The DUBA-SLC7A11-c-Myc axis is critical for stemness and ferroptosis

Ferroptosis is characterized by the accumulation of lipid peroxidation as a unique iron-dependent cell death. However, the interplay between stemness and ferroptosis remains unknown. Here, we demonstrate that undifferentiated cells are more sensitive to ferroptosis than differentiated cells, and cystine transporter SLC7A11 protein is highly up-regulated by deubiquitinase DUBA in differentiated cells. Additionally, DUBA promotes stemness by deubiquitinating SLC7A11. Moreover, SLC7A11 drastically increases the expression of c-Myc through cysteine, the combination of sorafenib and c-Myc inhibitor EN4 has a synergetic effect on cancer therapy. Together, our results reveal that enhanced stemness increases the susceptibility to ferroptosis, and the DUBA-SLC7A11-c-Myc axis is pivotal for differentiated cancer stem cells (CSCs) resistant to ferroptosis, providing a promised targets to eradicate CSCs through ferroptosis.


INTRODUCTION
Cancer stem cells (CSCs), a special cell population initiating tumor, are closely associated with cellular heterogeneity and therapeutic resistance of advanced cancers [1]. Four pluripotent inducers OCT4, SOX2, KLF4, and c-Myc (OSKM) named Yamanaka factors can reprogram somatic cells into induced pluripotent stem cells (iPSCs) which decreases the glutathione-mediated reactive oxygen species (ROS) levels, restrain the DNA damage response [2,3]. Additionally, MYC enhanced the chemotherapy resistance of breast CSCs by increasing mitochondrial oxidative phosphorylation (OXPHOS) and the production of ROS [4]. Targeting CSCs is a promising strategy to eradicate tumors [5].
As a newly defined regulated cell death in 2012, ferroptosis can be inhibited by eliminating lipid peroxides through Glutathione peroxidase (GPX4), which depends on system xc-(SLC3A2 + SLC7A11) importing cystine. Moreover, depletion of glutathione leads to accumulating ROS and ferroptosis, having the tremendous power to connect other areas of biology and medicine for cancer therapy [6,7]. Epithelial-mesenchymal transition (EMT) marker ZEB1 and Wnt receptor FZD7 activate the mesenchymal progression of cancer cells and mitigate the sensitivity to conventional therapy, while they increase the sensitivity to ferroptosis [8,9]. In addition, drug resistant cancer cells are vulnerable to GPX4 inhibition, resulting in ferroptotic death and preventing tumor relapse [10].
Compared to cancer cells, CSCs are addicted to more iron to maintain their self-renewal and plasticity. Given the requirement of iron is different from CSCs and others, it is potentially developed that personalized cancer therapies targeting CSCs for iron-regulated ferroptosis beyond traditional treatments [11].
Therefore, this motivates us to explore the susceptibility of stemness to ferroptosis and its concrete molecular mechanism in the study, which can be considered for effective cancer therapy.

RESULTS
Undifferentiated cells are more sensitive to ferroptosis than differentiated cells To investigate the sensitivity of stemness to ferroptosis, we first cultured different cancer cells with distinct proportion of side population (SP) cells. SP cells, a specific subgroup existed in most tumours, have the same characteristics stemness [12]. Subsequently, we treated these cells with indicated concentrations of various ferroptosis activators, including erastin directly blocking system Xc- [13] and RSL3 inhibiting GPX4 [14,15]. Compared to lung cancer cells H1299, liver cancer cells LM3, and breast cancer cells MCF-7, the results showed that A549 cells (Fig. 1A) and Hep3B cells (Fig. 1B) which possessed a higher percentage of SP cells (Fig. S1A, B) were more sensitive to erastin through cell viability assays. Similarly, glioma cells U87 and breast cancer cells MDA-MB-231 relative to U251 and MCF7 respectively were more sensitive to ferroptosis inducer erastin or RSL3 (Fig. S1C-E). Besides, compared to H1299 cells, A549 cells displayed a higher level of lipid ROS when they were exposed to RSL3 simultaneously ( Fig. S1F). Therefore, it was suggested that stemness enhanced the sensitivity to ferroptosis.
Compared to parent cells A549, the data showed that tumor sphere-formed cells (Fig. 1E) were more sensitive to erastin and RSL3, whereas this could be rescued by ferroptosis inhibitor ferrostatin-1 (Fer-1). Moreover, we found that the malondialdehyde (MDA) levels of A54 oncosphere are higher than the MDA levels of A549 cells when they were exposed to erastin (Fig. S1G). Likewise, compared to Hep3B cells, tumor sphere-formed Hep3B cells (Fig. S1H, I) were more sensitive to Erastin and RSL3 (Fig. S1J).
On the other hand, All-trans Retinoic Acid (ATRA) was utilized to induce the mouse embryonal carcinoma stem cells P19 differentiation [17] and reduced the protein levels of OCT-4 and SOX2 (Fig. 1F). We also observed that P19 cells relative to differentiated P19 cells had a lower proportion of cell viability under another system Xcinhibitor Imidazole Ketone Erastin (IKE) [18] and RSL3 treatment (Fig. 1G). Moreover, P19 cells exhibited a higher level of lipid ROS (Fig. 1H) when they were exposed to RSL3. Besides, the pictures derived from transmission electron microscopy (TEM) demonstrated that a higher degree of thickened crista and shrunken mitochondria were observed in P19 cells beyond differentiated P19 cells (Fig. 1I). And more numbers of the dead cells were observed in P19 cells without ATRA-induced differentiation after IKE and RSL3 treatment (Fig. 1J). Consequently, this evidence indicated that differentiated cells increased the resistance to ferroptosis.
Taken together, these results demonstrate that enhanced stemness promotes the sensitivity of cells to ferroptosis different from conventional radiotherapy and chemotherapy. It is possible that targeting ferroptosis provides a promising method for CSCs eradication.

SLC7A11 is up-regulated by DUBA in differentiated cells
Notably, the drug-resistant cells always stay at a mesenchymal state, but they have no sufficient antioxidative ability to oxidants like GPX4 inhibitor [19]. To elucidate the detailed mechanisms that CSCs were sensitive to ferroptosis, we first analyzed the protein expression of lipid peroxidation factors including SLC7A11, ACSL4, GPX4 and COX2 [20] between the A549 cells and H1299 cells (Fig.  S2A), and U87 cells and U251 cells (Fig. S2B), and found SLC7A11 was up-regulated in H1299 cells and U251 cells. Moreover, we treated P19 cells (Fig. S2C) and LM3 cells ( Fig. S2D) with ATRA and found that the mRNA levels of these ferroptosis-related factors have a slight decrease in differentiated cells. This could be explained by two reasons: low levels of ROS could maintain the growth of ATRA-induced cells [21] and the decrease of stemness factors like SOX2 reduced the mRNA levels of SLC7A11 [22]. However, ATRA greatly increased the protein level of SLC7A11 with increasing concentrations of ATRA in P19 cells ( Fig. 2A) and LM3 cells (Fig. 2B), which probably played a vital role in differentiated cells resistant to ferroptosis.
For the protein of SLC7A11 was dramatically up-regulated in differentiated cells, we wondered whether the critical posttranslational modification ubiquitin-proteasome system (UPS) involved this quality control of proteins [23]. To confirm that SLC7A11 protein was regulated by UPS, we treated ATRA-induced P19 cells (Fig. 2C) and LM3 cells (Fig. 2D) with MG132 respectively. The results demonstrated that MG132 could offset the difference of SLC7A11 protein between DMSO and ATRA treatment, suggesting that SLC7A11 was regulated by UPS in differentiated cells.
Protein ubiquitination which was reversed by several distinct families of deubiquitinases (DUBs) involved in this stabilization process [24]. To screen the DUBs up-regulating SLC7A11, about 90 DUBs including five families: UCHs, USPs, OTUs, Josephins, and JAMMs in the human proteome [25,26] were transiently transfected into 293T cells. Immunoblotting assays showed that the increasing level of SLC7A11 was the most obvious when DUBA was overexpressed (Fig. 2E). Moreover, we detected the protein levels of DUBA in ATRA-induced P19 cells and found that DUBA was increased with increasing concentrations of ATRA (Fig. 2F), indicating that ATRA could increase SLC7A11 through inducing DUBA. DUBA is a significant regulator of multiple factors including mTOR signaling, type I interferon and TRAF3 [27].
Iron is mainly stored in liver and the imbalance of iron homeostasis involves various liver diseases including liver injury and hepatocellular carcinoma (HCC) [28]. HCC is one of the leading causes of cancer-related death worldwide [29,30]. The report points out that triggering ferroptosis in iron-rich tumors (such as HCC and NSCLC, etc) may develop new therapeutic avenues or reverse drug-resistance in cancers [31]. Therefore, HCC cells were considered as main research object in our study. First, we detected the protein levels of DUBA and SLC7A11 in HCC cell lines and found that they both were highly expressed in HCC cell lines relative to normal liver cells 7701. Moreover, there was a positive correlation between DUBA and SLC7A11 in these cell lines (Fig.  S3A). Subsequently, a similar finding that the SLC7A11 was gradually up-regulated with increasing amounts of DUBA was also observed in Hep3B cells (Fig. 2G).
In summary, we discover that SLC7A11 is up-regulated by DUBA in differentiated cells, leading to inhibit cell ferroptosis. This may provide therapy strategy for CSCs through DUBA-SLC7A11.

DUBA stabilizes and interacts with SLC7A11
Next, to investigate the interaction between DUBA and SLC7A11, we first co-transfected DUBA and SLC7A11 into 293T cells and found DUBA could be immunoprecipitated by SLC7A11 antibody (Fig. 3A). Moreover, it was also indicated that DUBA was immunoprecipitated by SLC7A11 antibody from LM3 cell lysates (Fig. 3B). In addition, the 293T cells were co-transfection with SLC7A11 and DUBA and followed by immunofluorescence assay. The images displayed that there was a co-location between SLC7A11 and DUBA (Fig. 3C).
To determine whether DUBA stabilized SLC7A11 protein through deubiquitination, we first treated the DUBA-overexpressed Hep3B cells with the proteasome inhibitor MG132 [32]. The immunoblot displayed that DUBA overexpression increased the level of SLC7A11, whereas there was no obvious difference between Vector and Overexpression cells treated with MG132 (Fig. 3D). Furthermore, we treated Hep3B cells overexpressing DUBA using protein synthesis inhibitors cycloheximide (CHX) and examined SLC7A11 protein degradation for the indicated hours. It was indicated that DUBA had a stable effect on SLC7A11 protein (Fig. 3E). Conversely, knocking down DUBA with two separate short hairpin RNAs (shRNAs) promoted SLC7A11 degradation in LM3 cells treated with CHX for the indicated hours (Fig. 3F).
Above all, we screen and identify a DUBs DUBA that can maintain SLC7A11 stability through deubiquitination through interacting with SLC7A11. Fig. 1 Undifferentiated cells are more sensitive to ferroptosis than differentiated cells. A The viability of A549 cells and H1299 cells treated with erastin for the indicated concentrations was counted by CCK-8. B The viability of Hep3B cells and LM3 cells treated with erastin for the indicated concentrations was counted by CCK-8. C Representative images of A549 cells and A549 oncospheres. Scale bar, 100 μm. D Stemnessrelated proteins were detected by western blotting in A549 cells and A549 oncospheres. E The viability of A549 cells and A549 oncospheres treated with erastin, RSL3, and Fer-1 was counted by CCK-8. F Stemness-related proteins were detected by western blotting in P19 cells and P19 cells treated with ATRA for 48 h. G The viability of P19 cells and ATRA-induced P19 cells treated with erastin, RSL3, and Fer-1 was counted by CCK-8. H The lipid ROS level of P19 cells and ATRA-induced P19 cells treated with RSL3 was analyzed by flow cytometry using the probe BODIPY® 581/591. I Representative mitochondrion pictures in P19 cells and ATRA-induced P19 cells treated with RSL3 were assessed by transmission electron microscope. Scale bar, 2 μm. J Representative images of P19 cells and ATRA-induced P19 cells treated with IKE and RSL3. Scale bar, 100 μm. The data are presented as mean ± SEM. P values (*p < 0.05, **p < 0.01, ***p < 0.001) are indicated in each panel with twosided Student's t-test.

DUBA deubiquitinates SLC7A11
To examine the possibility that DUBA deubiquitinated SLC7A11, in vitro ubiquitylation assays showed that DUBA could directly remove the ubiquitin chain of SLC7A11 in Hep3B cells (Fig. 4A) and HepG2 cells (Fig. 4B). Since different ubiquitin chains have diverse roles in regulating protein location and function [33], we next co-transfected 293T cells with SLC7A11, Ub or its site mutants, and Vector or DUBA to investigate the type of SLC7A11 ubiquitination mediated by DUBA. The results exhibited that DUBA could significantly reduce the ubiquitination level of A Stemness-and ferroptosis-related proteins were detected by western blotting in P19 cells treated with ATRA for the indicated concentrations. B Stemness-and ferroptosis-related proteins were detected by western blot in LM3 cells treated with ATRA for the indicated concentrations. C The protein levels of SLC7A11 were detected by western blotting in P19 cells which were treated with ATRA for 48 h and then treated with MG132. D The protein levels of SLC7A11 were detected by western blotting in P19 cells which were treated with ATRA for 48 h and then treated with MG132. E The indicated OTU subfamily DUBs were each transfected into 293T cells and SLC7A11 expression was detected by western blotting. F DUBA was detected by western blotting in P19 cells treated with ATRA for the indicated concentrations. G Increasing amounts of DUBA were transfected into Hep3B cells and the protein levels of SLC7A11 were detected by western blotting. SLC7A11 in 293T cells transfected with WT ubiquitin and its mutant K48 (Fig. 4C), suggesting that DUBA cleaved the K48linked ubiquitination of SLC7A11 rather than other forms of polyubiquitin chains.
Subsequently, we constructed two different DUBA mutations including a phosphosite mutant DUBA S177A and a catalytic cysteine mutant DUBA C224S [34] and respectively transfected them into Hep3B cells. Immunoblot showed that DUBA S177A and DUBA C224S relative to DUBA wild type (WT) could not upregulate SLC7A11 dramatically (Fig. 4D). Moreover, the cell lysates from 293T cells co-transfected DUBA WT, DUBA S177A, or DUBA C224S with SLC7A11 were precipitated by anti-flag antibody. Consequently, both the two mutations of DUBA did not affect the interaction with SLC7A11 (Fig. 4E). However, the ubiquitylation results indicated that DUBA C224S could not evidently decrease the ubiquitination level of SLC7A11 in 293T cells, while DUBA S177A did not affect the ubiquitination level (Fig. 4F).
Taken together, these studies corroborate that DUBA can interact and stabilize SLC7A11 through removing the K48-linked ubiquitin chains of SLC7A11, but DUBA C224S could not decrease the ubiquitination level of SLC7A11 relative to DUBA WT, which provides multiple intervention measures for the stabilization of SLC7A11 through DUBA.

DUBA inhibits ferroptosis and enhances stemness
Since the critical role of SLC7A11 in ferroptosis, we confirmed the similar function of DUBA on ferroptosis. Consequently, we found that overexpression of DUBA could resist RSL3 treatment in Hep3B cells (Fig. S4A) and decrease the lipid ROS level of Hep3B cells (Fig.  S4B). Reversely, when DUBA was inhibited in LM3 cells, DUBA inhibition could enhance the sensitivity of cells to ferroptosis (Fig.  S4C) and increase the lipid ROS level of LM3 cells (Fig. S4D).
To explore the role of SLC7A11 in cancer stemness, we found that SLC7A11 overexpression facilitated the oncosphere formation ability of Hep3B cells (Fig. S5A). Moreover, SLC7A11 also increased the proportion of SP cells (Fig. S5B) and CD338+ cells (Fig. S5C) in Hep3B cells, and CD338+ cells population in HepG2 cells (Fig.  S5D). Then, we investigated the effect of SLC7A11 on key stemness factors (Fig. S5E) and surprisingly found that its overexpression up-regulated c-Myc evidently and had a little effect on SOX2 expression, etc. Conversely, the ALDH activity was reduced in LM3 cells (Fig. S5F) after the knockdown of SLC7A11 with two shRNAs. This problem that the up-regulation of SLC7A11 promotes cancer stemness in differentiated cells seems contradictory. We consider that SLC7A11 is not a core stemness factor and cannot change the ATRA-induced differentiation. Moreover, the combination of stemness inhibitor and SLC7A11 may show a great therapeutic potential. To demonstrate the role of DUBA in stemness maintenance, it was found that overexpression of DUBA promoted cell proliferation (Fig. S6A) and facilitated sphere formation frequency (Fig. 5A) in Hep3B cells. Subsequently, we found that the ALDH activity (Fig. 5B) and the proportion of SP cells (Fig. 5C) and CD338+ cells (Fig. S6B) were increased in Hep3B cells after overexpression of DUBA. Similarly, overexpression of DUBA greatly up-regulated the protein level of c-Myc (Fig. 5D). Consistent with the results from in vitro experiments, and in vivo study indicated that overexpression of DUBA increased incidence of Hep3B xenograft tumor and tumor weight in nude mice subcutaneously injecting Hep3B cells (Fig. 5E). Moreover, immunoblot analysis of Hep3B xenograft tumor also displayed that there was a positive correlation between DUBA, SLC7A11, and c-Myc (Fig. 5F). Furthermore, we explored the role of DUBA mutations in stemness maintenance and demonstrated that DUBA S177A and DUBA C224S relative to DUBA WT could not promote the growth of HepG2 cells (Fig. S6C). Compared to Fig. 4 DUBA deubiquitinates SLC7A11. A DUBA-overexpressed Hep3B cells were treated with MG132. And the cell lysates were subject to immunoprecipitate with anti-SLC7A11 antibody, which was followed by immunoblot with anti-Ub antibody to detect SLC7A11 ubiquitination. B DUBA-overexpressed HepG2 cells were treated with MG132. And the cell lysates were subject to immunoprecipitate with anti-SLC7A11 antibody, which was followed by immunoblot with anti-Ub antibody to detect SLC7A11 ubiquitination. C SLC7A11, HA-Ub, or its lysine residue K48 and K63 were transfected into 293T cells with or without Flag-DUBA, which were treated with MG132. The cell lysates were immunoprecipitated with anti-SLC7A11 antibody, which was followed by immunoblot with anti-Ub antibody to detect SLC7A11 ubiquitination. D Hep3B cells were transfected with Flag-tagged DUBA WT, DUBA S177A, or DUBA C224S mutants. And SLC7A11 expression was detected by western blotting. E 293T cells were transfected with SLC7A11 alone or in combination with Flag-tagged DUBA WT, DUBA S177A, or DUBA C224S mutants. And the cell lysates were subject to immunoprecipitation with an anti-SLC7A11 antibody. F Flag-tagged DUBA WT, DUBA S177A, or DUBA C224S mutants were transfected into 293T cells treated with MG132. And the cell lysates were subject to immunoprecipitation with an anti-SLC7A11 antibody.
Conversely, the growth (Fig. S7A) and migration (Fig. S7B) ability of LM3 cells were evidently attenuated when DUBA was inhibited.
Moreover, the proportion of SP cells was reduced in LM3 cells after the knockdown of DUBA (Fig. S7C). Consistently, we performed LM3 xenograft tumor into nude mice with subcutaneous injection of LM3 cells and found that knockdown of DUBA significantly reduced tumor size (Fig. S7D) and weight (Fig. S7E). immunoblot analysis of LM3 xenograft tumor also demonstrated DUBA could stabilize SLC7A11 and c-Myc (Fig. S7F).
To sum up, similar to SLC7A11, DUBA inhibits cell ferroptosis and promotes cancer stemness properties both in vitro and in vivo, while DUBA S177A and DUBA C224S are unable to achieve. DUBA has critical roles in ferroptosis and stemness through SLC7A11 and serves as a promising cancer therapy target.
The DUBA-SLC7A11-c-Myc axis enhances stemness and predicts a poor prognosis in HCC Patients To better confirm the DUBA-SLC7A11-c-Myc axis in stemness maintenance, we transfected SLC7A11 into LM3 cells with knockdown of DUBA. The results indicated that DUBA knockdown decreased the level of SLC7A11 and c-Myc (Fig. 6A) and inhibited the clone forming ability and metastatic potential (Fig. 6B) of LM3 cells were consistent with previous research. Moreover, SLC7A11 overexpression could reverse the status and had no effect on the protein level of DUBA, but significantly increased the protein level of c-Myc.
Since the effect that self-renewal of cells was inhibited by reduced expression of c-Myc-HIF-2α was blocked by N-acetyl cysteine (NAC) [35], we speculated that SLC7A11 could stabilize c-Myc through cysteine. Thus, we cultured LM3 cells (Fig. 6C) and HepG2 cells (Fig. S8A) with DMEM medium lacking cystine [36] for the indicated hours respectively. As expected, it was found that deprivation of cystine greatly increased the expression of c-Myc at 24 h and reduced the protein level of c-Myc significantly rather than other stemness factors at 48 h of treatment. Actually, this change was in accord with a previous study that c-Myc directly or indirectly regulated the expression of amino acid transporters at high levels, especially SLC7A11 and SLC1A5 [37]. Furthermore, the deficiency of cysteine dramatically inhibited LM3 cells growth (Fig.  S8B) and promoted the sensitivity to RSL3 inducer, which was blocked by the addition of cysteine (Fig. S8C). Next, we explored the role of cysteine in stemness via c-Myc. The results showed that ALDH intensity (Fig. 6D) was greatly inhibited and the SP population of LM3 cells was decreased when cysteine was deprived in medium (Fig. S8D).
To clarify the significance of DUBA and SCL7A11 in HCC patients, we first utilized IHC analysis (Fig. 6E) and western blot (Fig. 6F) to detect their expression in HCC patient tissues compared to paracancerous normal tissues. The results indicated that both DUBA and SLC7A11 were increased in HCC samples relative to normal samples. Next, Kaplan-Meier curves analysis was used to estimate the relationship between the level of DUBA and SLC7A11 and the survival rate of HCC patients. According to the TCGA database, we found that high expression of DUBA and SLC7A11 was in accord with a low survival rate respectively (Fig. 6G). Moreover, the combined high expression of DUBA and SLC7A11 predicted a lower survival ratio relative to the collectively low expression of them (Fig. 6H).
In a word, SLC7A11 is up-regulated by DUBA through deubiquitination and imported cystine stabilizing c-Myc protein mainly, thus resisting ferroptosis and enhancing stemness. Moreover, both DUBA and SLC7A11 are highly expressed in HCC patients, and their combined high expression predicts a poor prognosis.
The combined treatment of sorafenib and c-Myc inhibitor EN4 inhibits HCC Currently, though sorafenib is the only first-line chemotherapy for advanced HCC, it is far from satisfactory in HCC therapy efficacy and causes severe adverse effects [38,39]. Besides, sorafenib could induce HCC cells ferroptosis through SLC7A11 inhibition [40]. Thus, to investigate the effect of DUBA on sorafenib treatment, we first treated DUBA-overexpressed Hep3B cells with sorafenib and found that overexpression of DUBA could enhance the growth inhibition (Fig. 7A) and tumorspheres formation (Fig. 7B), while DUBA S177A and DUBA C224S could not resist to sorafenib. Conversely, after exposure to sorafenib, the knockdown of DUBA greatly restrained LM3 cell proliferation (Fig. S9A) and oncospheres formation (Fig. S9B) beyond without sorafenib.
For the stability of c-Myc by SLC7A11, a c-Myc inhibitor EN4 [41] was used to treat HCC in combination with sorafenib. First, we treated Hep3B cells with EN4 and found that DUBA could enhance the growth inhibition (Fig. 7C). To determine the significance of combination between sorafenib and EN4 in HCC, we first cotreated LM3 cells with EN4 and sorafenib. The results uncovered that their combination had an obvious growth inhibition beyond EN4 or sorafenib alone (Fig. 7D) and decreased ALDH intensity dramatically (Fig. 7E). Moreover, compared to EN4 or sorafenib treatment, the combination of sorafenib and EN4 significantly increased the lipid ROS level of LM3 cells (Fig. 7F). Importantly, the sorafenib or EN4 had no effect on the expression of SLC7A11 and c-Myc separately, while their combination significantly decreased the level of SLC7A11 and c-Myc (Fig. 7G).
Above all, we demonstrate an important pathway DUBA-SLC7A11-c-Myc increases the resistance to sorafenib and EN4. Moreover, the combination of sorafenib and EN4 has a synergetic inhibition role in HCC progression through stemness maintenance and ferroptosis, which could be a new and effective treatment modality for HCC (Fig. 7H).

DISCUSSION
Since the CSCs are responsible for cancer metastasis, recurrence, and heterogeneity that seriously threaten human life, chemotherapy, radiotherapy, and targeted therapy are failed to eradicate tumors [42]. Stem cells generally share two pivotal abilities to selfrenew indefinitely and differentiate into diverse cell types. Thus, exploring the control of embryonic stem cells, adult stem cells, and CSCs function may guide our understanding of human disease and tissue homeostasis and cancer grade and metastasis [43]. For example, pharmacologic targeting of circadian networks induced apoptosis which disrupted cancer stem cell growth [44]. On the other hand, impaired cancer cell stemness results in increased apoptosis and reduced survival potential [45]. Especially, two different research institutions recently have revealed chaperone-mediated autophagy (CMA) not only sustains hematopoietic stem-cell pluripotency [46], but embryonic stem cell's , and the numbers of oncospheres were recorded in the bar graph (right). B Representative images of ALDH activity in Hep3B cell stably expressing DUBA were analyzed by flow cytometry using the dye ALDEFLUOR (left). And the ALDH intensity was calculated in the bar graph (right). C Representative images of SP cells in Hep3B cell stably expressing DUBA were analyzed by flow cytometry using the dye Hoechst 33342 (left). And the percentage of SP cells was calculated in the bar graph (right). D Stemness-related proteins were detected by western blotting in Hep3B cells stably expressing DUBA. E Tumor pictures (up) of mice with subcutaneous injection of DUBA-overexpressed Hep3B cells with limiting dilution were taken, and tumor weight was recorded in the bar graph (down). F Stemness-related proteins were detected by western blotting in Hep3B xenograft tumors from E. G Representative images of CD338 positive cells in HepG2 cell with DUBA WT, DUBA S177A, and DUBA C224S mutants were analyzed by flow cytometry using CD338 antibody (left). And the percentage of CD338 positive cells was calculated in the bar graph (right). The data are presented as mean ± SEM. P values (*p < 0.05, **p < 0.01, ***p < 0.001) are indicated in each panel with two-sided Student's t-test.
function. Moreover, CMA is up-regulated in these activated cells [47]. In addition, current studies have indicated that high intracellular iron content is required for CSCs growth, and targeting ferroptosis can induce CSCs death [48]. Ferroptosis plays a significant role in radiotherapy-triggered and chemotherapy-caused cell death in tumor suppression [49]. Thus, combination therapeutic strategies of ferroptosis targeting CSCs are the promising way to effective cancer treatment.
In our study, we have demonstrated that enhanced stemness increases the sensitive of cells to ferroptosis and infer that the lack of antioxidant ability and the abundance of iron storage make CSCs sensitive to ferroptosis. Moreover, SLC7A11 protein is stabilized by DUBA via deubiquitination in differentiated cells. And the stabilizing efficiency of DUBA on SLC7A11 is greater than OTUB1 reported previously [50]. However, their binding sites remain to be investigated and the inhibitors of DUBA have not been developed.  Fig. 6 The DUBA-SLC7A11-c-Myc axis enhances stemness and predicts a poor prognosis in HCC Patients. A DUBA-knockdown LM3 cells using shRNA#1 were transfected with SLC7A11. B Representative images of cell migration and colony were derived from A. C LM3 cells were cultured in DMEM medium lacking cystine for the indicated hours, and stemness-related proteins were detected by western blotting. D Representative images of ALDH activity in LM3 cells cultured in DMEM medium lacking cysteine were analyzed by flow cytometry using the dye ALDEFLUOR (left). And the ALDH intensity was calculated in the bar graph (right). E Representative IHC images of DUBA and SLC7A11 in normal tissue and tumor tissue derived from HCC patients (Normal n = 10; Tumor n = 10). F DUBA and SLC7A11 expression was detected by western blotting in paracancerous tissue (N) and tumor tissue (T) derived from HCC patients G Kaplan-Meier overall survival curves were compared between HCC patients with a high or low SLC7A11 level (left) and a high or low DUBA level (right) (High n = 184; Low n = 184).
H Kaplan-Meier overall survival curves were compared between HCC patients with a combined high expression of SLC7A11 and DUBA or an integrated low SLC7A11 level and low DUBA level (High n = 94; Median n = 188; Low n = 94). The data are presented as mean ± SEM. P values (*p < 0.05, **p < 0.01, ***p < 0.001) are indicated in each panel with two-sided Student's t-test.
In addition, DUBA-SLC7A11 enhances the stem cell-like properties by increasing the level of c-Myc. About the problems that SLC7A11 is up-regulated in differentiated cells and promotes cell stemness, we suppose that SLC7A11 is not a core stemness factors and cannot change the state of differentiated cells. If the cancer cells are treated with SLC7A11 overexpression and differentiation inducers like ATRA, they are more resistant to ferroptosis. As a critical role in innate immune and T cells [51], DUBA not only increases the Fig. 7 The combined treatment of sorafenib and c-Myc inhibitor EN4 inhibits HCC. A The growth of HepG2 cells transfected with DUBA WT, DUBA S177A, and DUBA C244S mutants and treated with sorafenib was counted by CCK-8. B Representative oncosphere and colony pictures of HepG2 cells transfected with DUBA WT, DUBA S177A, and DUBA C244S mutants and treated with sorafenib. C The growth of HepG2 cells transfected with DUBA WT, DUBA S177A, and DUBA C244S mutants and treated with EN4 was counted by CCK-8. D The growth of LM3 cells treated with sorafenib and EN4 was counted by CCK-8. E Representative images of ALDH activity in LM3 cells treated with sorafenib or EN4 alone, and in combination with sorafenib and EN4, was analyzed by flow cytometry using the dye ALDEFLUOR (left). And the ALDH intensity was calculated in the bar graph (right). F The lipid ROS level of LM3 cells treated with sorafenib or EN4 alone, and in combination with sorafenib and EN4, was analyzed by flow cytometry using the probe BODIPY® 581/591. G SLC7A11 and c-Myc expression were detected by western blotting in LM3 cells treated with sorafenib or EN4 alone and in combination with sorafenib and EN4. H A schematic model of DUBA-SLC7A11-c-Myc axis in ferroptosis inhibition and stemness maintenance, increasing the resistance to sorafenib and EN4. The data are presented as mean ± SEM. P values (*p < 0.05, **p < 0.01, ***p < 0.001) are indicated in each panel with two-sided Student's t-test.
resistance of cells to sorafenib, but also to c-Myc inhibitor EN4. Importantly, the combination of sorafenib and EN4 greatly promotes ferroptosis and attenuates stemness, displaying a synergetic inhibition role in HCC progression.
Since the pluripotency of stem cells is determined by the core stemness factors, like SOX2 and Nanog, their regulations of them play an essential role in stemness maintenance. Recently, besides transcriptional and epigenetic regulation, extensive researches declared that post-translational modifications are also an important way to maintain stemness, especially ubiquitination [52]. For instance, OTUD7B increases the stability of SOX2 by removing polyUb from it during neural progenitor cells (NPCs) differentiation [53]. Conversely, Ube2s ubiquitinates and degrades SOX2 through the proteasomal system, which contributes to mouse ES cell maintenance [54]. In our study, we demonstrate that SLC7A11 also could be a critical stemness maintenance factor and identified deubiquitinase DUBA on SLC7A11 stability. Moreover, apart from transcriptional regulation like deubiquitinase BAP1 and P53 [55], the Gan professor proposes that the diverse post-translational mechanisms of SLC7A11 can be a meaningful strategy for ferroptosis, nutrient dependency, and cancer therapy [56]. In addition, we demonstrated that cysteine mediated by SLC7A11 had a more stable effect on c-Myc than SOX2 and KLF4, which enhanced the stem cell-like properties distinctly.
To sum up, we demonstrate the interplay between stemness and ferroptsis, and uncover the potential molecular mechanism. Based on the critical role of DUBA in ferroptosis and immune, we hope to combine DUBA inhibitor and conventional therapy targeting ferroptosis for CSCs elimination, providing a promising strategy for effective cancer therapy.

MATERIALS AND METHODS Cell culture
Human normal hepatocytes 7701, human hepatoma cell lines Hep3B, HepG2, LM3 and 7721, mouse embryonal carcinoma cells P19 and 293T Cells were obtained from the Cancer Research Institute of Central South University. All these cells were cultured in DMEM medium (Gibco, USA) supplemented with 10%FBS and maintained at 37°C with 5% CO2. DMEM medium lacking glutamine, methionine, and cystine were purchased from Gibco (Cat#21013024). All cell lines were authenticated prior to study and tested for negative mycoplasma contamination.
Cell viability assays 1 × 10 3 cells were seeded in DMEM medium (100 μl/well) into 96 well plates with five replicate wells. After an overnight incubation, each well was added into 10 μl Cell Counting Kit-8 (Bimake) and incubated at 37°C, 5% CO2 for 2 h. Absorption values were measured at 450 nm on a multiplewell plate reader (Bio-Tek) and at 48 h. The cell viability was expressed as A/B, where A was the absorbance value from the experimental group and B was that from the control cells.

Quantitative real-time PCR (qRT-PCR)
Primers are listed in Supplementary Tables S1. TRIzol reagent (Takara) was used to separate total RNA. PrimeScriptTM RT reagent kit (Takara) was utilized to generate cDNA. Real-time PCR was analyzed by FastStart Universal SYBR Green Master. Relative gene expression was standardized by β-actin.

Western blot
Cells were harvested and washed twice with 1×PBS, lysed in IP lysis buffer on the ice for 2 h. After centrifugation at 15,000 × g, 4°C for 15 min, a quantity of 50 μg total protein was used for Western blot analysis. The protein lysates were resolved on SDS-PAGE gels and transferred onto a PVDF transmembrane for 2 h. The membrane was incubated at 4°C overnight with different primary antibodies. The next day, after washing three times with PBST, the membrane was incubated at room temperature with a corresponding second antibody and exposed by ChemiDox XRS+ image-forming system.

Immunoprecipitation
Total proteins were extracted in cell lysis buffer supplemented with protease inhibitor cocktail (MCE). Soluble cell lysates (1 mg protein) were pre-cleared with 5 μl protein G magnetic beads (Invitrogen) for 2 h at 4°C. Protein G magnetic beads were removed, followed by incubation at 4°C with another 5 μl protein G magnetic beads with antibodies overnight. Unbound proteins were removed by washing three times with lysis buffer. Following SDS-PAGE, immunoprecipitated proteins were transferred onto PVDF membranes and probed with various antibodies.

Immunofluorescence
Cells were fixed with pre-cooled methanol and permeabilized with 0.2% (vol/vol) Triton X-100 for 20 min at room temperature. Cells were blocked with 1% (vol/vol) BSA-PBS solution for 1 h at room temperature and incubated with primary antibody (1:500) at 4°C overnight. The secondary antibody (1:2000) was diluted in 1% (vol/vol) BSA-PBS and incubated at room temperature for 1 h. The slides were counterstained with DAPI. The slides were imaged using a Leica confocal microscope. The antibodies used were, anti-mouse Alexa Fluor 594, and anti-rabbit Alexa Fluor 488 (Invitrogen).

Transmission electron microscopy
P19 cells were seeded onto 6 cm plates and were treated with 10 μM erastin for 48 h. And the cells were fixed with 2.5% Glutaraldehyde solution. Images were acquired using a transmission electron microscope (HITACHI, HT7700).

Total Lipid ROS
The details of these procedures have been described previously [57].

Tumor sphere formation
Tumor sphere formation analysis was conducted as described before. In short, cells were plated in ultra-low attachment six-well plates (Corning) and were cultured in serum-free DMEM-F12 medium, supplemented with B-27 supplement (Gibco), 20 ng/ml EGF (Sigma), 20 ng/ml b-FGF (Gibco) as well as N-2 Plus Media Supplement (Gibco). After culturing for 1-2 weeks at 37°C with 5% CO2, cells then formed floating tumorspheres. Eventually, spheres larger than 50 μm in diameter were counted and analyzed.
Aldefluor assay, SP analysis, and CD338 activity The details of these procedures have been described previously [16]. The Aldefluor assay (STEMCELL Technologies) was performed by the manufacturer's instructions and analyzed with FlowJo software.
For SP analysis, the cells were harvested, resuspended at 1 × 10 6 cells/ml in PBS buffer containing 2% FBS, and Hoechst 33342 was added at a final concentration of 5 μg/ml with or without 50 μM Verapamil (Sigma), which was used to inhibit ABC transporters, followed by incubating at 37°C for 90 min. Then, the cells were washed with ice-cold PBS buffer and resuspended with ice-cold PBS buffer containing 2% FBS. Ultimately, propidium iodide was added to the cell suspension at a final concentration of 1-2 μg/ml (PI, Sigma).
The SP cells were selected and sorted by FACS and analyzed with FlowJo software.

Immunohistochemistry (IHC)
The Department of Pathology at Xiangya Hospital verified and provided biopsies of lung cancer. The method used for IHC analysis of paraffin sections from lung cancer tissues can be found in previous literature. Differential quantification was performed by two pathologists from Xiangya Hospital, Changsha, China.

Nude mice and study approval
The 4-week-old female nude mice were purchased from Hunan SJA Laboratory Animal Co., Ltd. (Changsha, China) and were subcutaneously injected with 1 × 10 6 cells. Tumor volume and the body weight of the nude mice were supervised. After 28 days, the tumors were removed after the mice were sacrificed at the indicated time, and then the weight of tumors was calculated.
All experimental procedures for animal study were approved by the Institutional Animal Care and Use Committee of the Xiangya Hospital (No.2020sydw0117), Central South University and adhered to the legal mandates and national guidelines for animal care and maintenance.

Statistical analyses
All statistical analyses were performed using the Graphpad Prism 8.0 software package (Abbott Laboratories, USA) for Windows. The data are presented as mean ± SEM from multiple individual experiments each performed in triplicates. Student's t-tests (two-tailed) were applied to compare differences between two groups: ns is non-significant (p > 0.05), *p < 0.05, **p < 0.01, ***p < 0.001. If data were not normally distributed, data were analyzed by Kolmogorov-Smirnov test.

DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request.