Obesity is usually associated with an increased risk of nonalcoholic fatty liver disease that is characterized by accumulation of excessive triglyceride (TG) in hepatocytes. However, the factors involved in the obesity-induced hepatosteatosis are poorly defined. Here, we report that SRY-box containing gene 4 (Sox4), a transcription factor that regulates cell proliferation and differentiation, plays an important role in hepatic TG metabolism. Sox4 expression levels are markedly upregulated in livers of obese rodents and humans. Adenovirus-medicated overexpression of Sox4 in the livers of lean mice promotes liver steatosis, whereas liver-specific knockdown of Sox4 ameliorates TG accumulation and improves insulin resistance in obese mice. At the molecular level, we show that Sox4 could directly control the transcription of SREBP-1c gene through binding to its proximal promoter region. Thus, we have identified Sox4 as an important component of hepatic TG metabolism.
Nonalcoholic fatty liver disease (NAFLD), characterized by an excessive triglyceride (TG) accumulation in hepatocytes, is now the most common liver disorder (1–3). NAFLD has become a grave public health problem because of its high prevalence and association with type 2 diabetes and cardiovascular diseases (1–3).
NAFLD is particularly prevalent in the obese population. Liver steatosis develops as a result of an imbalance of liver TG metabolism with increased fatty acid synthesis and reduced fatty acid oxidation and VLDL secretion (4). It has been reported that de novo lipogenesis is substantially increased in patients with NAFLD (4,5), representing a key feature of hepatosteatosis. At the molecular level, de novo lipogenesis is mainly regulated by transcription factors, SREBP-1c and carbohydrate response element–binding protein (ChREBP), which are activated by insulin and glucose, respectively (6,7). Indeed, upregulated hepatic SREBP-1c mRNA expression has been observed in obese mice and humans (8,9), although its molecular determinants remain largely unknown.
To identify the critical genes that are dysregulated in obesity and might contribute to the development of obesity-associated NAFLD, we previously carried out gene expression profiling using Affymetrix arrays on RNAs isolated from livers of mice fed a high-fat diet (HFD) and mice fed a normal diet (ND) (10–12). According to the biological functions, those aberrantly expressed genes were classified into several categories, such as transcription factors (10), cytokines (11) and ubiquitin E3 ligases (12). In the current study, we focus on the roles of the Sox4 gene, which belongs to a member of a highly conserved transcription factor sex-determining region Y (SRY)-box (Sox) family (13). Previous studies have shown that Sox4 plays an important role in many biological processes, including embryogenesis, neurodevelopment, and cell proliferation, differentiation, and apoptosis (14–16). For example, Sox4-deficent mice die at embryonic day 14 owing to heart malformation, suggesting its critical role in embryonic development (17). Besides, increasing reports indicate that Sox4 is implicated in tumor cell proliferation, metastasis, and epithelial-mesenchymal transition (18,19). Sox4 is overexpressed in many types of human malignancies, including hepatocellular carcinoma (HCC), and is correlated with increased cancer cell survival, tumor progression, and metastasis (20,21). Moreover, recent studies found that Sox4 has a role in insulin secretion in adult β-cells and that mutations in the Sox4 gene are associated with an increased risk of developing type 2 diabetes (22–24). Mechanistic studies further showed that Sox4 allows β-cell proliferation and replication through direct repression of p21CIP1 (25). In addition, expression of Sox4 was also shown to participate in skeletal muscle differentiation (26). However, whether Sox4 plays a role in other metabolic organs, such as the liver, remains poorly understood. Besides, considering that Sox4 is upregulated in HCC tissues and that NAFLD is one of the most common risk factors for HCC, we speculate that Sox4 might have a role in NAFLD.
Research Design and Methods
Animal Experiments and Human Liver Tissues
Male C57BL/6 mice aged 8–12 weeks were purchased from the Shanghai Laboratory Animal Company (Shanghai, China). ob/ob and db/db mice were obtained from Nanjing Biomedical Research Institute of Nanjing University (Nanjing, China). All mice were housed at 21 ± 1°C with a humidity of 55 ± 10% and a 12-h light/dark cycle. The HFD contained 60% kcal from fat, 20% kcal from carbohydrate, and 20% kcal from protein. The ND contained 10% kcal from fat, 70% kcal from carbohydrate, and 20% kcal from protein. The animal protocol was reviewed and approved by the Animal Care Committees of Zhongshan Hospital, Fudan University. For analysis of hepatic Sox4 expression and TG content in humans, the liver tissues were collected in subjects who donated their partial livers for liver transplantation as previously described (10,11). The human study was approved by the Human Research Ethics Committee of Zhongshan Hospital. Written informed consent was obtained from each subject.
For generation of liver-specific overexpression of Sox4, C57BL/6 mice were injected via tail vein with adenoviruses expressing the Sox4 gene or GFP as a negative control. With the BLOCK-IT Adenoviral RNAi Expression System (Invitrogen, Shanghai, China), Sox4 expression in obese mice was silenced via generation of two short hairpin (sh)RNA knockdown adenoviruses that target Sox4 or express a scrambled negative control shRNA. The two shRNAs designed to knockdown Sox4 had the following target sequences: Ad-S1, 5′-GCGACAAGATTCCGTTCATCC-3′, and Ad-S2, 5′-GGTACAACAGACCAACAAC GC-3′. The negative control virus, Ad-NC, had a core-scrambled sequence of 5′-GTTCTCCGAACGTGTCACGTTT-3′. All viruses were purified by the cesium chloride method and dialyzed in PBS containing 10% glycerol prior to animal injection.
Hepatic and Cellular Triglyceride Measurement
Liver tissues (weighed ∼100 mg) were harvested and homogenized in chloroform:methanol (2:1 v/v) using a polytron tissue grinder. The extracts were dried under nitrogen flow and resuspended in isopropanol. For the in vitro model of cellular steatosis, Hep1-6 cells were infected with adenovirus containing (Ad-)Sox4 or Ad-GFP. After 48 h, cells were harvested and TG concentrations were measured using commercial kits (BioVision, Milpitas, CA) according to the manufacturer’s instructions.
Glucose and Insulin Tolerance Tests
Glucose tolerance tests were performed by intraperitoneal injection of d-glucose (Sigma, St. Louis, MO) at a dose of 2.0 mg/g body wt after a 16-h fast. For insulin tolerance tests, mice were injected with regular human insulin (Eli Lilly, Indianapolis, IN) at a dose of 0.75 units/kg body wt after a 6-h fast. Blood glucose was determined using a portable blood glucose meter (LifeScan; Johnson & Johnson, New Brunswick, NJ).
Luciferase Reporter Assays
For the luciferase reporter assays, Hep1-6 cells were transfected with SREBP-1c reporter vectors and Sox4 expression plasmids using Lipofectamine 2000 (Invitrogen). Cells were harvested 36 h after transfection, and luciferase activity was measured using the Dual Luciferase Reporter Assay System (Promega).
Chromatin Immunoprecipitation Assays
A chromatin immunoprecipitation (ChIP) assay kit was used (Upstate, Billerica, MA). In brief, Hep1-6 cells or homogenized mouse liver tissues were fixed with formaldehyde. DNA was sheared to fragments at 200–1,000 bp using sonications. Chromatin was incubated and precipitated with antibodies against SOX4 or IgG.
RNA Extraction and Quantitative Real-time PCR
Total RNA was isolated from hepatic tissues or cell lysates using the standard TRIzol method according to the manufacturer’s instructions (Invitrogen). First-strand cDNA synthesis was performed for each RNA sample using the Reverse Transcription System (Promega). Oligo dT was used to prime cDNA synthesis. In order to quantify the transcripts of the genes of interest, quantitative real-time PCR was performed using an SYBR Green Premix Ex Taq (Takara Biotechnology Co., Ōtsu, Japan) on LightCycler 480 (Roche, Basel, Switzerland). Relative quantitation analysis of gene expression data was performed according to the 2−△△Ct method. The results of relative expression were normalized to 36B4 mRNA levels. The primer sequences used are available upon request.
Hepatic tissues and cells were lysed in radioimmunoprecipitation buffer containing 50 mmol/L Tris-HCl, 150 mmol/L NaCl, 5 mmol/L MgCl2, 2 mmol/L EDTA, 1 mmol/L NaF, 1% NP40, and 0.1% SDS. The protein concentrations were quantified using a BCA Protein Assay kit (Pierce). Proteins were equally loaded on 10% SDS-polyacrylamide gels, transferred to polyvinylidene difluoride membranes by electrophoresis, incubated with primary and secondary antibodies, and finally visualized by a chemiluminescence detection kit (Millipore). The following primary antibodies were purchased: Sox4 (ab86809; Abcam); SREBP-1 (SC-13551; Santa Cruz), phosphorylated AKT (4051; Cell Signaling), total AKT (9272; Cell Signaling), Fasn (3180; Cell Signaling), Acc1 (3662; Cell Signaling), and β-actin (ab8227; Abcam).
Values are shown as mean ± SEM. Statistical differences were determined by the Student t test. Statistical significance is displayed as P < 0.05, P < 0.01, or P < 0.001.
Upregulation of Sox4 in Obesity
Our Affymetrix arrays showed an increased expression of lipogenic genes associated with obesity-induced NAFLD, including SREBP-1c, Fasn, and Scd-1 (10,11). Besides, this screen additionally revealed a pronounced overexpression of Sox4 in the livers of HFD mice compared with controls (Fig. 1A). In contrast, other members of the Sox family remained unchanged between the two groups of mice (Fig. 1A). Increased Sox4 expression in HFD mice was further confirmed by quantitative real-time PCR and Western blot analysis (Fig. 1B and C). Sox4 mRNA and protein expression was also increased in the livers of leptin-deficient ob/ob mice, a genetic model of severe obesity and NAFLD, compared with expression in livers of lean control mice (Fig. 1D and E). The increased expression levels of lipogenic genes in HFD and ob/ob mice are shown in the Supplementary Fig. 1A–C. Besides, we found that Sox4 expression was gradually increased by age in leptin receptor–deficient db/db mice (Supplementary Fig. 1D and E). Importantly, we found that mRNA levels of Sox4 are correlated with the body weight and hepatic TG contents at different ages in db/db and lean mice (Supplementary Fig. 1F and G). Moreover, an increase of hepatic Sox4 mRNA was also confirmed in mice with HFD-induced obesity (Fig. 1F). Finally, C57BL/6 mice were fed with an ND or HFD for 3 days. Consistent with previous reports (27,28), hepatic TG contents and expression levels of lipogenic genes were markedly increased in mice fed an HFD for 3 days (Supplementary Fig. 2A and B), while proinflammatory cytokines were unaffected (Supplementary Fig. 2C). As a result, expression levels of Sox4 were also upregulated (Supplementary Fig. 2D), suggesting that the function of Sox4 in hepatic lipid metabolism might be independent of inflammation.
In agreement, mRNA levels of Sox4 were significantly increased in NAFLD patients and correlated with liver TG contents (Fig. 1G and H). Taken together, hepatic expression of Sox4 is upregulated in dietary and genetic rodent models of obesity as well as in NAFLD subjects, suggesting it may play a role in the progression of hepatosteatosis.
Regulation of Sox4 by Nutrients
Next, to clarify mechanisms that potentially lead to the upregulation of Sox4 in obese livers, we explored the effects of nutrient supply on its expression. We found that mRNA levels of Sox4 were significantly inhibited during fasting conditions and induced upon refeeding (Supplementary Fig. 3A). Infusion with high levels of glucose or lipid via the jugular vein caused a significant increase in Sox4 expression in the livers of C57BL/6 mice (Supplementary Fig. 3B and C). Moreover, addition of high levels of glucose or palmitate dramatically increased Sox4 expression in Hep1-6 cells, a mouse hepatoma cell line (Supplementary Fig. 3D and E). However, insulin treatments downregulated it both in vivo and in vitro (Supplementary Fig. 3F and G). Thus, our data suggest that increased Sox4 expression in the obese liver is due, at least in part, to overnutrition.
Sox4 Promotes Liver Steatosis in Lean Mice
For exploration of the effect of Sox4 on liver TG metabolism, Ad-Sox4 or Ad-GFP was generated and administrated into C57BL/6 mice via tail vein injection. Plasma markers of liver damage (alanine aminotransferase and aspartate aminotransferase) and expression of proinflammatory cytokines (TNFα, IL-1β) were not affected by the Ad-GFP or Ad-Sox4 treatment in C57BL/6 mice compared with saline treatment, suggesting that our adenovirus did not induce acute liver inflammation or injury (data not shown). As shown in Fig. 2A, SOX4 was overexpressed in the livers (Fig. 2A) but not in other metabolic tissues such as skeletal muscles and white adipose tissues (data not shown). Although we detected no differences in body weight, body fat, and food intake (Supplementary Fig. 4A–C), at day 10 after injection, acute overexpression of Sox4 led to a significant increase in liver weight and TG content (Fig. 2B and C), which was also confirmed by Oil Red O and hematoxylin-eosin (H-E) staining (Fig. 2D). Similarly, plasma TG levels were also increased (Fig. 2E). Plasma cholesterol levels (Fig. 2F) and alanine aminotransferase and aspartate aminotransferase levels (Supplementary Fig. 4D and E) were comparable between the two groups. To determine the rate of VLDL secretion, we injected tyloxapol, an inhibitor of circulating lipases, into mice overexpressing GFP or Sox4. As a result, hepatic VLDL secretion was unaffected by Sox4 overexpression (Supplementary Fig. 4F).
Sox4 Promotes Hepatic TG Accumulation Through Upregulation of SREBP-1c
Liver TG homeostasis is determined by the balance of TG synthesis, oxidation, and transport (4). For exploration of the molecular basis for the increase in TG content by Sox4, a quantitative analysis of a panel of genes involved in TG metabolism was conducted. As a result, overexpression of Sox4 markedly increased expression levels of genes that participate in the fatty acid and TG synthesis (Fig. 3A), whereas those genes involved in β-oxidation, transport, and cholesterol metabolism were not changed (Fig. 3B–D).
We next sought to investigate the molecular basis responsible for upregulation of lipogenic genes by Sox4. Hepatic lipogenic genes are regulated by several nuclear transcription factors, including SREBP-1c, USF-1, FXR, ChREBPα/β, and LXRα/β (6,7,29–31). Our quantitative PCR results showed that SREBP-1c was significantly increased in the livers of mice expressing Sox4, while other regulators remained unchanged (Fig. 3E). Besides, mRNA levels of SREBP-1a, SREBP-2, Scap, and Insig1 were not altered (Fig. 3F). Furthermore, the increased expression of SREBP-1c was further detected at protein levels by Western blots (Fig. 3G).
For confirmation of the in vivo experiments in an independent setting, Hep1-6 cells were transfected with Ad-Sox4 or Ad-GFP (Fig. 4A). As expected, cellular TG content and expression levels of SREBP-1c were increased after Sox4 overexpression in a dose-dependent manner (Fig. 4B and C). Knockdown of SREBP-1c in Hep1-6 cells largely prevented the role of Sox4 overexpression on lipogenic enzyme expression and cellular TG contents (Supplementary Fig. 5), suggesting that the roles of Sox4 were, at least in part, dependent on its direct upregulation of SREBP-1c.
On the other hand, Hep1-6 cells were transfected with two siRNA oligos targeting Sox4 (Fig. 4D). Knockdown of Sox4 led to a dramatic decrease of TG content and downregulation of SREBP-1c (Fig. 4E and F). Besides, Sox4 depletion also reduced glucose or palmitate-induced TG retention and SREBP-1c expression (Fig. 4G–J). Therefore, our data suggest that Sox4 is not only required for the basal expression of SREBP-1c but also required for induced SREBP-1c expression by nutrients. In addition, mRNA levels of SREBP-1c were correlated well with the expression of Sox4 in livers of human subjects (Fig. 4K). Taken together, our data demonstrate that SREBP-1c might act as an important downstream target gene involved in Sox4-induced TG accumulation.
Sox4 Regulates SREBP-1c Transcription Through Binding to Its Proximal Promoter Region
To address the molecular basis for this regulation, we firstly investigated whether SREBP-1c is a direct transcriptional target of Sox4. Sox4 regulation of SREBP-1c was not sensitive to cycloheximide, a eukaryote protein synthesis inhibitor, suggesting that it was a direct transcriptional effect (Fig. 5A). We thus constructed a luciferase reporter plasmid containing the mouse SREBP-1c promoter and determined its transcriptional activity by SOX4. The binding site sequences for SOX4 have been identified by previous studies, which are 5′-AACAAAG-3′ (32), 5′-AACAAT-3′ (33), and WWCAAWG (20). First, we found that the transcriptional activity of the SREBP-1c promoter from position −1,000 to 1 (transcription start site) was upregulated by SOX4 in a dose-dependent manner (Fig. 5B). Next, using a series of truncated promoters, we defined a minimal SOX4-responsieve region from position −600 to −400 (Fig. 5C) and further identified a potential SOX4 motif at −537 to −532, which is located upstream of the DNA elements for LXR (Fig. 5D), the most established transcription factor for SREBP-1c gene expression (34). Indeed, mutation of this site substantially prevented SOX4 from activating the SREBP-1c promoter (Fig. 5E). Similar results were also observed in human embryonic kidney 293T cells (Data not shown).
In addition, ChIP assays revealed that SOX4 proteins could bind to this motif in Hep1-6 cells and mouse livers (Fig. 5F–H). The binding affinity was enhanced in HFD mice and ob/ob mice compared with that in lean mice (Fig. 5G and H). Consistently, high levels of glucose or palmitate treatment also enhanced this binding in Hep1-6 cells (Fig. 5I and J). All these data suggest that the SREBP-1c gene represents a direct target for SOX4.
Knockdown of Sox4 Improves Hepatosteatosis in ob/ob Mice
A key prediction from these observations is that Sox4 deficiency might protect obese mice from liver steatosis. To test this, we generated two independent shRNA adenoviruses (S1 and S2) to knockdown hepatic Sox4 expression in ob/ob mice. Injection of Sox4 shRNA significantly reduced hepatic Sox4 mRNA and protein levels compared with those in negative control shRNA-treated littermates (Fig. 6A and B). As a result, knockdown of Sox4 dramatically improved hepatosteatosis and decreased liver weight and plasma TG contents in ob/ob mice (Fig. 6C–F). Accordingly, we found that expression of SREBP-1c mRNA and protein was significantly reduced in Sox4-shRNA treated livers (Fig. 6G and H). Expression of lipogenic genes, including Fasn, Scd-1, and Acc1, was also reduced in this group (Fig. 6G and H). However, there was no change in body weight, food intake, fat mass, and plasma cholesterol levels (Supplementary Fig. 6).
Deficiency in Hepatic Sox4 Improves Systemic Insulin Sensitivity in ob/ob Mice
Because hepatic lipogenesis and steatosis are closely associated with insulin resistance and hyperglycemia (35,36), we next investigated the impact of hepatic Sox4 ablation upon glucose metabolism in ob/ob mice. Insulin sensitivity was enhanced in ob/ob mice with Sox4 deficiency, as shown by lower blood glucose and insulin concentrations (Fig. 7A and B), improved insulin and glucose tolerance (Fig. 7C and D), increased phosphorylated AKT signaling, and reduced expression levels of gluconeogenic genes (Fig. 7E and F).
Finally, mice with HFD-induced obesity were administrated with shRNA adenoviruses to knock down endogenous Sox4 expression in the liver. Consistent with the results observed in ob/ob mice, depletion of Sox4 by two shRNA adenoviruses (S1 and S2) also decreased hepatic TG retention, reduced lipogenic gene expression, and improved glucose metabolism in HFD-induced obese mice (Supplementary Fig. 7). Taken together, these results demonstrate that knockdown of hepatic Sox4 expression is beneficial for glucose and lipid homeostasis in obese mice.
In the current study, we define a previously unknown mechanism through which obesity may promote liver steatosis. This is suggested by multiple lines of evidence. First, Sox4 expression is increased in the liver of obese rodents and NAFLD patients. Second, liver-specific overexpression of Sox4 results in massive TG accumulation, while silence of Sox4 significantly ameliorates the severity of fatty liver and insulin resistance in obese mice. Third, our results showed that SREBP-1c could be a transcriptional target to mediate the steatotic effects of Sox4. Therefore, we propose that Sox4 may be an important mediator for obesity-associated liver steatosis. Interestingly, our microarray data and subsequent quantitative real-time PCR analysis showed that among the Sox family members, Sox4 expression was uniquely induced in obese livers. Although the exact mechanism of this high selectivity of Sox4 mRNA is currently unknown, our data suggest that dysregulation of Sox4 has a critical role in obesity-associated metabolic diseases. Until now, 20 different Sox genes have been identified in mammals, which are divided into eight subgroups named A to H (18,37). These proteins contain an evolutionary conserved high-mobility group (HMG) domain with high amino acid similarity to that of the HMG domain of SRY protein (18,37). It has been shown that individual members of the Sox family, such as SOX4, SOX11, and SOX12, may share similar biochemical properties and biological functions, due to overlapping expression patterns and recognizing the same DNA consensus motif (37,38). On the other hand, recent studies showed that the crystal structure of the SOX4 HMG domain has conformational differences compared with the structures of the HMG domains in other SOX proteins (18,39), suggesting that the biological roles and downstream target genes of SOX4 might be specific under different physiological or pathological processes. However, whether other members of the Sox family are involved in maintaining hepatic lipid homeostasis is worth studying. Especially, recent studies demonstrated that ubiquitination-mediated protein stability and degradation have an important role for the aberrant expression of SOX9 in human malignancy (40,41). Besides, a methylation-phosphorylation switch controls SOX2 protein stability in embryonic stem cell differentiation (42). These reports demonstrate that posttranscriptional regulation plays an important role in the regulation of SOX expression and function in different biological processes. Therefore, we could not exclude the possibility that protein levels of other Sox family members might be dysregulated in the development of obesity-associated NAFLD, which needs to be determined in future studies. In addition, our results indicate that Sox4 expression in hepatocytes could be regulated by glucose, palmitate, and insulin. Recent studies identified Sox4 as a downstream target of inflammatory cytokines (43,44). Therefore, whether chronic and persistent inflammation in obesity contributes to the upregulation of Sox4 remains to be addressed.
The SREBP-1c transcription factor is a principal regulator of lipogenesis, and its excessive expression has been observed in mouse models of fatty liver as well as in NAFLD patients (8,9). Transgenic mice that overexpress a truncated version of human SREBP-1 that enters the nucleus without a requirement for proteolysis exhibited severe hepatosteatosis due to markedly increased TG synthesis (45). On the other hand, SREBP-1–deficient mice showed reduced expression of lipogenic genes and decreased lipogenesis (46). Besides, adenovirus-mediated overexpression of SREBP-1c in mice with streptozotocin-induced diabetes induces lipogenic enzyme gene expression and represses the expression of gluconeogenic enzyme, leading to increased hepatic TG content and a marked decrease in hyperglycemia in diabetic mice (47). Therefore, additional mechanisms might be responsible for the improved glucose metabolism in our Sox4-deficient obese mice. In the process of revision, we found that tribbles homolog 3 (Trb3), a pseudokinase that impairs hepatic insulin signaling by binding to AKT and blocking its activation (48–50), was also a target gene of Sox4. Our data showed that mRNA expression of Trb3 was upregulated and downregulated by overexpression or inhibition of Sox4, respectively (Supplementary Fig. 8). In agreement, phosphorylated AKT levels were enhanced in ob/ob mice depleted of Sox4 even in the absence of insulin stimulation (Fig. 7E). Therefore, we speculate that hepatic Sox4 might regulate glucose metabolism and insulin sensitivity through regulation of TG metabolism and Trb3 expression.
Although recent advances have increased our understanding of how SREBP-1c is regulated at the transcriptional and posttranscriptional levels (51,52), an important question that remains to be answered is how SREBP-1c is upregulated in the obese liver. It has been well-established that SREBP-1 promoter was activated by LXR, SP1, and a feed-forward mechanism medicated by itself (51,52). Therefore, our results add Sox4 as a novel regulator of SREBP-1c expression. In summary, our in vivo gain- and loss-of-function studies demonstrate a critical role of Sox4 in obesity-associated NAFLD. We propose that dysregulation of Sox4 is a feature of the liver steatosis. Thus, inhibiting the expression or function of Sox4 might be a promising therapeutic target for fatty liver diseases and related metabolic syndrome.
Funding. This study is supported by grants from the Natural Science Foundation of China (31530033, 81722013, 81570769, and 81773018), the Shanghai Rising-Star Program of the Science and Technology Commission of Shanghai Municipality (17QA1400800 and 16JC1404100), and the Chenguang Program supported by the Shanghai Education Development Foundation and Shanghai Municipal Education Commission (15CG11).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. Y.J., J.Z., Z.Z., M.L., X.Y., and Y.Y. performed animal experiments and analyzed the data. Y.J., J.Z., and Y.L. wrote the manuscript. J.L., S.L., D.L., Y.W., D.Z., and Y.C. performed cellular experiments. G.S., B.L., Y.L., and X.L. conceived the research ideas, supervised the project, and handled the funding. X.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.