Cancer stem cells (CSCs) may contribute to chemoresistant sub-fractions in a variety of malignancies (1–4) and underpin the post-therapeutic recovery potential of tumors. In mouse embryonic stem cell identification, cell surface markers, such as, CD15, CD184, and c-kit are frequently exploited in flow cytometric studies (5), whereas for CSCs in solid tumors various markers have been investigated including ABCG5, ALDH1, CD24 (HSA), CD44 (PGP1), CD90 (THY1), CD133 (promin1), EPCAM, ABCG2, and Hoechst side population (SP) phenotype (3). The ABCG2 transporter (MXR/BCRP) (5, 6) is linked with the CSC phenotype (3, 6–9) contributing to the definition of the pluripotential SP fraction using flow cytometry (10–12) and slide-cytometry (13). SP cells show efficient exclusion of the DNA minor-groove-binding ligand (MGL) Hoechst 33342, detected by changes in the fluorescence emission spectrum of nuclear bound dye (14). Hoechst 33342 has cytotoxic, mutagenic, and DNA damaging properties (15) arising from both minor groove binding and interactions with DNA topoisomerase I (16). An early study recognized the differences between MGLs in their ability to generate topoisomerase I-cleavable complexes (e.g., in order of efficiency: Hoechst 33342 and 33,258 >> distamycin A > berenil > netropsin), and that cleavable complex formation did not correlate with DNA binding efficiency alone (16). In the case of the drug tallimustine (FCE24517), which contains a benzoic acid nitrogen mustard appended to MGL distamycin A, there is evidence of a cooperative DNA binding mode in which noncovalent residence in the DNA minor groove perturbs the helix DNA structure and allows another drug molecule to alkylate the N7 position of guanines located on the periphery of distamycin A consensus sequences (17). The impact of SP-like ABCG2 expression on FCE24517 toxicity is not known.
In murine bone marrow ABCG2 expression appears to be exclusively able to allow the visualization by cytometry of the SP, although the bone marrow of Mdr1a/1b−/− mice display an elevated SP, reversible by the ABCG2 inhibition, suggesting that ABCG2 expression can over-compensate for Mdr1a/1b loss (18).
Some human cancer cell lines have SP fractions containing stem-like cancer cells (19, 20), although the study of SP resistance patterns is complicated by the need for FACS-separation of SP and nonSP fractions and the impact of dye toxicity, dye photosensitization, and clonal instability through asymmetric division (12). ABCG2 expression may reflect fast-cycling tumor progenitors in SP fractions (21), presenting a challenge for therapies that target active S-phase and comprise agents susceptible to active cellular efflux. This is highlighted in the case of the camptothecin class of S-phase targeting antineoplastic agents (e.g., topotecan; TPT) for which the ABCG2 transporter is functional (22–24). TPT acts by stabilizing a covalent topoisomerase I-DNA complex, thereby generating a hindrance to DNA replication fork progression with a subsequent formation of potentially lethal DNA lesions (25, 26). CSC/SP evasion of such cell cycle targeting may involve both modified cell cycle status and “micropharmacokinetic” protection of nuclear targets by changes in the active drug concentrations available within cellular sub-compartments (27). Further, it is not clear how additional features expressed by SP-like cells may augment such resistance to a defined agent.
Although a murine cell line cannot substitute for the likely complexity associated with SP dynamics in tumor cell populations, we hypothesized that we could assess the degree of nuclear protection minimally afforded by a transporter-driven SP phenotype, and its cellular consequences without sorting, by using cells in which the majority of the population expresses Hoechst 33342 resistance. Here we have exploited an established clonal variant murine cell line, previously derived by one-step selection resulting in resistance to the MGLs Hoechst dye 33258 (28) and the more lipophilic 33,342 (29). This approach allows for a direct assessment of the extent to which the SP phenotype recruits co-resistance to MGLs, specifically TPT, given that these agents share noncovalent DNA minor groove-binding and topoisomerase I inhibitory properties (16, 30, 31). This unique murine system obviates the need for SP isolation facilitating microarray analysis for gene expression patterns that contribute to a basal SP-like phenotype. Here we have compared the murine SP profile with the SP pattern operating in the A549 human lung cancer cell line which demonstrates a functional over-expression of the ABCG2-encoded efflux pump (19, 20) and maintains a resident CSC/SP fraction (20).
The results provide evidence of a high specific protection against noncovalent MGLs attributable to ABCG2 function in the murine system, establishing a stable clonal system for the functional definition of the SP-like phenotype and basal drug crossresistance patterns. Hoechst 33342 resistance and evasion of TPT-induced cell cycle arrest was attributable to efflux alone with no evidence of modulation of target availability, endogenous competitors for nuclear DNA binding or augmentation by any ancillary gene expression changes. A novel, nonfluorometric, photo-crosslinking method was used to demonstrate directly that residence of Hoechst 33342 molecules on DNA was greatly limited by efflux, whereas covalent linkage of another minor groove ligand circumvented the impact of efflux. The murine models reveal a strong linkage between ABCG2 function and enhanced TPT resistance via drug exclusion with implications for the predicted efficient evasion of TPT therapy by SP fractions of human tumors.
MATERIALS AND METHODS
Cell Lines and Cell Culture
The thymidine kinase deficient murine fibroblast cell line (Ltk−; (32) was derived from L-M (TK−) clone 1D cells. The Hoechst dye 33258-resistant isolate (HoeR415) is a clonal outgrowth of cells from a Hoechst 33258-treated parental culture and its routine maintenance in RPMI medium has been described previously (28, 29, 33, 34). Resistance was confirmed by a single round of reselection in Hoechst 33342 (1 μM Hoechst 33342 for an 8-day exposure) followed by culture maintenance in dye-free medium. The A549 human lung alveolar cell carcinoma [American Type Culture Collection (ATCC)] cells were grown in DMEM medium (Sigma-Aldrich, Dorset, UK) with 10% FCS, 1 mM glutamine, and antibiotics, and incubated at 37°C in an atmosphere of 5% CO2 in air. Sub-lines of A549 cells (3B11 and 3C5) were obtained by limiting dilution and expansion from single cells for 20–22 population doublings.
Reagents and UVA-Irradiation
Hoechst 33342 (Sigma-Aldrich, Dorset, UK) was prepared as an aqueous stock solution and added directly to the culture media. FCE 24,517 (Tallimustine; a kind gift from Farmitalia Carlo Erba) and topotecan (Hycamptin, SmithKline Beecham/Merck Pharmaceuticals, UK) were prepared in sterile water and the stock was stored at −80°C until required. Fumetiremorgin C (FTC) was obtained from Sigma (UK) and DRAQ5 from Biostatus (Leicestershire, UK). Attached cells, over-layered with PBS in six-well plastic dishes, were UVA-irradiated from below using two black-light-blue fluorescent tubes (General Electric F20T12/BLB tube emitting at 368 nm; 2.5 kJ m−2 at 12 Jm−2 s−1 monitored at 365 nm wavelength) with dishes being supported on a glass sheet.
The K+-SDS precipitation method was used for 14C-deoxycytidine radiolabeled cells incubated with Hoechst 33342 as indicated, washed then UVA-irradiated according to the schedule described. Briefly, the method involves the precipitation of radiolabeled cellular DNA covalently crosslinked to protein, from cell lysates using concentrated salt solutions by a modification of the method described by Rowe et al. (35).
Clonogenic Potential and Timelapse Imaging
Murine cells were plated at a density of 250 cells/well in six-well plates and incubated overnight to allow attachment before drug addition followed continuous exposure or by washing with a return to culture medium. After 14 days incubation, colonies were fixed, stained, and counted. The proliferation of murine cells similarly exposed to TPT was also tracked using timelapse microscopy and events analyzed for the entry of cells into mitosis as described previously (36).
Polyamine Content Analysis
The method was adapted from that described previously for cellular polyamine content (37, 38) and outlined in the Supporting Information. Briefly samples were prepared for reversed-phase high pressure liquid chromatography (RP-HPLC) analysis based on a dansylation procedure (38).
A FACS Vantage flow cytometer was used (Becton Dickinson, Cowley, UK), equipped with a Coherent Enterprise II argon ion laser having 488 nm and multiline UV (350–360 nm; regulated at 30 mW output) outputs (Coherent, Santa Clara, CA). CELLQuest Pro software (Becton Dickinson Immunocytometry Systems) was used for signal acquisition and analysis for Hoechst 33342 uptake and TPT accumulation. DRAQ5 staining and analysis of DNA in live cells (20 μM × 10 min) has been described previously (39). Forward and side scattered light was collected for 10,000 cells, analyzed in linear mode and used to exclude any cell debris by gating.
Side Population (SP) Analysis
Single cell suspensions, with or without FTC pretreatment (10 μM × 1 h) in complete medium supplemented with 20 mM HEPES were incubated with Hoechst 33342 (5 μM). Hoechst 33342 fluorescence was assessed at various time points on the same sample using flow cytometry. Samples were kept at 37°C in a dry-block heater (Techne, Barloworld Scientific, Stone, UK) between time points. Hoechst 33342 fluorescence signals derived from UV excitation were detected via a 660/20 nm band-pass filter (red) and a 405/20 nm band-pass filter (violet) as originally described for the tracking of spectral changes in Hoechst 33342 fluorescence emission by flow cytometry (14, 40).
TPT Uptake Analyzed by Flow Cytometry and Two-Photon (2-P) Laser Scanning Microscopy
TPT loading analyzed by flow cytometry used the system described above and used excitation at both 488 nm (emission at 530/30 nm) and UV. Analysis of TPT excited by the delayed UV beam comprised a long pass 510 nm dichroic and a bandpass barrier filter of 530/30 nm. Two-photon (2-P) laser scanning microscopy was used to image uptake after 10 min TPT exposure as described previously (41).
Mouse (Affymetrix) cDNA Microarrays
Adherent cells in 75 cm2 flasks were treated with TRIZOL for total RNA isolation (42) according to the manufacturer's instructions (Invitrogen, Carlsbad, CA). Samples were processed for quality assessment on Agilent RNA Chips (Agilent, Santa Clara, CA) and analyzed on triplicate Affymetrix mouse chips (MOE430 2.0) using a Affymetrix GeneChip expression profiling service (CBS, School of Medicine, Cardiff). Data were analyzed using GeneSpring software. Details in Supporting Information.
Over-Expression of ABCG2 in a Hoechst-33342 Resistant Cell Line
The variant murine cell line HoeR415 is resistant to the AT-specific DNA minor groove ligand (MGL) Hoechst 33342 with crossresistance to the noncovalently binding MGL (distamycin A) but not a covalently binding MGL (anthramycin) or DNA intercalators (adriamycin, mitoxantrone, and mAMSA) or a nonintercalating topoisomerase II poison (VP16-213) (33). Figure 1A shows the >10-fold resistance to Hoechst 33342 of HoeR415 cells, for continuous low-dose dye exposure, relative to the Ltk− parent. HoeR415 showed no difference in the ability of the intercalating agent DRAQ5 (39) to accumulate in cells and bind to cellular DNA, also revealing a similar DNA content distribution to that of Ltk− cells (Fig. 1B) and equivalent nuclear binding sites.
Flow cytometry also revealed a moderate reduction in accumulation of Hoechst 33342 in HoeR415 cells but a greatly enhanced rapidity of clearing upon removal of exogenous dye (Fig. 1C). An SP-like phenotype for HoeR415 cells was apparent because there was a relative lag (Fig. 1D) in red fluorescence acquisition (14) as accumulation of nuclear dye is opposed by ongoing clearing from the cell (10). An initial microarray survey was undertaken to assess whether HoeR415 cells showed constitutive differences in the expression of ATP-binding cassette transporter genes whose products include those with potential to participate in the exclusion of xenobiotic molecules (9) (Fig. 1E). The results showed unremarkable differences between HoeR415 cells compared with Ltk− cells except for the enhanced expression of ABCG2 in the variant cell line.
Hoechst 33342 Resistance Confers an SP-like Phenotype
The SP-like and functional ABCG2 dependency of the resistance phenotype of HoeR415 cells is shown in the bivariate red/violet Hoechst 33342 fluorescence plots in Figure 2. The A549 cell line served as a positive control (19). Heterogeneity for transporter function within the parent A549 population was apparent, yielding an SP fraction of ∼6%. Two clones, derived from the A549 parent line, were used to reduce the impact of heterogeneity and were found to yield a minimal SP of ∼0.6% (clone 3B11) and a well-defined SP of ∼15% (clone 3C5), respectively. The A549-derived data provided a region of interest for the SP fraction no longer occupied when cells were pretreated with the ABCG2 inhibitor fumetiremorgin (FTC) (8). The intersection of the quadrant overlays on the contour plots in Figure 2D indicates the position of the SP region as defined by HoeR415 cells. This region also mapped to the SP regions indentified in A549 controls. The majority of HoeR415 cells (up to 80%) cells, compared with <1% of Ltk− cells, demonstrated the SP-like phenotype. The results showed that FTC also enhanced cellular uptake of Hoechst 33342 in HoeR415 cells, whereas Ltk− cells showed no significant FTC-modifiable Hoechst 33342 uptake response.
Hoechst 33342 Resistance is Associated with Reduced Residence of Dye Molecules on Cellular DNA
Because Hoechst 33342-DNA binding is noncovalent, we have used dye-enhanced photo-crosslinking of protein to DNA (43) to determine directly whether the residence of dye molecules on DNA in contact with protein was reduced in HoeR415 cells. We initially excluded the possibility that any reduced residency was a reflection of intracellular competition for MGL-DNA binding or chromatin-access associated with abnormal polyamine content (44) (Fig. 3A).
Pretreatment with Hoechst 33342 photosensitized both Ltk− and HoeR415 cells to DNA–protein crosslinking, whereas HoeR415 cells showed a more rapid loss of crosslinking potential upon incubation in dye-free medium (Fig. 3B). This is consistent with a reduced residency of dye molecules on DNA, at locations permissive for DNA–protein interactions, during post-treatment recovery. Clonogenic analyses confirmed that rapid loss of dye-enhanced photo-crosslinking potential also conferred a survival advantage to HoeR415 cells (Fig. 3C).
MGL Resistance Pattern and Evasion of TPT-Induced Cell Cycle Arrest
The polyamide FCE24517 (a benzoyl nitrogen mustard derivative of desformyldistamycin) binds both reversibly and irreversibly to DNA with cytotoxic activity in human and murine tumor cell lines (45). The variant showed the same sensitivity as the parental cell line irrespective of the FCE24517 treatment schedule (Fig. 4A). However, Figure 4B shows that HoeR415 cells were resistant to continuous exposure to the MGL TPT. Timelapse microscopy was used to track the delivery of cells to mitosis (36) (Figs. 4C and 4D), an approach that allows for the identification of individual events within the same population and, therefore, has a high resolution capacity for changes with time that can be achieved by conventional cell cycle distribution analyses. HoeR415 cells were found to be resistant to the inhibitory effects of low-dose continuous TPT exposure. The sensitive timelapse method detected a small inflection (arrowed in Fig. 4D) in the response of HoeR415 cells, compatible with a minor delay being associated with cells that have traversed a full S-phase before delayed delivery to mitosis although recovery is apparent, given by the acceleration in rate after the inflection point to match the control response. HoeR415 also showed a greater fraction of cells escaping arrest in the first cycle following a high-dose pulsed TPT exposure after 22 h (0.3 vs. 0.1 cumulative events). We excluded the possibility that evasion of arrest by HoeR415 cells reflected reduced levels of DNA topoisomerase I in S-phase (Supporting Information Table 1). The results suggested that the selective resistance of HoeR415 to MGLs primarily arises from reduced residency on DNA, driven by MGL efflux in the majority of cells. Minor groove binding that leads to adduct formation or direct base damage precludes resistance.
Reversal of Enhanced TPT Nuclear Exclusion and Cytotoxicity Resistance
Micropharmacokinetic considerations (46) suggest that cellular concentrations of TPT might be greatly influenced by ongoing ABCG2 function given the low affinity of TPT than Hoechst 33342 for minor groove binding. As predicted, HoeR415 revealed significant resistance to the initial loading of TPT irrespective of the TPT excitation wavelength used (Fig. 5A). The progressive loss of signal at >10 min in Ltk− can be attributed to ongoing lactone ring hydrolysis of extracellular drug acting to reduce the pool of the active drug available for passive uptake (46). Two-photon excitation laser scanning microscopy of TPT in single murine cells (Fig. 5B) provided direct evidence of the significant protection of the nuclear compartment for TPT binding/residence in the variant. A functional and primary involvement of ABCG2 in the TPT resistance of HoeR415 cells was supported by the ability of FTC to enhance cellular uptake of TPT in HoeR415 cells (Fig. 5C) at least to the extent observed for A549 cells. Ltk− cells showed a greatly reduced FTC-modifiable TPT uptake response suggesting a basal level of dependence on ABCG2 function for TPT exclusion. The results shown in Figure 5D revealed the ability of FTC to abolish HoeR415 resistance to growth inhibition by both Hoechst 33342 and TPT.
Gene Expression and the Murine SP Phenotype
A wide microarray gene expression data survey was undertaken to reveal any linked expression with previously recognized CSC markers or genes products that could act to augment the ABCG2-linked TPT resistance shown by HoeR415. Figures 6A and 6B show the gene symbol ranking for >twofold over- or under-expression. Gene expression changes that could modify cellular capacity to cope with chronic stress were: Mgst1 (HoeR415/Ltk− = 2.519) encoding microsomal glutathione S-transferase 1; Plk2 (HoeR415/Ltk− = 0.353) encoding polo-like kinase 2; and Nnt (HoeR415/Ltk− = 0.0597) encoding nicotinamide nucleotide transhydrogenase. A potential candidate for a CSC-linked co-marker was Aldh1a1 (HoeR415/Ltk− = 108.8) encoding aldehyde dehydrogenase Family 1, subfamily A1. Augmentation of TPT resistance via co-expression of ALDH was excluded for the murine system—further confirmed by the ALDH inhibitor disulfiram (47) showing no effect on the reduced TPT uptake shown by HoeR415 compared with Ltk− (data not shown).
The HoeR415 variant showed SP-like characteristics apparent in the majority of cells and a significantly reduced residence of MGLs on DNA. Protection is driven primarily by a selective and functional over-expression of the drug transporter ABCG2, identified in both microarray and inhibitor studies. The Hoechst 33342 clearing observed in HoeR415 was at a similar level to that achieved by the SP fraction in human A549 cell populations. Enhanced efflux recruited a moderate twofold clonogenic resistance to TPT, some fivefold less than the relative resistance achieved for continuous Hoechst 33342 exposure. Resistance permitted complete evasion of early cell cycle arrest associated with the TPT targeting of S-phase at low-drug concentrations—both Hoechst 33342 and TPT resistance being FTC-reversible. HoeR415 showed that a dramatic reduction in initial TPT loading is also reversible by FTC and again similar in extent to that achieved in A549 cell populations. The apparent stability of the HoeR415/SP phenotype allows for whole population analyses, overcoming the problem of generation of a nonSP sub-fractions shown to occur upon isolation of tumor-derived SP cells (12).
HoeR415 and parental cells had similar responses to Hoechst 33342 presensitization to UVA radiation (generating DNA–protein crosslinks; Fig. 3) and FCE24517 (a base-damaging agent; (17) indicating that the intrinsic responses to a range of genotoxic damage remains unchanged in the variant. UVA sensitization was lost more rapidly by HoeR415 at both the clonogenic survival and this paralleled the loss of Hoechst 33342 sensitization to UVA-induced DNA–protein crosslinking. The findings strongly suggest that dye residence on DNA limits MGL sensitization mechanisms. Hoechst 33342 sensitization to UVA-induced DNA–protein crosslinking paralleled the fluorometrically tracked rapid loss of dye—providing a nonfluorescence based identification of SP-like cells and a means of selection of SP cells through dye-targeted UVA-toxicity for nonSP cells. The unique murine cell system removes the need to fractionate populations (12) with implications for rapid drug screening and could provide a functional benchmark for the quantification of SP fractions by flow cytometry.
Microarray analyses revealed 56 recognized gene transcipts were over-expressed and 17 under-expressed in HoeR415 relative to parental cells. Searches were undertaken for candidate genes, whose change expression in HoeR415 might impact directly on drug resistance. Other than ABCG2, we identified ALDH, MGST1, Plk2, and Nnt. Increased MGST1 could contribute to the drug resistance phenotype via direct detoxification effects with indirect evidence previously reported of a positive correlation for camptothecin resistance (48). A minor reduction in Plk2 expression observed in HoeR415 cells is unlikely to impose resistance to the S-phase targeting agent TPT because silencing Plk2 sensitizes cells to replicative stress (49). Likewise, the reduction in Nnt expression observed in the variant is unlikely to contribute to the SP-like phenotype because this encoded mitochondrial inner membrane protein normally affords indirect protection against mitochondrial oxidative stress (50).
The increase in gene expression in HoeR415 cells of ALDH1A1 is intriguing because it is a potential candidate for a CSC-linked marker. CSCs may express collateral drug resistance through the aldehyde dehydrogenase (ALDH) (3). The enzyme family oxidizes a wide range of reactive and toxic aldehydes to their corresponding carboxylic acids (51) and facilitates resistance to cyclophosphamide, reversible by the ALDH inhibitor disulfiram (47). There is currently no evidence to functionally link the operation of the ABCG2 transporter with that of other CSC-associated markers in augmenting innate resistance to a defined antineoplastic agent. Importantly co-expression of ALDH and ABCG2, which can occur in stem-like cell fractions (3) does not appear to augment TPT export in the murine system—a view supported by our preliminary studies that show TPT accumulation in HoeR415 is not blocked by the ALDH inhibitor disulfiram and despite the additional potential for disulfiram metabolites to inactivate P-glycoprotein (52). To be able to establish that this murine cell line can be a surrogate for human SPs, comparative gene expression analyses and pharmacologic studies would be required profiling different human tumors and cell lines. However, the basal SP-like phenotype described here may impose selective MGL resistance due to the degree of expression of the transporter or potentially a mutated form. A recent report has noted that acquisition of moderate levels of enhanced transporter expression in mouse models, below those levels found in some normal tissues, can be sufficient to cause doxorubicin resistance (53).
We suggest that the basal ABCG2-driven efflux process, in underpinning the micropharmacokinetic SP phenotype, gives rise to selective resistance to noncovalent binding MGLs because they are preferred substrates and can dissociate from intracellular targets. Our results show that complete first cycle evasion of TPT action can be achieved via enhanced efflux despite the exquisite sensitivity of S-phase, with the implication that low-dose continuous exposure would favor CSC survival if primarily determined by a basal SP-like process. Higher levels of functional ABCG2 activity may be required to give rise to additional resistance for substrates, such as, mitoxantrone, that show reduced dissociation capacity from binding at intracellular targets (54) biasing their location at target. Effectively, a hierarchy of micropharmacokinetic resistance is imposed on the SP fraction together with a pharmacodynamic impact that varies according to the drug potency during residency at target. In a CSC model, the associated loss of SP characteristics during asymmetric division will generate cohorts of cells with changing drug resistance profiles with respect to different agents, even when the impact of a single transporter is considered. Intracellular stability of the α-hydroxylactone E-ring of camptothecins in SP cells may be critical if ring-opened products are favored substrates for ABCG2-mediated export. Enhanced SP-targeting using the camptothecins could be achieved by the deployment of stabilized E-ring camptothecin keto analogs (e.g., S38809 and S39625) given that such analogues can retain topoisomerase I inhibition without being substrates for either the ABCB1 (multidrug resistance-1/P-glycoprotein) or ABCG2 drug efflux transporters (55). Furthermore, the use of agents actively transported by human ABCB1A/ ABCB1B, but not by ABCG2 (BCRP1) (56), could be used to target CSC fractions carrying intrinsically drug resistant fractions, as recently demonstrated for the polyether organic anion salinomycin (57).
EF was supported by an EPSRC Industrial Case Studentship award in partnership with Biostatus Ltd (UK). The authors thank Sally Chappell, Amy Brook, and Janet Fisher for technical support.