DC120, a novel AKT inhibitor, preferentially suppresses nasopharyngeal carcinoma cancer stem-like cells by downregulating Sox2.

Side population (SP) contains cancer stem-like cells (CSLCs). In this study, we characterized SP cells from nasopharyngeal carcinoma (NPC) cell lines and found that SP cells had a higher self-renewal ability in vitro and greater tumorigenicity in vivo. The AKT pathway was activated in NPC SP cells. DC120, a 2-pyrimidyl-5-amidothiazole inhibitor of the ATP binding site of AKT, inhibited phosphorylation of FKHRL1 and GSK-3β. DC120 inhibited SP fraction, the sphere-forming ability in vitro and growth of primary xenografts as well as secondary xenografts’ tumor recurrence. This inhibition was accompanied by reduced expression of stem-related gene Sox2 due to induction of p27 and miR-30a. A combination of DC120 and CDDP more effectively inhibited NPC cells compared with monotherapy in vitro and in vivo. Clinical evaluation of DC120 is warranted.


INTRODUCTION
Nasopharyngeal carcinoma (NPC) is the most frequent head and neck tumor in South China with the highest incidence rate of up to 54.7/100,000/year [1]. Despite improvements in conventional treatment, approximately 30% of patients with locoregionally advanced disease will subsequently relapse, frequently with associated metastases, and develop resistance to first-line drugs [2]. Accumulating evidence suggests that side population cells [3], a rare population of cells from primary tumors or cancer cell lines, are enriched in a subset of tumor-initiating or cancer stem-like cells (CSLCs) [4], which contribute to tumor metastasis and recurrence as well as tumor resistance to both radiation and chemotherapy [5].
It is well established that the PI3K/AKT signaling pathway is involved in a wide variety of biological processes, including cell proliferation, differentiation, apoptosis, autophagy, glucose metabolism, DNA doublestrand break repair and tumorigenesis [6][7][8][9][10]. Loss or mutation of tumor suppressor PTEN, amplification or mutation of PI3K, activation or mutation of growth factor receptors and oncogenes, and amplification of AKT itself are involved in the activation of AKT in tumors [10][11][12].
Recently, increasing evidence suggests that the activation of AKT has a key role in the function of CSLCs and modulates the percent of SP cells in a variety of cancers, including esophageal carcinoma [13], glioma tumors [14], lung cancer [15] and breast cancer [16]. Further studies demonstrate that constitutive activation of AKT promotes CSLC resistance to treatment with chemotherapy and/ or radiation therapy partially by downregulating the expression of p27 [17], a well-known tumor suppressor.
AKT kinase has become an attractive target for small molecular drug discovery. To date, researchers have developed many AKT inhibitors, including targeting the ATP binding site, pleckstrin homology domain [18], or protein substrate binding site of AKT. Several of these inhibitors, such as MK-2206, GDC0068, and perifosine are currently in phase I and II trials alone or in combination to treat multiple types of cancers [3,19]. To find compounds that target AKT kinase, we designed and synthesized a series of 2-pyrimidyl-5-amidothiazole compounds based on the ATP binding site of AKT. We had previously reported that DC120, which was screened out from these compounds, exhibited an inhibitory effect on cancer cells in vitro and in vivo through the inhibition of AKT kinase activity and the blockade of the AKT downstream signaling pathway [20,21]. In this study, we sought to explore the involvement of DC120 in the regulation of NPC cancer stem-like SP cells. Our data showed that DC120 inhibited the proliferation of human NPC CNE-2-S-18/SP and CNE-1/SP cells in vitro and in vivo and significantly reduced the self-renewal and tumor-initiating capacities of cancer stem-like SP cells via the induction of cell apoptosis. Additionally, we observed that DC120 suppressed the cancer stem-like SP cells through the inhibition of AKT kinase activity and the blockade of the PI3K/AKT downstream signaling pathway, further regulating Sox2 expression. Moreover, we found that the combination of DC120 and cisplatin (CDDP) has a significant synergistic effect, and DC120 could sensitize the inhibitory effect of CDDP on NPC cells.

NPC SP cells have the characteristics of cancer stem-like cells (CSLCs)
It is believed that certain ATP-binding cassette (ABC) transporters (e.g. ABCG2/BCRP) can pump out the fluorescent dye Hoechst 33342, which may be why the SP phenotype exhibits a low level of Hoechst fluorescence intensity [14]. Using a FACS assay, we sorted SP cells in human NPC cell lines CNE-2-S-18 and CNE-1, which were characterized by a low fluorescent "tail" in the flow cytometry histogram ( Figure 1A). In the present study, we found that the average percent of SP cells was approximately 60.0% in the CNE-2-S-18 cell line and approximately 2.0% in the CNE-1 cell line; which was consistent with the results of previous studies [22], however, 5 μM FTC, the ABCG2-specific inhibitor, could significantly decrease the SP proportion to 0.2% (P < 0.01) and 0.1% (P < 0.01), respectively. We also examined whether SP cells sorted through a FACS assay displayed abilities associated with human CSLCs. We observed that not only the size of the spheres increased by 8-to 125-fold (P < 0.01; Figure 1B), but also the number of spheres of SP cells increased by approximately 5-fold (P < 0.01; Figure 1C) relative to matched NSP cells when grown in suspension cultures, an in vitro measure of CSLC self-renewal activity. The result of colony formation assay indicated that SP cell proliferation were better than that of NSP cell ( Figure 1D and 1E). We next directly estimated the tumor-initiating capacity by injecting sorted CNE-2-S-18/SP cells and CNE-2-S-18/NSP cells into NOD/ SCID mice. Tumors were generated with 1,000 SP cells, which was 10-fold less than was required for tumor seeding by NSP cells and grew at a faster rate compared with CNE-2-S-18/NSP cells ( Figure 1F, Table 1). DC120 down-regulated the activated PI3K/AKT pathway in NPC cancer stem-like SP cells As reported, the activation of the PI3K/AKT pathway plays an important role in the maintenance of cancer stemlike SP cells [4,23]. Among the cancer cell lines used in this study, both CNE2-S18 and CNE1 cell lines were previously confirmed to have hyper-activated PI3K/AKT signaling due to the PIK3CA and HRAS mutation, respectively. Our results indicated that the phosphorylation status of AKT on Thr308 and Ser473 and the phosphorylation levels of AKT downstream targets (FKHRL1 and GSK-3β) were much higher in SP cells than those in NSP cells (Figure 2A), suggesting that the PI3K/AKT pathway was activated in NPC cancer stem-like SP cells. We also verified the expression of stem cell transcription factors in SP and NSP cells, and found that the expressions of C-myc, klf4, Sox2 were higher in SP than in NSP, which further confirmed that the SP cells has the characteristics of stem cells ( Figure 2B). As the inhibition of substrate phosphorylation can reflect the inhibition of AKT activity, we examined whether DC120 ( Figure 2C) could inhibit AKT and its downstream targets. Figure 2D and 2E showed that the phosphorylation levels of FKHRL1 and GSK-3β were all partially attenuated by DC120 dose and time dependently without affecting the amount of total proteins. However, the phosphorylation of Thr308 and Ser473 on AKT increased concomitantly, although AKT kinase activity was inhibited, the conformational change of AKT led to its self-hyperphosphorylation. More precisely, phosphorylation of FKHRL1 and GSK-3β was reduced within 30 minutes after exposure to 10 μmol/L DC120 in CNE-2-S-18/SP and CNE-1/SP cells. These data suggested that the down regulation of the PI3K/AKT self-renewal pathway might contribute to the inhibitory effects of DC120 on NPC cancer stem-like SP cells.

DC120 inhibited NPC cancer stem-like SP cells in vitro
Using an MTT assay, we determined the effect of DC120 on the proliferation of human NPC SP and NSP cells. S-18-SP/NSP and CNE-1 SP/NSP cells were sorted by FACS analysis assay and then exposed to DC120 (0.625-40 μmol/L) for 48 hours. As shown in Figure 3A, exposure to DC120 resulted in a dose-dependent inhibition of cell viability, and compared with NSP cells, SP populations were more sensitive to DC120 especially at low doses for S-18 and CNE2 cells.
To examine whether DC120 could inhibit the SP phenotype in vitro, we performed a FACS analysis assay.
As shown in Figure 3B, 2.5 μmol/L DC120 significantly decreased the number of SP cells by over 76% in the CNE-2-S-18 cell line (P < 0.01) and 23% in the CNE-1 cell line (P < 0.05), and 10 μmol/L produced a greater than 89% reduction of SP cells in the CNE-2-S-18 cell line (P < 0.01) and 71% in the CNE-1 cell line (P < 0.01). To evaluate whether DC120 could suppress the formation of nasospheres in vitro, we exposed freshly sorted CNE-2-S-18/SP and CNE-1/SP cells to varying concentrations of DC120 and then cultured them for 7-14 days in the presence of the compound. As shown in Figures 3C  and 3D, DC120 inhibited the formation of spheres. Not only did the number of spheres decline by 75% to 99% (P < 0.01; Figure 3D) but also the size of the spheres was reduced by 8-to 132-fold (P < 0.01; Figure 3C). The IC 50 values were approximately 0.5-1 μmol/L for both the CNE-2-S-18/SP and CNE-1/SP spheres. Another recognized AKT inhibitor GDC0068 was employed, the same effect was obtained (Figure S1A-S1C). These data showed that DC120 inhibited the cancer stem-like SP cells at similar concentrations to those that inhibited nasosphere formation and at approximately 7-fold lower concentrations than those that inhibited cancer cells as determined by the MTT assay.

DC120 effectively sensitized NPC cells to CDDP therapy in vitro
Cisplatin (CDDP) is a clinically active agent against NPC. However, in contrast to DC120, NSP cells were more sensitive to CDDP than SP cells in CNE-2-S-18 and CNE-1 cells ( Figure 5A). Moreover, CDDP treatment increased the percentage of SP cells ( Figure 5B), which was similar to other findings and may play a role in resistance [24]. We further examined the combination of CDDP and DC120, which had a significant synergistic effect ( Figure 5C) and produced a combination index (CI)  value of < 1 in both CNE-2-S-18 and CNE-1 cells ( Table 2 and Table 3). Although treatment with CDDP alone increased the proportions of cancer stem-like SP cells ( Figure 5B), 5 μmol/L or 10 μmol/L DC120 plus 5 μmol/L CDDP significantly decreased the proportions of SP cells by over 76.9% (P < 0.01) or 81.7% (P < 0.01) in CNE-2-S-18 cells, and 66.9% (P < 0.01) or 97.8% (P < 0.01) in CNE-1 cells, respectively ( Figure 5D).

DC120 and CDDP combination therapy effectively diminished cancer stem-like SP cells in vivo
To determine whether DC120 could target NPC cancer stem-like SP cells in vivo, we established primary CNE-2-S-18 xenografts and gave an i.p. administration of 8% solvent (negative control), DC120, CDDP, or DC120+CDDP on day 5 after implantation. Treatment with DC120 noticeably suppressed the tumor growth, and the tumor growth inhibition (T/C %) was approximately 48.7% (P < 0.05; Figure 6A, Table 4). Meanwhile, the inhibitory rate of the CDDP group was 31.4% ( Figure  6A, Table 4). DC120 plus CDDP treatment in vivo led to an even greater reduction in tumor growth, and the inhibitory rate was 78.7% (P < 0.01; Figure 6A, Table 4).
No obvious toxicity was observed in mice receiving the above treatments ( Figure S2). Then, tumors were isolated from the animals, and single tumor cells were analyzed by FACS assay or seeded into ultra-low adhesion 6-well plates in serum-free media. The proportion of SP cells isolated from DC120-treated tumors demonstrated a 44.6% reduction (P < 0.01; Figure 6B, 6C), and the number and size of the spheres significantly decreased 5 to 10 fold and 16 to 128 fold compared with controls (P < 0.05; Figure 6D, 6E). However, the tumors treated with CDDP alone increased the proportion of SP cells by 9.1% (P < 0.05; Figure 6B, 6C), and had a similar nasosphere-forming ability compared with untreated controls ( Figure 6D, 6E). Additionally, primary xenografts treated with DC120 and CDDP in combination displayed a 64.8% reduction in SP cells (P < 0.01; Figure 6B, 6C) and decreased nasosphere formation ( Figure 6D, 6E), Furthermore, we injected 100,000 cells obtained from 4 groups of primary tumor cells into secondary recipient mice and examined the growth of the secondary tumors. We observed that cancer cells obtained from DC120treated mice resulted in the slowest tumor recurrence, reaching a final tumor size ranging from 300 to 500 mm 3 in the secondary mice ( Figure 6F). But the cancer cells from DDP-treated mice showed the most rapid tumor regrowth, reaching a final tumor size ranging from 600 to 800 mm 3 in the secondary mice ( Figure 6F). And the tumor regrowth in DC120 combined with CDDP treatment was the slowest among the four groups ( Figure 6F). These data illustrated that DC120 in combination with CDDP showed a significant synergistic effect on the inhibition of NPC cells, and DC120 effectively sensitized NPC cells to CDDP therapy in vivo.

DC120 repressed NPC cancer stem-like SP cells through downregulating Sox2 expression
The subsequent studies focused on the signaling pathways involved in the DC120-mediated repression of stem-like SP cells in NPC cancer cells. As mentioned earlier, as AKT can phosphorylate p27 T157 to impair the nuclear import and function of p27, constitutive activation of AKT promotes CSLC resistance to treatment with chemotherapy and/or radiation therapy partially by down-regulating the expression of p27. Inhibition of PI3K/AKT pathway by a PI3K inhibitor LY294002 or a kinase dead dominant-negative AKT mutant can upregulate the levels of p27 in cancer cells [25,26]. The embryonic stem cell gene SRY (sex determining region Y)-box 2 (Sox2), a transcription factor, is involved in the regulation of embryonic stem cells [24] and plays a key role in maintaining the pluripotent properties of stem cells. As a known tumor suppressor, P27 can directly repress Sox2 expression in induced pluripotent stem cells (iPSCs) through binding to the Sox2 promoter [27]. Our results showed that the protein expression level of p27 was upregulated dose and time dependently in CNE-2-S-18 cells after treatment with DC120, while the protein expression and mRNA level of Sox2 was down-regulated dose and time dependently (Figure 7A, 7B). To further demonstrate the involvement of DC120 in the repression of Sox2 in NPC cells, two different p27 siRNAs were employed to knock down p27 in CNE-2-S-18 cells. As shown in Figure  7C, a transient transfection with 50 nM siRNA downregulated p27 expression by at least 90%. Accordingly, the decrease of the Sox2 expression levels caused by DC120 was blocked when the expression of p27 was down-regulated. To further verified Sox2 was the target gene to regulate the efficacy of the DC120, we found overexpression of Sox2 could prevent from decrease of the SP population in NPC cells by DC120 treatment ( Figure  7D, 7E). Furthermore, knockdown of the Sox2 by siRNA could decrease the proportion of SP cells in NPC ( Figure  S3A, S3B). Collectively, these results demonstrated that DC120 repressed Sox2 through increasing p27 expression.
Growing evidence has shown that microRNAs (miRNAs) are extensively involved in gene regulation by binding to the mRNAs of protein-coding genes to direct their posttranscriptional repression [28]. As a key regulator in development and carcinogenesis, Sox2 also displayed close associations with microRNAs. According to reports previous, at translational level, activities of Sox2 were controlled by several miRNAs, namely microRNA-145, microRNA-126, microRNA-9 and microRNA-21 [29,30]. To explore the role of miRNAs in DC120-mediated Sox2 down-regulation, we analyzed the differences in miRNAs profiles between control (DMSO solution) and DC120 treatment in CNE-2-S-18-SP cells from miRNAs microarrays. The cells were separately cultured with a DMSO solution or 10 μmol/L DC120 for 15 h, and total RNA from all of the cells was extracted. The expressed fold changes of miRNAs from DC120-treated cells were normalized to the DMSO-treated control cells. From the results, 7616 miRNAs with significantly different expression ratios (p < 0.05) were selected. Among these, 80 miRNAs were up-regulated and 78 were downregulated, date was not shown. We then chose 20 miRNAs through the qRT-PCR, and our results showed that 9 miRNAs were up-regulated after treatment with DC120, which was consistent with the results of the microarray ( Figure S4A). Among these, miR-30a exhibited a high fold change (3.5-fold) according to the microarray data. Similarly, DC120-treated cells showed a higher expression of miR-30a (5.3-fold) than control cells in real-time PCR (RT-PCR) analysis ( Figure S4B). These results indicated that miR-30a expression was up-regulated by DC120. To investigate the role of miR-30a in DC120-treated cells, we explored its potential targets using a bioinformatics approach of complementary base pairing ( Figure  S4C). The relative luciferase activity of the reporter that contained the wild-type 3' UTR was significantly suppressed when miR-30a was cotransfected ( Figure  S4D). In contrast, the luciferase activity of the mutant reporter was unaffected by the cotransfection of miR-30a ( Figure S4D), indicating that miR-30a suppressed Sox2 gene expression using the miR-30a-binding sequence at the 3' UTR of the Sox2 gene. Additionally, the ectopic expression of miR-30a caused a decrease in Sox2 protein expression and mRNA levels in CNE-2-S-18 cells ( Figure  S4E-S4G). These data demonstrate that miR-30a may directly target the Sox2 gene via seed matches on both 3' UTR. Sox2 was down-regulated and p27 was up-regulated when CNE-2-S-18 cells were treated with another an AKT inhibitor, GDC0068, which also induced the mRNA level of Sox2 down-regulated as well as miR-30a mRNA level up-regulated ( Figure S5A-S5C). Taken together, our results demonstrate that DC120 regulates Sox2 via upregulation of p27 and miR-30a.

DISCUSSION
Here we report that a new compound DC120 selectively targeted NPC cancer stem-like SP cells and sensitized NPC cells to conventional chemotherapy in vitro and in vivo by blocking the AKT signaling pathway and inhibiting Sox2 expression.
Cancer stem-like cells (CSLCs) refer to a subpopulation of tumor cells with important properties, including self-renewal (by symmetric and asymmetric division), differentiation capacity, chemoresistance to standard cancer therapies, cancer invasion and metastasis [5,[31][32][33]. The therapy of specifically targeted cancer stem cells may be the breakthrough for tumor therapy [34,35]. Diverse cell surface markers such as CD44 and CD133 have been used for the identification of CSCs in human tumors. So far, the markers for nasopharyngeal carcinoma stem cells is not clear. Hoechstexcluding side populations sorted from hematological and solid malignancies, including ovarian cancer [24], breast cancer [36], esophageal carcinoma [13] and nasopharyngeal carcinoma [37], have been reported to be enriched for CSLCs. In our study, we also found that NPC cancer stem-like SP cells had the ability of self-renew and initiate tumors through nasosphere formation assays, colony formation assays and NOD/SCID xenograft assays ( Figure 1B and 1F).
Activation of the PI3K/AKT pathways is critical for the regulation and maintenance of cancer stem-like cells in breast cancer, prostate cancer, esophageal cancer, glioblastoma multiforme, and non-small cell lung cancer [4,14,15,36,38,39]. Furthermore, PI3K/AKT signaling is also associated with chemoresistance in a subset of cancer cells called SP cells, which contribute to drug resistance [40]. Therefore, targeting PI3K/AKT signaling may be beneficial in cancer treatment by eliminating cancer stem-like SP cells. In this study, we first found that AKT kinase and its signaling pathway were hyperactivated in NPC cancer stem-like SP cells (Figure 2A). Additionally in the presence of DC120 phosphorylation of AKT downstream targets such as FKHRL1 and GSK-3β were markedly decreased in cancer stem-like SP cells. Simultaneously, a concomitant increase in the Thr308 and Ser473 phosphorylation of AKT was also observed ( Figure 2D, 2E), which might be caused by a feedback loop induced by DC120 or a direct consequence of DC120 binding to the ATP binding site of AKT, as we previously reported [20]. Importantly, we found that DC120 could preferentially target cancer stem-like SP cells in vitro and in vivo. Firstly, we found that SP cells were more sensitive to DC120 by MTT assays (Figure 3A), then we confirmed that DC120 noticeably suppressed cancer stem-like SP cells' self-renewal and tumor-initiating abilities by reducing the cancer stem-like SP cell fraction and nasosphere formation in vitro ( Figure 3B and 3D). It is worth noting that the concentrations of DC120 that were capable of suppressing nasosphere formation and the proportion of cancer stemlike SP cells were much lower than those that exhibited anti-proliferative effects in CNE-2-S-18 and CNE-1 cells ( Figure 3B and 3D). Another recognized AKT inhibitor GDC0068 was employed and the same effect was obtained ( Figure S1A-S1C). In addition, DC120 decreased the SP percentage in xenograft tumors, and single cells isolated from these tumors lost their nasosphere-forming capacity and reduced the regrowth of tumors in secondary mice ( Figure 6B and 6F), which is the more compelling evidence that DC120 is able to effectively deplete cancer stem-like SP cells in vivo. However, it should be noted that although DC120 preferentially targeted cancer stem-like SP cells, the compound was able to inhibit proliferation in most of the cancer cells ( Figure 3A) and primary tumor growth of CNE-2-S-18 xenografts ( Figure 6A).CDDP alone mainly targeted at the NSP population of NPC cells, and DC120 mostly aimed at the SP population, the combination treatment was actually more potent in suppressing NPC cell growth ( Figure 5C). Further study confirmed that DC120 decreased cancer stem-like SP cell viability mainly due to induction of apoptosis, as demonstrated by an increased sub-G1 population, Annexin V-positive cells and cleaved PARP in tumor samples ( Figure 4A-4D). Therefore, we suggested that DC120 could inhibit AKT kinase activity and block its signaling pathway via induction of apoptosis in NPC cancer stem-like SP cells.
The negative cell cycle regulator p27 has previously been reported as a commonly downregulated tumor suppressive protein in NPC [41]. Chan et al. [42] found that stable p27 protein expression may be due to downregulation of the p27 ubiquitination mediator Skp2 through downregulating AKT and pAKT in NPC cells. In our present study, we found that DC120 targeted AKT in cancer stem-like SP cells, and the effect was correlated with the reduction of pAKT and the increased expression of p27 ( Figure 7A). Recently, a new study reported that p27 directly repressed the expression of stem cell marker Sox2 during embryonic stem cell differentiation [27]. The embryonic stem cell gene SRY (sex determining region Y)-box 2 (Sox2), a transcription factor, is involved in the regulation of embryonic stem cells [24] and plays a key role in maintaining the pluripotent properties of stem cells [43]. Further studies revealed that Sox2 was a crucial player in maintaining the stemness of GSC (Glioma stem cells) through miR-9 [44], CSCs in HNSCC (Head and Neck squamous cell carcinoma) through miR-302 [45], CSCs in breast tumors [25], mammalian neural stem cells [46], CD44+ cancer stem-like cells in EBV-associated nasopharyngeal carcinoma [47], cancer stem-cell from squamous-cell carcinoma [48] and SP (side population) cells in NSCLC [15]. All these studies revealed that Sox2 expression promoted and maintained the stemness of CSCs through similar manner.
In this study, we also proved that knockdown of the Sox2 by siRNA could decrease the proportion of SP cells in NPC ( Figure S3A, S3B). Also, we found that DC120 could inhibit Sox2 expression through increasing p27 expression ( Figure 7A). Significantly, Sox2 was restored when p27 was knocked down by specific siRNAs after treatment with DC120 ( Figure 7C). Apart from p27, miRNAs are one of the most common regulatory factors in NPC [49][50][51], and they play physiological and pathophysiological roles in cell proliferation, differentiation and apoptosis by regulating protein expression in a post-transcriptional manner [52,53]. Wu et al. [54] identified miR-30d as a tumor suppressor, which was regulated by a new AKT/FOXO/miR-30d/ MTDH signaling transduction pathway. In our study, we found that miR-30a was up-regulated after DC120 treatment ( Figure S4B). Meanwhile, our results revealed that Sox2 was directly targeted by miR-30a to induce Sox2 down-regulation ( Figure S4F, S4G). Above all, our study proved that DC120 inhibited Sox2 via increasing p27 and miR-30a expression. Overexpression of Sox2 prevented from decrease of the SP population in NPC cells by DC120 treatment (Figure 7D, 7E). Combined with DC120mediated Sox2 downregulation and repression of the SP population and knockdown of Sox2-induced decrease of the SP percentage, this result suggested that DC120 could repress SP population via down-regulation of Sox2.
Combining AKT inhibitors with other cancer therapeutics is a promising way to improve the tumor therapeutic window. In this study, we found that DC120 could enhance NPC cells' chemosensitivity to CDDP therapy. Although CDDP could inhibit cancer cell proliferation, NSP cells were more sensitive to CDDP than SP cells ( Figure 5A), and it increased the percentage of cancer stem-like SP cells ( Figure 5B), which further proved that conventional treatments target the bulk of cancer cells but do not affect CSCs [24]. However, DC120 effectively reversed the resistance of NPC cells to CDDP treatment ( Figure 5C) and showed a synergistic effect in combination with CDDP (Table 2 and Table 3), which supports further exploration of the combination efficacy of DC120 with other anticancer agents in anticancer therapy.
In summary, our data provide evidence of DC120's sustained anti-cancer effect in stem-like SP cells and a promising development of future therapies with DC120. DC120 preferentially suppressed the proliferation of cancer stem-like SP cells and the bulk of cancer cells. The combination of DC120 and CDDP had a synergistic effect in targeting cancer stem-like SP cells. These data provide validation for the development of DC120 to treat cancer stem-like SP cells and offer a novel treatment strategy for NPC in the future.

DC120 preparation
For all of the in vitro studies, compound DC120 was dissolved in DMSO at a concentration of 50 mM and stored at −20°C. For the tumor xenograft studies, DC120 was formulated in 8% solvent diluent (DMSO/ (Cremophor EL+ethanol), 1:3) at a concentration of 50 mg/ml. Its structure was reported previously [20,21]. www.impactjournals.com/oncotarget

Cell culture and reagents
The human NPC cell line CNE-1 was cultured and conserved by Sun Yat-sen University cancer center from 1982 and have been used in previous study [41], and the S18 cell line is a single cell clone of CNE-2 (a kind gift from Dr. Chao-Nan Qian in 2011, China) [55] and has also been used in previous study [22]. Both cell lines were authenticated by Applied Biosystems on Nov 16, 2012 via STR analysis. 293T cell line was obtained from the American Type Culture Collection in 2010. The cells were cultivated in DMEM medium supplemented with 10% fetal bovine serum in a 5% CO2 humidified atmosphere at 37°C. GAPDH, AKT, phospho-AKT (Ser473), phospho-AKT (Thr308), GSK3α/β, cleaved PARP, p27, Sox2, α-tubulin primary antibodies and a horseradish peroxidase-conjugated secondary antibody were purchased from Santa Cruz Biotechnology (California, CA, USA). Anti-phospho-FKHRL1, phospho-GSK3β, and chemiluminescence reagents were obtained from Cell Signaling Technology (Beverly, MA, USA). Hoechst 33342, Fumitremorgin C (FTC), DAPI, MTT and DMSO were acquired from Sigma-Aldrich (St. Louis, MO, USA).

MTT assay
The cells were seeded in a 96-well plate (Falcon) at 3,000 to 5,000 cells per well. Cell viability was determined by MTT assay. The IC 50 and combination index (CI) values were calculated using CalcuSyn software (Biosoft) as described previously [20,56].

Fluorescence-activated cell sorting (FACS) assay
The CNE-2-S-18 and CNE-1 cells or single cells obtained from primary xenograft tumors were harvested and resuspended at a concentration of 1 × 10 6 cells/ml. Hoechst 33342 was then added to a final concentration of 5 μg/ml and incubated for 90 min at 37°C in the dark with interval mixing. After washing twice with PBS, the cells were kept at 4°C in the dark before flow cytometry analysis (EPICS ALTRA Flow Cytosorter, Beckman Coulter). Meanwhile, a subset of the cells was incubated with 5 μM FTC for 5 min at 37°C prior to adding Hoechst 33342. Flow cytometry data were analyzed using FlowJo software.

Suspension culture
CNE-2-S-18/SP, CNE-2-S-18/NSP, CNE-1/SP and CNE-1/NSP cells or single cells obtained from primary xenograft tumors were counted, then plated at 1,000 cells per well in ultra-low attachment 6-well plates (Corning) with DMEM/F-12 medium mixed with 20 ng/ ml epidermal growth factor (BD Biosciences), 20 ng/ml basic fibroblast growth factor (Invitrogen), and B-27 supplement (Invitrogen). Then, different concentrations of DC120 were added to the SP cells, and the cells were cultured for approximately 2 weeks. The spheroids, also called nasospheres, were photographed and counted using a microscope.
Colony formation assay S18-SP, S18-NSP or CNE1-SP, CNE1-NSP cells were counted, plated in triplicate at 200 cells per well in 6-well plates (Corning), and cultured in DMEM (supplemented with 10% fetal bovine serum) for approximately 7 days. Then, the cells were washed twice with PBS and fixed in methanol for approximately 10 minutes. After two additional washes with PBS, the cells were dyed with crystal violet for 30 minutes. Then, the crystal violet was washed out and the numbers of the colonies were counted.

Non-obese diabetic/severe combined immunodeficient (NOD/SCID) mouse model
The procedures involving mice and their care were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals with the UKCCCR (UKCCCR, 1998). NOD/SCID mice were purchased from the animal institute of the Chinese Academy of Medical Science. CNE-2-S-18 cells were suspended in serum-free DMEM media after FACS, and both CNE-2-S-18/NSP and CNE-2-S-18/SP cells were injected subcutaneously into the left and right flank of the mice at 1 × 10 3 , 1 × 10 4 , 1 × 10 5 , and 1 × 10 6 cells. The tumors were measured with a caliper, and the volume was calculated using V = 1/2(width 2 × length).

Flow cytometry
Cells were cultured in a six-well plate and exposed to DC120. The cells were harvested and stained with Annexin-V-FLUOS Solution and PI Solution provided by the Annexin-V-FLUOS Staining Kit (Roche, Switzerland) for 10-15 min or fixed overnight in 75% ethanol at −4°C and resuspended in 1 ml of staining solution (50 μg/ml PI, 50 μg/ml RNase) for 15 min. Then, apoptotic or sub-G1 phase cells were analyzed by flow cytometry (Beckman Coulter, USA).

Western blot analysis
The cells were harvested and lysed in 1× cell lysis buffer (Cell Signaling Technology) with 1 mM phenylmethanesulfonyl fluoride (PMSF) added immediately before use. The protein concentration was estimated using a Pierce BCA protein assay kit. Equal amounts of protein (20 μg-40 μg) were separated electrophoretically on 8%-15% SDS-polyacrylamide gels, transferred onto PVDF membranes (Millipore), and analyzed as previously described [20]. www.impactjournals.com/oncotarget Luciferase reporter assay A reporter assay using luciferase was performed as previously described [57]. 293T cells (2 × 10 4 ) plated in a 24-well plate were cotransfected with internal control renilla luciferase (Rluc), reporter firefly luciferase (Fluc) with a wild-type or mutated Sox2 3' UTR, and either miR-30a minics or a miR-30a inhibitor using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Fluc and Rluc activities were assayed using the Dual Luciferase Reporter Assay System (Promega, Fitchburg, WI, USA) 48 hours after transfection, and Fluc activity was then normalized to Rluc activity. The Fluc-Sox2 3' UTR construct was generated from 293T cell cDNA using 5ʹ CCGCTCGAGGGGCCGGACAGCGAACTGGAGG 3ʹ and 5′ATAAGAATGCGGCCGCTCAGTGTCCATATTT CAAAAATTTATTT 3ʹ primers. Each treatment was performed in triplicate and repeated in three independent experiments.

Silencing of protein expression by siRNA transfection
siRNA against p27 and negative control siRNA were purchased from Santa Cruz Biotechnology. Cells were transfected with siRNA using Lipofectamine RNAiMAX Transfection Reagent (Invitrogen) according to the manufacturer's instructions. The final concentration of p27 siRNA was 50 nM. Protein expression was measured 48 h after transfection by western blotting.

miRNA microarray
The freshly sorting CNE-2-S-18-SP cells were cultured in 6-well plates with either DMSO or 10 μmol/L DC120, and allowed to reach logarithmic growth phase. After 15 hours, total RNA was extracted and analyzed by miRNA microarrays (miRCURY LNA™ microRNA Array (v.18.0), KangChen Inc, shanghai,China). The expressed fold changes of miRNAs from DC120-treated cells were normalized to the DMSO-treated control cells. Microarray data have been deposited in GEO (Series accession number GSE59503).

In vivo antitumor activity
BALB/c nude mice were obtained from Hunan SlacJingda Laboratory Animal Co. Ltd and were 4 to 6 weeks old. All manipulations were performed under sterile conditions. Primary tumor xenografts were established by injecting 2 × 10 6 CNE-2-S-18 cells into mice. The mice were randomly divided into 4 groups, and each group contained 6 mice. The treatments were initiated on day 5 after inoculation, by which time the tumor volume had reached approximately 50 mm 3 . The treatments, including (i) 8% solvent diluent (negative control), (ii)CDDP × 4 days, (iii) DC120 daily or (iv)CDDP × 4 days+DC120 daily, were administered intraperitoneally for 16 days for each group. The tumor volumes and body weights of the mice were recorded. When all of the control tumors developed to more than 2,000 mm 3 , the nude mice were sacrificed. Tumor growth inhibition (T/C %), which was used to evaluate the tumor response to the drugs, was calculated using the ratio of the average tumor weight of the treated group (T) to the average tumor weight of the control group (C). Fresh cells from the dissociated primary tumors were suspended and counted. All of the 6 secondary nude mice were inoculated with 100,000 cells from each group of primary mouse tumors. The growth of the tumors was monitored, and the tumor volumes were measured twice weekly.

Statistical analysis
Student's t-test was used to evaluate the statistical significance of the result at the 95% confidence level, and a P value less than 0.05 was considered statistically significance.