addition, cells sorted for cancer stem cell markers and cells growing as spheroids were significantly more resistant to chemotherapeutic agents than were parental cells, and were highly tumorigenic in mice. Two conclusions can be derived from these studies. First,
Brca1-deficient mouse mammary tumors contain heterogeneous populations of cells that share cancer stem cell properties. Second, some
Brca1-deficient tumors contain CD44
+/CD24
-/Low cells, previously associated with human breast cancer stem cells [
10,
11,
28], whereas others contain CD133
+ cells previously associated with tumors in other organs [
29]. Which population provides a closer correlate to human disease requires further study.
Human and murine cancer cell lines contain a small fraction of cells that have cancer stem cell properties. These cells are drug resistant, express stem cell associated genes, and have high capacity for reconstituting tumors
in vivo [
9,
28,
30,
31]. Because they constitute a very small cell fraction, their characterization requires isolation, usually by sorting based cell surface markers, and expansion
in vitro. Our data confirm that cells sorted for cancer stem cell markers reconstitute the parental population after a limited number of passages
in vitro. Repopulation is indicative of self-renewal, presumably by asymmetric division, and is consistent with stem cell-like properties of these cells. However, the expansion of sorted stem cells
in vitro results in rapid reduction in the cell fraction that expresses stem cell phenotype, which complicates
in vitro studies of cancer stem cell biology and development of specific therapies.
Other investigators have reported that cancer-initiating cells sorted from established cell lines or primary tumors form spheroids when plated in semisolid support or serum-free media supplemented with epidermal grwoth factor and basic fibroblast growth factor [
9]. Breast cancer-derived spheroid structures have been termed mammospheres and have similarities to those derived from normal mammary glands. Neurospheres derived from neural tissues or gliomas grow in similar conditions [
32].
Brca1 tumor cells sorted for expression of putative stem cell markers formed spheroids in the absence of attachment without supplementation with growth factors. Furthermore, neither sorting nor growth factor supplementation was required for generation of spheroids from unsorted
Brca1 cell lines that had a substantial population of cells expressing stem cell markers. The unsorted cells that grew spheroids in long-term culture were significantly enriched in stem cell makers and were highly resistant to DNA-damaging agents after multiple passages
in vitro. Whether other cancer cells that ordinarily grow in monolayer can survive in the absence of attachment and form spheroids, and become enriched in cancer stem cells remains to be established.
In contrast to the orderly transition of normal stem cells through differentiation with generation of progenitor and terminally differentiated cells, multistep carcinogenesis is likely to generate heterogeneity in cancer-initiating population, which is reflected in different cell surface markers that identify cancer stem cells in different tumor types. Previous studies corroborate that
BRCA1 deficiency results in genetic instability associated with centrosome amplification, defective cell cycle checkpoint control, and impaired DNA damage repair [
33,
34]. We found that that same
Brca1-deficient genetic background gave rise to mammary tumors with two distinct and non-overlapping populations of cells that bear cell surface markers previously assigned to tumor-initiating cells from human breast and other organs. Differences in cancer stem cell populations, which constitute a minute fraction of total tumor, may underlie some of the difficulties in using the genome-wide approach to characterizing molecular profiles for these tumors.
Previous studies showed that 200 CD44
+/CD24
-/Low human breast cancer cells reconstitute the entire population of cancer cells
in vitro and form tumors
in vivo, whereas larger numbers of parental cells or cells sorted for the absence of these markers are needed to generate smaller, slower growing tumors [
10,
11,
28]. We found that CD44
+/CD24
- Brca1-deficient mouse mammary tumor cells have cancer stem cell characteristics
in vitro, and 50 of these cells are sufficient to initiate tumors in mice.
CD44 is a complex, multispanning, transmembrane glycoprotein whose expression correlates with drug resistance and poor prognosis in many malignancies [
35]. In addition to hyaluronic acid, CD44 binds fibrinogen, fibronectin, collagen, laminin, fibroblast growth factor-2, other heparin-binding growth factors, and osteopontin, an inflammatory cytokine that is associated with metastatic progression. Recent reports suggested heterogeneity in human breast cancer stem cells that express CD44 [
36,
37]. Whether CD44 expression plays a direct role in drug resistance by activating multiple survival pathways via growth factor receptors or integrin-mediated 'inside-out' signaling is not known and warrants further investigation.
Although expression of CD24 negatively correlated with stem cell characteristics in human breast cancer [
10,
11,
28], the situation is more complex in mouse mammary tumors. In murine mammary gland the CD24
low cells correspond to myoepithelial cells and have high mammary fat pad reconstitution capacity, whereas the CD24
+ population has low capacity for reconstitution and is devoid of normal mammary stem cells [
34]. Furthermore, Weinberg and colleagues [
39] recently reported that mammosphere-forming and tumor-initiating capacities reside within CD24
+ freshly isolated normal mammary cells, but when these cells are briefly cultured the CD24
- population was enriched with cancer stem cells. These previously reported differences in expression of CD24 in different tumor types and in normal mouse mammary epithelium support our search for the role of CD24
- cells in combination with a more established marker, CD44, in
BRCA1-deficient tumors.
Expression of CD133 in cancer-initiating cells is well documented for brain, prostate, and colon cancers [
29] but has not been described in breast cancer. We detected 2% to 4% of CD133
+ cells in multiple cell lines derived from one
Brca1 tumor with characteristics similar to those found in CD44
+/CD24
- cells, including drug resistance, the ability to form spheroids with further 30% enrichment in CD133
+ cells, expression of stem cell genes, and
in vivo reconstitution of tumors with as few as 100 cells. CD133, also known as prominin-1, is a cell-surface glycoprotein with five transmembrane domains and two large glycosylated extracellular loops that localize to membrane protrusions [
29]. The function of CD133 in cancer stem cells has not been established, but one alternatively spliced form binds cholesterol and thus may be involved in Hedgehog signaling, which is required for primitive cell differentiation and epithelial-mesenchymal interactions [
40].
We previously described generation of tumors from cell suspensions from multiple genetically engineered mouse mammary tumors and their expansion by transplantation in naïve recipients
in vivo [
16]. Gene expression analysis of individual
Brca1 mammary tumors and their subsequent passages
in vivo revealed substantial heterogeneity in gene expression [
40], as predicted by differences in frequency and identity of cancer stem cell populations in cell lines derived from these tumors. Thus, in contrast to other models, such as MMTV-PyMT and MMTV-wnt1, in which pooling individual mouse tumors can generate sufficient material for basic and translational studies [
16,
42],
Brca1 tumors will need to be analyzed individually and these studies are limited by the size of each original tumor. Thus, generation of cell lines provided valuable reagents for these studies.
A fundamental characteristic of cancer stem cells is their resistance to chemotherapeutic agents. Much understanding of
BRCA1 drug resistance comes from studies of a single human cell line, namely H1937, which is null for the
BRCA1 gene. Although replacement of this gene increases resistance to vinorelbine and cisplatin, it does not change sensitivity to other agents, such as docetaxel [
43,
44]. This suggests that other mechanisms may determine drug resistance in that cell line. Consistent with these observations, we found that H1937 cells contain a significant (2% to 3%) population of cells expressing CD44
+/CD24
- and no detectable CD133
+ (data not shown). The contribution of these cells to drug sensitivity remains to be determined.
Over-expression of several ABC transporters has been linked to drug resistance. Our analysis showed that
Abcb1b expression correlated with an increase in cells having stem cell markers in A1.8 and RP.1 cell lines. Expression of
Abcb1b was further enriched in A.8 cells sorted for stem cell markers. However, this did not occur in RP.1 cells, because CD133
- cells, but not CD133
+ cells, exhibited greater expression of
Abcb1b. Increased
Abcg2 expression, previously associated with cancer stem cells, was evident in all
Brca1 cell lines, regardless of presence or absence of putative stem cell fraction. Furthermore, the highest expression of
Abcg2 was detected in B.15, a cell line that was not enriched in stem cell markers and did not form spheroids. Thus, expression of other transporters, or different drug resistance mechanisms, such as aldehyde dehydrogenase, which was over-expressed in both CD44
+/CD24
- and CD133
+ cell types, may be operational in
Brca1 cancer stem cells. Aldehyde dehydrogenase-1 participates in oxidation of retinol to all-
trans retinoic acid, and confers drug resistance to chemotherapeutic agents by an uncertain mechanism [
45].
We previously demonstrated schedule-dependent synergy of the HSP90 inhibitor 17-DMAG with doxorubicin for lymphoma cells [
20]. Here, we found that addition of 17-DMAG simultaneously or after chemotherapeutic agents sensitizes
Brca1 cancer stem cells to three types of DNA-damaging agents: doxorubicin, cisplatin, and etoposide. Because HSP90 inhibitors impair multiple signal transduction pathways, resulting in decreased cell survival and DNA repair [
20,
24,
46], our data showed that functional inactivation of
Brca1 is particularly vulnerable to this combination. Whether HSP90 inhibitors sensitize other cancer stem cells to chemotherapeutic agents remains to be established. Development of cancer stem cell-directed therapies has been hampered by inability to expand cancer stem cells
in vitro. Enrichment of cancer stem cells in spheroids formed by
Brca1 cell lines provides ample material for studies of cancer stem cell biology and preclinical testing.
Authors' contributions
MHW developed all cell lines, and performed the flow cytometry labeling, RNA extraction and analysis, as well as drug testing in vitro. AMC performed all studies on expression of ABC transporters. CDS performed additional cell labeling, some of the in vivo injections into mouse fat pad, and tumor measurements. MDC performed tumor measurements and tumor dissections. SVA advised on testing and data interpretation for ABC transporter expression and other drug resistance mechanisms. LV performed tumor dissection, in vivo injections of cells into mammary fat pad, overall coordination, and design of the study.