Introduction

Drug resistance, either intrinsic or acquired, is omnipresent in the clinical treatment of cancer, limiting durable therapeutic benefits and even accelerating tumor recurrence or metastasis (Szakacs et al, 2006; Andrei et al, 2020; Dallavalle et al, 2020). Multidrug resistance, one of the most difficult problems occurring during chemotherapy, is commonly related to the expression of ATP-binding cassette (ABC) transporters, especially transporter—multidrug resistance protein 1 (MDR1) encoded by ABCB1 (Hall et al, 2009). Along with the emergence of appropriate analytical tools, advanced medicinal techniques, and multidisciplinary research, ample evidence not only identifies high expression of MDR1 portending a poor response to chemotherapy and adverse outcomes in patients with cancers but also confirms the necessity of applying new drug or drug delivery protocols to prevent cancer MDR in the clinic (Fan et al, 2017).

Unfortunately, the growing realization that cancer cells, rather than being either sensitive or resistant, can be dynamic and transient in nature within the context of treatment is detracting from such studies (Qin et al, 2020). There is actually, a particular state lying between sensitivity and resistance, termed the “drug-tolerant persister (DTP) state,” in which a cell population is endowed with a dormant, slow-cycling state and a stem-like signature (Shen et al, 2019, 2020). The concept of DTP originates from an early observation of bacterial response to antibiotics, where the existence of residual bacteria exposed to antibiotics is due to non-genetic variations and resumption of their initial characteristics upon interruption of treatment (Balaban et al, 2019). In the context of cancer, tumor cells share a similar situation in which drug resistance is, in a similar fashion to antibiotic resistance, driven by epigenetic inheritance of variant gene expression patterns. This could result in not only inhibiting the therapeutic efficacy but also providing a reservoir for further evolution (Shen et al, 2019, 2020). Given that this DTP state has been hypothesized to be part of an alternative approach toward a bona fide drug resistance mechanism, this line of research is attracting considerable attention. In addition to the recognition that DTP could serve as a target for therapy (i.e., incorporation of an epigenetic modulator), studies focusing on DTP have deepened our understanding of the mechanisms driving DTP. This encourages the development of methods aimed at disposing of these cells, including sustaining a harmless dormant state, and reactivating proliferation to enhance response to anti-proliferative drugs to eradicate them (Recasens & Munoz, 2019). However, there remains some confusion on how best to target DTP cells, indicating the requirement for a further understanding of both DTP itself and DTP-mediated drug tolerance in order to devise countermeasures against these persisters.

To date, multiple studies have emphasized the impact of stress response on persister generation, and several teams have highlighted the capacity to adapt to oxidative stress as a common characteristic of cancer cells in a DTP state (Raha et al, 2014; Sahu et al, 2016; Hangauer et al, 2017; Anand et al, 2019; Dhimolea et al, 2021). Based on this characteristic, increasing numbers of drug targets involved in oxidative stress defense of DTP have been investigated. Among them the phospholipid glutathione peroxidase 4 (GPX4) is crucial for the survival of cancer cells in a therapy-resistant state. In this context, subsequent work has shown that targeting GPX4-dependent oxidative stress defense can almost completely eradicate persister cells by induction of ferroptosis, an oxidative cell death, suggesting a potential treatment strategy by disturbing the redox homeostasis (Hangauer et al, 2017). Additionally, activated NF-E2-related factor 2 (NRF2), detected in HER2-inhibited persistent breast cancer, has been found to drive the re-establishment of redox homeostasis in a glutathione metabolism-dependent manner, thereby promoting the reactivation of dormant tumor cells. This reactivation triggered by NRF2 signaling can be prevented by glutaminase inhibition. This has also been shown to impair the growth of recurrent tumors with high levels of NRF2, suggesting a novel approach to treat NRF2 high dormant and recurrent cancer (Fox et al, 2020). Such findings confirm persister population adaption to therapy-induced oxidative stress. Furthermore, MDR cancer cells might raise a robust antioxidant system for resisting oxidative stress caused by chemotherapeutic agents (Trachootham et al, 2009). We therefore questioned whether persister cancer cells arising from MDR cancer cells prefer to orchestrate the evolutionarily conserved antioxidant system against treatment.

In this study, we investigated how MDR cancer cells characterized by overexpression of MDR1 underwent a combinatorial therapy (chemotherapeutic agent with MDR1 inhibitor verapamil)-induced DTP state. To gain insight into the alterations of gene expression profile in the course of non-mutationally acquired resistance, we performed RNA-seq comparing MDR persister cells to MDR cancer cells. We also investigated the function of screened genes in MDR cancer cells with a DTP status. This revealed that NPC1L1, an important regulator for redox homeostasis promoting uptake of vitamin E which can interact directly with lipid peroxyl radicals, thus preventing oxidative stress, was highly expressed in DTP. By adding the NPC1L1 inhibitor ezetimibe into the combinatorial therapy, the function of NPC1L1 on vitamin E absorption was compromised and additional cell death was observed as a consequence of macropinocytosis induction. Mechanistically, our results demonstrated a link between both NRF2 transcriptional activation and decreased DNA methylation with NPC1L1 expression. Using a nanomedicine approach to alleviate side effects of verapamil in vivo, we further confirmed the anti-tumor effect of a triple-combinatorial therapy strategy (a combination of chemotherapeutic agents, verapamil, and ezetimibe) that prevented tumor recurrence in vivo. Together, our study progresses our understanding of MDR persistence from a redox perspective, repositioning the potential of MDR therapy for treating cancer cells.

Results Oncotherapy induces the transformation of MDR cancer cells into a DTP state

To elucidate the molecular mechanisms underlying the established drug-resistant cancer cell models (Du145TXR resistant to taxol; MCF-7ADR resistant to adriamycin) (Appendix Fig S1A and B), we first examined the expression of MDR1 (also known as p-glycoprotein), which has been extensively studied in the context of drug-resistant phenotype (Gouaze et al, 2005; Takeda et al, 2007; Robey et al, 2018) and confirmed the marked upregulation of its protein expression (Appendix Fig S1C). We further determined the role of MDR1 played in regulating the MDR phenotype using rhodamine123 (a substrate of MDR1) and MDR1 inhibitor verapamil. Fluorescence analysis indicated that MDR1 high expressing cancer cells showed increased efflux of rhodamine123 compared to control cancer cells, and verapamil could partially reverse this phenotype (Appendix Fig S1D). To investigate which combination of chemotherapeutic agent and verapamil contributed most to the anti-cancer ability and synergistic effects, we examined cell viability using the Chou–Talalay method (combination index (CI) < 1 representing synergism) in different drug combination-treated MDR cancer cells. Our results showed that 50 μM verapamil combined with 20 nM taxol or 200 nM adriamycin exhibited beneficial anti-cancer effects in Du145TXR cells or MCF-7ADR cells (Appendix Fig S1E and F). Furthermore, the proliferation of MDR cancer cells was significantly inhibited under the combination treatment, as evidenced by reduced colony formation (Appendix Fig S2A and B). Consistently, knockdown of ABCB1 clearly decreased both cell viability and clonogenic growth capacity in chemotherapy-treated MDR cancer cells (Appendix Fig S2C–G). Together with previous reports, our data support MDR1 as the key factor driving the MDR phenotype of these cancer cells and establish that targeting MDR1, at least in part, enables MDR cancer cells to recover chemosensitivity.

Notably, we observed the occurrence of residual MDR cancer cells following combination treatment. These cells were similar to those described for drug-tolerant persister (DTP) cells, which survive cytotoxic exposure via reversible and non-mutational mechanisms. To evaluate whether these residual MDR cancer cells were associated with the DTP state, we examined the cell viability and observed morphological changes in Du145TXR cells and MCF-7ADR cells following combination treatment (50 μM verapamil/20 nM taxol or 200 nM adriamycin). Surprisingly, compared with MDR cancer cells (MCs), which had not been treated with the combinatorial therapy, residual MDR cancer cells (RMCs) showed dramatic morphological alterations following combination treatment for 3 days and obvious tolerance to a second exposure of treatment applied 1, 3, 6, and 9 days after treatment withdrawal as shown by a cell viability assay (Fig 1A–D). Subsequently, a long-term “drug holiday” (30 days) allowed residual MDR cancer cells to regrow (regrown cells, RCs) and resulted in re-acquisition of sensitivity to combination treatment (Fig 1E). The reversibility of further drug resistance in residual MDR cancer cells implies that oncotherapy drives MDR cancer cells into a DTP state, termed as MDR persister cells (MPCs). To compare the differences between MPCs and persister cancer cells (PeCs), we used a similar protocol to generate DTP cells in control cells (CCs) (Fig EV1A). Our results indicated that prolonged drug exposure in control cells for > 3 days could generate a small population of DTP cells (Fig EV1B–D). In addition, DTP cells progressively re-acquired drug sensitivity following withdrawal of chemotherapy and eventually became non-differential compared with control cells by Day 6 as evidenced by cell viability (Fig EV1E and F). Accordingly, we converted the above data obtained by cell viability assays (Figs 1C–E and EV1C–F) into a drug-tolerant rate. As shown in Fig 1F and G, MDR cancer cells can further evolve upon treatment compared with control cells as evidenced by faster induction and a longer duration of the DTP state. Taken together, our results demonstrate that MDR cancer cells can transform into a DTP state resembling control cells, possibly leading to a worse outcome such as modest MDR reversal and tumor recurrence.

Figure 1. Oncotherapy induces the transformation of MDR cancer cells into a DTP state

A. Schematic of residual MDR cell generation and subsequent analyses. The circular arrow represents the DTP state. B. Phase contrast images of the morphological changes in MDR cells, residual MDR cells, and regrown cells for Du145TXR or MCF-7ADR. Red circles mark magnified areas (bottom). Scale bars, 50 μm (low magnification images). C, D. Cell viability of residual MDR cells (RMCs) and MDR cells (MCs) of (C) Du145TXR or (D) MCF-7ADR cells using the same combination treatment for 24 h on the indicated days. Student’s t-test was used to analyze statistical differences. Mean with ± SD. E. Cell viability of regrown cells (RCs) and Du145TXR (left) or MCF-7ADR (right) cells with the same combination treatment for 24 h. Student’s t-test was used to analyze statistical differences. Mean with ± SD. F, G. Drug-tolerant rate of (F) Du145/Du145TXR (20 nM taxol/20 nM taxol plus 50 μM verapamil) or (G) MCF-7/MCF-7ADR (200 nM adriamycin/200 nM adriamycin plus 50 μM verapamil) cells treated with drugs on the indicated days. Drug-tolerant rate defined by comparing cell viability followed treatment, RMCs vs. MCs or CCs vs. PeCs, > 1 represents drug tolerant. Induction time indicates the time of entering DTP state; and sustaining time indicates the time of maintaining DTP state. Blue colors represent control groups; Red colors represent MDR cancer cells. Mean with ± SD.

Data information: Results are representative of three independent experiments.

Source data are available online for this figure.

Click here to expand this figure.

Figure EV1. Control cancer cells transform into DTP states in response to chemotherapy

A. Schematic of drug persister cancer cell generation and subsequent analyses. B. Phase contrast images of the morphological change in Du145 or MCF-7 cells followed by chemotherapy treatment for 72 h. Red circles mark magnified areas (bottom). Scale bar, 50 μm (low-magnification images). C, D. Cell viability of control cancer cells (CCs), 3 days treated control cancer cells (CCs-3), and persister cancer cells (PeCs) of (C) Du145 or (D) MCF-7 cells treated with the indicated concentrations of taxol (TAX) or adriamycin (ADM) for 24 h. Mean with ± SD. E, F. Cell viability of CCs, PeCs, 3 days drug withdrawn PeCs (PeCs-3), and 6 days drug withdrawn PeCs (PeCs-6) of (E) Du145 or (F) MCF-7 cells treated with the indicated concentrations of taxol (TAX) or adriamycin (ADM) for 24 h. Mean with ± SD.

Data information: Results are representative of three independent experiments.

Source data are available online for this figure.

Upregulation of NPC1L1 supports cell survival in response to cytotoxic stress in MPCs

Given the well-known characteristics of DTP cells, such as induction of cell cycle arrest, increased expression of stemness markers, as well as activation of epithelial–mesenchymal transition (EMT) (Shen et al, 2019, 2020), we investigated whether MPCs had similar phenotypes. Our data indicated that MPCs seemed to be in a state of cell cycle arrest, as evidenced by significant increases in p27 and p21 expression (Appendix Fig S3A). In addition, we further performed flow cytometry analysis which demonstrated that cell cycle arrest at G0/G1 phase was induced in MPCs, showing a non-proliferative or slowly proliferative state (Appendix Fig S3B–D). We also observed the upregulation of stemness markers ALDH1A1, Oct-4A, Sox2, KLF4, and CD44 in MPCs (Appendix Fig S3A). Consistently, sphere formation assays showed that MPCs enhanced self-renewal ability due to the increased stemness, as evidenced by a marked increase in both the number and size of MPCs spheres (Appendix Fig S3E–G). Interestingly, cell viability and colony formation assays revealed that ALDH1A1 inhibitor CM10 was most selectively lethal to MPCs cells and sensitized MCs to the combination treatment consistent with previous persister cell studies (Raha et al, 2014; Appendix Fig S3H–J). In addition to these characteristics, downregulation of epithelial marker E-cadherin and upregulation of mesenchymal marker Vimentin were detected by Western blotting, implying induction of EMT (Appendix Fig S3A). Taken together, these data support our model for a similar phenotype to persister cells.

The attempt to target persister cells, especially MDR persister cells, as a therapeutic approach for overcoming the clinical bottleneck has long been enigmatic. To identify cellular processes or key effectors that may play a role in MPCs survival, RNA sequencing (RNA-Seq) was conducted by comparing the gene expression profiles between MCs and MPCs, through which 3421 and 4000 differentially expressed genes (DEGs) were filtered for statistical significance (|fold change| ≥ 2, P ≤ 0.05) in Du145TXR and MCF-7ADR cell lines, respectively (Fig 2A). These upregulated or downregulated gene sets were then grouped by pathway enrichment analysis using the ConsensusPathDB Bioinformatic tool (http://cpdb.molgen.mpg.de/). The results suggested that various genes, for example, those involved in cell cycle arrest and Myc pathway inactivation, which had been widely reported in previous studies (Dhimolea et al, 2021; Rehman et al, 2021), were significantly enriched in Du145TXR or MCF-7ADR cells (Figs 2A and EV2A–E). In order to further identify the shared patterns of expression across two cell lines, 109 consistently upregulated or downregulated genes in MPCs compared with their corresponding MCs (between Du145TXR vs. persister Du145TXR and MCF-7ADR vs. persister MCF-7ADR) were identified (Fig 2B–D). Given that cancer cells in the DTP state are dormant with a necessarily global reduction in protein neosynthesis, we mainly focused on translationally upregulated mRNAs that might be involved in persistence. As shown in Fig 2D, Niemann-Pick C1 Like 1 (NPC1L1) was the top scoring candidate of the upregulated genes. Notably, NPC1L1 is a direct target of anti-lipemic agent ezetimibe, and moreover, a promising target for clinical treatment (Bays, 2002).

Figure 2. Upregulation of NPC1L1 supports cell survival in response to cytotoxic stress in MPCs

A. Volcano plot showing expressed genes identified by RNA-seq of MCs vs. MPCs in Du145TXR or MCF-7ADR. Dotted lines represent the screening criteria (|fold change| ≥ 2, P ≤ 0.05). B. Venn diagram indicating the overlaps of expressed genes in (A). C. Volcano plot revealed 109 consistently expressed genes identified by RNA-seq of MCs vs. MPCs in Du145TXR and MCF-7ADR. Dotted lines represent the screening criteria (|fold change| ≥ 2, P ≤ 0.05). D. Heat map based on scoring rank indicated consistently expressed genes identified by RNA-seq of MCs vs. MPCs in Du145TXR and MCF-7ADR. E. qRT-PCR analysis of NPC1L1 in MCs and MPCs of Du145TXR or MCF-7ADR. Student’s t-test was used to analyze statistical differences. Mean with ± SD. F. Immunoblotting of NPC1L1 in MCs, MPCs, and RCs of Du145TXR or MCF-7ADR. G. Immunoblotting of NPC1L1 in Du145TXR or MCF-7ADR cells transfected with siNPC1L1 or siScramble. H, I. Colony formation assay and quantification of (H) Du145TXR or (I) MCF-7ADR cells transfected with siNPC1L1 or siScramble followed by treatment with indicated agents. One-way ANOVA was used to analyze statistical differences.

Data information: Results are representative of three independent experiments.

Source data are available online for this figure.

Click here to expand this figure.

Figure EV2. Signaling pathways identified in MPCs by performing RNA-seq

A. KEGG/PID pathway analysis showed the changed signaling pathways of MCs vs. MPCs in Du145TXR and MCF-7ADR. B, C. GSEA indicated GO Myc targets of MCs vs. MPCs in (B) Du145TXR and (C) MCF-7ADR. D, E. GSEA indicated GO cell cycle and DNA replication of MCs vs. MPCs in (D) Du145TXR and (E) MCF-7ADR.

To validate the RNA-seq results presented above, we confirmed the upregulated expression of NPC1L1 in MPCs compared with their corresponding MCs by quantitative real-time PCR analysis (Fig 2E). In addition, NPC1L1 was examined in MCs, MPCs, and RCs by Western blotting, which demonstrated that protein levels of NPC1L1 were significantly increased in MPCs and returned back to their basal levels in RCs, consistent with the reversible drug-resistant capacity of MDR cancer cells during the treatment (Fig 2F). We hypothesized that NPC1L1 inhibition in combination with chemotherapy agents/verapamil might usefully improve outcomes of current drug-resistant reverse therapy in MDR cancer cells by directly targeting MDR persister cells. We therefore evaluated the efficacy of this strategy using a colony formation assay and found that siRNA-mediated silencing of NPC1L1 resulted in a further decrease in colony formation in chemotherapy agents/verapamil-treated MDR cancer cells (Fig 2G–I). Collectively, these results reveal that NPC1L1 is a crucial regulator of DTP state and a potentially rational target for tumor patients with MDR disease in the clinic.

NPC1L1 compromises oxidative stress in MDR cancer cells by promoting vitamin E uptake during chemotherapeutic agents/verapamil treatment

Based on the current knowledge of MDR cancer cells, these special populations are acknowledged to have specific dependency on the antioxidant system (Trachootham et al, 2009; Viswanathan et al, 2017; Wang et al, 2018a,b). In addition, a number of studies have recently reported targeting persistence based on redox modulation (Sahu et al, 2016; Hangauer et al, 2017; Takahashi et al, 2018; Fox et al, 2020). Accordingly, these findings directed our attention toward whether NPC1L1 was involved in oxidative stress defense mechanisms in MPCs. To test whether MDR cancer cells underwent oxidative stress during the combination treatment, reactive oxygen species (ROS) analysis using the fluorescence probe DCFH-DA was performed. The results revealed that the intracellular ROS levels rapidly accumulated in 12 and 36 h following combination treatment, while pharmacological inhibition of NPC1L1 using ezetimibe (25 μM) further enhanced combination treatment-induced ROS accumulation. This tended to be sustained at a high level for a relatively long time (Fig 3A–C). We next examined ROS levels in Du145TXR and MCF-7ADR cells following triple-combinatorial treatment with or without the antioxidants (5 mM N-Acetyl-L-cysteine (NAC) and 50 μM vitamin E (VE)) by investigating DCF fluorescence intensity with flow cytometry. As shown in Fig EV3A–C, triple-combinatorial treatment induced marked accumulation of ROS levels, while this accumulation could be compromised by NAC or vitamin E. Consistently, knockdown of NPC1L1 by siNPC1L1 significantly enhanced combination treatment-induced ROS accumulation. In contrast, either NAC or vitamin E visibly decreased ROS levels following combination treatment in siNPC1L1-transfected cancer cells (Fig EV3D–F). Moreover, we also observed that ezetimibe induced an obvious upregulation of the ratio of oxidized glutathione (GSSG) to reduced glutathione (GSH) in combination treatment-stimulated cancer cells (Fig EV3G and H). As expected, NPC1L1 knockdown induced a similar outcome to ezetimibe (Fig EV3I and J). Importantly, either ezetimibe or siNPC1L1-mediated knockdown led to both remissive cell viability and colony formation in combination treatment. In contrast, NAC compromised triple-combinatorial treatment-induced growth inhibition (Figs 3D–G and EV3K and L). These results indicated that NPC1L1 might contribute to ROS neutralization which was necessary for MDR cancer cell survival.

Figure 3. NPC1L1 compromises oxidative stress in MDR cancer cells by promoting vitamin E uptake during chemotherapeutic agents/verapamil treatment

A–C. Flow cytometric analysis of ROS levels in Du145TXR or MCF-7ADR cells treated with indicated agents for 12 and 36 h. (A) Representative images and quantification of ROS in (B) Du145TXR or (C) MCF-7ADR cells are shown. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. D, E. Cell viability of (D) Du145TXR or (E) MCF-7ADR treated with indicated agents for 72 h. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. F, G. Colony formation assay and quantification of (F) Du145TXR or (G) MCF-7ADR cells treated with indicated agents. One-way ANOVA was used to analyze statistical differences. H, I. Cellular vitamin E analysis of medium supernatant in MCs and MPCs of (H) Du145TXR or (I) MCF-7ADR cells treated with 50 μM vitamin E (VE) followed by treatment with or without 25 μM ezetimibe for 12 h. Student’s t-test was used to analyze statistical differences. Mean with ±SD. J, K. Cell viability of (J) Du145TXR or (K) MCF-7ADR cells cultured in normal or serum-free medium followed by treatment with indicated agents for 72 h. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. L, M. Colony formation assay and quantification of (L) Du145TXR or (M) MCF-7ADR cells treated with indicated agents. One-way ANOVA was used to analyze statistical differences.

Data information: Results are representative of three independent experiments.

Click here to expand this figure.

Figure EV3. Inhibition of NPC1L1 induces oxidative stress-mediated cell death

A–C. Flow cytometric analysis of ROS levels in Du145TXR or MCF-7ADR cells treated with indicated agents for 24 h. Chemo represents 20 nM taxol or 200 nM adriamycin for Du145TXR or MCF-7ADR, respectively. (A) Representative images and quantification of ROS in (B) Du145TXR or (C) MCF-7ADR cells are shown. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. D–F. Flow cytometric analysis of ROS levels in Du145TXR or MCF-7ADR cells transfected with siNPC1L1 or siScramble followed by indicated treatment for 24 h. Chemo represents 20 nM taxol or 200 nM adriamycin for Du145TXR or MCF-7ADR, respectively. (D) Representative images and quantification of ROS in (E) Du145TXR or (F) MCF-7ADR cells are shown. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. G, H. GSSG/GSH ratio measurement of (G) Du145TXR or (H) MCF-7ADR cells treated with indicated agents for 24 h. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. I, J. GSSG/GSH ratio measurement of (I) Du145TXR or (J) MCF-7ADR transfected with siNPC1L1 or siScramble followed by indicated treatment for 24 hours. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. K, L. Cell viability of (K) Du145TXR or (L) MCF-7ADR cells transfected with siNPC1L1 or siScramble followed by treatment with the indicated agents for 72 h. One-way ANOVA was used to analyze statistical differences. Mean with ± SD.

Data information: Results are representative of three independent experiments.

Previous studies have demonstrated that clathrin-mediated endocytosis induced internalization of NPC1L1 when it functioned as a transporter (Ge et al, 2011; Li et al, 2014). To confirm this observation, immunofluorescence analysis revealed that combination treatment indeed significantly triggered the expression and internalization of NPC1L1 (Fig EV4A). Next, we examined total cellular cholesterol (CHOL) (a well-known substrate of NPC1L1), and found marked accumulation of cholesterol in MPCs compared with MCs. By contrast, ezetimibe significantly induced the reduction in cholesterol in MPCs (Fig EV4B and C). We also performed vitamin E (a key substrate of NPC1L1; Narushima et al, 2008) analysis to evaluate whether vitamin E was increasingly absorbed by NPC1L1. Our data demonstrated that the content of vitamin E in MPCs medium was prominently decreased following exogenous addition of 50 μM vitamin E. In turn, ezetimibe blocked the uptake of vitamin E in MPCs (Fig 3H and I). Cholesterol and especially vitamin E which has a powerful anti-oxidative ability by scavenging lipid ROS are inextricably linked with redox homeostasis (Singh et al, 2005; Kopecka et al, 2020). Thus, we hypothesized that NPC1L1 was involved in redox regulation in MPCs due to substrate absorption. To investigate our hypothesis, we treated MDR cancer cells with cholesterol or vitamin E to examine which substrate could rescue MDR cancer cells from triple-combinatorial treatment-elicited oxidative stress and cell death. Using BODIPY 581/591 C11 reagent, we detected the generation of lipid ROS in Du145TXR and MCF-7ADR cells following triple-combinatorial treatment with or without 50 μM cholesterol or 50 μM vitamin E by flow cytometry. We observed that ezetimibe treatment further elevated lipid ROS accumulation induced by combination therapy. Both vitamin E and cholesterol combined with triple-combinatorial therapy alleviated lipid ROS in Du145TXR. In MCF-7ADR cells, vitamin E but not cholesterol led to decreased lipid ROS (Fig EV4D–F). In line with the results from ezetimibe, knockdown of NPC1L1 by siNPC1L1 also resulted in a similar outcome (Fig EV4G–I). Furthermore, utilization of either ezetimibe or siNPC1L1 significantly promoted malondialdehyde (MDA) levels (a specific product of lipid peroxidation) in combination therapy-treated cells, while vitamin E alone compromised triple-combinatorial treatment-induced upregulation of MDA levels (Fig EV4J–M). In addition, cell viability and colony formation assays revealed that 50 μM vitamin E significantly rescued tri-combinatorial treatment-induced cell death. In contrast, 50 μM cholesterol showed only a modest effect (Fig 3J–M). Notably, we found chemotherapy agents/verapamil treatment further induced cell death of MDR cancer cells that were cultured in serum-free medium where cholesterol and vitamin E were not present, compared with normal culture conditions (Fig 3J–K), suggesting that MDR cancer cells survive treatment with chemotherapy agents/verapamil due to absorption of vitamin E from serum by activating NPC1L1. Expectedly, NPC1L1 knockdown induced a similar outcome to ezetimibe (Fig EV4N–Q). Moreover, we also measured ROS levels of MPCs and MCs, and found that the intracellular ROS levels of MPCs were slightly higher than those of MCs in both Du145TXR and MCF-7ADR (Fig EV4R and S), implying that elevated expression of NPC1L1 likely protected from “damaging” species such as lipid peroxides, while the observation of higher ROS levels might support ROS-mediated survival signaling in MPCs. Taken together, these data demonstrate that NPC1L1 plays a key role in maintaining redox homeostasis mainly by enhancing vitamin E absorption, preventing cell death from chemotherapy agents/verapamil treatment-induced oxidative stress in a timely manner.

Click here to expand this figure.

Figure EV4. Lipotoxicity from NPC1L1 inhibition-induced oxidative stress may partially contribute to cell death of MPCs

A. Immunofluorescence analysis of NPC1L1 in Du145TXR and MCF-7ADR treated with indicated agents for 12 h. Scale bar, 10 μm. B, C. Cellular total cholesterol analysis in MCs and MPCs of (B) Du145TXR or (C) MCF-7ADR cells treated with 25 μM ezetimibe for 12 h. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. D–F. Flow cytometric analysis of lipid ROS using BODIPY 581/591 C11 reagent in Du145TXR or MCF-7ADR cells treated with indicated agents for 24 h. Chemo represents 20 nM taxol or 200 nM adriamycin for Du145TXR or MCF-7ADR, respectively. (D) Representative images and quantification of lipid ROS in (E) Du145TXR or (F) MCF-7ADR cells are shown. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. G–I. Flow cytometric analysis of lipid ROS using BODIPY 581/591 C11 reagent in Du145TXR or MCF-7ADR cells transfected with siNPC1L1 or siScramble followed by indicated treatment for 24 h. Chemo represents 20 nM taxol or 200 nM adriamycin for Du145TXR or MCF-7ADR, respectively. (G) Representative images and quantification of lipid ROS in (H) Du145TXR or (I) MCF-7ADR cells are shown. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. J–K. MDA levels measurement of (J) Du145TXR or (K) MCF-7ADR cells treated with indicated agents for 24 h. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. L, M. MDA levels in (L) Du145TXR or (M) MCF-7ADR cells transfected with siNPC1L1 or siScramble followed by indicated treatment for 24 h. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. N, O. Cell viability of (N) Du145TXR or (O) MCF-7ADR cells transfected with siNPC1L1 or siScramble followed by treatment with the indicated agents cultured in normal or serum-free medium for 72 h. One-way ANOVA was used to analyze statistical differences. Mean with ± SD. P, Q. Colony formation assay and quantification of (P) Du145TXR or (Q) MCF-7ADR cells transfected with siNPC1L1 or siScramble followed by treatment with indicated agents. One-way ANOVA was used to analyze statistical differences. R, S. Flow cytometric analysis of ROS levels in MCs and MPCs of Du145TXR and MCF-7ADR cells. (R) Representative images and (S) quantification of ROS are shown. Student’s t-test was used to analyze statistical differences. Mean with ± SD.

Data information: Results are representative of three independent experiments.

Source data are available online for this figure.

NPC1L1 inhibitor ezetimibe induces lethal macropinocytosis in MPCs

Specific therapy targeting of MDR persister cells in the clinic may be applied by two alternative strategies: inhibition of NPC1L1 function by using ezetimibe or complete deprivation of vitamin E in cancer cells. The former is a readily available option. Following triple-combinatorial treatment, surprisingly we observed a striking cytoplasmic vacuolization, which is not likely caused by autophagy (Movie EV1). To further exclude the role of autophagy, we first examined the expression of autophagy-related proteins (Atg5 and LC3B) after chloroquine (CQ) treatment. As shown in Appendix Fig S4A, Atg5-deficient Du145TXR cells displayed a failed conversion of LC3B-I to lipidated LC3B-II (an established autophagosome marker) and a lack of Atg5 expression after CQ treatment which is consistent with previous studies (Ouyang et al, 2013). Therefore, we further investigated whether triple-combinatorial treatment induced autophagy in MCF-7ADR cells. As shown in Appendix Fig S4B, triple-combinatorial therapy resulted in marked autophagy induction, as evidenced by increased LC3B turnover and levels of Atg5 and Atg7. In contrast, an early autophagy inhibitor (3-Methyladenine (3-MA), 1 mM) prominently decreased LC3B lipidation, Atg5, and Atg7 expression in triple-combinatorial therapy-treated MCF-7ADR cells. In addition, to investigate whether autophagy was involved in the anti-cancer effect of triple-combinatorial therapy, Du145TXR and MCF-7ADR cells were treated with triple-combinatorial therapy combined with 3-MA. As shown in Appendix Fig S4C, 3-MA showed no obvious influence on cell viability. Consistently, no obvious change in clonogenic survival was observed following 3-MA treatment in triple-combinatorial therapy-treated cancer cells (Appendix Fig S4D and E). Taken together, our results suggest that triple-combinatorial treatment-induced cell death may not be dependent on autophagy induction.

Subsequently, we paid our attention to macropinocytosis, which has previously been reported as an endocytic adaptive process in the context of various stress situations (Marques et al, 2017). Bafilomycin A1 (Baf-A1), which plays an important role in inhibiting early- and late-phase macropinocytosis (Yoshimori et al, 1991), was used in a preliminary investigation to understand the nature of induction of these vacuoles. Phase-contrast microscopy indicated that treatment induced multiple large single membrane bounded empty vesicles, which were almost completely abrogated by Baf-A1 (100 nM; Fig 4A and B). In addition, we observed an increased uptake of fluorescently labeled high molecular weight dextran (FITC-dextran, a specific substrate of macropinocytosis) by immunofluorescence. Expectedly, co-treatment of Baf-A1 compromised this effect (Fig 4C and D). These results indicate that the NPC1L1 inhibitor ezetimibe induces macropinocytosis in MPCs.

Figure 4. Utilization of NPC1L1 inhibitor ezetimibe induces lethal macropinocytosis in MPCs

A, B. Phase contrast images showing the morphological change in (A) Du145TXR or (B) MCF-7ADR cells treated with indicated agents for 72 h. Scale bars, 25 or 50 μm (low-magnification images). C, D. Endocytosis analysis of (C) Du145TXR or (D) MCF-7ADR cells treated with indicated agents for 8 h, followed by staining with 250 μg/ml FITC-DEX for 4 h. Scale bars, 10 μm. E, F. Phase contrast images of the morphological changes in (E) Du145TXR and (F) MCF-7ADR cells cultured in normal or serum-free medium followed by treatment with indicated agents for 12 h. Scale bars, 10 or 25 μm. The red arrows indicate vacuoles.