Nedd4 ubiquitylates VDAC2/3 to suppress erastin-induced ferroptosis in melanoma

Ferroptosis is a newly defined form of regulated cell death characterized by the iron-dependent accumulation of lipid hydroperoxides. Erastin, the ferroptosis activator, binds to voltage-dependent anion channels VDAC2 and VDCA3, but treatment with erastin can result in the degradation of the channels. Here, the authors show that Nedd4 is induced following erastin treatment, which leads to the ubiquitination and subsequent degradation of the channels. Depletion of Nedd4 limits the protein degradation of VDAC2/3, which increases the sensitivity of cancer cells to erastin. By understanding the molecular mechanism of erastin-induced cellular resistance, we can discover how cells adapt to new molecules to maintain homeostasis. Furthermore, erastin-induced resistance mediated by FOXM1-Nedd4-VDAC2/3 negative feedback loop provides an initial framework for creating avenues to overcome the drug resistance of ferroptosis activators.

1. Initial data presented by authors suggest that VDAC2 and VDAC3 need to cooperate in order to induces ferroptosis following erastin stimuli. This is evident in Fig 1e, where overexpression of either alone has limited effect. Conversely, KD of either of t he two was shown to be sufficient to affect degree of ferroptosis. This seemingly contradiction need to be addressed. Further and as important is the notion that most if not all subsequent experiments shown in this manuscript were performed with a single VDAC construct, and not with both, a point that needs to be further addressed experimentally.
2. Data suggesting that Nedd4 is the E3 ligase that limits VDAC2/3 stability (Fig 2/3) need to be substantiated by half-life studies for the endogenous proteins prior and following erastin addition, in the presence and absence of Nedd4.
3. Figure 4g-I demonstrate the effect of erastin on ferroptotic cell death in the presence of Nedd4 overexpression. Yet, the effect shown is rather modest and need to be discussed relative to the changes shown in GSH and Fe levels. Would scavengers for Fe and ROS ablate these effects? 4. Figure 5c-e claims for rescue of cell sensitivity ot erastin upon OE of KR mutants of VDAC2/3yet the effect is really partial. The text in legend and results and related sections of the ms need to be properly adusted. 5. Likewise, changes hown in Fig 6h-j are rather modest and claims should therefore be adjusted in respective sections in text. 6. Experiments performed in Fig 7a, 7c should be also carried out using constructs that will impair the levels of VDAC2/3 and FOXM1, in order to substantiate the model proposed by authors.
previously reported observation -VDAC2/3 expression decreases upon induction of ferroptosis by cystine transporter inhibitor, erastin (note that erastin has been reported previously to be able to bind to VDACs). Following this lead, by using several melanoma cell lines, the authors found that erastin, and possibly more generally, oxidative stress, upregulates transcription factor FOXM1; FOXM1 in turn upregulates the transcription of an E3 ubiquitin ligase Nedd4, which the authors demonstrated to be able to promote VDAC2/3 proteasomal degradation. The authors further showed that VDAC2/3 are positive regulators of ferroptosis, and thus FOXM1-Nedd4-VDAC2/3 form a negative feedback loop to regulate ferroptosis. Relevant to cancer, they showed Nedd4 inhibition can sensitize the anti-melanoma effect of erastin in both cell culture and xenograft experiments. Overall, the results are convincing and of high quality, the presentation is clear and logic, and the work is of general interest to related fields. This reviewer has the following comments: Major points: 1. Is the FOXM1-Nedd4-VDAC2/3 loop specific to erastin/cystine starvation or a general response to oxidative stress? This is an important question to address, since if the mechanism is validated to be a general one (testing a few additional stresses by western blot will do), its significance will certainly be much higher. 2. It is well known in the ferroptosis field that erastin is not effective in vivo due to issues including its rather poor solubility. This reviewer is curious how the authors could make it work (Fig. 7C) in this study when delivered peritoneally. The authors need to describe very clearly their experimental condition, such as how they reconstituted erastin for the injection, etc. 3. A thorough discussion of the potential link between the role of VDACs in mitochondria and their role in ferroptosis will be helpful, especially considering the recent publication in Molecular Cell "Role of mitochondria in ferroptosis". A convenient analysis monitoring the effect of VDAC2/3 knockdown on mitochondrial activity (change of mitochondrial membrane potential by using appropriate mitotrackers; or change of oxygen consumption) will be important here. 4. Is the mechanism they describe here is due to the previous reported erastin-VDAC interaction? A careful discussion and clarification about this will be helpful and informative to the field. Minor points: 1.  Fig 6c, d, e: detailed information about the reporter construction should be included, i.e., how do they get the sequence of FOXM BS1, BS2 (reported before or predicted by themselves)? How did they construct BS1/BS2 mutated reporter? 5. Fig 7: IHC of VDAC2/3, Nedd4, and a ferroptosis/redox marker (e.g., COX2) should be included for the tumor sections 6. For the in vivo xenograft model, the experimental procedure description is not consistent. In the Methods part, the author described that "To generate murine subcutaneous tumors, melanoma cells (5X106 cells per mouse) were injected subcutaneously into the right posterior flanks of 7week-old immunodeficient nude female mice", while in the figure legend, the author described that "C57BL/6 mice were injected subcutaneously with indicated A375 cells (2X106 cells/mouse)". The inconsistency should be corrected. In this manuscript, the authors characterized the molecular mechanism by which ferroptosis activator erastin induces ubiquitination and proteosomal degradation of VDAC2/3 proteins in melanoma cells. The authors demonstrated that erastin induces the expression of Nedd4 E3 ligase, which directly interacts with and ubiquitinates VDAC2/3. In addition, the induction of Nedd4 by erastin was shown to be mediated by FOXM1-dependent and ROS-dependent transcriptional control of Nedd4 expression. This is an interesting study and may have therapeutic implications on targeting ferroptosis in cancer cells. Overall, this study is well executed with sufficient experimental controls. Data are quite convincing and well presented. I only have a few comments, mostly moderate. 1) Suppression of ferroptosis by erastin through degradation of VDAC2/3 was proposed by the authors as a negative feedback mechanism to explain "erastin-induced cellular resistance". However, the melanoma cell lines used in this study appear to be quite sensitive to erastin (Fig.  7), and no significant treatment-induced resistance was presented. Additional melanoma cell lines with lower sensitivities to erastin could be used to test this hypothesis.
2) Does knockdown of Nedd4 affect the protein levels of VDAC2/3 in the absence of erastin ( Figure  3b)?
3) The authors showed that Nedd4 regulated erastin-induced, but not RSL3-induced ferroptosis in melanoma cells (Fig. S4). Does RSL3 induce ROS-dependent activation of FOXM1 and Nedd4 expression in these cells?

RESPONSE TO REVIEWERS
We are genuinely appreciative of the reviewer's constructive and insightful comments, according to which the manuscript has been carefully and rigorously revised. We hope the new version of our manuscript is now appropriately suited for publication in Nature Communications. A detailed response to the Reviewer's critiques and a description of the new experiments (in italic) follow:

Point-by-point Response to Reviewer #1
1. Initial data presented by authors suggest that VDAC2 and VDAC3 need to cooperate in order to induces ferroptosis following erastin stimuli. This is evident in Fig 1e, where overexpression of either alone has limited effect. Conversely, KD of either of the two was shown to be sufficient to affect degree of ferroptosis. This seemingly contradiction needs to be addressed. Further and as important is the notion that most if not all subsequent experiments shown in this manuscript were performed with a single VDAC construct, and not with both, a point that needs to be further addressed experimentally.

Response:
We truly appreciate the reviewer's comments. Both VDAC2 and VDAC3 are expressed in the external membrane of the mitochondrion, and regulate the entry and exit of numerous ions and metabolites between the cytosol and the mitochondrion. The mitochondrion plays a crucial and proactive role in erastin-induced ferroptosis but not in GPX4 inhibition-induced ferroptosis 1 . VDAC2 and VDAC3 may work together to mediate ferroptosis through regulating the homeostasis of different ions and metabolites in the mitochondrion. Knockdown of either VDAC2 or VDAC3 can disrupt the function of the mitochondrion, and suppress erastin-induced ferroptosis. However, overexpression of either VDAC2 or VDAC3 cannot promote erastin-induced ferroptosis, suggesting either VDAC2 or VDAC3 is necessary but not sufficient for keeping the sensitivity of melanoma cells to erastin. This phenomenon is also universal when two or more genes participate in a process together. Knockout of either one can disrupt the pathway and generate a mutant phenotype, while overexpression of either one is insufficient to activate the pathway, and two or more genes must be activated at the same time to produce the overexpression phenotype. Our results are also consistent with the previous studies 2 . As shown in figure 3d,3e and supplementary figure S14,S15, knockdown of VDAC2 or VDAC3 in HT-1080 cells is sufficient to suppress the sensitivity of cells to erastin 2 . Also, overexpression of VDAC3 in BJ-TERT cells yields no change in sensitivity to erastin 2 .
We performed more function assay experiments with VDAC2 and VDAC3 constructs together. As shown in figure S8, we overexpressed VDAC2 and VDAC3 to rescue erastin-induced ferroptosis, which was suppressed by Nedd4 overexpression. Meanwhile, we also knocked down VDAC2 and VDAC3 simultaneously to suppress ferroptosis which enhanced by knockdown of Nedd4. Similarly, when we test whether VDAC2 and VDAC3 are essential for FOXM1 mediated ferroptosis, we simultaneously overexpressed or knocked down VDAC2 and VDAC3 (figure S10). In the main figures, there are also some function assay experiments with VDAC2 and VDAC3 constructs together. For example, when we use the wild type and mutants of VDAC2 and VDAC3 to rescue Nedd4 mediated ferroptosis, we transfected the cells with both VDAC2 and VDAC3 constructs ( figure 5c-e, 5h). Similarly, we also knocked down both VDAC2 and VDAC3 to rescue FOXM1 mediated ferroptosis ( figure 6h-j) 2 2. Data suggesting that Nedd4 is the E3 ligase that limits VDAC2/3 stability (Fig 2/3) need to be substantiated by half-life studies for the endogenous proteins prior and following erastin addition, in the presence and absence of Nedd4.

Response: We truly appreciate the reviewer's comments. We have completed the experiments suggested by the reviewer. The half-life of VDAC2/3 was reduced after erastin treatment, and knockdown of Nedd4 inhibited the degradation of VDAC2/3 either prior or following erastin treatment (figure S3c).
3. Figure 4g-I demonstrate the effect of erastin on ferroptotic cell death in the presence of Nedd4 overexpression. Yet, the effect shown is rather modest and need to be discussed relative to the changes shown in GSH and Fe levels. Would scavengers for Fe and ROS ablate these effects?

Response:
We genuinely appreciate the reviewer's enthusiasm and great comments. In addition to the accumulation of lipid ROS, the intracellular Fe 2+ also increased in ferroptosis 3,4 7 and figure S11). Because nearly half of melanoma patients carry BRAF mutants, we use A375 and G-361 cells in our experiments. Also, the effect of Nedd4 on ferroptosis is statistically significant in both A375 and MeWo cells. We have integrated these discussions into the revised manuscript. Figure 5c-e claims for rescue of cell sensitivity to erastin upon OE of KR mutants of VDAC2/3 -yet the effect is really partial. The text in legend and results and related sections of the ms need to be properly adusted.

4.
Response: Thank the reviewer for this comment. The rescue effect of KR mutants of VDAC2/3 is partial. One possible reason is that overexpressed Nedd4 may target other genes, which also regulate ferroptosis. As shown in figure 5c-e, overexpression of KR mutants of VDAC2/3 rescued more than half of the phenotypes mediated by Nedd4, suggesting VDAC2/3 is the most critical target of Nedd4 in erastin-induced ferroptosis. We genuinely appreciate the reviewer's excellent comments and have modified the related sections in the revised manuscript. Fig 6h-j are rather modest and claims should therefore be adjusted in respective sections in text.

Response:
Thank the reviewer for this comment. We have modified the related sections in the revised manuscript. Fig 7a, 7c should be also carried out using constructs that will impair the levels of VDAC2/3 and FOXM1, in order to substantiate the model proposed by authors. Figure S11a

Point-by-point Response to Reviewer #2
Specific concerns: 1. Is the FOXM1-Nedd4-VDAC2/3 loop specific to erastin/cystine starvation or a general response to oxidative stress? This is an important question to address, since if the mechanism is validated to be a general one (testing a few additional stresses by western blot will do), its significance will certainly be much higher.  Figure S5a and figure below). However, VDAC2/3 does not affect RSL3-induced ferroptosis (figure S5b-e). The specific function of VDAC2/3 on erastin-induced ferroptosis may be caused by its role in mitochondria, which play a pivotal role in cysteine-deprivation induced ferroptosis but not in GPX4-inhibition induced ferroptosis 1 . Although other oxidative stress can activate the FOXM1-Nedd4-VDAC2/3 loop, it mainly regulates cysteine-deprivation induced ferroptosis.

Response
2. It is well known in the ferroptosis field that erastin is not effective in vivo due to issues including its rather poor solubility. This reviewer is curious how the authors could make it work (Fig. 7C)  Erastin was used successfully in other mouse xenograft models in previous works 5,6 . For our in vivo experiments, we dissolved the erastin in 5% DMSO+Corn oil (Sigma C8267). The corn oil is used as a delivery vehicle for fat-soluble compounds 7,8,9,10 . To better dissolve it, we warmed the tube at 37℃ water bath and shook it gently. We have included this information in the methods.

3.
A thorough discussion of the potential link between the role of VDACs in mitochondria and their role in ferroptosis will be helpful, especially considering the recent publication in Molecular Cell "Role of mitochondria in ferroptosis". A convenient analysis monitoring the effect of VDAC2/3 knockdown on mitochondrial activity (change of mitochondrial membrane potential by using appropriate mitotrackers; or change of oxygen consumption) will be important here.

Response:
We thank the reviewer for this insightful comment. We have measured the mitochondrial membrane potential (MMP) using MitoTracker (tetramethylrhodamine ethyl ester, TMRE) in figure 5fh. Knockdown of VDAC2/3 suppressed erastin-induced MMP hyperpolarization. In addition, overexpression of Nedd4 also suppressed erastin-induced MMP hyperpolarization, and VDAC2/3 mutants can rescue the inhibition effect of Nedd4. As mentioned in the Molecular Cell paper, MMP hyperpolarization increases lipid ROS generation, knockdown of VDAC2/3 suppresses MMP hyperpolarization and subsequent intracellular MDA level and ferroptotic cell death ( figure 1c). There is a strong possibility of a potential link between the role of VDACs in mitochondria and their roles in ferroptosis, and we discussed it in the manuscript.