MicroRNA-mediated epigenetic targeting of Survivin significantly enhances the antitumor activity of paclitaxel against non-small cell lung cancer

Elevated expression of Survivin correlates with poor prognosis, tumor recurrence, and drug resistance in various human cancers, including non-small cell lung cancer (NSCLC). The underlying mechanism of Survivin upregulation in cancer cells remains elusive. To date, no Survivin-targeted therapy has been approved for cancer treatment. Here, we explored the molecular basis resulting in Survivin overexpression in NSCLC and investigated the antitumor activity of the class I HDAC inhibitor entinostat in combination with paclitaxel. Our data showed that entinostat significantly enhanced paclitaxel-mediated anti-proliferative/anti-survival effects on NSCLC cells in vitro and in vivo. Mechanistically, entinostat selectively decreased expression of Survivin via induction of miR-203 (in vitro and in vivo) and miR-542-3p (in vitro). Moreover, analysis of NSCLC patient samples revealed that the expression levels of miR-203 were downregulated due to promoter hypermethylation in 45% of NSCLC tumors. In contrast, increased expression of both DNA methytransferase I (DNMT1) and Survivin was observed and significantly correlated with the reduced miR-203 in NSCLC. Collectively, these data shed new lights on the molecular mechanism of Survivin upregulation in NSCLC. Our findings also support that the combinatorial treatments of entinostat and paclitaxel will likely exhibit survival benefit in the NSCLC patients with overexpression of DNMT1 and/or Survivin. The DNMT1-miR-203-Survivin signaling axis may provide a new avenue for the development of novel epigenetic approaches to enhance the chemotherapeutic efficacy against NSCLC.


INTRODUCTION
Lung cancer remains the leading cause of cancerrelated deaths worldwide [1,2]. The current treatment for lung cancer patients who have been diagnosed at an early stage is surgical resection followed by chemotherapy. However, majority of the patients will eventually experience disease progression and require further treatment [3].
Although a number of new therapies have been developed for the patients with non-small cell lung cancer (NSCLC) [4], which accounts for 85% to 90% of all lung cancer cases [5], the 5-year survival rate of patients with advanced NSCLC remains low [6].
Survivin, the smallest member of IAP (inhibitor of apoptosis) family, is overexpressed in tumors, but not normal tissues [7,8]. Increased Survivin correlates with Research Paper poor prognosis, tumor recurrence, and drug resistance in a wide variety of human cancers, including NSCLC [9,10]. Several approaches, such as the transcriptional inhibitor of Survivin YM155, antisense oligonucleotides, immunotherapy, and gene therapy have been designed to inhibit Survivin for cancer treatment [10,11]. YM155 is the most advanced studied and actively under clinical evaluations (http://www.clinicaltrial.gov/ct2/results?term= YM155&Search=Search). A recent Phase II trial indicates that YM155 is safe, but it fails to increase the response rates to chemotherapy in patients with advanced NSCLC [12]. This failure is likely due to YM155's insufficient inhibition of Survivin. It is believed that novel strategy effectively downregulating Survivin is required to enhance chemotherapeutic efficacy in the treatment of NSCLC patients. Nonetheless, the underlying mechanism of Survivin upregulation in NSCLC remains unclear.
Currently, the platinum-based doublet regimens containing paclitaxel and cisplatin are the standard of care for advanced NSCLC [13]. Paclitaxel, either as monotherapy or combined with other agents, has shown potent anti-NSCLC activity [14,15]. However, both de novo and acquired resistance to paclitaxel frequently occurs and represents a huge clinical problem [16]. We showed that overexpression of erbB3 upregulated Survivin to confer paclitaxel resistance in erbB2-overexpressing breast cancer cells [17]. Inhibition of Survivin via a shRNA or an anti-erbB3 antibody (Ab) significantly increased the cytotoxicity of paclitaxel [17,18]. In identifying novel approach targeting of erbB3, we discovered that the class I histone deacetylase (HDAC) inhibitor entinostat specifically increased miR-125a, miR-125b, and miR-205, which acted in concert to inhibit erbB3, and subsequently induced apoptosis in erbB2overexpressing breast cancer cells [19,20]. Entinostat exerts potent antitumor activity in a number of cancers [21]. Its clinic activity against NSCLC is being tested in combination with DNA methyltransferase (DNMT) inhibitor Azacitidine or EGFR inhibitor Erlotinib (Tarceva), but not conventional chemotherapy (http://www.clinicaltrial.gov/ct2/results?ter m=entinostat&Search=Search). While epigenetic therapy emerges as a new strategy to overcome drug resistance and re-sensitize cancer cells to chemotherapy [22,23], we wondered whether entinostat might possess such activity as a chemo-sensitizer via inhibition of Survivin. In the current report, we investigated the mechanism of action of entinostat in potentiation of paclitaxel-mediated antitumor activity against NSCLC. We have also explored the molecular basis responsible for Survivin overexpression in NSCLC.

Entinostat enhances the anti-proliferative/antisurvival effects of paclitaxel on NSCLC cells
To explore the therapeutic potential of entinostat against NSCLC, we first studied the inhibitory activity of entinostat combined with paclitaxel in NSCLC cell lines. While entinostat alone slightly reduced proliferation of A549 and H460 cells, it significantly increased paclitaxel-mediated growth inhibition in both lines ( Figure 1A). This data was supported by clonogenic assays showing that combinations of entinostat and paclitaxel as compared to either agent alone dramatically reduced the colony numbers ( Figure 1B). To evaluate whether the combinatorial inhibition via a similar mechanism observed in breast cancer and multiple myeloma cells [24,25], we examined the effects of entinostat and/or paclitaxel on induction of apoptosis. Treatment of A549 or H460 cells with both entinostat and paclitaxel, but not either agent alone, clearly induced PARP cleavage, a hallmark of apoptosis, and activation of caspase-8 and caspase-3 ( Figure 1C). An ELISA measuring histone-associated DNA fragments further confirmed that the cells underwent apoptotic cell death upon the combinatorial treatments ( Figure 1D). These data indicate that entinostat significantly enhances paclitaxel-induced anti-proliferative/anti-survival effects on NSCLC cells associated with caspase-dependent apoptosis.

Entinostat selectively inhibits Survivin independent of PI-3K/Akt/mTOR signaling
To explore the mechanism by which entinostat potentiates paclitaxel-induced apoptosis in NSCLC cells, we first examined several anti-apoptosis proteins, including Bcl-xL, Mcl-1, and Survivin in A549 and H460 cells. Entinostat markedly decreased both protein and mRNA levels of Survivin, but not Bcl-xL and Mcl-1, in a timeand dose-dependent manner (Figure 2A and 2B). Although entinostat inhibited PI-3K/Akt signaling and activation of the PI-3K/Akt pathway led to Survivin upregulation in breast cancer cells [17,19], it had no effect on phosphorylation of Akt (P-Akt) and MAPK (P-MAPK) in NSCLC cells ( Figure 2C). Consistently, specific inhibitor of PI-3K (LY294002), Akt (Akt1/2 inhibitor VIII), or mTOR (rapamycin) did not alter Survivin expression ( Figure 2D). We next studied whether the reduction of Survivin played a causal role for entinostat to potentiate paclitaxel-induced apoptosis. Ectopic expression of Survivin in A549 or H460 cells abrogated entinostat enhancement of paclitaxelinduced PARP cleavage, activation of caspase-8 and -3 and DNA fragmentation ( Figure 3). Finally, we tested whether Survivin reduction was able to enhance the cytotoxicity of paclitaxel. Two verified shRNA sequences were used to downregulate Survivin [17]. As expected, specific knockdown of Survivin not only enhanced paclitaxelmediated growth inhibition, but also significantly increased paclitaxel-induced apoptosis (Supplementary Figure S1). Collectively, our data demonstrate that entinostat elicits a PI-3K/Akt/mTOR pathway-independent mechanism to downregulate Survivin, which is essential for entinostat to potentiate paclitaxel-mediated cytotoxicity in NSCLC cells. www.impactjournals.com/oncotarget (A) A549 and H460 cells were plated onto 96-well plates. After 24 h, the culture media were replaced with fresh media or the same media containing indicated concentrations of either entinostat (Ent) or paclitaxel (Pac) or both entinostat and paclitaxel for another 72 h. The percentages of surviving cells from each cell line relative to controls, defined as 100%, were determined by reduction of MTS. (B) A549 and H460 cells were grown in triplicates in the absence or presence of either entinostat (1 μmol/L for A549 and 0.2 μmol/L for H460) or paclitaxel (4 nmol/L) alone or both entinostat and paclitaxel for 2-3 weeks. The pictures and numbers of the cell colonies were obtained by the QuantiOne software of Fluor-S ™ Multimager. (C and D) A549 and H460 cells were treated with either entinostat (3 μmol/L and 0.5 μmol/L for A549 and H460, respectively), or paclitaxel (6 nmol/L and 3 nmol/L for A549 and H460, respectively) alone, or both entinostat and paclitaxel for 24 h. Cells were collected and subjected to western blot analyses of PARP (F-PARP, full length PARP; C-PARP, cleaved PARP), caspase-8 (Pro-Casp-8, pro-caspase-8; C-Casp-8, cleaved caspase-8), caspase-3 (Pro-Casp-3, pro-caspase-3; C-Casp-3,  As miRNAs emerge as important regulators of gene expression by targeting mRNA for degradation or translational repression [26], we asked if entinostat might modulate specific miRNAs to control Survivin expression in NSCLC cells. Quantitative real-time (qRT)-PCR showed that entinostat significantly induced miR-203 and miR-542-3p, two Survivin-targeting miRNAs [27,28], in a dose-and time-dependent manner ( Figure 4). It had no effect on Let-7c and miR-29b, which have been reported to target Bcl-xL and Mcl-1, respectively [29,30]. It appeared that chemotherapy alone had no effect on the miRNAs in vitro, as treatment of A549 or H460 cells with paclitaxel did not alter the expression of either miR-203 or miR-542-3p (Data not shown). We next explored whether the induction of miR-203 and/or miR-542-3p was sufficient to inhibit Survivin. While transfection with the mimic of either miR-203 or miR-542-3p slightly decreased Survivin, the combinations of both miRNA mimics profoundly reduced Survivin in A549 and H460 cells ( Figure 5A). It seemed that miR-542-3p or two miRNA mimic(s) were more effective than miR-203 mimic to enhance paclitaxel-induced growth inhibition ( Figure 5B). Moreover, two mimics exhibited a more potent activity than single mimic to significantly potentiate paclitaxel-induced apoptosis ( Figure 5C and 5D). The miRNA mimics largely recapitulated entinostat's activity to enhance paclitaxel cytotoxicity. We then sought to understand if the induction of miR-203 and/or miR-542-3p was required for entinostat-induced downregulation of Survivin. The lentiviral miRZip-203 or miRZip-542-3p was applied to specifically inhibit miR-203 or miR-542-3p, respectively. High efficiency of lentiviral infection of NSCLC cells was evidenced by visualization of GFP (Supplementary Figure S2). While either miRZip-203 or miRZip-542-3p had little effect on entinostat-mediated reduction of Survivin, inhibition of both miR-203 and miR-542-3p attenuated entinostat's inhibitory effect on Survivin ( Figure 6A), suggesting that miR-203 and miR-542-3p worked cooperatively in entinostat action to downregulate Survivin. As a consequence, simultaneous inhibition of miR-203 and miR-542-3p significantly reversed entinostat enhancement of paclitaxel-induced growth inhibition and apoptosis ( Figure 6B-6D). Taken together, our data indicate that both miR-203 and miR-542-3p contribute to entinostat-mediated reduction of Survivin; and upregulation of the two miRNAs is necessary and sufficient for entinostat potentiation of paclitaxel-induced growth inhibition and apoptosis in NSCLC cells.

Entinostat enhances the antitumor activity of paclitaxel against NSCLC in vivo
To determine whether entinostat holds potential to enhance paclitaxel's antitumor activity in vivo, we took advantage of tumor xenograft models established from A549 or H460 cells. The firefly luciferase-labelled A549 (A549-Fluc) and H460 (H460-Fluc) cells were first tested in vitro, and found that their luciferase activity was in a linear association with the cell numbers seeded (Supplementary Figure S3). We then injected A549-Fluc or H460-Fluc cells into nude mice to establish tumor xenografts. Tumor-bearing mice were then treated with intraperitoneal injection of either PBS, or entinostat, or paclitaxel, or both entinostat and paclitaxel. Paclitaxel slightly reduced tumor growth, whereas entinostat showed clear antitumor effects, consistent with previous findings [18,21]. Importantly, the combinations of entinostat and paclitaxel exhibited a more potent antitumor activity than either agent ( Figure 7A). The bioluminescence signals were monitored by IVIS-200 image system, which showed that treatment with both entinostat and paclitaxel profoundly reduced the luciferase activity (Supplementary Figure S4).
To examine the effects of entinostat and/or paclitaxel on proliferation and apoptosis in vivo, the tumors were subjected to IHC analyses of Ki-67, cleaved caspase-3, and Survivin. Mice treated with paclitaxel retained a similar number of tumor cells positively staining for Survivin or Ki-67 as control mice, whereas entinostat significantly reduced Survivin-and Ki-67-staining cells. Combinations of entinostat and paclitaxel further decreased Survivin-and Ki-67-positive cells. The tumors from mice treated with either entinostat or paclitaxel had a minor, but significant induction of cleaved caspase-3. However, entinostat   combined with paclitaxel markedly increased cleaved caspase-3 ( Figure 7B and 7C). Interestingly, entinostat significantly upregulated miR-203 (not miR-542-3p); and this induction was further enhanced when entinostat was combined with paclitaxel ( Figure 7D). Thus, our studies indicate that entinostat, likely through upregulation of miR-203, potentiates the antitumor activity of paclitaxel against NSCLC via inhibition of cell proliferation and induction of apoptosis in vivo.

Entinostat elicits hypomethylation of miR-203 promoter and downregulation of DNMT1
The expression of miR-203 is frequently downregulated due to promoter methylation in cancers of breast, prostate, liver, and hematologic malignancies [31][32][33][34][35][36]. To date, there is no such report in NSCLC. Although the expression levels of miR-203 are much lower in lung cancer cell lines than that in normal bronchial epithelial cells [37], it is unclear if this reduction is due to miR-203 promoter methylation. Methylation-specific PCR (MSP) showed that treatment with entinostat lowered miR-203 promoter methylation and increased unmethylation levels ( Figure 8A), implying that entinostat might induce miR-203 in NSCLC cells via demethylation of miR-203 promoter. DNA methylation is primarily controlled by a group of enzymes called DNA methyltransferases (DNMTs), among which only DNMT1, DNMT3A, and DNMT3B possess catalytic activity [38,39]. While DNMT3A and DNMT3B are de novo enzymes methylating previously unmethylated DNA, DNMT1 is mainly used to maintain DNA methylation status [40]. It has been shown that DNMT1 stability is regulated by a mechanism involving in acetylation-driven ubiquitination [41], and HDAC inhibition promotes ubiquitin-dependent degradation of DNMT1 in breast cancer cells [42]. We therefore investigated the effects of entinostat on DNMTs in A549 and H460 cells, and discovered that entinostat reduced the protein and mRNA levels of DNMT1, but not DNMT3A and DNMT3B, in a time-dependent manner ( Figure 8B-8D). To test if entinostat exhibited a similar activity in vivo, we performed western blot analysis using the tumors described in Figure 7. DNMT1 was clearly decreased in the tumors from mice treated with entinostat (not paclitaxel). Strikingly, the combinatorial treatment of entinostat and paclitaxel resulted in most significant reduction of DNMT1 ( Figure 9). This additive/ synergistic inhibition on DNMT1 was also observed in vitro (Supplementary Figure S5). Taken together, our data suggest that entinostat elicits miR-203 promoter hypomethylation via reduction of DNMT1 in vitro and in vivo, and thereby enhances expression of this miRNA in NSCLC cells.

Positive correlation of DNMT1 and Survivin expression is observed in NSCLC samples
Since intrinsic resistance to paclitaxel frequently occurs and causes treatment failure in NSCLC [16], it is conceivable to hypothesize that overexpression of DNMT1 may lead to epigenetic silencing of miRNAs to upregulate Survivin, and thereby compromise the efficacy of paclitaxel in NSCLC patients. To test this hypothesis, we first examined the expression of miR-203 and miR-542-3p as well as Survivin and DNMT1 in 20 freshly-obtained NSCLC samples and their self-paired normal adjacent lung tissues (Table 1). We noticed that the mRNA levels of both Survivin and DNMT1 were significantly elevated in NSCLC as compared to normal lung tissues ( Figure 10A). Independent-samples t-tests revealed that the means of relative mRNA values of Survivin and DNMT1 in tumors were 8.73-fold and 3.85-fold higher than that of normal tissues, respectively ( Table 2). Increased protein levels of Survivin and DNMT1 were also found in 8 of 20 NSCLC samples as compared to normal lung tissues ( Figure 10B). To confirm the positive correlation between Survivin and DNMT1, we then retrospectively examined the expression of DNMT1 and Survivin in a set of tissue sections consisting of 61 patients with lung adenocarcinoma. We observed that the staining rates for DNMT1 in NSCLCs and their selfpaired normal adjacent lung tissues were 47.5% (29/61) and 36% (22/61), respectively. The staining rates for Survivin in NSCLCs and normal lung tissues were 47.5% (29/61) and 37.7% (23/61), respectively (Supplementary Table S4). Consistent with the western blot data, elevated expression of DNMT1 and Survivin in NSCLC tumors was supported by IHC analyses ( Figure 10C). Evaluation of the IHC scores showed that expression of both DNMT1 and Survivin was significantly higher in lung adenocarcinoma than that in normal lung tissues ( Figure 10D). The positive correlation between DNMT1 and Survivin was highly significant in both tumors (r = 0.919, P = 0.000) and normal tissues (r = 0.983, P = 0.000) ( Figure 10E). Interestingly, the expression levels of DNMT1 were shown to be significantly correlated with TNM stages as well as age of the patients (r = 1, P < 0.0001 by Spearman rank test), while weakly correlated with gender of the patients (r = 0.30175, P = 0.0181 by Spearman rank test). As such, a strong correlation between the expression of Survivin and TNM stages, but not age or gender of the patients were also revealed by Spearman rank test (Supplementary  Table S6). Overall, our clinical analyses suggest that elevated expression of DNMT1 significantly correlates with increased Survivin in NSCLC, especially lung adenocarcinoma.

Hypermethylation of miR-203 promoter is associated with lower levels of miR-203 and higher levels of both DNMT1 and Survivin in NSCLC tumors
We next focused our clinical studies on miR-203 and miR-542-3p. QRT-PCR analyses found that the expression levels of miR-203, but not miR-542-3p were significantly decreased in NSCLCs as compared to that in normal lung tissues ( Figure 11A). To explore if miR-203 reduction  might be attributed to its promoter methylation, we examined the methylation status of miR-203 promoter in the clinical samples. The results of MSP and/or USP were summarized in Table 1 and partially shown ( Figure 11B). Hypermethylation of miR-203 promoter was discovered in 11 of 20 NSCLC tumors. The hypermethylation was inversely correlated with the levels of miR-203 in NSCLC (r = −0.668, P = 0.001; Figure 11C). Independent-samples t-test showed that the expression of miR-203, but not miR-542-3p, was significantly reduced in NSCLCs as compared to that in normal lung tissues ( Table 2). More importantly, a significant inverse correlation was detected between miR-203 (not miR-542-3p) and the expression of Survivin (r = −0.561, P = 0.01) or DNMT1 (r = −0.502, P = 0.024). In contrast, a positive correlation between Survivin and DNMT1 (r = 0.453, P = 0.045) was observed ( Figure 11D), consistent with our IHC analysis of Survivin and DNMT1 in lung adenocarcinoma ( Figure 10E). Collectively, our data demonstrate that elevated expression of DNMT1 may give rise to miR-203 promoter hypermethylation, and thereby lead to Survivin overexpression in NSCLC.

DISCUSSION
Paclitaxel has long been used to treat NSCLC either as single agent or combined with other therapeutics [14,15]. However, resistance to paclitaxel remains one of the main causes of treatment failure in NSCLC [16]. It is of particular significance to elucidate the underlying mechanism of paclitaxel resistance and identify new   miR-203, miR-542-3p, Survivin, and DNMT1,   Notes: Average value for the miRNA or mRNA abundance in self-paired normal adjacent lung tissues was taken as 1 and the relative miRNA or mRNA abundance in NSCLC tumors was presented as 2^− ΔΔCt . Here, we showed that epigenetic silencing of miR-203, leading to Survivin overexpression, played a key role in determining the sensitivity of NSCLC cells to paclitaxel. The HDACi entinostat selectively reduced Survivin via induction of miR-203 and miR-542-3p, and thereby significantly enhanced paclitaxelinduced apoptosis in NSCLC cells. However, induction of miR-542-3p by entinostat was observed in vitro, not in vivo. Clinical analysis found no difference of miR-542-3p levels between NSCLC tumors and the normal lung tissues. There was no correlation between the expression of miR-542-3p and mRNA levels of either Survivin or DNMT1 ( Figure 11A and 11D). These data suggest that 1) increased DNMT1 may not alter miR-542-3p expression in NSCLC; 2) although miR-542-3p contributes to entinostat-induced growth inhibition and apoptosis in vitro, it has little effect on entinostat potentiation of paclitaxel efficacy in vivo. Collectively, our studies indicate that inhibition of DNMT1 is at least one of the major mechanisms for entinostat to lift epigenetic silencing of miR-203, which results in downregulation of Survivin and thereby enhances paclitaxel-mediated antitumor activity against NSCLC ( Figure 12). Our data is supported by a recent report showing that the pan-HDACi vorinostat or panobinostat significantly promoted cisplatin-induced apoptosis in A549 and H460 cells [43]. Further studies revealed that inhibition of HDAC 1, 2 and 3 with entinostat potentiated cisplatininduced apoptosis in A549 cells [43]. Our conclusions seemed to be different from the findings of another report [44], which showed that epigenetic therapy (azacitidine plus entinostat) did not alter selected NSCLC cells' responsiveness to cisplatin, docetaxel, gemcitabine, and vinorelbine in vitro and in vivo. In addition to the fact that paclitaxel was not tested in the later study [44], this disparity might be explained by the distinct treatment regimens used. 1) We and Riley et al. [43] combined entinostat with chemotherapy (paclitaxel in our study and cisplatin in Riley's) simultaneously, whereas sequential treatment -epigenetic therapy followed by chemotherapy -was evaluated by Vendetti et al. [44]. 2) While epigenetic therapy consisting of both azacitidine and entinostat was used by Vendetti et al. only entinostat was applied in our study and Riley's. Thus, the doses of entinostat in Vendetii's study were much lower than that used in ours and Riley's. Nonetheless, the doses of entinostat we used are still within the clinically-relevant range [21]. Although the data by Vendetti [44] call into question whether epigenetic agents could potentiate chemotherapy in cancer treatment, it is possible that their models and sequential treatments do not fully reflect the genetic and/or epigenetic alterations in a subset of NSCLC patients who may benefit from epigenetic priming. Our data support that further evaluation in both preclinical models and clinical trials to test the therapeutic potential of entinostat in combination with paclitaxel against NSCLC is warranted.
Elevated expression of Survivin is observed in almost all types of human malignancies and positively correlates with poor prognosis, tumor recurrence, and drug resistance [9,10]. However, the molecular mechanisms controlling Survivin expression in cancers have not been fully elucidated. Recent studies show that a number of miRNAs, including miR-203 and miR-542-3p, play critical roles in regulating Survivin expression in cancer cells [45]. Nonetheless, majority of the studies are carried out in preclinical models, it is currently unknown whether the expression of miR-203 and/or miR-542-3p is altered in NSCLC tumors. Our analysis suggests that miR-203 is frequently downregulated due to promoter methylation in NSCLC, and the reduced miR-203 is inversely correlated with the expression of Survivin and DNMT1 ( Figure 11D). These findings support the importance of specific miRNA in regulation of Survivin expression. Thus, it is conceivable to hypothesize that novel therapies which are able to induce expression of miR-203 will effectively downregulate Survivin, and therefore sensitize the cancer cells to paclitaxel treatment. Our data presented in this manuscript with entinostat to enhance miR-203 expression strongly support this notion. Our studies with the second set of clinical samples ( Figure 10 and Supplementary Table S5 & S6) seem to suggest that there is a significant correlation between DNMT1 and Survivin in adenocarcinoma. However, after careful analysis of all of our clinical data, we could not rule out the possibility that this correlation may also occur in squamous cell carcinoma. Our first set of clinical samples (20 freshly-obtained NSCLC samples) contains 10 squamous cell carcinomas ( Figure 11D, Table 1 and Supplementary Table S3). Among these squamous cell carcinomas, increased mRNA expression of both DNMT1 and Survivin has been shown in 4 of them (#6, #11, #15, and #20). Thus, to determine whether the correlation of DNMT1 and Survivin may have preference towards certain histology subtypes of NSCLC, we need to expand the population of patients with all subtypes in our future work.
DNA methylation has long been recognized as a key player in tumor initiation and progression [46]. Yet, epigenetic and genetic mechanisms often intertwine and take advantage of each other during tumorigenesis [47]. Our discovery that increased DNMT1 negatively correlates with miR-203 expression in NSCLC offers a representative paradigm regarding the cross-talk between genetic and epigenetic alterations. As a global enzyme maintaining DNA methylation status, how does the dysregulated DNMT1 link to specific silencing of miR-203 in NSCLC is an unsolved question. A recent study reveals that a long noncoding RNA termed ecCEBPA is critical for regulation of DNA methylation at a particular site via interaction with DNMT1 [48]. Whether a similar mechanism accounts for miR-203 promoter methylation in NSCLC by DNMT1 awaits further investigation.
In summary, we demonstrate that entinostat downregulates Survivin via induction of miR-203 and miR-542-3p in vitro and/or in vivo, and thereby potentiates paclitaxel-mediated antitumor activity against NSCLC. Clinical studies show that increased DNMT1 positively correlates with miR-203 promoter methylation and overexpression of Survivin in NSCLC tumors. These findings shed new lights on the underlying mechanism of Survivin upregulation in NSCLC. Our data suggest that the addition of entinostat to paclitaxel-based regimens in selected NSCLC patients with increased DNMT1 and/or Survivin may exhibit significant survival benefit.

Cells and cell culture
NSCLC cell lines A549 and H460 were from ATCC (Manassas, VA) and maintained in RPMI1640 medium supplemented with 10% fetal bovine serum (FBS). HEK293T human embryonic kidney cells were maintained in DMEM/F12 medium containing 10% FBS. The cells were free of mycoplasma contamination, determined by the MycoAlert™ Mycoplasma Detection Kit (Lonza Group Ltd. Basel, Switzerland) every three months. All cell lines were cultured in a 37°C humidified atmosphere containing 95% air and 5% CO 2 and were split twice a week.

Cell proliferation, western blots, apoptosis, and quantitative real-time PCR (qRT-PCR)
The CellTiter96AQ cell proliferation kit (Thermo Fisher Scientific) to determine cell viability, western blots to examine protein expression and activation, an apoptosis ELISA kit (Roche Diagnostics Corp., Indianapolis, IN) to measure histone-associated DNA fragments, and qRT-PCR with the FastStart Universal SYBR Green Master Mixes (Roche) by a Real-Time PCR system (Applied Biosystems, Foster City, CA) were described previously [17,18,20].

Clonogenic assay
Clonogenic assays were performed as described previously [49,50]. The colony was defined to consisting of at least 50 cells. The colony numbers were quantified with QuantiOne software of Fluor-S™ Multimager (Bio-Rad Laboratories, Inc., Hercules, CA).

Construction of lentiviral expression vector pLEX-hSurvivin
The coding sequence of human Survivin was amplified from pcDNA3-hSurvivin (provided by Dr. Dario Altieri at The Wistar Institute Cancer Center) by PCR with the following forward primer: 5′-ATA GCG GCC GCA TGG GTG CCC CGA CGT TGC-3′ and reverse primer: 5′-GCG ACG CGT TCA ATC CAT GGC AGC CAG CTG-3′. The amplified fragments were inserted into the lentiviral vector pLEX-MCS (Open Biosystem, Huntsville, AL). After sequencing verification, the recombinant was nominated as pLEX-hSurvivin.

Tumor xenograft model
Athymic nu/nu mice (Shanghai SLAC Laboratory Animal Co. Ltd., Shanghai, China) were maintained in accordance with the IACUC procedures and guidelines. Five × 10 6 A549-Fluc or H460-Fluc cells were suspended in 100 μL of PBS, mixed with Matrigel (BD Biosciences) and injected subcutaneously into the flanks of 5-week-old female mice. Tumor formation was assessed by palpation and measured with fine calipers three times a week. Tumor volume was calculated by the formula: Volume = (Length × Width 2 )/2, which was statistically analyzed as we described previously [18,52]. When tumors reach ~150 mm 3 , mice were randomly assigned into four groups (n = 4 and n = 3 for A549-Fluc and H460-Fluc cells, respectively): 1) control mice received intraperitoneal (i.p.) injection of 100 μl of PBS; 2) mice received i.p. injection of entinostat (25 mg/kg for A549-Fluc or 12.5 mg/kg for H460-Fluc) in 100 μl PBS twice a week; 3) mice received i.p. injection of paclitaxel (7.5 mg/kg) in 100 μl PBS twice a week; 4) mice received i.p. injection of entinostat and paclitaxel in 100 μl PBS twice a week. At the end of study, mice were euthanized according to approved protocol. Tumors were excised and subjected to immunohistochemistry analyses and qRT-PCR measurement of miRNA expression.

Bioluminescence imaging
The IVIS200 instru ment (Caliper Life Sciences, PerkinElmer, Waltham, MA) was used. For cultured cells, we imaged luciferase signal by adding PBS or d-luciferin at a concentration of 0.15 mg/mL. For monitoring luciferaselabeled tumor cells in vivo, mice to be imaged were i.p. injected with 150 mg/kg body weight of d-luciferin in 200 μl of PBS and then anesthetized with continuous flow of isofluorane. Imaging of the mice was carried out 10 min later. All images were processed by manu facturer-supplied software for quantitative analysis.

Methylation-specific PCR (MSP)
MSP for DNA methylation was performed as described by Chim et al. [36]. Treatment of DNA with bisulfite for conversion of unmethylated cytosine to uracil was performed with the EpiTect Bisulfite Kit (Qiagen). Primers used for MSP and unmethylation-specific PCR (USP) were shown in Supplementary Table S1. PCR products were loaded onto 1.5% agarose gels and electrophoresed. The images were obtained by exposure of the gels to an ultraviolet trans-illuminator and recorded with Fluor-S™ Multimager (Bio-Rad Laboratories).

Statistical analysis
All results were confirmed by at least three independent experiments. Data are presented as mean ± SD. Statistical analysis was performed using SPSS 19.0 software (SPSS Institute). Student's t tests were used for comparisons of means of quantitative data between groups. The correlations between miR-203 promoter methylation and miR-203 expression, between miR-203 and Survivin expression, and between DNMT1 and Survivin expression were evaluated by Pearson's correlation coefficient (r). Values of P < 0.05 were considered significant. www.impactjournals.com/oncotarget