Dihydroartemisinin promotes CHAC1 transcription to induce ferroptosis in primary liver cancer cells: activation of unfolded protein responses

Background: This study aimed to explore whether dihydroartemisinin (DHA), an artemisinin derivate drug, eliminates primary liver cancer (PLC) cells by inducing ferroptosis. Methods: Four PLC cell lines were treated with varied concentrations of DHA. RNA interference was performed to knock down the expression of unfolded protein response (UPR) sensors in vitro. Results: DHA-caused PLC cell death was irrelevant to p53 status. PLC cells exposed to DHA displayed classic ferroptosis features – increased lipid ROS, MDA and iron ions, and decreased activity or expression of GSH, GPX4, SLC7A11 and SLC3A2. The anti-tumor effects of DHA were signicantly weakened by ferrostatin-1 and deferoxamine mesylate salt, but augmented by iron overload. DHA activated all three UPR branches, including PERK/eIF2/ATF4, IRE1α / XBP1 , and ATF6, in vitro . Further, to deactivate UPRs, exclusive siRNA was used to silence the expression of ATF4, XBP1 or ATF6 in PLC cells. Unexpectedly, ferroptosis induced by DHA was signicantly attenuated when ATF4, XBP1 or ATF6 was knocked down. The transcription of CHAC1, a molecule that is capable of degrading GSH, was enhanced by DHA, but weakened when the above three UPR transcription factors were silenced. Conclusion: DHA effectively induces ferroptosis in PLC cells, which involves the activation of anti-survival UPRs. GSH-PX activity (g) PLC (0 ± DHA, μM μM for vitalities CCK8 assay (a & and intracellular total ROS (b & e) and lipid ROS (c & f) levels were with ow cytometry (b-c, e-f) (n = 3). ***p < 0.001 vs DHA + siNC. Values were presented as mean values ± standard deviation. DHA, dihydroartemisinin; PLC, primary liver cancer; UPR, unfolded protein response; PERK, protein kinase R-like ER kinase, eIF2, eukaryotic initiation factor 2; ATF4, activating transcription factor 4; ATF6, activating transcription factor 6; IRE1α, inositol-requiring transmembrane kinase/endoribonuclease 1α; unXBP1, unsliced X box-binding protein 1; sXBP1, sliced X box-binding protein

cancer therapy [11,12]. To identify drugs capable of inducing ferroptosis will provide insights for the potential of ferroptosis as a new promising target for PLC treatment.
Artemisinin is a sesquiterpene trioxane lactone originally extracted from Artemisia annua L, and its derivates are effective anti-malarial agents [13]. Currently, besides the well-known application in antimalaria eld, artemisinins are being evaluated for treating multiple cancers due to their established safety recorded in thousands of malarial patients [13]. Ooko and colleagues treated 60 cancer cell lines with 11 artemisnin derivates, and analyzed the expression pro les of 30 iron-related genes in these cell lines via microarray hybridization [14]. They found that the log 10 (50% inhibition concentration, IC 50 ) values of analyzed artemisnins signi cantly correlated to the expression of at least 20 iron-related genes [14]. Their ndings imply the involvement of artemisinins in ferroptosis of cancer cells. The most frequently reported derivate of artemisnin that triggers ferroptosis in malignant cells is artesunate [15,16]. Recently, dihydroartemisinin (DHA), a semisynthetic derivative of artemisinin, has been demonstrated to exhibit anti-tumor activity in PLCs in vitro and in vivo [17,18]. The role of DHA in inducing ferroptosis in cancer cells was rst reported in head and neck carcinoma cells by Lin et al. [19], and later in leukemia cells by Du et al. [20] and in glioma cells by Chen et al. [21]. Up to date, no direct evidence has proved DHA's role in inducing ferroptosis in PLCs. Nonetheless, an earlier study from Wang and co-workers has shown that DHA is able to increase intracellular ROS in LM3 HCC cells [17], suggesting that DHA may trigger ferroptosis in PLC cells.
Cancer cells can thrive under hostile microenvironmental conditions. Within tumor masses, endoplasmic reticulum (ER) stress is provoked by nutrient deprivation, oxygen limitation and high metabolic demand [22]. Malignant cells initiate unfolded protein response (UPR) to cope with ER stress [23]. Three ERresident sensors, protein kinase R-like ER kinase (PERK), inositol-requiring transmembrane kinase/endoribonuclease 1α (IRE1α) and activating transcription factor (ATF) 6, coordinate UPR if misfolded proteins accumulate and aggregate beyond a tolerable threshold [22,23]. ATF4 upregulation induced by ferroptosis inducers, such as Erastin and Artesunate, is considered as a compensatory effect, and its knockdown is demonstrated to further augment ferroptosis in cancer cells [15,24]. In glioma cells, PERK/ATF4 signaling pathway was activated by DHA and negatively regulated DHA-induced ferroptosis [21]. Studies on revealing the roles of ATF6-and IRE1α-mediated UPR branches in ferroptosis are scarce, and how DHA affects these two UPR mediators remains largely unknown.
In this study, four PLC cell lines, Hep3B, Huh7, PLC/PRF/5 and HepG2, were treated with different concentrations of DHA in presence or absence of ferroptosis inhibitors or iron ions. Our results demonstrated that DHA was capable of inducing ferroptosis and activating all the three UPR branches in PLC cells. Interestingly, our date for the rst time demonstrated that DHA-induced ferroptosis was weakened if these UPR signaling pathways were suppressed.

Chemicals
Ferrostatin-1 and DHA both from Aladdin (Shanghai, China) were dissolved in dimethyl sulfoxide (DMSO) into a stock concentration of 5 mM and 50 mM, respectively. Deferoxamine mesylate salt (DFOM; MedChemExpres, Monmouth Junction, NJ, USA) and iron chloride hexahydrate (Aladdin) were dissolved in distilled H 2 O into stock concentrations of 50 mM and 2 mg/L, respectively.

PLC cell xenografted tumor mouse models
Male nude mice (BALB/c) of 6-8 weeks old (18-20g) were obtained from HFK Bioscience (Beijing, China), and housed in speci c pathogen free facility. A total of 32 mice were subjected into the xenograft mouse model experiment, and randomly divided into four groups (n = 8/group). PLC cells were subcutaneously injected into the nude mice. The tumor volumes were determined by using the following formula: 0.5 × tumor length × (tumor width) 2 . When the xenografted tumor grew into a size of approximately 80-100 mm 3 , half mice in each group were given 100 mg/kg DHA for 5 d per week by gavage. Twenty-one days later, all mice were sacri ced by iso urane anaesthesia and cervical dislocation before collecting the tumor tissues.
Hematoxylin and eosin (H&E) staining Tumor tissues were collected, xed in 4% paraformaldehyde, and embedded into para n. The tissue blocks were sliced into 5-μm sections. Following the depara nating and rehydration, the samples were incubated with H&E staining agents according to the manufacture's protocols.

ROS measurements
Contents of total ROS and lipid ROS were determined with an ROS assay kit (NJJCBio, Nanjing, China; Invitrogen, Carlsbad, CA, USA) based on DCFH-DA and BODIPY 581/591 C11 uorescence as per the supplier's protocols. The DCF and C11-BODIPY uorescence intensities were analyzed with the Tecan M200 PRO automatic microplate reader or a ow cytometer.

Iron concentration
Intracellular ferrous iron levels were analyzed by using an iron assay kit obtained from Leagene Biotech. (Beijing, China) according to the manufacturer's instructions. The output was measured on the Tecan M200 PRO reader at optical density (OD) of 562 nm.

Western blotting analysis
Total proteins from the whole cell lysates were isolated by using RIPA lysis buffer containing 1% PMSF (SolarBio, Beijing, China). Nuclear proteins were isolated with a Nuclear and Cytoplasmic Protein Extraction kit (Beyotime, Shanghai, China). Protein concentrations were analyzed with a BCA kit (SolarBio). After separating via SDS-PAGE, proteins were transferred to PVDF membranes (Millipore, Billerica, MA, USA), and blocked with 5% non-fat milk for 1 hr. The membranes were treated with primary antibodies at 4℃ overnight, and then with secondary antibodies at 37℃ for 1 hr. The information of primary antibodies were shown in Table 1. Colors of protein blots were developed by incubating the membranes with ECL agent (SolarBio).

pGL3 luciferase reporter assay
The promoter (-2000bp to +30 bp) of CHAC1 gene was inserted into pGL3 luciferase reporter, and the promoter activity was determined by analyzing Fire y/Renilla luciferase ratio according to the manufactory's instructions (Promega, Madison, WI, USA) Statistical analysis Data were expressed by mean values ± standard deviation. GraphPad Prism version 8.0 software was utilized to compare the data. One-way analysis of variance (ANOVA) and two-way ANOVA followed by Bonferroni's multiple comparison test were used to compare data from multiple groups. A p-value < 0.05 was considered signi cant.

DHA induced-ferroptosis in PLC cells is irrelevant to p53 status
Considering the involvement of p53 in ferroptosis [25], four PLC cell lines with discrepant p53 statuses, Hep3B (p53 null), Huh7 and PLC/PRF/5 (both p53 mutant) and HepG2 (p53 wild-type), were treated with DHA of increased concentrations for 24 hrs. Our data showed that DHA was able to inhibit the survival of all analyzed PLC cell lines, irrelevant to their p53 statuses (Fig. 1). Two different ferroptosis inhibitors, DFOM (an iron chelator; 10 μM for Huh7 cells and 50 μM for the others) and ferrostatin-1 (a lipid peroxidation inhibitor; 5 μM for HepG2 and 1 μM for the others) were utilized to treat PLC cells in presence of DHA. The concentrations of DFOM and ferrostatin-1 used here were determined by the preliminary experiments. We found that the cytotoxic effects of DHA on PLC cells were attenuated by both ferroptosis inhibitors (Fig. 1). In contrast, addition of exogenous iron ions (10 μg/mL iron chloride hexahydrate) further augmented DHA's effects (Fig. 1

DHA induces ferroptotic GSH synthesis in PLC cells
After incubation with DHA for 1, 6, 12 or 24 hrs, PLC cells were harvested to analyze total and lipid ROS content ( Fig. 2a-b), MDA levels ( Fig. 2c), and iron concentrations (Fig. 2d). The results showed increased ROS, MDA and iron levels in PLC cells exposed to DHA. These above data collectively depicted DHA as ferroptosis inducer in PLC cells in vitro.
Blocking GSH synthesis is known to facilitate toxic ROS accumulation [7]. Therefore, we analyzed GSH contents with a commercial available kit in DHA-treated PLC cells. As compared to the blank cells, GSH/GSSG (oxidized form of GSH) ratios markedly decreased in cells exposed to DHA (Fig. 2e). The expression of GPX4, SLC7A11 and SLC3A2, and the activity of GSH-PX were inhibited by DHA as well ( Fig. 2e-f). In contrast, CHAC1 expression increased in response to DHA treatment (Fig. 2f). These data demonstrate DHA as a negative regulator for GSH synthesis in PLC cells.

DHA limits xenografted PLC tumor growth in vivo
We next investigated the effects of DHA on PLC growth in vivo. No weight loss or health problem was observed in mice. PLC cells were subcutaneously injected into immune-de cient mice. DHA was given to these mice when the tumor volumes reached 80-100 mm 3 . The tumor volumes were recorded every three days, and three weeks later, the xenografted tumors were collected. As shown in Fig. 3a-c, DHA limited, but did not terminate, the formation of xenografted PLCs. Further, DHA augmented ROS accumulation, and downregulated GSH-PX activity in the tumor masses. These data together with earlier results from cell experiment con rmed that DHA could induce ferroptosis both in vitro and in vivo.

DHA activates three UPR branches in PLC cells
As DHA induced ferroptosis in all analyzed PLC cells in vivo and in vitro, we determined the alterations in UPR-associated molecules only in PLC/PRF/5 and HepG2 cells. Our data indicated that all the three UPR signaling pathways were activated by DHA-the expression of phosphorylated PERK, eIF2α, and IRE1α increased, and the levels of ATF4, nuclear ATF6 and spliced XBP1 upregulated (Fig. 5). DHA activated PERK/eIF2α/ATF4 signaling earlier in PLC/PRF/5 cells than in HepG2 cells (Fig. 4). To determine the role of activated UPR in DHA-mediated ferroptosis, ATF4, XBP1 and ATF6 were respectively silenced with their exclusive siRNAs. PLC cell vitalities increased (

DHA promotes the transcription of CHAC1 by activating UPRs in PLC cells
The potential binding sites for ATF4, XBP1 and ATF6 on CHAC1 gene promoter were predicted with JASPER (http://jaspar.genereg.net/) and PROMO (http://alggen.lsi.upc.es/cgibin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3). Multiple sites on CHAC1 promoter could be binding by these transcription factors (Fig. 7a). Since DHA could effectively increase the expression of ATF4, XBP1 and ATF6, it is possible that this agent can promote CHAC1 transcription. Results from pGL3 luciferase reporter assay con rmed that DHA enhanced the promoter activity of CHAC1 gene (Fig. 7b). In addition, to further validate the role of CHAC1 in DHA-induced ferroptosis, siRNA was used to knock CHAC1 expression down (Fig. 7 d & e). We found that ferroptosis induced by DHA was suppressed by CHAC1 siRNAs (Fig. 7 f-i).

Discussion
Inducing ferroptosis is being explored as an alternative approach to eradicate cancer cells resistant to apoptosis [10]. Several artemisnin derivates have been demonstrated to kill cancer cells by inducing ferroptosis [14]. A few recent studies revealed DHA as ferroptosis inducer in cancer cells [19][20][21]. In consistent with these previous studies, our work demonstrated that DHA killed PLC cells by inducing ferroptosis (sFig. 8).
The classic way to determine whether a drug can induce ferroptosis is to co-treat cancer cells with ferroptosis blockers or iron ions [11,12]. The anti-tumor effects of ferroptosis inducers, such erastin and sorafenib, can be attenuated by iron chelators (such as DFOM) and inhibitors of lipid peroxidation inhibitor (such as ferrostatin-1), but augmented by adding exogenous iron ions [7,26]. Based on the IC 50 results, we concluded that the PLC cells were more resistance to DHA-induced cytotoxicity when ferroptosis was suppressed by DFOM or ferrostatin-1. In contrast, addition of iron chloride hexahydrate sensitized PLC cells to DHA. Of note, treatment of DFOM or ferrostatin-1 does not completely abolish the anti-tumor effects of DHA, suggesting that triggering ferroptosis is not the only way for DHA to induce cytotoxicity in PLC cells. Although DHA has been shown to induce ferroptosis in other types of cancer cells [19-21], our work demonstrates such effects in PLC cells for the rst time.
P53 is a key tumor suppressor, whose mutation or loss directly contributes to tumorigenesis [27]. SLC7A11 (also known as xCT) and SLC3A2 controls the import of extracellular cysteine and cystine, and regulates GSH synthesis [28,29]. Interestingly, p53 of wild-type and acetylation-defective mutant can suppress SLC7A11 expression, thereby inhibiting GSH synthesis and sensitizing tumor cells to ferroptosis [25]. In light of the key role of p53 in SLC7A11-associated ferroptosis, we determined the cytotoxicity of DHA in four PLC cell lines with different p53 statuses. Our data demonstrated that DHA could effectively induce ferroptosis in the analyzed PLC cells, even in the p53 null cells (Hep 3B cells). It is worth noting that DHA also exhibited anti-tumor activity by activating Caspase-3, a key apoptosis regulator, against PLCs, regardless of p53 status [30]. This study together with our work suggests that DHA exerts its antitumor effects in an p53 independent manner. Thus, DHA is an attractive drug for treating p53 mutationor de ciency-associated cancer.
GPX4 was rst identi ed by Urisini et al. in 1982 [31], and is now recognized to possess a unique capability of reducing reactive phosphatidylcholine hydroperoxides and suppressing lipid peroxidation [8]. Direct or indirect targeting mechanism such as GSH depletion induces GPX4 repression [32]. In cancer cells, ferroptosis can be provoked by GPX4 inactivation [10]. To further explore how DHA induces ferroptosis in PLC cells, the contents of GSH and the expression and activity of GPX4 were analyzed. Our data showed that DHA reduced GSH synthesis and inhibited GPX4 expression in PLC cells. The decreased activity of GPX4 observed in PLC cells exposed to DHA may result from the reduced GPX4 expression. While a study from Lin et al [19]. supported our ndings, a recent work from Chen and co-workers did not [21]. The latter study demonstrated that DHA induced ferroptosis in glioma cells and it also induced a compensatory upregulation of GPX4 in U251 and U373 cancer cells [21]. At present, we cannot nd a reasonable explanation for such paradoxical phenomenon. Nonetheless, these ndings, at least, suggest that DHA's effects on regulating GPX4 expression in different solid tumors are inconsistent. The mechanisms underlying such interesting ndings require to be further investigated.
Next, we explored whether UPR-activated by DHA played a role in ferroptosis. UPR is coordinated by PERK/eIF2α/ATF4, IRE1α/XBP1 and ATF6 branches [22,23]. In response to ER stress, PERK is selfphosphorylated to induce the phosphorylation of eIF2α, thereby promoting ATF4 translation [33]. Once phosphorylated, IRE1α splices XBP1 into a short form [33]. ATF6 translocates into cell nucleus to function as a transcription factor after cleaving from ER [34]. ATF4 is suggested a protective molecule against ferroptosis because of its ability to activate the transcription of SLC7A11 [24]. Our prior work also demonstrated ATF4 as a pro-survival factor in erastin-induced ferroptosis in PLC cells (accepted). Therefore, we assumed that blocking PERK/eIF2α/ATF4 signaling pathway would further augment the anti-tumor effects of DHA in PLC cells. However, by determining cell vitality and ROS contents, we unexpectedly found that knockdown of ATF4 attenuated DHA's cytotoxic effects. GPX4 expression was upregulated in response to ATF4 silencing. Interestingly, we also found that DHA-triggered ferroptosis was attenuated when XBP1 or ATF6 was silenced. Unlike most previous studies revealing anti-ferroptosis roles of these UPR proteins [21, 24], our work demonstrates them as pro-ferroptosis molecules in DHA-treated PLC cells (sFig. 8).
Due to the signi cant role of ATF4 in inducing SLC7A11 expression [24], it is hard to nd previous evidence to support the pro-ferroptosis role of ATF4 observed in our study. Nonetheless, an earlier study evaluating the effects of artesunate, another classic ferroptosis inducer, in Burkitt's lymphoma cells [15] indirectly supported our ndings. This study showed that ATF4 activated the transcription of CHAC1 to augment ferroptosis induced by artesunate in DAUDI and CA-46 cells [15]. CHAC1 function as γ-glutamyl cyclotransferase to induce GSH degradation [35]. Such ndings imply that ATF4 upregulation also leads to GSH degradation. It is possible that DHA, like artesunate, activates mechanisms important for GSH degradation mediated by UPR proteins, rather than GSH synthesis. By performing dual luciferase reporter assay, we demonstrated that DHA enhanced the promoter activity of CHAC1 gene. Unlike ATF4, to our knowledge, XBP1 and ATF6 were demonstrated to promote CHAC1 expression for the rst time.
The rst limitation of our study was that we only knocking down the expression of key UPR regulators, ATF4, XBP1 and ATF6, in PLC cells. Since the endogenous expression levels of them are already abundant in PLC cells stimulated with DHA, our group believe that knocking them down is more appropriate. Nonetheless, to demonstrate the roles of activated UPR in DHA-induced ferroptosis, overexpressing them in PLC cells is still needed. There are several binding sites predicted for XBP1 and ATF6 on CHAC1 promoter. Although the emphasis of this study is to demonstrate how DHA affects the promoter activity of CHAC1, there is still a need to identify the speci c locations that these UPR sensors bind to. This is the second limitation.

Conclusion
Our study demonstrates that DHA effectively induces ferroptosis and activates UPRs in PLC cells. Inhibition of UPRs suppresses the cytotoxic effects of DHA on PLC cells, which involved a reduction in CHAC1, a key gene responsible for GSH degradation. DHA promotes the transcription of CHAC1. These ndings provide novel insights into the anti-tumor effects of DHA, suggesting UPR triggered by DHA is pro-death in PLC cells.

Ethics approval
The present work was approved by the First A liated Hospital of Zhengzhou University, and the animal procedures conformed to the Guideline for the Care and Use of Laboratory Animals.

Consent to participate
Not applicable Availability of data and materials All data generated or analyzed during this study are included in this published article.

Competing interests
The authors declare that they have no competing interests.    Administration of DHA limits PLC tumor growth in nude mice PLC cells were subcutaneously injected into immune de cient mice. DHA (100 mg/kg/d, 5 d/wk) was given to these mice when the tumor volumes reached 80-100 mm3. The tumor volumes were recorded every three days (a-b), and three weeks later, the xenografted tumors were collected for H&E staining (c; bar, 100 μm). The ROS contents (d) and GSH-PX activities (e) in tumor tissues were determined (n = 4). *p < 0.05, **p < 0.01, ***p < 0.001 vs control.
Values were presented as mean values ± standard deviation. DHA, dihydroartemisinin; PLC, primary liver cancer; ROS, reactive oxygen species; GSH-PX, glutathione peroxidase  Page 27/36 The cytotoxic effects of DHA are attenuated when the UPRs are blocked Three speci c siRNAs were synthesized to target ATF4, XBP1 and ATF6 in PLC cells. Knockdown of UPRs partly alters the expression of molecules associated with ferroptosis in PLC cells in presence of DHA Three speci c siRNAs were synthesized to targeting ATF4, XBP1 and ATF6 in PLC cells.   DHA enhances the promoter activity of CHAC1 gene (a) Potential binding sites for ATF4, XBP1 and ATF6 on CHAC1 gene promoter. (b) pGL3 luciferase reporter assay was carried out to determine the promotor activity of CHAC1 gene. (c) Two siRNAs were used to knock CHAC1 expression down in PLC cells. The protein levels of GPX4 and CHAC1 were determined with Western blot. Cell vitalities were determined with CCK8 assay (f & g), and intracellular lipid ROS (h & i) levels were determined with ow cytometry (n = 3).