Introduction
Metastatic recurrence in advanced breast cancer is a major cause of patient morbidity and mortality. A substantial body of evidence indicates that metastatic foci are initiated by a latent sub-population of cancer cells that are capable of surviving adjuvant chemotherapy and colonising distal sites, where they may remain dormant for several years before emerging to reinitiate tumour growth [
1]. The inability to remove this population, often referred to as cancer stem cells (CSCs), has proved to be the major obstacle in developing truly curative treatments for advanced breast cancer, with 15-year recurrence rates remaining as high as 40% [
2].
Given that traditional chemotherapeutic agents are reliant on mitotic activity to initiate cytotoxicity, it has been proposed that CSCs, like their non-malignant counterparts, exist in a slow-cycling state of relative mitotic quiescence that confers an inherent ability to survive chemotherapy. Indeed, the slow-cycling status of adult stem cell populations identified in the brain, skin and intestine has been transitively linked to their observed survival during chemotherapy and their subsequent ability to regenerate tissue following chemotherapy withdrawal [
3‐
5]. Analogous populations of slow-cycling cells have been identified in multiple cancer types. Yumoto et al. [
6] described the identification of slowly cycling sub-populations in a number of human prostate cancer cell lines based on their persistent retention of fluorescent lipophilic dye. Expanding on these findings, Wang et al. [
7] demonstrated that label-retaining prostate cancer cells differentially up-regulate expression of various haematopoietic stem cell niche-associated markers and were significantly more metastatic in vivo when compared to the rapidly dividing cell population isolated from the same parental culture. Dembinski and Krauss [
8] similarly reported that a slow-cycling population of label-retaining pancreatic adenocarcinoma cells demonstrated morphological and transcriptomic changes indicative of epithelial-to-mesenchymal (EMT) transition and were ten-fold more tumourigenic in vivo than the non-label retaining population. Label-retaining populations have also been identified in melanoma, brain, ovarian, colon and breast cancer cell lines [
9‐
12]. However, the link between these slow-cycling, label-retaining populations and chemoresistance has remained largely unstudied and it remains unclear whether label-retaining cells represent a tumour recurrence-initiating population.
In this study, we demonstrate the novel application of Vybrant® DiD for identification, isolation and characterisation of a latent, slow-cycling, label-retaining cell population in the oestrogen receptor-positive MCF-7 and triple-negative MDA-MB-231 human breast cancer cell lines. We demonstrate that these slow-cycling cells are significantly more resistant to conventional chemotherapeutic agents than their rapidly dividing counterparts and importantly, that label-retaining cells are exclusively capable of active proliferation following removal of chemotherapeutic drugs, implying their ability to drive tumour recurrence. Our data support that label-retaining cells can serve as a model for identification of molecular mechanisms driving tumour cell quiescence and de novo chemoresistance.
Materials and methods
Cell culture
MCF-7 and MDA-MB-231 human breast cancer cell lines were sourced directly from the American Type Culture Collection (ATCC) (Manassas, Virginia, USA) as fully authenticated cryogenically frozen cultures. Cells were routinely maintained in vitro as adherent cultures grown in complete growth medium composed of RPMI-1640 basal medium (11 mM glucose, 2 mM L-glutamine) (Life Technologies Ltd., Paisley, U.K.) supplemented with 10% (v/v) foetal bovine serum (FBS) (Life Technologies Ltd.).
Fluorescent dye labelling, label retention assays and fluorescence-activated cell sorting
Vybrant® DiD labelling was performed according to the manufacturer’s instructions for labelling cells in suspension (Molecular Probes MP22885) using 1.0 × 106 cells per ml of serum-free basal culture medium.
Labelled samples were grown as adherent cultures for up to six consecutive passages post-staining, with sub-culture being undertaken at approximately 80% of the duration of the logarithmic phase of culture growth. At each sub-culture interval, samples of 1.0 × 106 cells were prepared in serum-free basal medium and used to assess the degree of Vybrant® DiD staining present within each sample. Cytofluorimetric analyses were undertaken using the BD™ LSR-II™ platform (Beckton, Dickenson and Co. Plc., Oxford, UK). Fluorescence activated cell sorting (FACS) was undertaken using the BD™ FACSAria™ platform (Beckton, Dickenson and Co. Plc.) after six consecutive passages of culture post-labelling. Cytofluorimetric analysis and FACS platforms were calibrated at the outset of each experiment to enable detection of Vybrant® DiD-negative (DiD-) and -positive (DiD+) cells. Negative control samples consisting of cells that had never been exposed to Vybrant® DiD were first used to set the gate for the detection of DiD− events. All cells subsequently identified as DiD− had therefore completely lost their initial label and possessed a fluorescence intensity equivalent to that of an unlabelled cell population. Positive control samples consisting of cells that had been freshly labelled with Vybrant® DiD were then used to set the gate for detection of DiD+ cells. All cells subsequently identified as DiD+ (dye-retaining) therefore possessed a fluorescence intensity equivalent to that of a dye-saturated cell population immediately post-labelling. The general gating tree for identification, differential flow cytometric analysis and isolation of DiD− and DiD+ cells is shown in Supplementary Fig. 1.
Flow cytometric assays
In order to analyse the cell cycle profile of live DiD− and DiD+ cells, unsorted samples were prepared at 1.0 × 106 cells per ml in serum-free basal culture medium from labelled cultures at six passages post-staining. Hoechst 33342 stock solution (Life Technologies Ltd.) was added to samples at a final concentration of 5 µg/ml and incubation undertaken at 37 °C for 45 min in total darkness prior to analysis.
Aldehyde dehydrogenase (ALDH) activity was determined using the non-immunological ALDEFLUOR™ assay kit (STEMCELL™ Technologies U.K. Ltd., Cambridge, U.K.) according to the manufacturer’s instructions (STEMCELL™ Technologies Document 29888) using unsorted cultures at six passages post-staining with Vybrant® DiD. Samples of unlabelled cells treated with the specific ALDH inhibitor diethylaminobenzaldehyde (DEAB) were prepared in an identical manner concomitantly as control samples used to allow definition of the gating strategy and to control for autofluorescence by setting the appropriate photomultiplier tube voltage for detection of negative events within each channel.
In order to analyse the expression of breast CSC markers, unsorted Vybrant® DiD-labelled samples at passage six post-labelling were stained with mouse monoclonal IgG1 anti-human CD24-phycoerythrin (Abcam Plc., Cambridge, U.K., clone SN3, product code ab77219) and mouse monoclonal IgG2b anti-human CD44-Brilliant Violet 421™ (Beckton, Dickenson and Co. Plc., clone G44-26, product code 562890) primary antibodies concomitantly. The final working concentration of each antibody was 6 µg/ml and 10 µg/ml, respectively. Matched isotype control samples were prepared using phycoerythrin-conjugated mouse monoclonal IgG1 (Abcam Plc., clone B11/6, product code ab91357) and Brilliant Violet 421™-conjugated mouse monoclonal IgG2b (Beckton, Dickenson and Co. Plc., clone 27–35, product code 562748) isotype control antibodies. Each isotype control antibody was used at the same working concentration as the primary antibody to which it was matched. Briefly, samples of 1.0 × 106 live cells were prepared from harvested cell cultures and washed twice by resuspension in flow cytometry buffer (5% (v/v) FBS in PBS). Washed cell pellets were resuspended in 100 µl of primary or isotype control antibody solution (pre-diluted to the desired concentration in flow cytometry buffer) and incubated for 1 h at 4 °C in total darkness under constant agitation. Samples were then washed three times, resuspended in 1 ml of flow cytometry buffer and passed through a 40 µm cell strainer prior to immediate cytofluorimetric analysis.
Immunofluorescence
Vybrant® DiD− and DiD+ cells were sorted at passage six post-labelling of cultures and 2000 sorted cells deposited onto Superfrost Plus™ glass microscopy slides by cytocentrifugation (150xg for 3 min using medium acceleration) using the Shandon™ Cytospin™ 3 cytocentrifuge (Thermo Fisher Scientific, Paisley, UK). Samples were fixed in 4% (w/v) paraformaldehyde on ice for 10 min, washed in two changes of PBS, and permeabilised in 0.1% (v/v) Triton™ X-100 in PBS. Samples were washed three times using PBS-Tween® 20 (PBST) (0.01% (v/v) Tween® 20 in PBS) and blocked in a solution of 10% (v/v) normal goat serum + 1% (w/v) bovine serum albumin (BSA) in PBST at ambient temperature for 1 h. Immunostaining for Ki67 expression was undertaken using an unconjugated rabbit polyclonal IgG anti-human Ki67 primary antibody (Abcam Plc., product code ab15580) diluted in in 1% (w/v) BSA in PBST to a final working concentration of 1 µg/ml. Matched isotype control samples were prepared using an unconjugated rabbit polyclonal IgG isotype control antibody (Abcam Plc., product code ab171870) and were used at the same final working concentration as the primary antibody. Incubation was undertaken inside of a humidified slide tray overnight at 4 °C. Following three more washes in PBST, primary antibody staining was localised using an AlexaFluor®488-conjugated goat polyclonal anti-rabbit IgG secondary antibody (Abcam Plc., product code ab150077) diluted in in 1% (w/v) BSA in PBST to a final working concentration of 4 µg/ml. Incubation was undertaken for 1 h at ambient temperature. Three further washes in PBST were undertaken before samples were mounted using ProLong® Gold anti-fade mounting medium containing 4ʹ,6-diamidino-2-phenylindole (DAPI) (Life Technologies Ltd.).
Chemotherapy resistance assays
Cultures were labelled with Vybrant
® DiD and grown for five consecutive passages post-staining before being seeded into triplicate wells of six-well cluster plates. All plates were placed into incubation for a period of 24 h post-seeding prior to treatment with either pegylated liposomal doxorubicin (Clinical Research Pharmacy, Weston Park Hospital, Sheffield, U.K.) or paclitaxel (Abcam Plc.) at the pre-determined IC
95 concentration (2.36 µM and 46.45 nM, respectively, for MCF-7 cells and 0.32 µM and 20.80 nM, respectively, for MDA-MB-231 cells). These concentrations were established using drug dose–response curves generated using the MTT assay (Supplementary Fig. 2) as previously described [
13]. After 72 h, drug-containing medium was removed from cultures and Vybrant
® DiD content analysed cytofluorimetrically. Vybrant
® DiD− and DiD+ populations were sorted by FACS and deposited directly into triplicate wells of six-well plates containing drug-free culture medium at a clonogenic density of 200 cells per well. Plates were incubated for a period equivalent to at least six log-phase doubling times prior to colony formation being assayed. Colonies were fixed in a 4% (w/v) paraformaldehyde in PBS for 15 min at ambient temperature. Fixative was then removed and colonies stained with 0.05% (w/v) aqueous crystal violet solution (Merck Chemicals Ltd., Nottingham, U.K.) for 30 min at ambient temperature. Plates were imaged using the Pixera Professional 1.2 megapixel digital camera system (Pixera UK Ltd., Bourne End, UK) and colonies counted by way of semi-automated image analysis using GeneTools software (Syngene U.K. Ltd., Cambridge, UK).
Real-time qPCR
Total RNA was isolated from DiD− and DiD+ cells at passage six post-labelling of cultures using the miRNeasy Micro Kit (Qiagen UK, Manchester, UK) according to the manufacturer’s instructions. During FACS, each population was sorted directly into separate tubes containing 700 µl of QIAzol® lysis reagent. Following extraction of RNA, cDNA was synthesised using the RT2 First Strand Kit (Qiagen) and real-time qPCR carried out using RT2 SYBR Green ROX™ qPCR Mastermix (Qiagen) according to the manufacturer’s instructions. The complete PCR reaction mixture (Mastermix + cDNA synthesis product) was dispensed as 10 µl volumes into the wells of a 384-well RT2 custom PCR array containing pre-dispensed primers (information for primers is available in Supplementary Table 1). A total of 1.25 ng of cDNA was used per 10 µl reaction. The real-time qPCR amplification protocol was run on an Applied Biosystems 7900-HT real-time cycler and was as follows: 95 °C for 10 min (1 cycle) followed by 15 s at 95 °C and 1 min at 60 °C (40 cycles). Dissociation (melting) curve analysis was carried out at the end of each run using the default melting curve programme of the cycler.
Statistical analyses
All data are expressed as the mean ± the standard error of the mean (SEM). All statistical analyses were undertaken using GraphPad Prism 7 (GraphPad Software Inc., La Jolla, USA). The statistical analysis applied to data, where undertaken, is indicated within the respective figure legend. Statistical significance was attributed when P < 0.05.
Discussion
In this report we have demonstrated that Vybrant® DiD can be used for pulse-chase identification and characterisation of an intrinsic sub-population of slow-cycling cells in breast cancer cell lines based on their ability to retain their initial Vybrant® DiD label for an extended duration. Compared to the rapidly dividing bulk cell mass, the reduced mitotic capacity of the label-retaining sub-population was shown to be associated with an array of biologically and clinically relevant differential characteristics. These included alterations in the expression of CD44 and CD24 surface markers, increased ALDH activity, transcriptomic profiles indicative of an intermediate EMT-MET phenotype, a marked reduction in chemosensitivity and an exclusive capacity for re-initiation of culture growth following cessation of exposure to chemotherapeutic agents. Collectively, these traits could signify that label-retaining cells are a population of cancer stem cells.
A number of previous studies have reported a significant expansion of the supposed cancer stem cell compartment following chemotherapy [
8,
20‐
23]. Such enrichment would require either a large-scale expansion of the cancer stem cell population, or effective reduction of the bulk cell mass. A similar pattern of preferential proliferation or cytotoxicity would also be required to account for the significant increase in the relative proportion of slow-cycling, DiD+ cells surviving following exposure to chemotherapy agents reported here. Calculation of the absolute number of DiD+ cells present in MCF-7 cultures at the time of seeding and following exposure to a high concentration (IC
95) of chemotherapeutic agents demonstrated an increase in the absolute label-retaining DiD+ cell number from ~ 1000 cells at the time of seeding to ~ 3600 cells at the end of the assay period. Similarly, in the MDA-MB-231 cell line the number of label-retaining DiD+ cells increased from ~ 1600 to ~ 2800 across all treatment groups. These increases in label-retaining cell numbers most likely occurred due to mitotic division generating partially labelled daughter cells. However, such expansion alone could not account for the significantly increased relative proportion of DiD+ cells within each of the drug-treated cultures compared to untreated cultures. These data not only demonstrate that label-retaining cells continued to proliferate when exposed to standard chemotherapeutic agents, but that the non-label-retaining DiD− cell population was significantly more susceptible to drug-induced cell death than their slow-cycling DiD+ counterparts, resulting in a net enrichment for DiD+ cells. These conclusions mirror those drawn from similar in vitro studies reported by Moore et al. [
11], who also demonstrated effective enrichment of a slow-cycling fluorescent label-retaining population in vivo using clinically relevant doses of oxaliplatin and fluorouracil to treat tumours derived from the HCT116 human colon cancer cell line.
While the intrinsic ability of cancer cell populations to survive anti-neoplastic chemotherapy is clinically important, cells that are able to initiate tumoural relapse or disease recurrence must also be capable of subsequent proliferation. We therefore established that the DiD+ population were not avoiding the cytotoxic effects of doxorubicin or paclitaxel by entering a permanently non-dividing state of senescence, or that the onset of the cytotoxicity exerted by these drugs was simply delayed due to the slow-cycling nature of the label-retaining population. The formation of new clonal populations by label-retaining cells following cessation of exposure to chemotherapeutic drugs indicated that a proportion of this sub-population was completely resistant to chemotherapy. These findings mirror those previously reported by Moore et al. [
11], which showed that a sub-population of label-retaining cells could actively proliferate in vitro and in vivo shortly after halting oxaliplatin and fluorouracil treatment. Moreover, the proportion of label-retaining cells that were able to form clonal populations following withdrawal of chemotherapy in this study (~ 2% for both MCF-7 and MDA-MB-231 across both chemotherapy drugs used) was several orders of magnitude larger than the proportion of cells previously demonstrated as being required to successfully establish metastasis (0.001–0.02%) [
29‐
31]. When taken together, these results indicate that slow-cycling cells not only survive during drug exposure but also are capable of reactivation after withdrawal of chemotherapy and could therefore potentially initiate either local tumoural relapse or formation of a secondary metastatic lesion.
Comparative live cell cycle profiling of DiD− and DiD+ populations indicated that DiD+ cells in both MCF-7 and MDA-MB-231 cell lines were enriched for cells in the G
2/M-phase of the cell cycle. A number of instances of label-retaining cells having an increased G
2/M-fraction compared to their rapidly proliferating counterparts have been reported in the existing literature across various cancer types (breast, prostate, intestinal, myeloid cell, brain and ovarian) [
7,
10,
11]. It has also previously been reported that such G
2/M-phase arrest is associated with multi-drug chemoresistance and a propensity to evade apoptosis [
32,
33]. The latter observation implies that, while reduced mitotic activity itself is likely to contribute to the survival of label-retaining cells in response to chemotherapy, it may not be the sole means by which de novo drug resistance occurs; growth arrest or slowed cell cycle transition could effectively increase the time for drug efflux, drug metabolism or repair of drug-induced cellular stress, and thereby enable evasion of pro-apoptotic signals in the relatively quiescent cell fraction. Indeed, this may have been a significant contributory factor in the enhanced survival and enrichment of DiD+ cells in MCF-7 and MDA-MB-231 cultures treated with anti-neoplastic drugs. In support of this hypothesis, the expression of anti-apoptotic proteins has frequently been observed in quiescent normal and cancer stem cell populations and has been shown to contribute to their enhanced survival following chemotherapy [
34]. Increased DNA damage repair pathway activation, up-regulation of xenobiotic drug pumps, and elevated enzymatic drug metabolism (particularly by ALDH) have also been reported in quiescent stem cell populations [
35,
36], and we similarly found ALDH activity to be significantly up-regulated within the slow-cycling, chemotherapy-resistant DiD+ fraction of both MCF-7 and MDA-MB-231 cell lines.
Given the slow-cycling and therapy-resistant nature of the DiD+ cell fraction, it appears that there could be an associated between slow-cycling cells and the purported cancer stem cell population. A number of studies across various cancer types have transitively linked quiescence to the cancer stem cell phenotype through retrospective functional validation of putative cancer stem cell populations isolated using cell surface markers [
12,
20,
37]. In contrast, studies in which the prospective identification of a quiescent cell population was undertaken through label retention assays have demonstrated only partial overlap with cell surface marker signatures associated with the supposed cancer-type-specific stem cell population [
7,
8]. Our findings are in agreement with the latter reports; we observed only a modest increase in ALDH activity differentiated the DiD− and DiD+ populations, and DiD+ cells were not enriched for the CD44
+CD24
−/low putative breast CSC marker signature. Notably, the DiD+ sub-population in both MCF-7 and MDA-MB-231 cell lines did show a significant enrichment with cells expressing the CD44
+CD24
+ phenotype. The functional implications of enrichment for CD44
+CD24
+ cells in the DiD+ population and the nature of other cell sub-populations (e.g. DiD+ ALDH
−) remain to be established. While the CD44
+CD24
−/low surface marker signature and high ALDH activity have been widely used to identify the putative CSC population since the pioneering studies of Al-Hajj et al. [
26] and Ginestier et al. [
24], respectively, more recent studies have demonstrated that the CD44
+CD24
−/low and ALDH
+ marker profiles associated with putative breast CSCs identify minimally overlapping, spatiotemporally distinct populations across different breast cancer sub-types [
38]. These studies exemplify that the concept of exclusive tumour cell sub-sets possessing increased nascent capacity for tumour propagation is still evolving. Another very notable example of this situation in breast cancer is that of the MDA-MB-231 cell line; data reported here and by others has shown that in excess of 99% of the total cell population express the CD44
+CD24
−/low marker signature that is supposed to describe highly tumourigenic breast CSC, yet only a minority fraction of the cell population possesses the ability to initiate population re-growth [
39]. Indeed, an often overlooked finding of the landmark study undertaken by Al-Hajj et al. [
26] was that CD44
+CD24
+ breast cancer populations also remained viable and exhibited tumourigenicity in xenotransplantation studies but seemingly possessed reduced proliferative capacity compared to the CD44
+CD24
−/low cells that were taken to represent the stem-like fraction. Based on these findings it seems plausible that the CD44
+CD24
−/low signature was implicitly linked to enhanced tumourigenicity simply due to the rapid expansion of this population coupled with an insufficient follow-up period in mice injected with CD44
+CD24
+ cells, leading to the conclusion that the former exclusively possessed tumourigenic potential. In support of this, a number of other studies have indicated that stem cell activity in breast cancer is not exclusively limited to the CD44
+CD24
−/low phenotype but that CD44
+CD24
+ can be equally tumourigenic in mouse xenograft models, most notably illustrated by Meyer et al. [
40] in oestrogen receptor-negative disease. Moreover, evidence from a more recent study indicates that CD24 status has little bearing on tumourigenic potential in breast cancer and that CD24
−/low status can in fact reduce tumour initiation in some murine models [
41]. Given the potential for a majority of cancer cells or numerous intrinsic sub-populations to display tumourigenicity, we believe that the identification of a slow-cycling and inherently therapy resistant cellular sub-set capable of leading to tumor recurrence is both biologically and clinically significant, independent of a cancer stem cell model.
The use of Vybrant® DiD to identify slow-cycling cancer cell populations is a simple and highly reproducible process that yields an easily identifiable population for isolation and further characterisation. In addition, the use of Vybrant® DiD and analogous dyes offers the additional distinct advantage over other pulse-chase techniques (e.g. BrdU or EdU) of allowing live cells to be isolated, allowing functional studies to be carried out and comparisons to be made between the different populations. One drawback of this method is that sufficient time in culture (several weeks) must elapse before label-retaining cells can be identified. This factor has a significant knock-on effect on it’s prospective utility for primary cultures, in that passaging this material multiple times could potentially alter the characteristics from those of the original tissue. In addition, the capability of the dye-retaining, therapy-resistant cell population to form new tumours can ultimately only be demonstrated by implantation in vivo. On balance, however, this method does offer the possibility of obtaining much more material for characterisation when compared to traditional means of identifying prospective stem-like cells, and the progressive dilution or retention of dye is reliant on a functional cellular phenotype that is independent of the expression of protein markers with poorly understood functional roles in mitotic dynamics.
In summary, the work reported here details a highly reproducible and user-friendly method to enrich, isolate and characterise a live population of chemotherapy-resistant tumour cells based on a functional, mitotically quiescent phenotype. Further characterisation of this population could reveal biological programmes associated with mitotic quiescence and de novo drug resistance and thereby yield novel targets for therapies enabling elimination of the cells responsible for breast cancer recurrence.