Afatinib Enhances the Efficacy of Conventional Chemotherapeutic Agents by Eradicating Cancer Stem–like Cells

Cancer stem cells (CSC) are thought to be responsible for tumor metastasis, chemoresistance, and recurrence after therapy (1). An intriguing property of CSCs is their high expression of ABC efflux transporters, especially ATP-binding cassette subfamily G member 2 (ABCG2), which actively extrudes anticancer drugs and contributes to chemoresistance of CSCs. Moreover, ABCG2 is responsible for the “side population” (SP) phenotype, which is often used for the identification and isolation of cancer stem-like cells (CSLC). Furthermore, ABCG2 also plays a crucial role in stem cell proliferation and self-renewal (2). Therefore, ABCG2 has been proposed to be a promising biomarker for the identification and a new therapeutic target for the eradication of CSCs. Novel strategies targeting ABCG2, including the downregulation of ABCG2 expression and/or function, may effectively circumvent drug resistance and eliminate CSCs to achieve better chemotherapeutic effect (3).

Tyrosine kinase inhibitors (TKI) are highly promising agents for specific inhibition of cancer cell growth and metastasis. A large number of TKIs targeting different oncogenic signaling pathways have been developed, and are currently used in the clinic or being tested in clinical trials. Interestingly, it has been reported that several TKIs, such as gefitinib, imatinib, and lapatinib can interact with ABCG2, thereby inhibiting its drug transport activity and enhancing the anticancer efficacy of conventional chemotherapeutic agents (4). Importantly, various recent studies have reported favorable clinical outcome by combining TKIs with conventional cytotoxic agents or monoclonal antibody in cancer chemotherapy (5, 6).

Afatinib (BIBW 2992) is a novel irreversible multitargeted TKI targeting ErbB family members EGFR, HER2, and HER4 (7). It exhibits superior anticancer activity in patients with lung cancer that harbor the gefitinib/erlotinib-resistant mutant EGFR (including T790M, exon 20 insertion, and T790M/L858R double mutant). Afatinib is also currently in clinical development for the treatment of other solid tumors, such as breast, head and neck carcinomas, and gliomas (8, 9). For now, afatinib has shown promising clinical activity and a higher overall response rate compared with trastuzumab or lapatinib in patients with HER2-positive advanced breast cancer (8). In addition, a phase II study showed that afatinib had clinical activity in a subset of patients with refractory and/or metastatic head and neck squamous cell carcinoma (10). However, the effect of afatinib alone or in combination with other chemotherapeutic agents on drug-resistant CSCs has not been investigated.

In this study, we found that afatinib effectively eliminated cancer stem cell–like SP cells and inhibited their self-renewal ability in vitro and in vivo. In addition, afatinib dramatically enhanced the efficacy of chemotherapeutic agents against SP cells by decreasing the expression of ABCG2. More importantly, we identified a unique mechanism by which afatinib increases DNA methyltransferases (DNMT) activity and leads to methylation of the ABCG2 promoter and downregulation of ABCG2 mRNA. Our findings may provide a novel strategy for overcoming chemoresistance of CSCs and a unique way to eradicate cancer cells by afatinib and probably other selected TKIs.

Afatinib (BIBW2992), whose molecular structure was shown in Supplementary Fig. S1A, was purchased from Boehringer Ingelheim Pharma GmbH & Co. and dissolved in dimethyl sulfoxide for use at indicated concentrations.

The following cell lines were kindly provided by Dr. S.E. Bates (NCI, NIH, Bethesda, MD): human colon carcinoma cell line S1 and its mitoxantrone-selected ABCG2-overexpressing subline S1-MI-80 (11), human breast carcinoma cell line MCF7 and its flavopiridol-resistant, ABCG2-overexpressing subline MCF7/FLV1000 (12). The human nasopharyngeal carcinoma cell line CNE2 and its high-metastatic clone s-18 (13) were a kind gift from Prof. M.S. Zeng (Cancer Center, Sun Yat-Sen University, Guangzhou, China). Cell lines used in this study were thawed from early passage stocks and were passaged for less than 6 months. Cell lines were periodically monitored for mycoplasma by Hoechst staining. All cell lines were cultured in DMEM or RPMI1640 medium supplemented with 10% FBS and with 1% antibiotic solution (penicillin-streptomycin).

Bone marrow blood was obtained from 30 patients with new diagnosed acute myeloid leukemia (AML) or acute lymphoblastic leukemia (ALL) after informed consent and with the approval of the Ethics Review Committee at Sun Yat-Sen University (Guangzhou, China). Patient characteristics are summarized in Supplementary Table S1.

Leukemia blasts were isolated using Ficoll–Hypaque density gradient centrifugation and cultured in RPMI1640 medium containing 20% FBS.

Cytotoxicity was determined using MTT assay as previously described (14). The IC₅₀ value was defined as the drug concentration resulting in 50% cell death. Both the fitted sigmoidal dose response curve and IC₅₀ were calculated by use of the Bliss method (14).

The intracellular accumulation assay of Hoechst 33342 was performed as described by Goodell and colleagues with some modifications (15). In brief, cells were incubated in 6-well plates with exposure to 1, 2, and 4 μmol/L afatinib for 24 hours. Then Hoechst 33342 dye (0.5 μg/mL) was added and further incubated for 30 minutes. Finally, cells were washed with ice-cold PBS three times and resuspended in PBS for flow cytometry analysis.

The SP sorting and analysis were performed using the methods described by Goodell and colleagues with some modifications (15). Briefly, cells were resuspended at a density of 1 × 10⁶ cells/mL in prewarmed DMEM containing 2% FBS and 10 mmol/L HEPES. Hoechst 33342 dye was added at a final concentration of 5 μg/mL in the presence or absence of Fumitremorgin C (FTC, 10 μmol/L). Cells were then incubated at 37°C for 90 minutes with intermittent shaking. At the end of incubation, cells were washed twice with ice-cold PBS. The Hoechst dye was excited at 355 nm, and the fluorescence profile was measured in dual wavelength analysis (blue, 402–446 nm; red, 650–670 nm).

SP and non-SP (NSP) cells were sorted from CNE2-s18 cells as described above. Sphere cell culture was preformed according to published protocol with minor modifications (16). Briefly, single-cell suspensions were plated in six-well ultra-low attachment plates (Corning Inc.) at a density of 1,000 cells/mL. Cells were maintained in serum-free DMEM/F12 containing different concentrations of afatinib. Subsphere-forming assay was performed by harvesting the primary spheres and replating after dilution. Subspheres were cultured in serum-free DMEM/F12 without afatinib treatment. The number and size of tumor spheres formed were evaluated by light microscopy after 7 days.

In vivo experiments were done in accordance with the guidelines for the use of laboratory animals of the Sun Yat-Sen University Institutional Animal Care and Use Committee.

Athymic nude mice (BALB/c-nu/nu) of both sexes, 5- to 6-week-old, were used to establish S1-MI-80 cell xenograft model for the drug intervention experiment. The S1-MI-80 cell xenograft model was established as previously described (17) with minor modification. Briefly, 1 × 10⁶ cells were inoculated subcutaneously into the flanks of the nude mice. The mice were randomized into four groups (10 in each group) after the tumors reached a mean volume of about 100 mm³, and then received various treatments: (i) saline (every 3 days for 11 times, i.p.); (ii) topotecan (every 3 days for 11 times, 3 mg/kg, i.p.); (iii) afatinib (every 3 days for 11 times, 20 mg/kg, orally); and (iv) topotecan (every 3 days for 11 times, 3 mg/kg, i.p.) plus afatinib (every 3 days for 11 times, 20 mg/kg, orally; afatinib was given 1 hour before topotecan administration). Throughout the study, mice were weighed and tumors were measured with a caliper every 4 days. Tumor volumes (V) were calculated using the formula: V = (length × width²/2). When the tumors grew to proper size, the mice were euthanized and the tumors were excised and weighed. The ratio of growth inhibition (IR) was estimated according to the following formula:

CNE2-s18 xenograft models were used to evaluate whether the antitumor effect of afatinib alone or its combination with topotecan could be sustained after stopping the treatment regimen. Mice were grouped randomly and treated with different drugs as described above. Drug treatment was stopped at 12 days after the initial start date. Thereafter, tumor volume and mice body weight were measured every 2 days.

Real-time PCR was performed with a Bio-Rad CFX96 Real-Time System (Bio-Rad). The amount of each target gene in a given sample was normalized to the level of GAPDH in that sample. The amplification products were analyzed on 1% agarose gels. The 2^−ΔΔC_t method was used to analyze the relative changes in gene expression (18). The PCR primers used in this study are summarized in Supplementary Table S2.

Cells were treated with afatinib (2 μmol/L) alone, 5-Aza-dC (5 μmol/L) alone, or in combination for 72 hours. The drug-containing medium was replaced every 24 hours. After treatment, genomic DNA was extracted using the TIANamp Genomic DNA Kit (TIANGEN) followed by bisulfite conversion of 2 μg of genomic DNA using the EpiTect Bisulfite Kit (Qiagen) according to the manufacturer's instructions.

The methylation level of ABCG2 and TP53 promoter was detected by methylation-specific PCR (MSP). The primer pairs for MSP were listed in Supplementary Table S1. PCR products were electrophoresed on a 2% polyacrylamide electrophoresis gel. For the analysis of methylation level of CpG islands of ABCG2 promoter, the PCR products from MSP were cloned into the pMD18-T vector (Takara) and sequenced. DNA methylation levels were calculated as ratio of the methylated CpG sites to all available CpG sites within the sequence analyzed. The sequence results were analyzed by BiQ Analyzer software.

Cells were treated with afatinib (2 μmol/L) and 5-Aza-dC (5 μmol/L) for 72 hours. Drug-containing medium was replaced every 24 hours. After the drug treatment, the cells were harvested and the nuclear fraction was collected using the Nuclear Extraction Kit (Sangon Biotech). DNMT activity of nuclear extracts was assayed according to the manufacturer's protocol using the DNMT Activity Assay (ActiveMotif).

All the experiments were performed at least in triplicate. The statistical significance of differences was assessed using t test or nonparametric statistics (SPSS16.0) for comparison between groups. Values represent mean ± SD. For all tests, P < 0.05 was considered as statistically significant.

The in vitro cytotoxicity of afatinib was first assessed in various cancer cell lines by the MTT assay. More than 80% cell survival was observed in the presence of afatinib up to 4.0 μmol/L in ABCG2-overexpressing S1-MI-80, MCF7/FLV1000, and CNE2-s18 cells (Supplementary Fig. S1). Therefore, all subsequent assays were performed with afatinib at a concentration below 4.0 μmol/L.

To verify whether afatinib downregulated cellular expression of ABCG2, we performed a Western blot analysis of ABCG2 in whole-cell lysate collected from CNE2-s18 and S1-MI-80 cells with or without afatinib treatment. The results indicated that afatinib decreased the total cellular expression of ABCG2 in a time- and concentration-dependent manner (Fig. 1A and B). In addition, immunofluorescence microscopic analysis also showed a concentration-dependent decrease of cell surface staining of ABCG2 in afatinib-treated S1-MI-80 cells (Fig. 1C).

It has been reported that the SP phenotype and the chemoresistance of SP cells are attributed primarily by cell surface expression of ABC transporters (in particular, ABCG2 and ABCB1; refs. 19, 20). To investigate whether SP cells derived from CNE2-s18 have elevated ABC transporters expression, we analyzed the cell surface expression of ABCG2 and ABCB1 by flow cytometry. It was found that SP cells exhibited significantly higher expression of ABCG2 than NSP cells (Fig. 1D and E). On the other hand, ABCB1 was not detected in both SP and NSP cells (Supplementary Fig. S2A and S2B). These results indicated that ABCG2 played a more important role in the SP phenotype in CNE2-s18 cells.

We then analyzed the cell surface expression of ABCG2 protein after treatment with a range of concentrations of afatinib in SP cells sorted from CNE2-s18 by flow cytometry. Afatinib was found to significantly decrease the cell surface expression of ABCG2 in a concentration-dependent manner in SP cells (Fig. 1F and G).

SP cells are known to be enriched in cancer stem–like cells, which represent a population of self-renewing and drug-resistant cells driving the tumor growth. Afatinib was found to decrease the SP fraction of CNE2-s18 cells compared with that of the control cells in a concentration-dependent manner (Fig. 2A and B). The proportion of SP cells was decreased from about 18.66% to 9.51%, 3.04%, and 1.21% after a 48-hour treatment with afatinib at the concentration of 1, 2, and 4 μmol/L, respectively. Significant decreases in the percentage of SP cells in response to treatment with afatinib were also observed in the parental cell line CNE2 (Fig. 2C and D). These results suggested that afatinib can eliminate SP cells. To further verify that the disappearance of SP cells was caused by the specific elimination by afatinib, SP cells were sorted out from CNE2-s18 cells and the purity was >97% (Supplementary Fig. S2C). We found that afatinib drastically reduced SP fraction and cell viability in sorted SP cells in a concentration-dependent manner (Fig. 2E and F and Supplementary Fig. S2D). Treatment with up to 4 μmol/L afatinib eliminated SP cell populations by about 67.6%.

Functional ABCG2 has been reported to be expressed on the surface of leukemic progenitor cells with variable level in AML and ALL (21). Therefore, we performed the detection of ABCG2 expression level and the analysis of SP cell populations in leukemia blast cells, which is used as a primary CSC model, with or without afatinib treatment. The results showed that 18 of 30 (60%) patient samples displayed detectable expression of ABCG2 with varying level and distinct SP fraction with a variable percentage (Fig. 3). Afatinib treatment at 2 μmol/L significantly decreased the expression level of ABCG2 and the proportion of SP cell populations in these 18 samples (Fig. 3; P < 0.001). These results further demonstrated that afatinib can selectively eliminate SP cells in primary leukemia cells via decreasing expression of ABCG2.

The ability to form spheres in nonadherent culture is widely used as a surrogate for the self-renewal property of CSCs. Therefore, we performed the sphere-forming assay to investigate the potential role of afatinib in the self-renewal capacity of SP cells. SP cells were found to form bigger and much more spheres than the NSP cells (Fig. 4A and B). More importantly, afatinib was found to inhibit the formation and growth of spheres in SP cells in a concentration-dependent manner, indicating its possible inhibition of self-renewal in SP cells (Fig. 4A and B). In addition, afatinib also inhibited the subsphere-forming ability of SP cells (Fig. 4C and D). Afatinib at 2 μmol/L resulted in a 3.5-fold decrease in the number of subspheres, further the size of the subspheres was also obviously reduced (Fig. 4C and D). On the other hand, afatinib also diminished the soft agar colony-forming ability of SP cells (Fig. 4E). These data support that afatinib can effectively inhibit the self-renewal capacity of SP cells in vitro.

To investigate whether the inhibition of stemness property by afatinib in SP cells could be extended to in vivo conditions, we used the NOD/SCID mouse model to evaluate the effect of afatinib on tumorigenic potential. The tumorigenicity of SP and NSP cells from CNE2-s18, with or without afatinib treatment, at cell dilution ranging from 1 × 10³ to 1 × 10⁶ cells per injection was evaluated. As expected, the SP cells displayed higher tumorigenicity than NSP cells, thereby validating the stemness property of sorted SP cells (Table 1). Importantly, SP cells treated with afatinib were found to exhibit lower tumorigenic capacity than the untreated SP cells (Table 1). While as few as 1 × 10³ untreated SP cells were sufficient to generate tumors, more than 1 × 10⁴ SP cells were necessary to initiate tumor formation after afatinib treatment (Table 1). The data clearly illustrated the impairment of tumorigenicity of SP cells by afatinib in vivo.

Besides, we also examined the effect of afatinib on the growth of tumors. A representative tumor growth curve in mice transplanted with 1 × 10⁵ cells was shown (Fig. 4G). Untreated SP cells were found to grow faster than the SP cells pretreated with afatinib (Fig. 4F and G; P < 0.05). Thus, the data indicated that afatinib can impair the tumorigenicity of SP cells and inhibit the tumor growth in vivo.

Therefore, when used alone, afatinib was shown to effectively eliminate the SP cells and reduce their self-renewal capability and tumorigenic properties, both in vitro and in vivo.

As indicated above by the decrease of ABCG2 expression by afatinib in SP cells, we hypothesize that afatinib can also sensitize SP cells to ABCG2-substrate chemotherapeutic agents. As shown in Fig. 5A and B, the SP cells were more resistant to topotecan and mitoxantrone than NSP cells. After exposure to afatinib, the IC₅₀ values for topotecan and mitoxantrone in SP cells were significantly reduced, indicating the enhancement of SP cells to topotecan and mitoxantrone.

In addition, we found afatinib could increase apoptosis (PI-negative, Annexin V-FITC–positive) induced by topotecan and mitoxantrone in SP cells (Fig. 5C and D and Supplementary Fig. S2E and S2F).

The potentiation of anticancer activity by transporter inhibitors is usually mediated by the inhibition of transporter-mediated efflux, thereby leading to an increase in the intracellular drug accumulation (22). To explore the potential mechanism by which afatinib sensitizes SP cells to chemotherapeutic drugs, we examined the intracellular accumulation of Hoechst 33342, known fluorescent substrates of ABCG2, by flow cytometry. As shown in Fig. 5E–G, in the presence of 1, 2, or 4.0 μmol/L afatinib, the fluorescence index of Hoechst 33342 in SP cells was elevated by 1.34-, 1.49-, 1.87-fold, respectively (Fig. 5E–G). These results suggested that afatinib, similar to a control potent ABCG2-specific inhibitor FTC, dramatically increases the accumulation of Hoechst 33342 in a concentration-dependent manner in SP cells (Fig. 5E and G). However, neither afatinib nor FTC affected the intracellular levels of Hoechst 33342 in NSP cells (Fig. 5F and G).

On the basis of the in vitro findings, we investigated whether afatinib can enhance the antitumor effect of topotecan in vivo. The results showed that the antitumor ability of topotecan was significantly enhanced when it was administrated in combination with afatinib compared with treatment with topotecan or afatinib alone in S1-MI-80 cell xenografts in nude mice (Fig. 5H; P < 0.01). The mean weights of tumors excised from mice were 0.95 ± 0.30, 0.86 ± 0.46, 0.83 ± 0.21, 0.56 ± 0.37 g for saline, afatinib, topotecan, and combination groups, respectively. And the inhibition rate (IR) of the combination group calculated using the formula described in Materials and Methods was up to 46.32%.

CSCs are also believed to cause tumor recurrence. Therefore, we performed 12 days of drug treatment followed by another 12 days of drug withdrawal on CNE2-s18 xenograft model to evaluate the duration of antitumor effect of afatinib. A better and more durable tumor growth inhibitory effect was observed in the combination group (afatinib and topotecan) compared with afatinib or topotecan alone (Fig. 5I; P < 0.01). Interestingly, afatinib alone also exhibited a longer lasting antitumor effect than topotecan alone after the drug treatment was terminated (Fig. 5I; P < 0.05). Neither significant body weight loss nor treatment-related deaths was observed throughout the in vivo study (Supplementary Fig. S3A and S3B). Thus, the more effective and durable tumor growth inhibitory effect exhibited by the combination of afatinib with topotecan may be due to the specific elimination of SP cells in vivo.

The downregulation of ABCG2 protein by afatinib is likely a universal effect in resistant cancer cells with ABCG2 overexpression because ABCG2 protein was also found to be decreased by afatinib in a concentration-dependent manner in MCF7/FLV1000 cells (Fig. 6A). To further understand the mechanism for ABCG2 protein downregulation by afatinib, the possible involvement of protein degradation systems by ubiquitin-proteasome, lysosome, and calpain were evaluated. MG132, Leupeptin, and ALLN were used as specific inhibitors of proteasome, lysosome, and calpain, respectively. None of the inhibitors can block the downregulation of ABCG2 protein after afatinib treatment (Fig. 6B–D). ABCG2 protein stability was also evaluated by using cycloheximide to stop protein synthesis and measuring the amount of remaining ABCG2 protein at various time points after 2.0 μmol/L afatinib treatment by Western blot analysis (Fig. 6E). The quantitative analyses shown that afatinib do not alter the half-life of ABCG2 protein (Fig. 6F). These results indicated that the decrease of ABCG2 protein caused by afatinib was independent of ubiquitin-proteasome-, lysosome-, and calpain-mediated protein degradation systems.

The regulation of ABCG2 mRNA expression by afatinib was then examined. The results showed that ABCG2 mRNA level was reduced by afatinib in a time- and concentration-dependent manner in S1-MI-80 cells (Fig. 6G and H). Similar findings were observed in CNE2-s18 cells (Fig. 6I and J). These data suggest that afatinib inhibited ABCG2 expression at the mRNA level.

Several mechanisms have been reported to regulate the mRNA expression of ABCG2 in CSCs, including epigenetic phenomena (23, 24). It has been shown that CpG methylation of the ABCG2 promoter could result in its transcriptional silencing (25). Therefore, we sought to determine whether afatinib could affect the methylation status of the ABCG2 promoter, which lead to a reduction in ABCG2 mRNA level. According to result from MSP, ABCG2 promoter was found to be more heavily methylated than untreated cells (Fig. 7A). Importantly, treatment of cells with 5-Aza-dC (72 hours), a well-known DNA methyltransferase I inhibitor, was found to elevate ABCG2 mRNA expression, which was aaccompanied by a significant demethylation of the ABCG2 promoter (Fig. 7A and B).

These results were further corroborated by direct bisulfite sequencing of a 125 bp fragment (ranging from -337 to -213 relative to the transcription initiation site), which was reported to be a functional CpG island suppressing the expression of ABCG2 (26). Surprisingly, afatinib evidently increased the level of CpG methylation in the ABCG2 promoter (Fig. 7C and D). Treatment with a demethylating agent, 5-Aza-dC, was able to reduce the extent of ABCG2 promoter methylation in afatinib-treated cells (Fig. 7D). These results indicated that afatinib downregulated ABCG2 by inducing methylation of the gene promoter.

Total DNMT activity was then evaluated to determine whether the increased ABCG2 methylation by afatinib is associated with a change in DNMT activity. The results showed a 1.6-fold increase in DNMT activity after 2.0 μmol/L afatinib treatment for 72 hours (Fig. 7E; P < 0.05). Consistent with the restoration of ABCG2 expression by 5-Aza-dC in afatinib-treated cells, 5-Aza-dC was found to decrease the basal DNMT activity and prevent the afatinib-induced increase in DNMT activity (Fig. 7E). Furthermore, afatinib concentration dependently increased DNMT activity (Supplementary Fig. S3B and S3C). Therefore, both increased ABCG2 promoter methylation and elevated DNMT activity are involved in the downregulation of ABCG2 by afatinib.

It has been reported that drug-induced DNA methylation or demethylation could be either global or site specific (27, 28). As a control for comparison, methylation status of the TP53 gene was also evaluated. Interestingly, afatinib did not affect the promoter methylation of TP53 (Fig. 7F). Therefore, the increased methylation at the ABCG2 locus after afatinib treatment was likely a site-specific event.

The existence of CSCs has attracted a lot of attention recently because they are the major contributing factor for cancer recurrence and metastasis, which severely hindered successful cancer chemotherapy (29, 30). Importantly, the multidrug resistance (MDR) phenotype of CSCs is responsible for chemotherapeutic failure (31, 32). The most common cause of MDR is the increased drug efflux from cancer cells mediated by ATP-binding cassette (ABC) transporters including ABCG2 and ABCB1 (P-gp). Recent reports indicate that ABCG2 is highly expressed in stem cells, and play an important role in the proliferation and self-renewal of stem cells and in the tissue regeneration process (2, 33, 34, 35). In pharmacologic assays, the ability of stem cells to efflux a fluorescent Hoechst dye 33342 (a ABCG2 substrate) has been adopted to prospectively identify and isolate stem cell population (20, 36). Therefore, ABCG2 may be used as a novel target for the development of chemosensitizing agents to target drug-resistant cancer cells and to eradicate CSCs.

TKIs represent a new class of highly specific anticancer agents, which target various oncogenic signaling pathways in different cancer types. They were demonstrated to be safe and effective antitumor drugs used in clinical trials and in clinical treatment. Recently, the combinations of TKIs with cytotoxic anticancer drugs have been shown to be effective in treating drug-resistant cancers. Several TKIs have been found to inhibit ABCG2 and to sensitize resistant cancer cells to the cytotoxic effect of ABCG2 substrate anticancer drugs. However, most reports including our previous studies showed that these TKIs (including lapatinib and gefitinib) reversed ABCG2-mediated MDR via direct inhibition of the transporter function, but with no obvious effect on the expression level of ABCG2, (37, 38). In this study, we found that afatinib can effectively sensitize cancer stem–like cells to conventional chemotherapeutic agents by suppressing the transport activity and also downregulating the expression of ABCG2. Importantly, this reduction in ABCG2 expression by afatinib was associated with a significant decrease in the proportion and the self-renewal capacity of SP cells in the cancer cell lines tested. Moreover, afatinib dramatically impaired the tumorigenicity of the SP cells and suppressed their tumor growth in vivo. It has been proposed that ABCG2 inhibitors can be divided into two categories: one inhibiting ABCG2 activity only (static) and the other one inhibiting ABCG2 activity as well as reducing the expression of ABCG2 (dynamic; ref. 39). To this end, the current data suggest that afatinib is a “dynamic” inhibitor of ABCG2, which makes it distinctive from other TKIs.

Like lapatinib, afatinib inhibits irreversibly EGFR and HER2 kinases, which regulate two important downstream PI3K and AKT pathways. Several reports indicated that these PI3K and AKT pathways regulate the activity of ABCG2 and subsequently mediate the SP phenotype (40). A recent report indicated that the phosphorylation status of AKT was associated with translocation of ABCG2 (41). Therefore, we also examined the effect of afatinib on the downstream AKT and ERK kinases by Western blot analysis. The result indicated that afatinib did not inhibit the activation of AKT and ERK1/2 up to 4 μmol/L in both S1 and S1-MI-80 cells (Supplementary Fig. S4A and S4B). Moreover, afatinib did not affect the expression of total AKT and ERK1/2 (Supplementary Fig. S4A and S4B). The lack of effect on AKT and ERK1/2 phosphorylation by afatinib at its sensitizing concentration was also noted in CNE2-s18 cells (Supplementary Fig. S4C). Therefore, the blockade of EGFR signaling is unlikely to play a significant role in the sensitizing effect of afatinib on the ABCG2-overexpressing cells.

The ABCG2 expression is known to be regulated at multiple levels. DNA methylation and histone modifications have been reported to play important roles in the epigenetic regulation of ABCG2 expression in human renal carcinoma and drug-resistant cells, respectively (23, 42). Among all the “dynamic” ABCG2 inhibitors previously identified, most of them reduced the expression level of ABCG2 by the degradation of protein mediated by lysosome and proteasome (43, 44). However, in our study, the decrease in ABCG2 expression by afatinib was not related to protein degradation mediated by ubiquitin-proteasome, lysosome, or calpain. On the other hand, the fact that afatinib also reduce ABCG2 mRNA level suggests transcriptional regulation.

The ABCG2 promoter is known to contain CpG islands. It has been reported that demethylation of the ABCG2 promoter is related to reactivation of the gene and the subsequent ABCG2-dependent MDR phenotype (26, 45). To this end, drug-induced epigenetic modulation of the ABCG2 gene has attracted increasing attention recently (45–47). Our data showed for the first time that afatinib induced methylation of the ABCG2 promoter, thus leading to a decrease in mRNA level of ABCG2 in S1-MI-80 cells, which has not been reported with other TKIs. Consistent with this finding, the reduced ABCG2 mRNA expression in afatinib-treated cells can be restored by treatment with 5-Aza-dC (a demethylating agent), which was accompanied by the demethylation of the ABCG2 promoter (Fig. 7B–D).

A schematic model illustrating the possible targeting of CSLCs by afatinib is proposed in Fig. 7G. Primary tumor consists of both highly proliferative tumor cells and the putative CSLCs. Conventional chemotherapeutic agents kill primarily the highly proliferative tumor cells. However, the CSLCs are spared by their high expression of multidrug transporters (especially, ABCG2), mediating their chemoresistance, and eventually regenerate the tumor. Our data showed that afatinib targets and eliminates the CSLCs by reducing ABCG2 expression. Therefore, the combination use of afatinib with conventional chemotherapeutic agents can eradicate the whole tumor mass by killing both the highly proliferative cells and the CSLCs. While the clinical application of the drug combination is advocated, further study is still warranted to understand fully how afatinib causes methylation of the ABCG2 promoter and whether there is any cell type/tissue specificity.

No potential conflicts of interest were disclosed.

Conception and design: K.K.W. To, L.-W. Fu

Development of methodology: K.K.W. To

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.-H. He, J.-H. Xu, S. Ye, H. Zhang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.-H. Xu, K.K.W. To

Writing, review, and/or revision of the manuscript: X.-K. Wang, F. Wang, K.K.W. To, L.-W. Fu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.-H. Xu, S. Ye, Z.-C. Huang, K.K.W. To

Study supervision: Z.-C. Huang, K.K.W. To

Other (performed all the experiments): X.-K. Wang

The authors thank Dr. Susan Bates (NCI, NIH, Bethesda, MD) for the ABCG2-overexpressing cell lines and M.S. Zeng (Cancer Center, Sun Yat-Sen University, Guangzhou, China) for the highly metastatic CNE2 clone s-18 cell line (CNE2-s18).

This work was supported by grants from the Major State Basic Research Development Program of China (973 Program, No. 2012CB967000), the National High Technology Research and Development Program of China (863 Program, No. 2012AA02A303), Ph.D. Programs Foundation of Ministry of Education of China (No. 20120171110090), and Excellent Doctoral Dissertations Foundation of Guangdong Province (No. 8500-3226202).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.