Molecular crosstalk between ferroptosis and apoptosis: emerging role of ER stress-induced p53-independent PUMA expression

Ferroptosis is a type of programmed cell death that depends on iron and is characterized by the accumulation of lipid peroxides. In the present study, we investigated the nature of the interplay between ferroptosis and other forms of cell death such as apoptosis. Human pancreatic cancer PANC-1 and BxPC-3 and human colorectal cancer HCT116 cells were treated with ferroptotic agents such as erastin and artesunate (ART) in combination with the apoptotic agent tumor necrosis factor-related apoptosis-inducing ligand (TRAIL). We observed synergistic interaction of erastin or ART with TRAIL as determined by cell death assay, caspase activation, poly [ADP-ribose] polymerase 1 (PARP-1) cleavage, flow cytometry analysis, and lipid peroxidation assay. Moreover, erastin and ART induced endoplasmic reticulum (ER) stress and promoted p53 upregulated modulator of apoptosis (PUMA) expression via C/EBP-homologous protein (CHOP). Synergy of erastin/ART and TRAIL was abolished in PUMA-deficient HCT116 cells and CHOP-deficient mouse embryonic fibroblasts, but not in p53-deficient HCT116 cells. The results suggest the involvement of the p53-independent CHOP/PUMA axis in response to ferroptosis inducers, which may play a key role in ferroptotic agent-mediated sensitization to TRAIL-induced apoptosis.


INTRODUCTION
Several processes such as ferroptosis, apoptosis, necrosis, and autophagy are the primary mechanisms of cell death. Each type of biological death leaves a biochemical and morphological fingerprint [1]. However, emerging evidence suggests that these different types of cell death often share common pathways [2].
Since its discovery in 1995, tumor necrosis factorrelated apoptosis-inducing ligand (TRAIL) has sparked growing interest among oncologists due to its remarkable ability to induce apoptosis in malignant human cells, but not in most normal cells [24,25]. TRAIL initiates the extrinsic pathway by binding to death receptors (DRs) such as DR4 and DR5 and induces the apoptotic signal. Activation of death domains (DDs) leads to formation of the death-inducing signaling complex (DISC) [26]. Caspase-8 recruitment and its activation at the DISC leads to further triggering of signaling molecules downstream, which results in the activation of executioner caspase-3, -6, and -7, which culminates in apoptotic death [27]. Activated caspase-8 also cleaves a pro-apoptotic molecule, Bid; truncated Bid (tBid) translocates to the mitochondria and induces Bax (Bcl-2-associated X protein) and Bak (Bcl-2 homologous antagonist killer) oligomerization [28,29]. Oligomerized Bax and Bak's insertion into the mitochondrial outer membrane, permeabilization, and depolarization of the mitochondria promote cytochrome c release [30]. Released cytochrome c facilitates the formation of the Apaf1 (apoptosis signal-regulating kinase)/caspase-9 apoptosome, which activates caspase-9 and subsequently, caspase-3 [31].
TRAIL-induced cytotoxicity can be modulated by various agents such as chemotherapeutic drugs [32][33][34], ionizing radiation [35], other cytokines [36], and matrix metalloprotease inhibitors [37]. In this study, we observed a synergistic interaction between TRAIL and ferroptotic agents. A combined treatment of ART/erastin with TRAIL markedly enhanced TRAIL-induced apoptosis. Our studies suggest that these synergistic effects are due to endoplasmic reticulum (ER) stress-induced p53-independent PUMA (p53 upregulated modulator of apoptosis) expression.

Ferroptosis inducers ART and erastin enhance TRAIL-induced apoptosis
To determine whether ferroptotic agents are cytotoxic to human cancer cells, human pancreatic cancer PANC-1 and BxPC-3 and human colon cancer HCT116 cells were treated with various doses of ART. We observed dosedependent effects of ART on cytotoxicity ( Figure 1). Next, we examined the effect of combinatorial treatment (ART + TRAIL) on cytotoxicity. As shown in Figure  1A-1C, a synergistic cytotoxic effect was observed with the combinatorial treatment compared with any single treatment (p<0.001). Confirmation was obtained by combination index (CI) analysis as shown in Table 1; there was strong or moderate synergy of ART in combination with TRAIL, especially in the high dose group in the three cell lines. We further investigated whether the combinatorial treatment-induced cytotoxicity was associated with apoptosis. Data from the immunoblotting assay demonstrate that the synergistic effects were due to an increased activation (cleavage) of caspase 8/9/3 and thus, the hallmark feature of apoptosis, PARP (poly (ADP-ribose) polymerase) cleavage ( Figure 1D-1F). Similar results were observed with erastin in combination with TRAIL-treated HCT116 cells. We observed a dose-dependent apoptotic effect of TRAIL and ferroptotic effect of erastin ( Figure 2A and 2B). The synergistic cytotoxicity was observed when cells were treated with erastin in combination with TRAIL ( Figure 2C and Table 1). Data from the immunoblotting assay and flow cytometry assay demonstrate that the combined treatment-induced cytotoxicity was associated with apoptosis as evident by increased PARP cleavage ( Figure 2C) and apoptotic cells at the upper right quadrant of the plots ( Figure 2D). An increase in apoptosis was also observed in BxPC-3 cells during treatment with TRAIL and erastin ( Figure 2E). These results indicate that synergistic induction of cytotoxicity by combined treatment of erastin or ART with TRAIL is mediated by an increase in apoptosis.

Ferroptotic agent-induced lipid peroxidation is not changed or influenced by the combination treatment of erastin or ART with TRAIL
Since we observed that the combinatorial treatmentinduced synergistic cytotoxicity was due to an increase in apoptosis, we examined whether the combinatorial treatment enhances ferroptosis by analysis of malondialdehyde (MDA), a lipid peroxidation marker. As shown in Figure 3A, 3B and 3D, ferroptotic agents, but not TRAIL, induced lipid peroxidation in a dose-dependent manner in HCT116 cells. The combined treatment of TRAIL and erastin/ART did not enhance ferroptotic agentinduced lipid peroxidation ( Figure 3C and 3E). Similar results were observed in BxPC-3 cells ( Figure 3F). Similar to the lipid peroxidation result, erastin, but not TRAIL, induced heme oxygenase-1 (HO-1) expression, a marker of oxidative stress, in a dose-dependent manner ( Figure  3B). The combined treatment of TRAIL and erastin did not enhance ferroptotic agent-induced HO-1 expression. www.impactjournals.com/oncotarget

Ferroptotic agents induce ER stress
Previous studies have shown that inhibition of cystine-glutamate exchange by ferroptotic agents leads to activation of an ER stress response and upregulation of the CHAC1 (glutathione-specific gammaglutamylcyclotransferase 1) gene [38,39]. Data from microarray assay studies ( Figure 4A) also reveal that the ferroptotic agent ART promotes the ER stress marker ATF4 (activating transcription factor 4)-dependent gene expression such as TRB3 (tribbles homolog 3) and CHOP (C/EBP-homologous protein) [40]. Ferroptotic agent-induced ER stress was confirmed by detecting the unfolded protein response (UPR) by pretreated with ART (10 or 50 μM) for 20 h and then exposed to TRAIL (PANC-1, 100 ng/ml; BxPC-3, 2 ng/ml; HCT116, 1 ng/ml) for an additional 4 h. The cells were stained with propidium iodide (PI). Phase-contrast images or fluorescence images were visualized under a light or fluorescence microscope, respectively (upper panels). Representative images are shown (magnification, X200). Cell death was determined by counting PI-stained cells and plotted (lower panels). Error bars represent the mean ± SD from triplicate experiments. For statistical analysis, Student's t-test (two-sided, paired) was used. p-values: * , 0.05; ** , 0.01; *** , 0.001. Cell lysates of PANC-1 (D), BxPC-3 (E), and HCT116 (F) cells were analyzed with immunoblotting assay using indicated antibodies.  24 h (A) or TRAIL for 4 h (B) and then the cell death rate was determined, respectively (upper panels). Cell lysates were analyzed with immunoblotting assay using indicated antibodies (lower panels). (C and D) HCT116 cells were pretreated with erastin (10 or 50 μM) for 20 h and then exposed to TRAIL (1 ng/ml) for an additional 4 h. Cell death was determined by counting and plotted. Whole-cell extracts were then analyzed with immunoblotting assay using indicated antibodies (C). The cells were stained with annexin V and PI and then analyzed using flow cytometry (D). (E) BxPC-3 cells were pretreated with erastin (10 or 50 μM) for 20 h and then exposed to TRAIL (6 ng/ ml) for an additional 4 h. Whole-cell extracts were then analyzed with immunoblotting assay using indicated antibodies.
using the anti-ubiquitinylated protein antibody FK2 antibody. Immunoblotting analysis showed that ARTor erastin-treated HCT116 cells contained a higher level of ubiquitin conjugates compared with untreated control cells ( Figure 4B). In particular, inhibition of proteasomal degradation flux with MG132 clearly revealed an increase in ubiquitinated proteins in ARTor erastin-treated cells ( Figure 4B). These results imply that the UPR occurs during treatment with the ferroptotic agents ART and erastin. Next, we detected ferroptotic agent-induced ER stress by using ER stress markers such as GRP78 (78 kDa glucose-regulated protein) and CHOP. As shown in Figure 4C-4E, a ferroptotic agent-induced ER stress response occurred in a time-and dose-dependent manner. A combined treatment of TRAIL and erastin did not alter the ER stress response ( Figure 4F).

Inhibition of ferroptotic agent-induced lipid peroxidation by ferrostatin-1 and liproxstatin-1 neither blocks the ER stress response nor suppresses the combined treatment of erastin or ART with TRAIL-induced synergistic cytotoxicity
We examined whether ferroptotic lipid peroxidation inducing agents play a role in ER stress and the combinatorial treatment-induced synergistic cytotoxicity. To examine this possibility, HCT116 cells were treated with erastin/ART in the absence/presence of the iron chelator deferoxamine (DFO), the lipid peroxidation inhibitors ferrostatin-1 (Fer-1), and liproxstatin-1 (Lip-1). Figure 5A and 5B show that Fer-1 and Lip-1 inhibited erastin/ART-induced lipid peroxidation, but not ER stress. DFO also inhibited erastin-induced lipid peroxidation, but not ER stress. However, unlike erastin, . Cells were pretreated with erastin (10 or 50 μM) for 20 h and then exposed to TRAIL (1 ng/ml) for an additional 4 h (C). Lipid peroxidation (upper panels) and heme oxygenase (HO-1, lower panels) levels were analyzed by malondialdehyde (MDA) assay and immunoblotting assay, respectively. (D and E) HCT116 cells were treated with various doses of ART (D) or pretreated with ART (10 or 50 μM) for 20 h and then exposed to TRAIL (1 ng/ml) for additional 4 h (E). (F) BxPC-3 cells were pretreated with erastin (10 or 50 μM) for 20 h and then exposed to TRAIL (1 ng/ml) for additional 4 h. MDA levels were determined and plotted. Error bars represent the mean ± SD from triplicate experiments. For statistical analysis, Student's t-test (two-sided, paired) was used. p-values: * , 0.05; ** , 0.01.
ART-induced ER stress was inhibited by treatment with DFO. DFO, Fer-1, and Lip-1, but not the caspase inhibitor Z-VAD, protected cells from ferroptosis (data not shown) and did not prevent the combinatorial treatment of erastin and TRAIL-induced synergistic cytotoxicity ( Figure 5C). Similar results were observed with Fer-1 and Lip-1 on ARTand TRAIL-treated cells ( Figure 5D). Unlike erastin and TRAIL-induced synergistic cytotoxicity, ART-and TRAILinduced synergistic cytotoxicity was inhibited by treatment with DFO. These results indicate that ER stress plays an important role in ferroptotic agent and TRAIL-induced synergistic cytotoxicity.

Ferroptotic agent induces PUMA expression
Previous studies show that CHOP binds to the proapoptotic protein PUMA (p53 upregulated modulator of apoptosis) promoter during ER stress and induces PUMA expression [41]. CHOP also induces several other proapoptotic proteins such as NOXA (Latin for damage) and BIM [42,43]. Figure 6 shows that the ferroptotic agent ART induces PUMA expression in three different cancer cell lines. The combinatorial treatment of TRAIL and ART either maintained or enhanced the ER stress response. However, unlike PUMA, we either failed to detect NOXA expression in BxPC-3 cells or did not observe an increase in NOXA expression during ART treatment in HCT116 cells ( Figure 6B and 6C). We did not observe an increase in BIM expression during erastin treatment in HCT116 cells. Unlike erastin, the DNA damaging agent mitomycin C induced three isoforms of BIM: BIM EL , BIM L , and BIM S ( Figure 6D).

Ferroptotic agent promotes TRAIL-induced apoptosis via the p53-independent CHOP/PUMA pathway
We further examined the role of ER stressassociated CHOP and PUMA expression in synergistic apoptosis during the combinatorial treatment of ART and TRAIL. For this study, we employed a CHOP knockout  Figure 7A and 7B, combinatorial treatmentinduced synergistic apoptosis was suppressed in CHOP -/-MEFs and PUMA -/cells. Moreover, the combinatorial treatment-induced synergistic apoptosis was not inhibited in p53 knockout (p53 -/-) cells ( Figure 7C). These results suggest that the p53-independent CHOP/PUMA pathway is responsible for the combinatorial treatment-induced synergistic apoptosis.  Fer-1, 10 μM), or liproxstatin-1 (Lip-1, 2 μM). Lipid peroxidation level was detected by MDA assay (left panels) and whole-cell lysates were analyzed with immunoblotting assay using indicated antibodies (right panels). (C and D) Cells were pretreated with erastin (C) and ART (D) for 20 h in the absence/presence of Z-VAD (1 μM), DFO (100 μM), Fer-1 (10 μM), or Lip-1 (2 μM) and then exposed to TRAIL (1 ng/ml) for an additional 4 h. Cell death was determined using trypan blue exclusion assay and plotted. www.impactjournals.com/oncotarget

Effect of ART and TRAIL on the growth of xenograft tumors
In vivo studies were performed to examine the effect of the combinatorial treatment with ART and TRAIL on the growth of luciferase-expressing HCT116 xenograft tumors. For xenograft tumor formation, five-week-old male nude mice (Balb/c nude) were subcutaneously inoculated with 1 × 10 6 HCT116-luc cells into the right hind leg. Prior to treatment with ART and TRAIL, tumor size was measured two to three times per week until the volume reached approximately 200 mm 3 . Tumor volume was calculated as W 2 × L × 0.52, where L is the largest diameter and W is the diameter perpendicular to L. After establishment of these tumor xenografts, mice were randomized into four groups of five mice per group. ART (200 mg/kg, oral gavage administration) and/ or rh-TRAIL (100 μg/kg, intratumoral injection) were administered twice per week for two weeks. TRAIL alone did not significantly affect tumor growth compared with the control group. ART alone caused a decrease of tumor growth (p < 0.05). The combinatorial treatment with ART and TRAIL was significantly more effective in inhibiting tumor growth compared with single treatment (p < 0.001) ( Figure 8C). Data from terminal deoxynucleotidyl transferase (TdT) dUTP nick-end labeling (TUNEL) assay and quantitative analysis ( Figure 8D and 8E) show that the combinatorial treatment effectively induced apoptosis. Although the tendency for weight loss was observed during the combined treatment of ART and TRAIL, it was not statistically significant ( Figure 8F).

DISCUSSION
Several conclusions can be drawn upon consideration of the data presented in the current study. First, the combined treatment of erastin or ART and TRAIL synergistically induced apoptosis, but not ferroptosis. Second, ER stress played an important role in the synergistic apoptosis through activating the p53independent CHOP-PUMA axis.
It is well known that the ER stress response mediated by the PERK (PKR-like ER kinase)-eIF2α (eukaryotic initiation factor 2α)-ATF4 pathway is involved in regulation of the expression of several target genes such as CHOP [44,45]. Data from immunoblotting assay confirmed an increase in expression of the ER stressrelated genes GRP78 and CHOP during ferroptotic agent treatment (Figures 4, 5, and 7). Several researchers have reported that ER stress induces apoptosis through inducing embryonic fibroblasts (MEFs) wild-type (WT) or MEF CHOP knockout (KO) cells were pretreated with ART (10 or 50 μM) for 20 h and then exposed to recombinant murine TRAIL (mTRAIL; 100 ng/ml) for an additional 4 h. Cell death was determined using trypan blue exclusion assay and plotted (left panel). Whole-cell extracts were analyzed with immunoblotting assay using indicated antibodies (right panel). (B) HCT116 WT or HCT116 PUMA KO cells were pretreated with ART (10 or 50 μM) for 20 h and then exposed to TRAIL (1 ng/ml) for additional 4 h. Cell death was determined by counting and plotted (left panel). Whole-cell extracts were then analyzed with immunoblotting using indicated antibodies (right panel). (C) HCT116 WT or HCT116 p53 KO cells were pretreated with ART (10 or 50 μM) for 20 h and then exposed to TRAIL (1 ng/ml) for an additional 4 h. Cell death was determined using trypan blue exclusion assay and plotted (left panel). Whole-cell extracts were then analyzed with immunoblotting assay using indicated antibodies (right panel). www.impactjournals.com/oncotarget pro-apoptotic proteins such as PUMA, NOXA, GADD34 (growth arrest and DNA damage-inducible protein), ERO1α (endoplasmic reticulum, oxidoreductin-1α), and BIM (Bcl-2-like protein 11) [41,42]. However, unlike previous reports, we observed an increase in expression of PUMA, but not NOXA and BIM, during treatment with ferroptotic agent (Figure 6). This discrepancy needs further investigation. Presently, we can only speculate on the differential expression in ER response-inducible proapoptotic gene expression during treatment with ferroptotic agent. Several transcription factors are known to regulate transcription of PUMA and NOXA gene expression: p53, c-Myc, forkhead box O3A (FoxO3A), p65 or p52 subunit of nuclear factor-κB, p73, CHOP, E2F1, C/EBPβ (CCAAT-enhancer-binding proteins), CREB (cAMP response element-binding protein), c-Jun, and Sp1 for PUMA gene expression [46], and HIF-1α, E2F1, c-myc, p53, p73, and ATF4 for NOXA gene expression [47]. They may share the same transcription factors for p53-dependent expression of these genes by genotoxic stress, but not in p53-independent expression of these genes by other stimuli.
In the present study, we report that Fer-1 and Lip-1 (inhibitors of ferroptosis), when combined with TRAIL, still suppressed ferroptosis but not synergistic apoptosis and ER stress. However, unlike Fer-1 and Lip-1, the different ferroptosis inhibitor DFO inhibited ARTinduced ER stress exclusive of that induced by erastin as well as synergistic apoptosis ( Figure 5). This was probably due to differences in mechanisms by which each ferroptosis inducer contributes to the various metabolic derangements that lead ultimately to ER stress. Since  3 ) in HCT116-luc tumor-bearing mice treated with PBS alone, ART alone, TRAIL alone, or the combination from day 0 to day 19. Error bars represent the mean ± SD from five mice. For statistical analysis, Student's t-test (two-sided, paired) was used. p-values: * , 0.05; ** , 0.01; *** , 0.001. (D and E) Tumor tissues were harvested on day 19 and subjected to TUNEL assay and DAPI (4',6-diamidino-2-phenylindole) staining. Cell nuclei were stained with DAPI. Apoptosis was detected using TUNEL assay (D) and percents of TUNELpositive cells were plotted (E). (F) Line graph illustrating the body weight (gram) in HCT116-luc tumor-bearing mice treated with PBS alone, ART alone, TRAIL alone, or the combination from day 0 to day 19. www.impactjournals.com/oncotarget DFO has pleiotropic effects (inhibition of iron-mediated lipid peroxidation, inhibition of NF-κB activation, and involvement in several peroxidative systems [48,49]), DFO may affect ART-induced ER stress and its activation of NF-κB by preventing IκBα degradation [48,50,51]. Obviously, further studies are necessary to understand the molecular mechanism of differences between each ferroptosis inducer-induced ER stress.
Although the ferroptotic agent erastin and ART induced pro-apoptotic protein PUMA expression ( Figures  6 and 7), unlike in previous reports, erastin and ART did not induce apoptosis; the caspase substrate PARP-1, a hallmark feature of apoptosis, was not cleaved during ferroptotic agent treatment (Figures 1 and 2). These results suggest that ferroptotic agent-induced PUMA remains inactive during treatment with ferroptotic agent alone and then switches to an active state during combinatorial treatment with ferroptotic agent and TRAIL. Previous biochemical studies indicate that PUMA induces apoptosis by activating the multidomain proapoptotic protein Bax and/or Bak through its interaction with antiapoptotic Bcl-2 family members such as Bcl-2 (B-cell lymphoma 2) and Bcl-xL (B-cell lymphoma-extra large), thereby triggering mitochondrial dysfunction, cytochrome c release, and caspase activation [46]. How ferroptotic agent-induced PUMA sustains a biochemically inactive state during treatment with ferroptotic agent alone is a question that remains unanswered. It is possible that ferroptotic agentinduced PUMA remains inactive through binding with anti-apoptotic Bcl-2 family members (Bcl-2, Bcl-xL, Mcl-1) as well as Beclin-1. During combinatorial treatment with ferroptotic agent and TRAIL, Beclin-1 is cleaved by TRAIL-activated caspase 8 [52] and the cleavage of Beclin-1 leads to PUMA changing from an inactive to an active state. The conversion of PUMA from an inactive to an active state results in an increase in apoptosis. Another possibility is the apoptotic threshold effect [53]. Apoptosis occurs when an apoptotic threshold is reached, owing to the inhibition of Bcl-2, Bcl-xL, and Mcl-1 by PUMA [53]. A combinatorial treatment of ferroptotic agent and TRAIL induces the level of PUMA at which apoptosis occurs above a threshold. Obviously, these possibilities need to be further examined to understand the role of PUMA in the combinatorial treatment-induced synergistic apoptosis.

Chemicals and reagents
For production of recombinant human TRAIL, a human TRAIL cDNA fragment (amino acids 114-281) obtained by RT-PCR was cloned into a pET-23d plasmid (Novagen, Madison, WI), and His-tagged TRAIL protein was purified using the Qiagen express protein purification system (Qiagen, Valencia, CA). Soluble recombinant murine TRAIL and cell-permeant pan caspase inhibitor Z-VAD-FMK were purchased from R&D systems (Minneapolis, MN). Liproxstatin-1, deferoxamine, and ferrostatin-1 were obtained from Sigma-Aldrich (St. Louis, MO). Artesunate, erastin, and mitomycin C were purchased from Selleckchem (Houston, TX).

Microarray assay
HCT116 cells were treated with 50 μM ART for 24 h and total RNA was extracted using the RNeasy mini kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. RNA quality and integrity were verified using the Agilent 2100 Bioanalyzer system (Agilent Technologies, Santa Clara, CA). Biotin-labeled cRNA samples for hybridization on Illumina HumanHT-12 v4 Expression BeadChip (Illumina, Inc., San Diego, CA) were prepared according to Illumina's recommended sample labeling procedure.

Lipid peroxidation assay
To measure the concentration of MDA, one of the end products of lipid peroxidation, colorimetric analysis of lipid ROS production was carried out using a Lipid Peroxidation (MDA) Assay Kit (#MAK085, Sigma-Aldrich) according to the manufacturer's instructions. MDA-TBA adduct was quantified colorimetrically at 532 nm using a spectrophotometer.

Apoptosis assay by flow cytometry
To detect the translocation of phosphatidylserine, one of the markers of apoptosis, from the inner to the outer leaflet of the plasma membrane, cells were stained with Annexin V according to the manufacturer's protocol of the fluorescein isothiocyanate Annexin V Apoptosis Detection Kit (BD Biosciences, San Diego, CA). Flow cytometry was performed using an Accuri C6 flow cytometer (BD Biosciences). www.impactjournals.com/oncotarget

Cell death and viability assay
Cell death was typically measured using trypan blue exclusion assay to detect plasma membrane integrity as previously described [54]. For quantification of cell death rate, cells were trypsinized and stained with trypan blue followed by counting with a hemocytometer under microscope. In some experiments, propidium iodide staining was performed. Red-stained cells were considered dead. Quantification of cell death was further confirmed using flow cytometry. Cell viability was typically assessed in 96-well format with Alamar Blue Cell Viability Assay Kit (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions.

Combination index (CI) analysis
CIs were calculated using the CompuSyn software (ComboSyn, Inc., Paramus, NJ, USA). Base on CI values, the extent of synergism/antagonism was determined. In general, CI values below 1 suggest synergy, whereas CI values above 1 indicate antagonism between the drugs. CI values in the 0.9-1.10 range mainly indicate additive effects, those between 0.9-0.85 suggest slight synergy, those in the range of 0.7-0.3 indicate moderate synergy, and those less than 0.3 suggest strong synergy.

Animal model
Human colon adenocarcinoma HCT116-Luc cells were established by subcutaneously injecting 10 6 cells into the right hind leg of five-week-old male nude mice (Balb/c nude) (Charles River Labs, Wilmington, MA, USA). Prior to treatment with ART and TRAIL, tumor size was measured two to three times per week until the volume reached approximately 200 mm 3 . Tumor volume was calculated as W 2 × L × 0.52 where L is the largest diameter and W is the diameter perpendicular to L. After establishment of these tumor xenografts, mice were randomized into four groups of five mice per group. Mice were fed ad libitum and maintained in environments with controlled temperature of 22-24°C and 12 h light and dark cycles. Artesunate group: 200 mg/kg -twice per week -oral gavage-two-week treatment rhTRAIL group: 100 ng/g -twice per week-intra-tumoral injection-twoweek treatment. Animals were anesthetized and subjected to NightOWL LB983 bioluminescence imaging (BLI) system (Berthold Technologies, TN, USA). D-luciferin sodium salt (BioVision Inc. Milpitas, CA) at 100 mg/kg was administered intraperitoneally as a substrate before BLI imaging. The captured images were quantified using the IndiGo™ software package. All animal procedures were carried out in accordance with guidelines approved by the Korea University Institutional Animal Care and Use Committee (IACUC).

Statistical analysis
All the values are represented as mean ± S.D. Statistical analysis was performed one-way analysis of variance (ANOVA) followed by Bonferroni's test or by the Student's t-test as indicated using GraphPad Prism 7 software. P value less than 0.05 was defined as statistical significance. Where indicated * p < 0.05; ** p < 0.01; ***