IL-8 confers resistance to EGFR inhibitors by inducing stem cell properties in lung cancer.

Epidermal growth factor receptor (EGFR)-targeted strategy is limited by resistance. We identify the potential genes involved in EGFR TKI (tyrosine kinase inhibitor) resistance and study the therapeutic mechanism in the non-small cell lung cancers. Potential genes involved in resistance were examined by analyzing datasets from a pair of EGFR TKI-sensitive (PC9) and TKI-resistant cells (PC9/gef). Blood specimens from patients taking EGFR TKI as first-line treatment were used to examine the correlation between drug's efficacy and IL-8 level. The effects of IL-8 on gefitinib-induced apoptosis, stemness, and in vivo tumorigenicity were investigated using established cell lines. We identified IL-8 was up-regulated in gefitinib-resistant cells, and high plasma IL-8 level was correlated with shorter progression-free-survival time. IL-8 overexpression suppressed gefitinib-induced apoptosis in gefitinib-sensitive cells. By contrast, suppression of IL-8 enhanced gefitinib-induced cell death in gefitinib-resistant cells. IL-8 also increased stem-like characteristics including aldehyde dehydrogenase activity, expression of stemness-related genes, clonogenic activity, side-population, and in vivo tumorigenicity. Consistently, knockdown of IL-8 leads to loss of stem cell-like characteristics in gefitinib-resistant cells. Our study demonstrates an important role for IL-8, and suggests IL-8 is a potential therapeutic target for overcoming EGFR TKI resistance.


INTRODUCTION
Lung cancer is the leading cause of cancerrelated deaths worldwide [1]. Epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (TKIs) are successfully used in non-small cell lung cancer (NSCLC) patients harboring EGFR-activating mutations [2]. Despite early responses to EGFR-TKIs, cancers develop resistance after around 10 months of therapy. The EGFR-T790M secondary mutation and c-MET amplification contribute to the majority of acquired resistance [3]. However, there are still some resistant mechanisms are incompletely understood. To facilitate the development of effective therapies against NSCLC, we elucidate the EGFR TKI-resistance machinery underlying tumor progression. www.impactjournals.com/oncotarget We previously demonstrated that epithelialmesenchymal-transition (EMT), a process in which epithelial cells lose cell polarity, confers resistance to gefitinib [4]. EMT is regulated by chemokines [5,6]. CCL18, inducing EMT and chemoresistance, was shown to elevate in adenocarcinoma patients and correlated with poor survival [7]. The chemokine, CXCL1/2, whose release was triggered by chemotherapeutic agents, conferred resistance to doxorubicin [8]. Trastuzumabresistant cells with an EMT phenotype were shown to secrete more IL-6 and be highly enriched with cancer stem cells (CSCs) [9]. CXCL12 evokes mobilization of lung cancer cells and confers a high capacity for self-renewal on these cells [10]. Blocking receptors for the chemokines CXCL1 and CXCL12 retards tumor growth, reduces invasion, eliminates CSCs, and restores drug sensitivity [10,11]. These suggest that chemokines are associated with the features of CSC and treatment resistance.
Accumulating evidence demonstrates that CSCs are capable of driving tumorigenesis, and resistance to chemotherapeutics and TKIs [12][13][14]. CSCs were identified and purified in lung cancer using functional assays including aldehyde dehydrogenase (ALDH) activity and side population [15][16][17]. We previously demonstrated that gefitinib-resistant cells (PC9/gef) exhibited stemlike characteristics, such as high potential for sphere formation, expression of stemness-related genes, and ALDH activity [18]. We previously isolated a small fraction of ALDH-positive cells from gefitinib-sensitive PC9 cells. These ALDH-positive cells have higher expression of stem cell features and are more resistant to gefitinib, and chemotherapeutic agents compared with ALDH-negative PC9 cells [18]. Side population has been shown to be enriched for CSCs [17]. Dr. Shien et al. reported that exposure to high-concentration of gefitinib resulted in expansion of side population and EGFR TKI resistance [19]. Hence, it was suggested that the subset of side population cells in the heterogeneous cancer cell population was resistant to EGFR TKIs.
In this study, we investigated the role of chemokine in modulating cellular sensitivity to gefitinib in lung cancer cells. We found that IL-8 plays a role in the gefitinib-resistance machinery through regulation of the CSC population.

IL-8 was identified as a potential chemokine with EGFR-TKI resistance through genetic screens
To investigate genes capable of conferring gefitinibresistance in lung cancer cells, we selected a pair of cell lines-one gefitinib-sensitive (PC9) and one gefitinibresistant (PC9/gef)-for analysis using oligonucleotide cDNA microarrays. An analysis of differentially expressed genes using the public DAVID bioinformatics resource was set up to identify the predominant molecular network listed in the Kyoto Encyclopedia of Genes and Genomes database (KEGG) (Supplementary Table S2). The differentially expressed genes of the predominant networkcytokine-cytokine receptor interaction included IL-1, IL-6 and IL-8 (Table 1). IL-1A, IL-1B, IL-6, and IL-8 are well-characterized cytokines involved in inflammation or chemoresistance [21]. We examined expression of IL-1A, IL-1B, IL-6 and IL-8 in two pairs of gefitinibsensitive (PC9, and HCC827) and gefitinib-resistant (PC9/ gef, and HCC827/gef) lung cancer cell lines to identify the specific cytokine involved in gefitinib resistance by RT-qPCR. We showed that IL-1A, IL-6, and IL-8 were up-regulated in PC9/gef, but only IL-8 mRNA was up-regulated in HCC827/gef ( Fig. 1a-b). IL-8 protein was significantly elevated in PC9/gef and HCC827/gef (Fig. 1c).
Studied has reported that IL-8 is elevated in the plasma of cancer patients, and IL-8 is associated with poor prognosis and resistance to chemotherapy [22,23]. Accordingly, we investigated whether IL-8 was involved in gefitinib resistance. Besides IL-8, IL-8-specific receptors, CXCR1, was significantly up-regulated in PC9/gef cells ( Supplementary Fig. S1a). CXCR1 is undetectable, but CXCR2 was up-regulated in HCC827/ gef cells (Supplementary Fig. S1b). We suggested that IL-8-CXCR1/2 signaling was involved in EGFR TKI resistance.

High plasma IL-8 level revealed a shorter progression-free-survival of EGFR TKI-treated EGFR-mutation positive lung adenocarcinoma patients
To investigate the association of IL-8 levels with EGFR TKIs responsiveness, we collected peripheral blood samples from 75 stage IV lung adenocarcinoma patients with EGFR-mutation positive tumors and receiving EGFR-TKIs only as the first-line treatment. The EGFR mutation status of these patients was summarized in Supplementary  Table S3. Of the 75 patients, 66 received gefitinib and nine received erlotinib. According to the median plasma IL-8 level (6.74 pg/mL), we divided patients into high-IL-8 and low-IL-8 groups. There were no significant differences in the clinical characteristics of high and low IL-8 groups (Table 2). However, median progression-free survival was longer in the low IL-8 group (13 months) than in the high IL-8 group (8.5 months; p = 0.02; Fig. 1d).

IL-8 increased stem cell-like characteristics in lung cancer cells
EMT is a biological process by which epithelial cells lose cell-cell adhesion, and have less E-cadherin expression. Both the EMT regulator (Slug) and IL-8 were up-regulated in PC9/gef cells and conferred resistance to gefitinib [4]. To investigate whether IL-8 initiated EMT to result in gefitinib resistance, we examined the expression of EMT-related genes. However, overexpression of IL-8 didn't induce EMT or up-regulation of EMT-related genes in PC9/IL-8 cells compared with PC9/mock cells ( Supplementary Fig. S4). The EMT regulator (Slug) and IL-8 are involved in cell motility, and invasion in previous reports and this study, respectively [4]. Slug initiates cell invasion through repression of E-cadherin and activation of  EMT, but IL-8 promotes cell motility by directly activating Rac GTPase (one member of Rho family) instead of regulating EMT mediators in previous studies [24,25]. We previously showed that PC9/gef presented a higher proportion of ALDH-positive compared with PC9 cells and ALDH-positive cells in PC9 cells were resistant to gefitinib [18]. Here, we isolated ALDH-positive and ALDHnegative sub-populations from PC9/gef to determine the correlation between IL-8 expression and ALDH activity. Intriguingly, the ALDH-positive sub-population from PC9/ gef showed simultaneously increased IL-8 ( Supplementary  Fig. S5), supporting a positive correlation between IL-8 and stem-like characteristic. To investigate whether IL-8 contributes to stem-cell like activity, we found that the ALDH-positive cell population was increased in PC9/IL-8 cells compared with PC9/mock cells (Fig. 4a). A number of stem cell-associated genes, including Nanog, Oct4, and Sox2, were also significantly up-regulated in PC9/IL-8 cells (Fig. 4b); others, including Bmi-1, c-Myc, Klf4 and Nestin, were not different. Side populations have been shown to be enriched for cancer stem-like cells. Using Hoechst 33342 dye and verapamil to characterize the side population, we found a significant increase in the side population fraction in PC9/IL-8 cells compared to PC9/mock cells (0.33% vs. 0.09%) (Fig. 4c). Using another pair of cells, we found a significant increase of side population fraction in HCC827/IL-8 cells compared with HCC827/mock Fig. S6a-c). Furthermore, we performed colony-forming assays to investigate whether IL-8 drives the development of stem cell-like properties. Compared with PC9/mock cells, PC9/ IL-8 cells gave rise to more and larger colonies in soft-agar cultures (Fig. 4d). We also found a significant increase of colonies in HCC827/IL-8 than HCC827/mock cells by clonogenic assay (Supplementary Fig. S6d), indicating IL-8 contributes to the inhibition of gefitinib-induced cell death and the development of stem cell-like characteristics. To further evaluate the effect of IL-8 on cell proliferation, we performed assays on growth rate, cell cycle and proliferation-related proteins (PCNA and cyclin D1). The cell proliferation rates of mock-infected and IL-8-infected cells were similar ( Supplementary Fig. S7a). Moreover, over-expression of IL-8 didn't result in change of cell cycle and proliferation-related proteins ( Supplementary Fig. S7b-c). Therefore, we excluded that IL-8 increased cell proliferation to lead to more colonies.
To determine the tumorigenicity in vivo, we subcutaneously injected cells into NOD.SICD mice and found that PC9/IL-8 cells gave rise to more visible tumors than PC9/mock cells (Fig. 4e). And, Hematoxylin-Eosin (H&E) stains of tumors from NOD.SCID mice injected with PC9/mock or PC9/IL-8 cells were shown in Fig. 4f. These results suggest that the IL-8-expressing cell population harbors a larger compartment of stem-like cancer cells with a higher potential for tumorigenicity.   (Continued ) www.impactjournals.com/oncotarget

Inhibition of IL-8 reduced the stemness in EGFR TKI-resistant cells
We previously showed that PC9/gef cells presented a higher proportion of ALDH-positive cells than PC9 cells [18]. To identify more stemness-like abilities, we made an attempt to conduct side population analysis and investigate stemness-related gene expressions.

IL-8 conferred resistance to chemotherapy
Since IL-8 induced stem-like properties, we suggested that IL-8 confers resistance to traditional chemotherapy (such as paclitaxol). Cell viability assays were performed to evaluate the paclitaxol-induced apoptosis between PC9/mock and PC9/IL-8 cells. Both the percentage of AnnexinV-positive cells and paclitaxol-induced cell death were inhibited in PC9/IL-8 cells (Supplementary Fig. S9a-b). Moreover, we made an attempt to suppress IL-8 activity using neutralizing IL-8 antibody and investigated whether it can promote paclitaxol-induced cell death. We showed that neutralizing IL-8 antibody enhanced www.impactjournals.com/oncotarget Location of the side population in a representative experiment is indicated by gate and dot plots. Right: Quantification of results represents the mean ± s.d. for n = 3 independent experiments. (***p < 0.001). B. The expression of stemness genes of PC9 and PC9/gef cells was quantified by RT-qPCR; the bar graph represents the mean ± s.d. for three determinations (***p < 0.001). C. Stem cell-related gene expressions in PC9/gef cells was examined by RT-qPCR after transient transfection with 50 nM IL-8 siRNA (siIL-8) for 48 hours. The bar graph represents the mean ± s.d. for three determinations (**p < 0.01, and ***p < 0.001). D. ALDH activity were examined by incubating with ALDH substrate in the presence or absence of DEAB after silencing IL-8 in PC9/gef cells.

Effect of exogenous IL-8 on stemness and EGFR-TKI resistance
To investigate whether exogenous IL-8 contributes to stemness and EGFR-TKI resistance, recombinant human IL-8 (rhIL-8) was used. Treatment with rhIL-8 significantly increased Nanog, Oct4, and Sox2 mRNA in PC9 cells, as determined by RT-qPCR (Fig. 6a-c). This result was consistent with the observation in Fig. 4b, as we showed that Nanog, Oct4, and Sox2 were up-regulated in PC9/IL-8 cells. Further, we evaluated whether exogenous rhIL-8 could confer resistance against gefitinib. PC9 cells were cultured with various concentrations of rhIL-8 for one day, and then exposed to gefitinib for one day. As shown in Fig. 6d, IL-8-treated PC9 cells displayed resistance to gefitinib. This result indicates that exogenous rhIL-8 exhibits same function as endogenous IL-8 does.

DISCUSSION
EGFR-TKIs have been widely used in treating nonsmall cell lung cancers with EGFR mutation-positive [26]. In the present study, IL-8 is highly associated with the cellular stemness and EGFR TKI resistance in lung cancer cells. We demonstrated that IL-8 elicits a protective effect against gefitinib-induced apoptosis and caspase activation. Knockdown of IL-8 circumvented EGFR TKI resistance and promoted gefitinib-induced apoptosis. Delineating the function of IL-8 in EGFR TKI resistance, we showed that IL-8 is capable of driving stem cell-like characteristics in lung cancer cells. Etopic introduction of IL-8 resulted in increase of side population, a greater potential for anchorage-independent growth, higher ALDH activity, and in vivo tumorigenic activity. In contrast, IL-8 inhibition in gefitinib-resistant cells diminished stem  cell-like characteristics, and increased gefitinib-induced apoptosis. Moreover, high plasma IL-8 was associated with poor-progression-free survival in EGFR mutationpositive NSCLC patients receiving EGFR TKI as first-line treatment.
In the study, we identified that several chemokines were up-regulated in gefitinib-resistant cells including CXCL-1, CXCL-2, CXCL-6, and IL-8 (Table 1). Accumulating evidences have revealed that chemokines promote cancer cell growth and are associated with treatment resistance [25,27]. Stromal cell-derived factor-1 is associated with tumor recurrence [28], and CXCL-1/2 mediate the chemoresistance in lung cancer [8]. Likewise, the role of IL-8 in mediating cancer cell migration, proliferation, and chemoresistance in an autocrine or paracrine manners has been demonstrated previously [27,29]. Elevated IL-8 levels have been demonstrated in several type of cancers, and was correlated with survival or relapse [30][31][32][33]. Previous reports showed that the serum levels of IL-8 in lung cancer patients were in the pg/mL range [34][35][36], which is similar to our study. Previous studies also demonstrated that the level of IL-8 in cell supernatant (100 pg/mL) is consistent with serum level of IL-8, such as HCT116 cells [27,36]. Furthermore, the level of IL-8 in cell line culture medium depended on cell type and absolute cell number [33,36]. In this study, we showed the level of IL-8 was consistent with previous reports.
Increased tumorigenic activity, and re-activation of survival signaling including PI3K/Akt, and Ras/Raf/ MAPK by IL-8 stimulation are demonstrated to confer resistance against chemotherapy [6,29]. We showed that Akt is hyper-phosphorylated in IL-8-overexpressing cells. The PI3K/Akt pathway is one of the principal downstream effectors of IL-8-dependent, CXCR1/2-mediated signaling, leading to tumor progression [37]. Activation of the CXCR1/2 signaling by IL-8 leads to activation of NF-κB, forming a positive feedback loop further promotes tumor development [38]. In contrast, shutdown of IL-8 signaling via siRNA or neutralizing antibodies leads to sensitizes cancer cells to chemotherapy [39]. IL-8 was previously supposed to transactivate EGFR signaling in NSCLC [40]. Here, we showed the first time that IL-8 plasma level in NSCLC patients is correlated with duration of EGFR TKI treatment in NSCLC (Fig. 1d), and IL-8 intervenes in resistance of EGFR TKIs (Figs. 2, 3, and Supplementary Fig. S3).
As a notable finding, soluble cytokines and chemokines may regulate stem cell-like characteristics, and emerging evidence has shown that CSCs are responsible for tumor growth and metastasis, and contribute to chemotherapy resistance and relapse. CSC's marker expression is a critical element to distinguish cancer cells into stem-like or non-stem-like cells. Side population, ALDH activity and Naong expression have been recognized as hallmarks of lung stem cell markers [41].
Knockdown of IL-8 resulted in inhibition of ALDH1A1, Bmi1, Nanog, Sox2 (> 30% reduction) both in PC9/gef and HCC827/gef cells (Fig. 5c and Supplementary Fig. 8d). However, we observed that Bmi-1 and Nestin were up-regulated only in HCC827/gef cells, but not in PC9/gef cells ( Fig. 5b and Supplementary Fig. 8c). It indicated that stemness-related genes may be differently regulated by IL-8 in PC9/gef and HCC827/gef cells. Previous studies showed that IL-8 and its cognate receptors, CXCR1/ CXCR2, played an important role in CSC activity and chemoresistance. Recently, several studies indicated that specific blockade of CXCR1 or CXCR2 with neutralizing antibodies resulted in different inhibition of CSC population and cancer cell growth [11,36]. It was suggested that the effect of IL-8 was mediated by CXCR1 and CXCR2, but the selection of receptor for IL-8 binding and actions were in a cell type-specific manner. In our study, we showed that PC9 cells are more abundant of CXCR1 than HCC827 cells (Supplementary Fig. S1). We suggest that the differential expression of CXCR1 and CXCR2 in these cell lines (PC9, and HCC827) may have different influence on CSC activity and other actions under IL-8 stimulation or knockdown of IL-8.
Besides IL-8, IGF1R (insulin-like growth factor 1 receptor) pathway was involved in gefitinib-resistance and stemness in gefitinib-resistant PC9 and HCC827 cells [42]. Activation of IGF1-IGF1R axis increased several stem cell markers (such as CD133, ALDH1A1, Oct4, Nanog, CXCR4, and Sox2) both in PC9/gef and HCC827/gef cells [42]. In our microarray database, we didn't observe up-regulation of IGF1 in PC9/gef cells. We suggested that different culture conditions and methods (step-wise escalation or high-concentration of gefitinib) used to establish gefitinib-resistant cell lines activate different pathways and stemness-related genes [42]. It should be concerned that the manner of EGFR TKI exposure influences the change of genotype and phenotype in EGFR TKI-resistant cells.
Studies have revealed that IL-8 secretion were up-regulated under chemotherapy, and IL-8 increased cancer stem cell populations [43,44]. Moreover, IL-8 receptor, CXCR1, is also specially expressed in an ALDH-positive cancer cell sub-population and these cells are resistant to FASL-induced apoptosis [11]. Blockade of the CXCR1 evoked FasL-mediated apoptosis in a CSC sub-population. It suggests that the IL-8/CXCR-1 axis is associated with stem cell renewal and setmness-related sphere formation is diminished upon IL-8 inhibition [11,45]. Interestingly, lung cancer cells surviving treatment with chemotherapeutic drugs in SCID mice propagated with CSCs-like characteristics and expressed elevated levels of CXCR1/2 [46]. Consistent with these findings, we showed that PC9/IL-8 cells have greater tumorigenesis potential in NOD.SCID mice and possess a larger fraction of side population cells compared to PC9/mock cells. We further found that www.impactjournals.com/oncotarget CXCR1/2 was up-regulated in gefitinib-resistant cells, which would enhance the sensitivity to IL-8-mediated signal activation. IL-8 expression was shown to be enriched in the side population of human melanoma cells which are capable of excluding anticancer drugs [47]. Side population cells are abundant of ATP-binding cassette (ABC) transporters [48]. Study reported that side population cells showed a great potential against EGFR TKI-induced apoptosis [19], and we consistently found ABCB1 transporter is up-regulated in EGFR TKI-resistant cells (data not shown). However, it is still unclear whether EGFR TKIs are substrates of ABC transporters, and the association between EGFR TKI resistance and ABC transporters is controversial.
In conclusion, our data suggest that diminishing the CSC sub-population through depletion of IL-8 in combination with gefitinib administration might potentially overcome EGFR-TKI resistance.

Oligonucleotide microarray
Messenger RNA (mRNA) expression profiles of PC9 and PC9/gef were generated using Affymetrix Human Genome U133 plus 2.0 GeneChips (Affymetrix; Santa Clara, CA), which contain 54,675 probe sets capable of analyzing the expression levels of 47,400 transcripts and variants. Oligonucleotide cDNA microarray analyses were performed by the Integrated Core Facility for Functional Genomics of National Science Council in Taiwan according to the manufacturer's protocols (Santa Clara, CA, http://www.affymetrix.com/). The expression profiles were deposited in the GEO database (GSE60189).

Real-time quantitative reverse transcription-PCR (RT-qPCR)
Complementary DNA (cDNA) was generated using a High-Capacity cDNA reverse transcription kit (Applied Biosystems; Foster city, CA). Gene specific primer sequences are listed in Supplementary Table S1.
Quantitative-PCR was carried out on an ABI 7500 system. Relative mRNA expression levels were calculated using the 2 −∆Ct method, where ∆Ct = (sample Ct -TBP Ct). It was considered "undetectable" if the value of Ct > 40.

Enzyme-linked immunosorbent assay (ELISA)
Cells were plated at a density of 5 × 10 3 cells/well in a 24-well plate. After cells had adhered, the culture medium was replaced with fresh medium and plates were incubated for 48 hours. Supernatants were collected for quantification of IL-8 protein using a Quantikine IL-8 ELISA kit (R&D Systems; Minneapolis, MN).

Patient population and study design
All treatment-naive patients with stage IV lung adenocarcinoma diagnosed between October 2010 and October 2013 were identified. Only those with EGFR mutation-positive and who received EGFR-TKIs as their first-line treatment were investigated in this study. Plasma IL-8 level in patients was determined by IL-8 ELISA kit and this study was approved by the Institutional Review Board of National Taiwan University Hospital.

Apoptosis assays
Apoptosis was detected using an Annexin-V-FITC detection kit (BD Biosciences). Briefly, cells were incubated with fluorescein isothiocyanate (FITC)-conjugated Annexin-V and propidium iodide (PI) for 15 minutes in the dark. Analyses were performed using flow cytometry and CXP analysis software (Beckman Coulter, CA).

Anchorage-independent growth
Cells were suspended in top agar (0.35% agarose) and plated in triplicate onto bottom layers of 0.5% agar in a 6-well plate. The cells were fed 2 mL of medium every 5-7 days. Colonies were photographed after 4 weeks and analyzed using Image J software.

Determination of tumorigenicity in NOD. SCID mice
Animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC No.20110492) of the National Taiwan University, Taiwan. NOD.CB17-Prkdc scid /NcrCrlBltw (NOD.SCID) male mice (BioLASCO; Taipei, Taiwan) at five weeks old were transplanted with cells. Briefly, cancer cells were suspended in HBSS buffer at density of 5 × 10 5 cells/mL, and subcutaneously injected into each mouse. All mice were examined regularly for the development of tumors, and cells were scored as tumorigenic if a palpable nodule appeared at the site of injection within 10 weeks and increased in size. www.impactjournals.com/oncotarget

ALDH activity
The ALDEFLUOR kit (Stemcell Technologies; Vancouver, Canada) was used to identify cell populations with ALDH enzymatic activity. Cells were suspended in ALDEFLUOR Assay Buffer containing ALDH substrate. Following a 30 minute incubation at 37°C and centrifugation, the cells were re-suspended for analysis using a FACSAria flow cytometer (BD Biosciences). Samples treated with the inhibitor, diethylaminobenzaldehyde (DEAB), were used as negative controls to gate the ALDH-negative region.

Side population
Cells were stained with 5 μg/mL Hoechst 33342 (Sigma-Aldrich; St. Louis, MO) at 37°C for 90 minutes. PI staining was used to isolate dead cells, and 150 μM verapamil (Sigma-Aldrich) was used to distinguish side population. Side population cells were sorted from the main cell population using a FACSAria through services provided by the Cell Sorting Core Facility.

Caspase-9 activity
Cells were re-suspended with density of 5 × 10 4 cells/mL and hung in 96-Well Hanging Drop Plates (Perfecta3D; Ann Arbor, MI) for 48 hours, and then incubated with gefitinib. After incubation, caspase-9 activity was measured with the caspase-Glo 9 luminescence kit (Promega; Madison, WI). Briefly, the cellular lysates were incubated with substrates to generate luminescent signal and the luminescence was recorded.

Statistical analysis
Student's t-test was used for comparing means of continuous variables between two groups. Progressionfree-survival curves were plotted using the Kaplan-Meier method and compared the differences between the groups using the log-rank test. Two sided p-values < 0.05 were considered statistically significant.

ACKNOwLEDGMENTS
This work was supported in part by the Department of Medical Research and the 3 rd core facility at National Taiwan University Hospital. We would like to acknowledge the services provided by the Cell Sorting Core Facility of the First Core Laboratory and the Center of Genomic Medicine, National Taiwan University. This project was supported by the National Science Council (NSC101-2314-B-002-170-MY3) and National Taiwan University (101R7601-3, 102R7601-3, and 103R7601-3), Taipei, Taiwan.

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest exist.

Disclaimers
The views expressed in the article are our own and not an official position of the institution or funder.