Generation of the LTLT-Ca cell line (long-term letrozole treated MCF-7 cells stably transfected with the human aromatase gene) was previously described (4). Briefly, LTLT-Ca cells were isolated from tumors of aromatase-transfected MCF-7 cells grown in ovariectomized BALB/c athymic mice after 56 weeks of letrozole treatment. The tumors start proliferating in the presence of the drug after long-term treatment. Human LTLT-Ca cells were generously provided by Dr Angela Brodie and were cultured in 75-cm² flasks in phenol red-free IMEM (Improved Minimum Essential Medium; Invitrogen; Thermo Fisher Scientific, Inc.), supplemented with 10% charcoal-dextran-stripped fetal bovine serum (FBS), 100 U/ml penicillin (Gibco; Thermo Fisher Scientific, Inc.), 100 U/ml streptomycin (Gibco; Thermo Fisher Scientific, Inc.), 25 µg/ml amphotericin B (Gibco; Thermo Fisher Scientific, Inc.), 7.5 µg/ml geneticin (Invitrogen; Thermo Fisher Scientific, Inc.) 1 µM letrozole (Sigma-Aldrich; Merck KGaA). The culture flasks were maintained in a humidified atmosphere of 5% CO₂ at 37°C. Letrozole-sensitive AC-1 cells were maintained in DMEM (Dulbecco's modified Eagle's medium; Invitrogen; Thermo Fisher Scientific, Inc.), supplemented with 5% FBS, 100 U/ml penicillin (Gibco; Thermo Fisher Scientific, Inc.), 100 U/ml streptomycin (Gibco; Thermo Fisher Scientific, Inc.), 25 µg/ml amphotericin B (Gibco; Thermo Fisher Scientific, Inc.), 7.5 µg/ml geneticin (Invitrogen; Thermo Fisher Scientific, Inc.) in a humidified atmosphere of 5% CO₂ at 37°C. T47D letrozole-sensitive (T47Darom) and T47D letrozole-resistant (T47DaromlR) cells were cultured as previously described by Gupta et al (21) and were a generous gift from IIT Research Institute. Mycoplasma testing was performed for all cell lines. All cells were authenticated by short tandem repeat (STR) profiling by American Type Culture Collection (ATCC), confirming that the AC-1 and LTLT-Ca cell lines shared more than 85% homology with the MCF-7 cell lines, while the T47Darom and T47DaromLR cell lines shared more than 85% homology with the T47D cell line. Cell lines with 80% match are considered related (derived from a common ancestry). In brief, 17 STR loci plus the sex-determining locus, amelogenin, were amplified using the commercially available PowerPlex^® 18D Kit (Promega). The cell line samples were processed using the ABI Prism^® 3500 ×L Genetic Analyzer. Data were analyzed using the GeneMapper^® ID-X v1.2 software (Applied Biosystems). Appropriate positive controls (MCF-7 or T47D cell lines) were run and confirmed for each sample submitted. Cell lines were authenticated using STR analysis as described in 2012 in the ANSI Standard (ASN-0002) by the ATCC Standards Development Organization (SDO), as well as Capes-Davis et al (22).

AC-1 and LTLT-Ca cells were grown in regular media to attain 80–90% confluency, and after media was removed, cells were rinsed twice with Hank's Balanced Salt Solution (HBSS; StemCell Technologies, Vancouver, Canada) to remove residual culture media. Then, cells were gently scraped and resuspended in 10 ml of MammoCult™ media (StemCell Technologies). Next, cells were centrifuged at 500 × g for 3 min at room temperature. The supernatant was aspirated, and the pellet was resuspended into a single cell suspension in 2 ml of MammoCult™ media. Cell viability and concentration were determined by the Trypan Blue exclusion assay. For mammosphere cultures, the seeding density for both cell lines were 100,000 cells per 25-cm² suspension flasks (CellTreat Scientific Products). All flasks were incubated in a 5% CO₂ humidified incubator at 37°C for at least 7 days. Once mammospheres were detected by light microscopy, the cells were harvested as detailed below in western blot analysis.

For primary mammosphere formation, AC-1 and LTLT-Ca cells were enumerated, and 20,000 cells/well were seeded in ultra-low adhesion 6-well plates. The cultures were incubated in a 5% CO₂ humidified incubator at 37°C for 7 days and spheres ≥60 µm were counted and recorded. After primary spheres were formed, secondary mammosphere formation was conducted for both AC-1 and LTLT-Ca cells by dissociating the primary spheres with Accutase (Sigma-Aldrich; Merck KGaA) according to the manufacturer's instructions. Next, 20,000 cells/well were seeded in ultra-low adhesion 6-well plates, with the remainder of the assay conducted as described for primary mammosphere formation.

Proliferation assays were conducted as previously described (23). Briefly, AC-1 or LTLT-Ca cells were seeded in 96-well plates at a density of 1 × 10³ cells/well in a total volume of 100 µl and allowed to attach overnight. Background levels were determined by preparing blank samples, with media added to wells in the absence of cells. On the following day, 10 µl of resazurin dye (Sigma-Aldrich; Merck KGaA) was added to each well and incubated for 4 h at 5% CO₂ and 37°C. Samples were agitated for 1 min and the fluorescence was measured at 24, 48, 72, 96 and 120 h using a Biotek Synergy H1 microplate reader (BioTek Instruments, Inc.) to measure fluorescence intensity at 550 nm excitation/590 nm emission background wavelengths. All experiments were performed with n≥3 and a total of 3 biological replicates were performed. The proliferative activity was determined and calculated as described below: Proliferative activity = [Fluorescence of viable cells-Fluorescence of blank (media only)].

AC-1 and LTLT-Ca mammospheres were dissociated and single cells were seeded at a density of 1.5×10⁵ cells/well in 6-well tissue culture plates until 100% confluency was attained. A scratch (wound) was made down the center of each well using a 10 µl pipette tip, and new media was added. Each scratch was immediately imaged (at 0 h) 3 times at different points using the 5X objective on a Zeiss AX10 fluorescence microscope. Images were captured using Olympus CellSens Standard 1.16 software. Then, cells were grown at 37°C, 5% CO₂, and images were captured at 24 and 48 h. The wounds were measured and analyzed using the WimScratch Wound healing Software.

Briefly, adherent cells were gently scraped and homogenized in cold RIPA buffer supplemented with 2X protease and phosphatase inhibitors (Thermo Fisher Scientific, Inc.). Mammospheres were centrifuged at 400 × g for 7 min at 4°C and then resuspended in cold RIPA buffer. All samples were incubated for 30 min on ice and then centrifuged for 20 min at 12,000 rpm. The protein extract was quantified in each sample using the Bradford assay. For gel electrophoresis, all samples were incubated with Laemmli protein sample buffer (Bio-Rad) at 70°C for 10 min. Then, 75 µg of denatured protein was separated using 4–20% Mini-PROTEAN^® TGX™ Precast Protein Gels (Bio-Rad) and transferred to polyvinylidene difluoride (PVDF) membranes. All blots were blocked for 1 h with 5% bovine serum albumin (BSA) in phosphate-buffered saline (PBS) and 0.1% Tween-20 (PBS-T) buffer. Following incubation with primary antibodies, anti-CD24 antibody at 1:1,000 dilution (catalog no. ab179821; Abcam) and anti-CD44 antibody 1:1,000 dilution (catalog no. ab157107; Abcam), the membranes were incubated with the anti-rabbit secondary antibody (catalog no. 7074S; Cell Signaling Technology). The protein bands were detected using the Clarity Max Western ECL Substrate (Bio-Rad) according to the manufacturer's instructions. Immunoreactive bands were visualized using the ChemiDoc XRS imaging system (Bio-Rad). The exposure time was automatically detected by the imaging system. The protein bands were analyzed using Image Lab software (Bio-Rad). Arbitrary densitometry units were quantified and expressed as mean ± standard deviation. The bands were normalized to the housekeeping protein bands (GAPDH), whereby the density of the target protein in each lane was multiplied by the ratio of the loading control density from the control sample (lane 1) to the loading control density of other lanes. The immunoblot images are representative of more than three independent experiments with a minimum of 2 duplicates per sample.

SCID female ovariectomized mice (29–32 days old) were obtained from Charles River Laboratories. All animals underwent an adaptation period of 5–7 days in a pathogen-free and sterile environment, with a phytoestrogen-free diet. AC-1 and LTLT-Ca cells in the exponential phase of growth were harvested using PBS supplemented with 2% EDTA solution and washed. Viable cells (5×10⁶) in a 50 µl sterile PBS suspension were mixed with 100 µl Matrigel Reduced Factors (BD Biosciences). AC-1 and LTLT-Ca cells were injected into the mammary fat pad (n=5 for each group). For AC-1 cells, estrogen pellets (0.72 mg, 60-day release; Innovative Research of America) were subcutaneously implanted in the lateral area of the neck, in the middle point between the ear and shoulder, using a precision trocar (10 gauge). All animal procedures were performed under anesthesia using a ketamine/xylazine mixture consisting of 80 mg/kg of ketamine and 10 mg/kg of xylazine. Tumors were allowed to form over 10 days. Tumor volume was measured weekly for 8 weeks using a digital caliper. Tumor volume was calculated using the following formula: 4/3LM², where L is the larger radius and M is the smaller radius. At necropsy, animals were euthanized by exposure to a CO₂ chamber with a flow rate of 2L of CO₂/min at a displacement rate of 30–70% of the chamber volume per minute. Death was confirmed by evidence of pale eyes and the absence of a heartbeat and lack of respiration for at least 1 min. For further analysis, tumors were removed and fixed in 10% formalin. All procedures involving animals were conducted in compliance with state and federal laws, standards of the US Department of Health and Human Services, and guidelines established by the Xavier University of Louisiana University Animal Care and Use Committee. The facilities and laboratory animal program of Xavier University of Louisiana are accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care.

For immunohistochemical analysis, AC-1 and LTLT-Ca tumors were grown as described above, using groups of 5 animals, each containing 2 tumors. Immunohistochemistry was performed as previously described (24). Briefly, a minimum of one tumor tissue per animal (n=5/group) was fixed, deparaffinized, and rehydrated. Anti-CD24 (catalog no. MAB5248, 1:100 dilution; R&D Systems) and anti-CD44 (catalog number MAB7045, 1:100 dilution; R&D Systems) antibodies were used as potential markers. Staining was performed using EXPOSE mouse and rabbit specific HRP/DAB detection IHC kit (catalog number: ab80436; Abcam). Cells were counterstained with hematoxylin for 3 min, dehydrated, and mounted. Once slides were prepared, at least 4 microscopic fields were randomly selected for each tumor and visualized using ×20 magnification. Data are presented as a semi-quantitative Histo-score, where the fractions were assigned as negative (score 0), weakly positive (score 1), positive (score 2), or strongly positive (score 3). All slides were scored blindly by three individual investigators.

Immunofluorescence was performed as previously described (24). Briefly, three replicates of LTLT-Ca cells were seeded in 8-well chamber slides (Thermo Fisher Scientific, Inc.) pre-coated with 2% gelatin and grown to 50% confluence. Cells were fixed with formaldehyde, permeabilized with 0.5% NP-40 in PBS, and rinsed with PBS. The slides were blocked with 10% goat serum (Invitrogen; Thermo Fisher Scientific, Inc.) in PBS and incubated with anti-CD24 (1:200) catalog no. sc-19585, anti-HCAM (1:200) catalog no. sc-7927, and anti-Ki67 (1:200) catalog no. sc-23900, (Santa Cruz Biotechnology Inc.). Then, samples were washed with 1% goat serum in PBS and incubated with Alexa Fluor goat anti-rabbit-488 (1:1,000), catalog number: A-11008, or Alexa Fluor goat anti-mouse-488 secondary antibodies (1:1,000), catalog number: A-11001 (Invitrogen; Thermo Fisher Scientific, Inc.) in 10% goat serum. The samples were washed and stained with 300 nM DAPI (Invitrogen; Thermo Fisher Scientific, Inc.). The slides were imaged using an Olympus Bx41 microscope (Olympus) and captured using DP72 CCD driven by DP2 software (Olympus); the color images were combined using ImageJ software.

Briefly, LTLT-Ca cells were cultured adherently in 75-cm² flasks in phenol red-free IMEM supplemented with 5% charcoal-stripped-FBS or as mammospheres as described above, cultured until 70%–80% confluency. Total RNA was extracted from cells using RNeasy (Qiagen) following the manufacturer's recommendations. Each array profiles the expression of a panel of 84 genes including 7 internal controls and 5 housekeeping gene controls. For each array, 2 µg of RNA was reverse-transcribed into cDNA in the presence of gene-specific oligonucleotide primers using the iScript cDNA Synthesis kit (Bio-Rad) as described in the manufacturer's protocol. The cDNA template was mixed with the appropriate ready-to-use PCR master mix (Bio-Rad). Equal volumes were measured (in aliquots) into each well of the same plate, and then the real-time PCR cycling program was run as described previously (7). RT-qPCR was performed using the manufacturer's protocols for Human Cell Motility (PAHS-128ZD) and Human Epithelial to Mesenchymal Transition (PAHS-090ZD) (EMT) RT² Profiler PCR Array (Qiagen). Relative gene expression was calculated using the 2⁻ΔΔ^Cq method (25), in which Ct indicates the fractional cycle number where the fluorescent signal reaches the detection threshold. The ‘delta-delta’ method uses the normalized ΔCt value of each sample, calculated using a total of five housekeeping gene control genes (18S rRNA, HPRT1, RPL13A, GAPDH and ACTB) (26). Fold change values are presented as average fold change = 2^−(average ΔΔ^Ct) for genes in treated samples, relative to control samples. Differences in gene expression between groups were calculated using Student's t-test, in which fold changes ≤3 were considered significant. All experiments were performed with a minimum of three biological replicates.

Studies involving more than 2 groups were analyzed by 1-way ANOVA with Tukey's posttest analysis; all others were subjected to unpaired Student's t test and are summarized as the mean ± standard error of the mean (SEM) using Graph Pad Prism V.6 (GraphPad Software, Inc.). Data are expressed as the mean unit ± SEM (****P<0.0001, ***P<0.001, **P<0.01, *P<0.05).

We have previously demonstrated that as breast cancer cells transition from letrozole-sensitive to letrozole-resistant, they are associated with estrogen independence, enhanced cellular motility, and an EMT-like phenotype (7,24). EMT is linked to the progression of cancer, as well as increased stemness of tumors (27). To examine the expression of two putative breast CSC markers in letrozole-sensitive (AC-1 cells) and letrozole-resistant (LTLT-Ca cells) breast cancer cell lines, we performed immunoblotting analysis. Considering that the mammosphere formation assay is employed as a surrogate reporter of cancer stem cell activity, both cell lines were cultured either adherently (2D) or in suspension (3D) as mammospheres. Our results demonstrated that when both AC-1 and LTLT-Ca cells were cultured as mammospheres, higher levels of CD24 and CD44 were expressed when compared with their adherently cultured counterparts (Fig. 1). On comparing CD24 expression between 2D and 3D cells, AC-1 3D cells exhibited an increase in CD24 expression exceeding 25%, while LTLT-Ca 3D cells exhibited a 500% increase in CD24 expression when compared with their 2D counterparts. When CD44 expression was examined, AC-1 3D and LTLT-Ca 3D cells exhibited increased CD44 expression (75 and 50%, respectively), suggesting that in both cell lines, mammosphere cultures presented enriched CSC characteristics when compared with cells cultured adherently, irrespective of their response to letrozole. Furthermore, a similar result was observed in T47D letrozole-sensitive and letrozole-resistant cell lines (Fig. S1).

Although in vitro human breast cancer cell models are useful screening tools, they can be limited by the absence of the breast tumor microenvironment. As in vitro cultured cells exhibit less complexity when compared with those in vivo, AC-1 and LTLT-Ca cells were inoculated into nude mice and allowed to form tumors (Fig. S2). The maximum diameter and volume of the AC-1 tumors were 6.36 × 6.78 mm and 365.66 mm³ respectively, while the maximum diameter and volume of the LTLT-Ca tumors were 5.24 ×5.85 mm and 214.17 mm³ respectively. Then, tumors were excised, and CD24 and CD44 protein expressions were examined by immunohistochemistry. In addition, hematoxylin-eosin staining was performed. Letrozole-sensitive tumors revealed markedly less CD44 expression when compared with letrozole-resistant tumors, scoring 1 and 3, respectively (Fig. 2), whereas CD24 expression was higher in letrozole-resistant tumors than in letrozole-sensitive tumors, scoring 2 and 1, respectively (Fig. 2). In summary, the expression of CSC markers in tumors was CD44⁺/CD24⁺ in letrozole-resistant cells (LTLT-Ca) and low CD44/CD24 in letrozole-sensitive cells (AC-1 cells); meaning that LTLT-Ca tumors had higher expression of both CD44 and CD24 and AC-1 tumors had low expression of CD44 and CD24.

As marked differences were observed in the CD44/CD24 expression profile between AI-sensitive and AI-resistant tumors, we examined whether these differences correlated with the self-renewal capacity by using the mammosphere self-renewal assay. AC-1 and LTLT-Ca cells were seeded at a low density in an environment that prevented adherence, thus enabling proliferation in suspension as spherical clusters (28). Interestingly, LTLT-Ca cells formed mammospheres at a 3.4-fold higher index when compared with AC-1 cell mammospheres (Fig. 3A). Culture images were obtained to examine their morphology, demonstrating that letrozole-resistant cells formed hollow mammospheres, while letrozole sensitivity was associated with solid mammospheres (Fig. 3B). Both cell lines formed symmetrical and tightly packed mammospheres. The mammospheres underwent a second passage, and LTLT-Ca mammospheres showed a 2.9-fold increase in mammosphere formation when compared with AC-1 cells. Compared with primary mammospheres, the total number of secondary mammospheres decreased; however, the ratio of secondary mammosphere formation between AC-1 and the LTLT-Ca mammospheres was similar to the primary mammosphere formation ratio. Based on this assay, letrozole-resistant cells were more highly associated with increased self-renewal capacity when compared with letrozole-sensitive cells. To determine whether the increase in LTLT-Ca mammosphere formation could be attributed to enhanced cell growth, proliferation assays were performed to compare both cell lines. AC-1 cells demonstrated a greater proliferative capacity than LTLT-Ca cells, suggesting that the increase in LTLT-Ca mammosphere formation was independent of cell proliferation (Fig. 4). To complement this finding, immunofluorescent analysis of LTLT-Ca mammospheres was conducted, and the LTLT-Ca mammospheres were CD44⁺/CD24 (Fig. S3), similar to the tumors and adherent cultures. Furthermore, LTLT-Ca mammospheres stained positive for Ki67, a common cell proliferation marker, indicating their proliferative capacity.

As previous reports have confirmed that LTLT-Ca cells demonstrate greater migratory ability than AC-1 cells (7), we measured the expression of genes involved in motility and EMT, to assess whether the presence of CSCs was associated with a migratory phenotype. Accordingly, LTLT-Ca cells were cultured either adherently or as mammospheres, with targeted gene expression arrays were conducted to measure the expression of genes involved in cellular motility and EMT (genes that were significantly altered where P<0.05 are shown in Table I). When compared with LTLT-Ca adherent cells, LTLT-Ca mammospheres displayed a −12.01-fold, −6.44-fold, and −8.14-fold decrease in the expression of caveolin-1, E-cadherin, and β-catenin, respectively. Interestingly, LTLT-Ca mammospheres exhibited a −3.33-fold and −3.29-fold decrease in epidermal growth factor receptor (EGFR) and integrin αV (CD51) expression. In addition to genes involved in motility, we also observed increased expression of TFPI-2 (4.42-fold) and STEAP1 (3.17-fold).

As proof of concept, a scratch assay (i.e., wound healing assay) was performed to assess whether the letrozole-resistant mammospheres impacted migratory behavior (Fig. 5). The results demonstrated that as early as 24 h, the letrozole-resistant mammospheres began migrating faster than the letrozole-sensitive mammospheres (Fig. 5). After 48 h, this effect was even more pronounced, and the LTLT-Ca cell wound closure was 70%, whereas the AC-1 wound closure was 39%, suggesting that as cells acquire resistance and express putative breast CSC markers, they become more aggressive through increased motility. Taken together, as letrozole-resistant cells acquire CSC characteristics, they are less associated with epithelial-like features, progressing toward a more mesenchymal phenotype.