VHL deficiency augments anthracycline sensitivity of clear cell renal cell carcinomas by down-regulating ALDH2

The von Hippel-Lindau (VHL) is deficient in ∼70% of clear-cell renal cell carcinomas (ccRCC), which contributes to the carcinogenesis and drug resistance of ccRCC. Here we show that VHL-deficient ccRCC cells present enhanced cytotoxicity of anthracyclines in a hypoxia-inducible factor-independent manner. By subtractive proteomic analysis coupling with RNAi or overexpression verification, aldehyde dehydrogenase 2 (ALDH2) is found to be transcriptionally regulated by VHL and contributes to enhanced anthracyclines cytotoxicity in ccRCC cells. Furthermore, VHL regulates ALDH2 expression by directly binding the promoter of −130 bp to −160 bp to activate the transcription of hepatocyte nuclear factor 4 alpha (HNF-4α). In addition, a positive correlation is found among the protein expressions of VHL, HNF-4α and ALDH2 in ccRCC samples. These findings will deepen our understanding of VHL function and shed light on precise treatment for ccRCC patients.

R enal cell carcinoma (RCC) is the most common type of kidney cancer, accounting for 2-3% of all adult tumours 1,2 . To date, surgery is the main treatment for this kind of cancer, while RCC is notoriously resistant to conventional chemotherapy, possibly through high expression of some multidrug resistance genes or inactivation of apoptotic pathways 3 . Also, most of these patients suffer from the side effects of chemotherapy. More recently, some new drugs targeting vascular endothelial growth factor receptor, such as sunitinib and sorafenib, have been proven to be beneficial for RCC 4 . Unfortunately, large portion of kidney cancer patients treated with these inhibitors will eventually experience disease progression. Drug treatment of kidney cancer is still unsatisfactory 5 . Therefore, it is urgent to identify the potential therapeutic targets and chemotherapeutic agents for RCC.
Clear-cell RCC (ccRCC), the most frequent and malignant type of RCC, is characterized by early loss of the von Hippel-Lindau (VHL) tumour-suppressor gene in most tumours 6 . The VHL gene encodes two isoforms, that is, a 24-30 kDa isoform with 213-amino-acid and a 19 kDa one with 160-amino-acid 3 . Early functional studies suggested that both isoforms have tumoursuppressor activity in vivo 6 . Functional studies indicated that VHL is an E3 ubiquitin ligase that targets the oxygen-sensitive a subunit of the hypoxia-inducible factor (HIF-a) for proteasomal degradation. When oxygen is available, HIF-a is hydroxylated on two proline residues by the oxygen-dependent HIF-a-specific prolyl hydroxylases. The hydroxylated HIF-a is ubiquitinated by VHL, followed by proteasomal degradation 4,7 . Under hypoxic conditions, HIF-a is stabilized due to the inhibition of hydroxylation and ubiquitination, and it subsequently translocates into the nucleus where it forms heterodimer with constitutively expressed HIF-1b (refs 7,8). The HIF-a/HIF-1b heterodimer binds to hypoxia-responsive elements (HREs) in the promoters and activates the transcription of its targeted genes, which are involved in energy metabolism, angiogenesis, cell proliferation/differentiation and invasion/ metastasis [9][10][11][12][13] . In addition to its role in HIF regulation, VHL is also implicated in a variety of HIF-independent processes, including regulation of the extracellular matrix, microtubule stabilization and maintenance of the primary cilium, control of cell senescence, and modulation of RNA polymerase II subunits [14][15][16][17][18] . However, whether VHL is a potential therapeutic target for ccRCC is currently unknown.
Aldehyde dehydrogenase belongs to a family of oxidizing enzymes that is responsible for the detoxification of aldehydes 19,20 . The mitochondrial aldehyde dehydrogenase 2 (ALDH2) is well known for its critical role in metabolism of acetaldehyde. In particular, ALDH2 plays a key role in oxidizing endogenous aldehydic products, such as 4-hydroxy-2-nonenal (4-HNE) and malondialdehyde, which arise from lipid peroxidation under oxidative stress 19,20 . ALDH2, which is expressed ubiquitously but abundant in liver, kidney, brain and heart 21,22 , has been regarded as a crucial enzyme in protecting the heart from oxidative stress 23 . ALDH2 dysfunction may contribute to a variety of human diseases such as cardiovascular diseases, neurodegenerative diseases, diabetes, stroke and cancer 19,[24][25][26][27] .
Anthracyclines, mainly including daunorubicin, doxorubicin and epirubicin, are still widely used in modern cancer treatments despite the advent of targeted therapy. In general, anthracyclines are applied to first-or second-line standard chemotherapy in treating childhood leukaemia, lymphomas and solid malignancies by intravenous administration to induce cancer-cell growth inhibition [28][29][30] .
Here we report that VHL-deficiency augments ccRCC sensitivity to anthracyclines by downregulation of ALDH2 in an E3 ubiquitin ligase-independent manner.

Results
VHL-deficiency augments cytotoxicity of anthracyclines. RCC4, a VHL-defective ccRCC cell line, was stably transfected with empty vector (RCC4/EV) or VHL (RCC4/VHL). As shown in Fig. 1a, RCC4/EV, but not RCC4/VHL cells, expressed higher levels of HIF-1a and pyruvate dehydrogenase kinase 1 (PDK1, a HIF-1a-responsive gene 9 ) under normoxia, while HIF-1a and PDK1 proteins were increased in RCC4/VHL cells under hypoxia (1% O 2 ). We compared the growth inhibition of RCC4/EV and RCC4/VHL under chemotherapeutic drugs. The results demonstrated that all three anthracycline drugs, but not nonanthracycline chemotherapeutic drugs tested, had higher growth inhibition rates to RCC4/EV than RCC4/VHL cells (Fig. 1b). Also, the cell growth inhibitory ability of doxorubicin at three different concentrations was always higher in RCC4/EV than in RCC4/VHL cells ( Supplementary Fig. 1a). The half maximal inhibitory concentrations (IC50) of all three anthracyclines tested in RCC4/VHL were over 2.9 folds more than those in RCC4/EV ( Supplementary Fig. 1b).
To evaluate the role of VHL on cytotoxicity of doxorubicin in vivo, 786-O/EV and 786-O/VHL or Caki-1/NC and Caki-1/ shVHL cells were subcutaneously implanted into the flanks of NOD/SCID mice. As depicted in Fig. 2, doxorubicin treatment significantly reduced tumour growth of VHL-deficient tumour cells compared with corresponding VHL-expressing tumour cells ( Fig. 2a-d). The same results were demonstrated in immunodeficient nude mice ( Supplementary Fig. 2a-d). Collectively, all these data propose that VHL-deficient ccRCC cells present higher sensitivity to anthracyclines than that of VHL-proficient cells.
HIFs do not contribute to the anthracycline cytotoxicity. Since HIF-a is a major substrate of VHL 4,32 , the potential role of HIFs in the regulation of cytotoxicity of anthracycline was tested. Towards this end, doxorubicin was applied to RCC4 and 786-O cells in the presence or absence of ectopic VHL expression under normoxia or hypoxia. Compared with normoxia, hypoxia, which accumulated HIF-1a and -2a proteins, failed to impact the drug cytotoxicity ( Fig. 3a,b, Supplementary Fig. 3a). In addition, the cell growth inhibition after doxorubicin treatment had no significant difference in the presence or absence of hypoxia mimetic agent CoCl 2 (Fig. 3c), which also accumulated the HIF-1a protein 33 . These results suggest that HIFs might not contribute to the enhanced cytotoxicity of anthracyclines. To consolidate this, we effectively silenced HIF-1a and HIF-1b or HIF-2a by their specific shRNAs in RCC4 or 786-O cells. Although these silences suppressed the proliferation ( Supplementary Fig. 3b,c), they did not impact on the growth inhibition rates of doxorubicin in RCC4 or 786-O cells ( Fig. 3d-f).
Proteomics reveals regulators of anthracycline cytotoxicity. Next, we sought to identify proteins mediating the enhanced cytotoxicity of anthracyclines in VHL-deficient RCC4 and 786-O cells by subtractive proteomics strategy. Figure 4a showed a schematic representation of the stable isotope dimethyl-labelling experiments. Briefly, cell lysates from each condition were digested and the peptides were labelled with light, medium and heavy reagents, respectively, and then were equally mixed and injected to Orbitrap Fusion liquid chromatography (LC)-mass spectrometry (MS/MS). Three independent experiments were performed in the indicated triplex configuration, which provided the direct comparisons between cells with EV and VHL transfection in normoxia, as well as cells with VHL transfection in normoxia and hypoxia.
As a result, we identified 4,282 and 3,585 proteins, respectively, in the pooled cell lysates of RCC4 and 786-O cells (Supplementary Data 1,2). All quantification data were normalized and natural log transformed before further analysis. Differentially expressed proteins were determined by applying paired t-test and corrected by Benjamini-Hochberg algorithm as multiple hypothesis testing correction 34 Data 9,10). To get an overview of these VHL alone-regulated proteins in RCC4, function enrichment and pathways analysis were performed using QIAGEN's Ingenuity Pathway Analysis tools. The bioinformatic analysis indicated complicated biological processes, such as organismal injury and abnormality, cancer, cell growth and proliferation, and multiple pathways were involved (Supplementary Fig. 4a-d).
The comparison of the bioinformatic data revealed that RCC4 and 786-O presented very different functional state (Supplementary Fig. 4a-d and Supplementary Data 11,12). The functional enrichment and pathway analysis suggested that RCC4 cells presented a series of suppressed functional state such as cellular movement, maintenance, cellular assembly and organization, whereas 786-O demonstrated a series of active functional state in the cellular processes mentioned above, which might contribute to the different biological characteristics of these two cell lines. 786-O was derived from a primary clear-cell adenocarcinoma and high tumorigenic, but RCC4 was low tumorigenic 35,36 .
However, these two VHL-deficient cell lines shared common effect on the enhanced cytotoxicity of doxorubicin. To further narrow down the candidate proteins involved in cytotoxicity of anthracyclines, these results of IPA analysis were compared between these two cell lines (Supplementary Data 11,12) and the common function enrichment and pathways were picked up (Supplementary Data 13,14). Totally, 107 function enrichment or pathway including cancer, cell death and survival, organismal injury and abnormalities were shared by RCC4 and 786-O cells (Supplementary Data 13,14). Among the proteins involved in these common function enrichment and pathway, 26 were shared by RCC4 and 786-O (Fig. 4a,b). Notably, five of these 26 VHL alone-regulated proteins, including vesicle-associated membrane protein-associated protein A, 4-hydroxy-2-oxoglutarate aldolase 1, aldehyde dehydrogenase 2 (ALDH2), vimentin and protein phosphatase 1 regulatory subunit 13 like, were found to be regulated consistently in RCC4 and 786-O cells (Fig. 4b). Hence, we silenced or ectopically expressed these five genes followed by doxorubicin treatment in RCC4/VHL (Fig. 4c,d) and 786-O/VHL (Fig. 4e,f). The results demonstrated that ALDH2 but not other four genes regulated the cytotoxicity of doxorubicin in both ccRCC cells.
ALDH2 regulates anthracycline cytotoxicity. ALDH2 was reported to regulate cytotoxicity of doxorubicin in cardiac cells, leukaemia cells and lung cancer cells [37][38][39] . Indeed, the cytotoxicity of doxorubicin was significantly higher in the primary mouse embryo fibroblast (MEF) cells from ALDH2 knockout mice than those from wild-type mice (Fig. 5a). To address the potential role of ALDH2 in cytotoxicity of anthracyclines in ccRCC cells, we silenced ALDH2 expression in RCC4/VHL cells, and found that the reduction of ALDH2 suppressed the proliferation of these cells ( Supplementary Fig. 5a), although the ALDH2 knockout did not affect the proliferation of the primary MEF cells ( Supplementary Fig. 5b). Then, these RCC4/VHL cells were treated with doxorubicin at different concentrations, and showed that cytotoxicity of doxorubicin significantly increased in ALDH2 silencing cells (Fig. 5b). Reciprocally, stable ectopic expression of ALDH2 decreased cytotoxicity of doxorubicin and daunorubicin in RCC4 cells (Fig. 5c). Because cytotoxicity of doxorubicin is contributed partially by apoptosis 30 , the apoptosis of these cells was examined. The doxorubicin-treated RCC4/EV cells had a higher apoptosis rate than RCC4/VHL ( Supplementary Fig. 5c), and cell apoptosis rate was higher in the ALDH2 À / À MEF cells ( Supplementary Fig. 5d) and RCC4/VHL with ALDH2 silencing ( Supplementary Fig. 5e).
To explore whether ALDH2 enzyme activity affected doxorubicin-treated cells, we used ALDH2 activator alda-1 (ref. 23) and inhibitor daidzin 40 to pre-treat RCC4/VHL cells. The results showed that alda-1 reduced, while daidzin increased, the cytotoxicity of doxorubicin in RCC4/VHL cells (Fig. 5d,e), suggesting that doxorubicin cytotoxicity might be negatively correlated with enzymatic activity of ALDH2.
It has been reported that doxorubicin could increase the level of intracellular ROS and 4-HNE (ref. 29). As the main metabolism enzyme of 4-HNE, ALDH2 may affect cell deaths by metabolism of 4-HNE. In agreement, ectopic VHL expression significantly reduced 4-HNE in RCC4 cells, while ALDH2 silence increased 4-HNE in the RCC4/VHL cells, under the treatment of doxorubicin but not vincristine ( Fig. 5f,g). Furthermore, doxorubicin combined with 4-HNE led to even higher growth inhibition than doxorubicin treatment alone in RCC4/VHL cells (Fig. 5h). All these data propose that a higher intracellular 4-HNE levels endow cells with higher cytotoxicity of doxorubicin.
VHL regulates the transcription of ALDH2. To validate the proteomic results that VHL regulates ALDH2 in ccRCC, VHL was overexpressed or knocked down in ccRCC cells. As shown in Fig. 6a, the VHL overexpression in RCC4 augmented the expression of ALDH2 in mRNA and protein levels. This is also true in 786-O cells (Fig. 6b). Visa versa, both mRNA and protein of ALDH2 were significantly downregulated by silencing VHL expression in RCC4/VHL and Caki-1 (Fig. 6c,d). Consistent with the above findings that HIFs do not contribute to the enhanced cytotoxicity of anthracyclines in VHL-deficient ccRCC cells, the ALDH2 protein levels were also not affected by HIF-1a and HIF-2a silencing in RCC4 cells and 786-O cells (Fig. 6e,f). These results suggest that VHL regulates the transcription of ALDH2 in ccRCC cells in HIF-independent manner. As documented 4,41 , the ubiquitin E3 ligase activity of VHL is dependent upon its a-domain and b-domain, among which the b-domain directly binds putative substrates such as HIF-1a, while the a-domain directly contacts elongin C in pVHL-elongin C-elongin B complex. We constructed an a-domain mutant C162F and a b-domain mutant Y98H of VHL, which lost its E3 ligase activity 41 , to examine the expression of ALDH2. The results showed that, like wild-type VHL, these two VHL mutants still upregulated ALDH2 expression, but these mutants failed to induce HIF-1a degradation ( Supplementary Fig. 6a). Of great interest, we sequenced VHL genes from 44 cases of ccRCC cancer tissues, and found that 13 cases of them carried VHL mutations, among which 8 cases were missense mutations with no E3 ligase activities ( Supplementary Fig. 6b). In spite of this, tissues with wild type and mutated VHL had the similar levels of ALDH2    3  VAPA  HOGA1  ALDH2  VIME  IASPP  ACTN1  4F2  RTN4  CD166  RABSC  ATLA3  XPO2  BASI  NSDHL  TMX1  SYIC  VAPB  AAAT  MAGD2  HUWE1  PRAF3  NBAS  GNAI3  TOP2B  SEC63  TLDC1   protein, as evaluated by immunohistochemistry ( Supplementary  Fig. 6c-e). Cumulatively, our results suggest that VHL regulates the transcription of ALDH2 in an E3 ubiquitin ligaseindependent manner.
HNF-4a mediates regulation of ALDH2 by VHL. It was reported that hepatocyte nuclear factor 4a (HNF-4a) is a transcription factor of ALDH2 (refs 42,43). Although HNF-4a was not detected in the proteomic analysis for its low abundance as a transcription factor, we put HNF-4a as well as all the proteins regulated by VHL but not hypoxia into the protein interaction analysis with IPA. Indeed, HNF-4a was found to be a central node in the network in RCC4 (Supplementary Fig. 7a) and 786-O ( Supplementary Fig. 7b). In addition, several HNF-4a target genes were found in these VHL alone-regulated genes (Supplementary Data 15). Thus, we extrapolated that HNF-4a might be involved in the regulation of ALDH2 by VHL. To confirm this, we showed that HNF-4a mRNA and protein levels were significantly upregulated in RCC4 and 786-O cells with ectopic VHL expression (Fig. 7a,b). Accordingly, silencing of VHL in RCC4/VHL cells suppressed HNF-4a expression on mRNA and protein levels (Fig. 7c). On the other hand, silencing of HNF-4a decreased ALDH2 expression in RCC4/VHL (Fig. 7d). Our results suggest that VHL could upregulate the mRNA and protein levels of HNF-4a to activate ALDH2 transcription. Furthermore, we found that anti-VHL antibody but not normal mouse IgG could precipitate the putative promoter of HNF-4a (Fig. 7e). Hence, VHL-expressing plasmids and the 1,500 bp of the HNF-4a promoter-driven luciferase reporter were cotransfected into 293T cells, and the results demonstrated that VHL transfection significantly increased the luciferase activity (Fig. 7f). In addition, we constructed 550 bp trunk of the HNF-4a  promoter, and five different mutants of this trunk 39,44 . The second mutant, from À 130 bp to À 160 bp, significantly decreased the luciferase activity in response to VHL transfection (Fig. 7g), which indicate that this site is essential for VHL regulating the transcription of HNF-4a.
VHL is positively correlated with HNF-4a and ALDH2. We further explored whether the relationships among VHL, HNF-4a and ALDH2 existed in clinical patient samples. As for this, 114 cases of RCC samples were collected to detect the expression of VHL, HNF-4a and ALDH2 in cancer and adjacent tissues by immunohistochemical assay. Significantly lower expression of VHL, HNF-4a and ALDH2 were found in cancer tissues than their corresponding adjacent tissues (Fig. 8a,b). On the other hand, as shown in Fig. 8c, high-VHL expression was often associated with high HNF-4a and ALDH2 expression, and vice versa (Po0.01). Taken together, these results suggest that VHL protein expression has a positive correlation with HNF-4a and ALDH2 in ccRCC. Furthermore, 72 cases of ccRCC patients with cancer and corresponding adjacent tissues in database GEO were used to verify the relationship among VHL, HNF-4a and ALDH2. As shown in Supplementary Fig. 8a-c, lower mRNA expression of VHL, HNF-4a and ALDH2 were found in cancer tissues compared to their corresponding adjacent tissues, and there was a positive correlation between VHL and HNF-4a mRNA (Supplementary Fig. 8d) and between HNF-4a and ALDH2 mRNA ( Supplementary Fig. 8e). But we could not find a correlation between VHL and ALDH2 mRNA ( Supplementary  Fig. 8f), which might be because VHL functional mutations interfere with protein stability or the alteration of the VHL start codon could result in loss of VHL protein expression 45,46 .
HNF-4a mediates the cytotoxicity of anthracyclines. To further confirm the role of HNF-4a in the cytotoxicity of anthracyclines in ccRCC, we silenced the expression of HNF-4a in Caki-1 cells by shRNA specifically against HNF-4a (Fig. 9a). The results demonstrated that the silence inhibited the cell proliferation (Fig. 9b), and significantly increased the cytotoxicity of doxorubicin (Fig. 9c). We also subcutaneously implanted Caki-1/NC and Caki-1/shHNF-4a cells into the flanks of nude mice. Tumour growth of Caki-1/shHNF-4a cells were significantly reduced compared with that of Caki-1/NC cells after doxorubicin treatment (Fig. 9d,e). Collectively, HNF-4a mediates the cytotoxicity of anthracyclines in ccRCC.
Regulation of anthracycline toxicity by VHL is conserved. VHL is a conservative key regulator under hypoxia condition. We further investigated its conservation on regulation of ALDH2 and chemotherapeutic toxicity. C.elegans is good system to verify the conservation of gene function as well as drug action 47 . The double-mutant strain of C.elegans acs-20;acs-22 increases penetration of several drugs 47 but has no effect on worm survivability. Thus this mutant worms were used to test the anthracycline toxicity and the role of these genes. As shown in Supplementary Fig. 9a, this mutant had no effect on the survival of the worms when no drugs were applied. Of note, this mutant strain of C.elegans had a higher toxicity of doxorubicin than its wild-type strain (Supplementary Fig. 9b). In addition, the RNAi of vhl-1 (corresponding to human VHL) and alh-1 (corresponding to human ALDH2) had no effect on the survival to this mutant worms ( Supplementary Fig. 9c,d). Consistent to our findings in ccRCC cell lines, the silence of either vhl-1 or alh-1 significantly increased the toxicity of doxorubicin in C.elegans compared with treatment of DMSO with the same osmolality ( Fig. 10a-d). Furthermore, vhl-1 silencing also inhibited alh-1 mRNA expression (Fig. 10e). Intriguingly, loss of function vhl-1 mutant could downregulate the mRNA of nhr-69 (HNF-4a in mammal) and alh-1 in C.elegans (Fig. 10f). Collectively, these results indicate the evolutionary conservation in the regulation of ALDH2 and HNF-4a expression and the anthracyclines toxicity by VHL.

Discussion
Here we found that VHL-deficiency made RCC cells more sensitive to anthracyclines. Previously, VHL overexpression has been found to synergize with doxorubicin to suppress hepatocellular carcinoma in mice 48 , suggesting different effects of VHL on anthracycline response in different cellular context between liver and kidney. Considering that HIF-a is the major substrate of VHL, we investigated whether HIFs are involved in the effect of VHL on anthracycline sensitivity in ccRCC cells. It has been reported that suppression of HIF-2a restores P53 activity and promotes the apoptosis induced by anthracyclines 49 , and it has also been reported that anthracyclines inhibited the binding of HIF heterodimer to the consensus HRE and thus impaires the transcriptional response of HIF 50 . However, our results show that the VHL-deficiency increased sensitivity of ccRCC cells to anthracyclines is independent upon HIF-1 or HIF-2. We demonstrate that ALDH2 is a target gene of VHL and participates in cytotoxicity of anthracyclines in ccRCC. The regulation of cytotoxicity of anthracyclines in cardiac cells by ALDH2 has been reported by several studies recently. DOX-induced myocardial cellular toxicity has been found to be further aggravated in ALDH2 knockout mice and ameliorated in ALDH2 transgenic mice 37 . Consistently, the cardiotoxicity is also found to be aggravated when the DOX plus ALDH2 inhibitor daidzin, while DOX plus ALDH2 agonist Alda-1 would partial or complete alleviate the cardiotoxity 39 . Notably, Moreb et al. have found that overexpression of ALDH2 in leukaemia cell K562 and lung cancer-cell H1299 could promote their proliferation and resistant to doxorubicin 38 . These findings suggested the sensitivity to anthracyclines is regulated by ALDH2, which is consistent to our finding in RCC cells. Considering the fact that an inactive mutant form of ALDH2 is found in 40% of East Asian populations 19 , it is of great interest to explore whether these findings can be extended to other cancers in the future.
Our further investigation show that the effect of VHL on cytotoxic sensitivity to anthracyclines in ccRCC cells is not dependent on its E3 ligase activity. VHL is located in both cytoplasm and nucleus, the nuclear form of VHL exhibits the anti-tumour properties 51 . There are a few reports that VHL plays a role in other ways except as an E3 ubiquitin ligase. It has been reported that VHL mediates the transcriptional suppression of the c-Myc gene by binding to the c-Myc promoter 52 . More intriguingly, here we also show that VHL could activate HNF-4a transcription through binding to the promoter of HNF-4a and revealed a novel nuclear function of VHL.
As a member of HNF-4, HNF-4a belongs to subfamily NR2 of the nuclear receptor superfamily 53 . HNF-4a is a key transcription factor for hepatocyte differentiation, proliferation control, cellular homeostasis, epithelial morphogenesis, glucose metabolism and insulin secretion [54][55][56][57][58] . Reduced expression of HNF-4a is found in hepatocarcinogenesis and confers advantages to tumour cells [59][60][61] . Thus, HNF-4a could be considered as a potential tumour suppressor in liver cells. The malfunction of HNF-4a has also been described in human RCCs 62 . Our results indicate that VHL could regulate HNF-4a, which partially explains the low expression of HNF-4a in RCC.
Despite the limitations in detecting low-abundance proteins, proteomics could reveal the most significant changes of highabundance proteins through an unbiased strategy, which could demonstrate some of the most important aspects of the drug effects. In this sense, proteomics could accelerate the drug development via uncovering the action mechanism of novel reagents 63 . On the other hand, proteomics, as drug targets ARTICLE discovery means, always encounter a bottleneck to conceive the real targets. Functional analysis of potential drug target proteins will be great helpful. As a model for drug target verification, Caenorhabditis elegans has its advantage to verify the functional conservation 64 . However, the surface barrier is essential for maintaining the internal environment to C.elegans and prevents drugs penetrating into the bodies of C.elegans 65 . acs-20;acs-22 mutant C.elegans has defective skin barriers, and thus drugs can easily penetrate into the bodies of C.elegans, which greatly reduces the lethal dose of drugs 47 . In our work, we have verified the drug action and candidate genes affecting cytotoxicity of anthracyclines by acs-20;acs-22 mutant and demonstrated the value of this mutant worm.
Conclusively, we found that VHL-deficiency augments anthracycline chemotherapy by downregulation of ALDH2 in ccRCCs. This work might provide clues for understanding the novel function of VHL and precise treatment for ccRCC patients. sectioned from each clinical sample. Then, each slide was deparaffinized in 60°C, followed by treatment with xylene and graded alcohol. After the antigen retrieval and being blocked with 5% bovine serum albumin, tissue slides were immunohistochemically stained by antibodies against VHL (1:50, Abcam, ab140989), HNF-4a (1:50, Abcam, ab181604) and ALDH2 (1:200, Abgent, AM1831a), respectively, then visualized by standard avidin-biotinylated peroxidase complex method. Then, hematoxylin was used for counterstaining and morphologic images were observed with Olympus BX51 microscope. All of staining was assessed by pathologists blinded to the origination of the samples and subject outcome. Each specimen was assigned a score according to the intensity of the staining (no staining ¼ 0; weak staining ¼ 1, moderate staining ¼ 2, strong staining ¼ 3) and the extent of stained cells (0% ¼ 0, 1-24% ¼ 1, 25-49% ¼ 2, 50-74% ¼ 3, 75-100% ¼ 4). The final immunoreactive score was determined by multiplying the intensity score with the extent of score of stained cells, ranging from 0 (the minimum score) to 12 (the maximum score). Cytotoxicity assay and cell proliferation. For cytotoxicity assays, 5,000 ccRCC cells were plated in 96-well plates in 100 ml of media, respectively. The following day, different concentrations of drugs was added in media for different time points at 37°C. Then each well was pulsed by addition of 10 ml of CCK-8 assay (WST-8; Cell Counting Kit-8 from Dojindo, Kumamoto, Japan) and incubated for 3 h. Absorbance readings at a wavelength of 450 nm were taken on Synergy H4 Hybrid Microplate Reader. The growth inhibition rate is calculated by (Ac-Ae)/ (Ac-Ab) Â 100%. Ac, Ae and Ab mean absorbance value in control, experiment (Ae) and blank. Cell proliferation was also evaluated by the CCK-8 assay. Briefly, 1,000 cells were plated in 96-well plates in 100 ml of media for different days, then each well was pulsed by addition of 10 ml of WST8 and incubated for 3 h. Absorbance readings at a wavelength of 450 nm were taken on Synergy H4 Hybrid Microplate Reader.
Luciferase assay. The indicated sequences in promoter of HNF-4a were obtained from National Center for Biotechnology Information, amplified by PCR from genomic DNA and subcloned into pGL3-Basic (Promega) to construct luciferase reporter plasmids. For the luciferase assay, 293T cells were seeded in a 12-well plate (Becton Dickinson, Franklin lakes, NJ, USA), and co-transfected with VHL expression vector, luciferase reporter plasmids driven by promoter fragments of HNF-4a and pRLSV40-Renilla. After 36 h transfection, cells were lysed and analysed by the Dual-Luciferase Assay system according to the manufacturer's instructions (Promega). The following oligonucleotides were used for HNF-4a promoter mutagenesis: the first mutant primers: -104 GGGTCGATGGTGGAT-CCGTCCCCCGCCGGTGGATAGGCTG -143; the second mutant primers: _-160 ATCCCTGCAGCCATGGCCAGCC TATCCACCG -130; the third mutant primers: -298 GGTGAGTCGACGCACAAAT GAGTGCCCGTGA -268; the forth mutant primers:-423 GCATTGAGGGTAGAA TCTAGAGATTTGGGAAGTTA-TTG -386; the fifth mutant primers: -419 AATGCTTTTGCAAAGCTTAGGCT-GCCCCATGGCCC -453 In vivo studies. Animal care and experiments were performed in strict accordance with the 'Guide for the Care and Use of Laboratory Animals' and the 'Principles for the Utilization and Care of Vertebrate Animals' and were approved by the Experimental Animal Ethical Committee at Fudan University. Male NOD/SCID or nude mice, 4 to 6 weeks old, were obtained from Shanghai Research Center for Model Organisms. 786-O/EV, 786-O/VHL, Caki-1/NC, Caki-1/sh-VHL or Caki-1/ sh-HNF-4a cells were resuspended in matrigel to a final volume of 200 ml and injected subcutaneously. Mice were randomized into vehicle control group or treatment group. When tumours reached an approximate average volume of 50 mm 3 , the mice were treated with 4 mg per kg of doxorubicin. Tumour growth was blinded to measure every 2 days after drug treatment was started. The volume of the tumour was calculated using the equation length Â width Â width/2.
Apoptosis assay. Apoptotic cells in the populations were measured with a FACScan flow cytometer (Becton-Dickinson) by the AnnexinV Fluos apoptosis detection kit (Roche Molecular Biochemicals, Mannheim, Germany). Cells were stained with Annexin-V-FITC for exposure of phosphatidylserine on the cell surface as an indicator of apoptosis, following the manufacturer's instruction (BD Biosciences). Data acquisition and analysis were performed using a BD Biosciences FASCalibur flow cytometer with CellQuest software. Positively stained by annexin-V-FITC only (early apoptosis) and propidiumiodide (late apoptosis) were quantitated, and both subpopulations were considered as overall apoptotic cells.
Trypsin digestion and stable isotope dimethyl labelling. Cells were dissolved in lysis buffer containing 7 M urea, 2 M thiourea with protease inhibitor cocktail (Roche), followed by centrifugation (15,000 g, 30 min) at 4°C. The supernatant was quantified using Brandford protein assay, and the precision was validated with SDS-PAGE electrophoresis.
In total, 100 mg of proteins were sequentially reduced with 10 mM DTT and alkylated with 12 mM iodoacetamide. Add trichloroacetic acid to 12% and let it sit for 2 h at room temperature to purify the protein samples. Spin for 5 min in a microfuge (15,000 g) and carefully discharge the supernatant and retain the pellet. Wash the pellet twice with one volume of cold acetone. Vortex and repellet the samples for 5 min at full speed between washes. Samples were reconstituted in 100 ml of 100 mM TEAB buffer. After 1 mg trypsin enzyme was added and digested for 4 h, 2 mg trypsin enzyme was added and incubated overnight. An aliquot of 100 ml of 100 mM TEAB, 8 ml of 4% (vol/vol) formaldehyde (CH 2 O, CD 2 O or 13CD 2 O), 9 ml of 0.6 M cyanoborohydride (NaBH 3 CN or NaBD 3 CN) were added and incubated for 1 h at room temperature. The reaction was quenched with 16 ml of 1% ammonia. Finally, 8 ml formic acid was added to terminate the reaction. The three differentially labelled samples were equally mixed and desalted using SPE-PAK C18(Waters) before LC-MS analysis.
Two-dimensional LC-MS/MS system. The peptide mixture was fractionated by high pH separation using a UPLC system (Waters Corporation, Milford, MA, USA). Fifteen fractions were collected, each fraction was dried in a vacuum concentrator. The successive low pH separation was achieved with a linear gradient starting from 5% ACN to 40% ACN in 40 min on an EASY-nLC-1200 system combined online with Orbitrap Fusion mass spectrometry (Thermo, San Jose, CA, USA). The mass spectrometer was operated in the data dependent mode collecting full MS scan from 350 to 1,200 m/z at 120 K resolution (after accumulation to a target value of 500,000). Ions above the intensity threshold of 2e5 were selected for tandem MS scan at 15 K resolution. Higher energy collisional dissociation (HCD) fragmentation was performed with normalized collision energy of 30.
Database search and data analysis. The raw data was search against Uniprot human database (2016-3 with 20,211 entries) using MaxQuant (version.1.5.3.7). The light, medium, heavy dimethyl label on both peptide N-terminus and lysine were selected as quantification tags. Peptides with two missed cleavages, fixed carbamidomethylated cysteines and variable acetylated protein N-terminus, and oxidized methionines were set as group-specific parameters. Initial MS and MS/MS tolerances were set at 20 p.p.m. Protein and peptide FDRs were 1%. Quantification results of three biological replications were normalized and the variations were calculated 66 . In each LC-MS run, Maxquant normalize the protein ratios so that the median of their logarithms is zero, which corrects for unequal protein loading, assuming that the majority of proteins show no differential regulation.
Differentially expressed proteins were determined by applying paired t-Test with unadjusted significance level Po0.05 and corrected by Benjamini-Hochberg algorithm 34 . Multiple test correction was performed by using Scaffold 4 software (version4.7.2, Proteome Software Inc., Portland, OR, USA).
The networks functional analyses were generated through the use of QIAGEN's Ingenuity Pathway Analysis (IPA, QIAGEN Redwood City).

Measurement of 4-HNE concentrations.
The concentrations of 4-HNE were measured using OxiSelect HNE-His Adduct ELISA Kit (Cell Biolabs) according to the manufacturer's instructions.
C.elegans experiments. C. elegans was maintained on nematode growth medium agar plates seeded with OP50 Escheria coli strains at 20°C. The wild-type strain Bristol N2 and the vhl-1-deficient mutant strain CB5602 was obtained from the Caenorhabditis Genetics Center. The double-mutant acs-20(tm3232);acs-22(tm3236) C.elegans were obtained from National BioResource Project for the Nematode in Japan. RNAi of candidate genes in C.elegans was carried out using standard bacterial feeding methods 67 . For RNAi feeding assays, Synchronized L4 worms were placed on each plate seeded with RNAi bacteria and incubated at 20°C until adulthood, then synchronized for off-spring. The C.elegans at L4 stage of the second generation were put into 96-well plates (50-100 per well) with M9 buffer. After treatment with doxorubicin or DMSO for 1-2 days, C.elegans deaths were observed with stereo microscope. We used the tip to slightly touch the worms when these worms stop moving, to make sure that these worms are really dead or just stop movements even if they are healthy.
Statistical analysis. All the statistical analyses were performed by the statistical package for social science (SPSS) (v. 13) (SPSS Institute). The Pearson's w 2 test was used to evaluate the correlation among the protein expression of VHL, HNF-4a and ALDH2 in immunochemical assay. Unless described otherwise the P values for comparison between line-linked groups were obtained by Student's two-sided t-test. The correlation of relative mRNA levels of VHL, HNF-4a and ALDH2 and its P values was analysed using linear correlation and regression. Multigroup comparisons of the means were carried out by one-way analysis of variance test followed by Bonferroni correction for post hoc test. Po0.05 was considered to be statistically significant.
Data availability. The authors declare that all the data supporting the findings of this study are available within the article and its Supplementary Information files and from the corresponding authors on reasonable request.