The long noncoding RNA HULC promotes liver cancer by increasing the expression of the HMGA2 oncogene via sequestration of the microRNA-186

The long noncoding RNA highly up-regulated in liver cancer (HULC) is aberrantly elevated in hepatocellular carcinoma (HCC), and this up-regulation is crucial for HCC pathogenesis. However, the underlying mechanism in HULC up-regulation is poorly understood. We hypothesized that HULC might modulate the oncogene high mobility group A2 (HMGA2) to promote hepatocarcinogenesis. Quantitative real-time PCR analysis showed that the expression levels of HULC were positively correlated with those of HMGA2 in clinical HCC tissues. Interestingly, we also observed that HULC could up-regulate HMGA2 in HCC cells. Mechanistically, we found that the microRNA-186 inhibited HMGA2 expression by targeting the 3′-untranslated region (3′-UTR) of HMGA2 mRNA. Strikingly, HULC acted as a competing noncoding RNA to sequester miR-186 and thereby relieved miR-186–mediated HMGA2 repression. Functionally, HMGA2 knockdown decreased the HULC-enhanced growth of HCC cells both in vitro and in vivo. We conclude that the long noncoding RNA HULC increases HMGA2 expression by sequestering miR-186 post-transcriptionally and thereby promotes liver cancer growth, providing new insights into the mechanism by which HULC enhances hepatocarcinogenesis.

Growing evidence demonstrates that many regulatory noncoding RNAs (ncRNAs) 3 play crucial roles in multiple biologi-cal cellular processes (1,2). The dysregulation of ncRNAs contributes to the development of numerous human diseases (3,4). At present, ncRNAs are operationally divided into two groups according to their sizes: small ncRNAs (Ͻ200 nt) and long ncRNAs (lncRNAs) (Ͼ200 nt). It has been reported that many identified lncRNAs can act as decoys, scaffold, guides, or microRNA sponges to affect tumorigenesis (5). Highly up-regulated in liver cancer (HULC) was identified as the specifically overexpressed long noncoding RNA in hepatocellular carcinoma (HCC) (6). Previously, our group reported that HULC could be up-regulated by hepatitis B virus X protein through transcription factor cAMP-response element-binding protein (7). HULC interacted with the 5Ј-UTR of CLOCK mRNA via complementary base pairing, resulting in the elevation of CLOCK expression (8). In addition, HULC could reduce the translational repression of PRKACB by sequestering miR-372 and up-regulated the expression of E2F1 by sequestering miR-107 (9,10). Functionally, the elevated HULC increased abnormal lipid metabolism and promoted angiogenesis, proliferation, migration, and invasion of HCC (11)(12)(13). However, the underlying mechanism by which HULC promotes the development of HCC is not well documented.
MicroRNAs (miRNAs) are a class of small ncRNAs that contain ϳ19 -25 nucleotides. MiRNAs act as post-transcriptional regulators of gene expression through directly binding to the 3Ј-UTR of the target mRNAs (14). It has been reported that the overexpression of miR-186 can lead to a reduction of Twist family basic helix-loop-helix transcription factor 1 (Twist1) expression, resulting in the elevation of cell apoptosis and the arrest of cell cycle in ovarian cancer (15). Extensive evidence suggests that miR-186 is a novel tumor suppressor gene (16). In addition, miR-186 targets Yes-associated protein 1 (YAP1) to suppress the Hippo signaling pathway and tumorigenesis of HCC (17). However, the roles of miR-186 in HULC-mediated HCC remain unclear.
High-mobility group A protein 2 (HMGA2), as an architectural transcription factor, plays a crucial role in the development of numerous cancers, such as colorectal cancer, lung cancer, and gastric cancer (18 -20). HMGA2 can influence the structure of chromatin by binding to AT-rich DNA sequences, leading to transcriptional regulation of genes by enhancing or suppressing transcription factors (21). In terms of functionality, it affects cell cycle, proliferation, epithelial-to-mesenchymal transition, and invasiveness of cancer cells (22,23). More importantly, the expression of HMGA2 is correlated with metastatic potential (24,25). HMGA2 expression is primarily increased in different cancers, and it can be regulated by different mechanisms. Up-regulation of HMGA2 often results from genetic alterations, such as gene amplification and translocation, which lengthen the promoter region or shorten the 3Ј-UTR of HMGA2 (26). It has been reported that HMGA2 is inhibited by let-7 in breast cancer and miR-107 in HCC (22,27). It also can be increased by the Wnt/␤-catenin pathway (28). However, whether HULC is involved in the elevation of HMGA2 in HCC is poorly understood.
In the present study, we are interested in whether lncRNA HULC enhances the development of HCC through HMGA2. Interestingly, our data showed that HULC increases HMGA2 expression by sequestering miR-186 post-transcriptionally and thereby promotes liver cancer growth. Our finding provides new insights into the mechanism of hepatocarcinogenesis mediated by lncRNA HULC.

HULC is positively correlated with HMGA2 in clinical HCC tissues and up-regulates HMGA2 in HCC cells
It has been reported that HULC plays important roles in the development of HCC (12). Given that HMGA2 is a crucial oncogene (25,29), we supposed that HULC might modulate the expression of HMGA2 to promote hepatocarcinogenesis. To demonstrate the significance of HMGA2 in liver cancer, we examined the expression of HMGA2 by immunohistochemistry (IHC) staining using microarrays of clinical HCC tissues. Our data showed that the positive rate of HMGA2 was 71.2% (141 of 198) in HCC tissues (Fig. 1A and supplemental Table S1). Interestingly, the mRNA levels of HMGA2 were significantly increased in 30 clinical HCC tissues relative to their corresponding peritumor tissues (p Ͻ 0.001, Wilcoxon's signed-rank test; Fig. 1B). Moreover, quantitative real-time PCR analysis revealed that the expression levels of HULC were positively associated with those of HMGA2 in 30 cases of clinical HCC tissues (r ϭ 0.7925, p Ͻ 0.001, Pearson's correlation; Fig. 1C), suggesting that HULC may up-regulate HMGA2 in HCC cells. Next, we evaluated the effect of HULC on HMGA2 in HCC cells. Interestingly, our data indicated that HULC could markedly up-regulate the expression of HMGA2 in HepG2 and Huh7 cells at the levels of mRNA and protein in a dose-dependent manner (Fig. 2, A-C). The transfection efficiency of HULC was validated in these cells. Furthermore, we observed that HULC increased the expression of interleukin-15 (IL15) and cyclin A, two downstream genes of HMGA2, in HepG2 cells in a dose-dependent manner ( Fig. 2A), indicating that HULC is able to up-regulate HMGA2. Conversely, inhibition of HULC by siRNA led to the reduction of the expression of HMGA2 in HepG2 and HepG2.2.15 cells (highly expressing endogenous HULC) in a dose-dependent manner at mRNA and protein levels. Meanwhile, the interference efficiency of si-HULC was validated in these cells (Fig. 2, D-F). In addition, si-HULC lowered the expression of IL15 and cyclin A in HepG2 cells in a dose-dependent manner as well (Fig. 2D). Therefore, we conclude that HULC is positively correlated with HMGA2 in clinical HCC tissues and up-regulates HMGA2 in HCC cells.

miR-186 inhibits the expression of HMGA2 by directly targeting the 3-UTR of HMGA2 mRNA
Given that lncRNAs can act as a molecular sponge to sequester miRNAs (30), we speculated that HULC might modulate HMGA2 through miRNAs. We first compared the sequences of HMGA2 mRNA 3Ј-UTR with those of HULC using the BLAST tool of NCBI (https://blast.ncbi.nlm.nih. gov/Blast.cgi#45219875). Interestingly, bioinformatics analysis identified that three sites of the HULC sequence were the same as those of the HMGA2 mRNA 3Ј-UTR (Fig. 3A). Furthermore, we identified that miR-186 could potentially bind to site 1 of the HMGA2 mRNA 3Ј-UTR using miRanda (http://www.microrna.org/microrna/home.do) 4 and Target-Scan (http://www.targetscan.org/). 4 We analyzed the expression levels of miR-186 in 30 paired clinical liver tissues. We found that the expression levels of miR-186 were lower in 30 HCC tissues relative to those in their corresponding peritumor liver tissues by quantitative real-time PCR analysis (Fig. 3B). Then we evaluated the correlation between miR-186 and HMGA2 in clinical HCC tissues. Quantitative real-time PCR analysis demonstrated that the expression levels of HMGA2 were negatively associated with those of miR-186 in 30 clinical HCC tissues (r ϭ Ϫ0.6002, p Ͻ 0.001, Pearson's correlation; Fig. 3C), suggesting that miR-186 may inhibit HMGA2 in HCC cells. Based on the bioinformatics prediction, we cloned the miR-186 binding site in the 3Ј-UTR of HMGA2 mRNA and its mutant into luciferase reporter gene vector pGL3-control, termed pGL-HMGA2 and pGL-HMGA2-mut (Fig. 3D). Luciferase reporter gene assays revealed that miR-186 could decrease the luciferase activities of pGL-HMGA2 in HepG2 and 293T cells in a dose-dependent manner, rather than pGL-HMGA2-mut (Fig. 3, E and F). Conversely, the inhibition of miR-186 increased the luciferase activities of pGL-HMGA2 in HepG2 and 293T cells in a dose-dependent manner, but the pGL-HMGA2-mut did not (Fig. 3, G and H), suggesting that  . Taken together, we conclude that miR-186 inhibits the expression of HMGA2 by directly targeting the 3Ј-UTR of HMGA2 mRNA.

HULC is able to increase the expression of HMGA2 by sequestering miR-186
Next, we evaluated the sponge effect of HULC on miR-186 in HCC cells. We predicted the interaction between HULC and miR-186 by using the Bielefeld Bioinformatics Service (http://bibiserv. techfak.uni-bielefeld.de/rnahybrid/submission.html). 4 Bioinformatics analysis showed that HULC could potentially interact with miR-186 through complementary base-pairing reactions (Fig.  5A). Moreover, quantitative real-time PCR analysis demonstrated that the expression levels of HULC were negatively associated with those of miR-186 in 30 clinical HCC tissues (r ϭ Ϫ0.5158, p Ͻ 0.001, Pearson's correlation; Fig. 5B), suggesting that HULC may interact with miR-186 in HCC cells. We observed that the overexpression of HULC could not inhibit the levels of miR-186 in HepG2 cells (Fig. 5C). However, quantitative real-time PCR analysis revealed that the overexpression of miR-186 could down-regulate the expression of HULC in HepG2 cells, whereas the interference with miR-186 by anti-miR-186 had the opposite effect (Fig. 5, D and E), suggesting that the interaction of miR-186 with HULC lead to the downregulation of HULC. To confirm whether HULC could enhance the expression of HMGA2 by sequestering miR-186, we constructed the mutant of HULC (HULC-186-mut), which carried a substitution of eight nucleotides within the binding sites of miR-186. Luciferase reporter gene assays showed that HULC could rescue the decrease of pGL-HMGA2 luciferase activities mediated by miR-186 in HepG2 and 293T cells, whereas HULC-186-mut failed to work (Fig. 5, F and G). Furthermore, the similar effect of HULC on the expression of HMGA2 at the protein level was obtained in HepG2 and HepG2.2.15 cells (Fig.  5H). Thus, we conclude that HULC is able to increase the expression of HMGA2 by sequestering miR-186 at the posttranscriptional level.

HULC contributes to the growth of HCC cells in vitro and in vivo through HMGA2
To test whether HMGA2 is involved in the HULC-mediated HCC cells proliferation, we performed 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazolium bromide (MTT), 5-ethynyl-2Јdeoxyuridine (EdU), colony formation, and flow cytometry assays in HepG2 cells. The expression of HMGA2 was evaluated by Western blot analysis in the cells after the treatments HULC up-regulates HMGA2 in liver cancer (Fig. 6A). Our data showed that HULC enhanced the proliferation ability of HCC HepG2 cells, whereas overexpression of miR-186 or inhibition of HMGA2 could suppress the proliferation (Fig. 6, B-E). Based on our data obtained in vitro, we verified our conclusion in vivo using a tumor xenograft model. As expected, we found that overexpression of HULC remarkably increased the volume and weight of the tumors, but the interference of HMGA2 mRNA could block the HULC-enhanced tumor growth in mice (Fig. 7, A-C). Meanwhile, we validated the expression levels of HULC, HMGA2, and miR-186 in the tumor tissues with the pretreatment, respectively. Consistently, the expression levels of HULC were increased in the HULC-overexpressed group, but not for HULC overexpression/si-HMGA2 group (Fig. 7D), which is the outcome of the decreased tumor size and tumorigenesis. The expression levels of HMGA2 were similar to those of HULC (Fig. 7E). The expression levels of miR-186 were opposite those of HULC and HMGA2 (Fig. 7F), which is the outcome of the decreased tumor size and tumorigenesis, suggesting that HULC enhances the growth of hepatoma through HMGA2, in which miR-186 is involved in the event. Accordingly, we examined the expression levels of Ki67, a cell proliferation marker, in the tumor tissues from mice. IHC assays indicated that inhibition of HMGA2 could decrease the expression of Ki67 (Fig. 7G), indicating that HMGA2 is involved in the HULC-mediated growth of HCC cells. Collectively, we conclude that HULC contributes to the growth of HCC cells in vitro and in vivo through HMGA2.

Discussion
lncRNAs play pivotal roles in a host of pathologic processes, especially in tumorigenesis (31)(32)(33). Our group has reported that lncRNA HULC contributes to the development of HCC by modulating HCC angiogenesis and abnormal lipid metabolism (10,11). As an oncogene, HMGA2 is highly expressed in many tumor tissues (34,35). However, the effect of HULC on HMGA2 in hepatocarcinogenesis is not well known. In this study, we are interested in whether HULC is involved in the regulation of HMGA2 in HCC.
To examine the effect of HULC on HMGA2 in the development of HCC, we first evaluated the correlation between expression levels of HULC and HMGA2 in clinical HCC tissues. Our data revealed that the mRNA levels of HMGA2 were much higher in the HCC samples than their corresponding peritumor liver tissues. In addition, the expression levels of HULC were positively associated with those of HMGA2 in clinical HCC samples. This suggests that HULC may increase HMGA2 expression in HCC. As expected, we observed that HULC was able to up-regulate the expression of HMGA2 at mRNA and protein levels in HCC cells. Therefore, this suggests that HULC is involved in the regulation of HMGA2 in liver cancer. Next, we tried to identify the underlying mechanism by which HULC up-regulates HMGA2. Given that lncRNAs can act as an miRNA sponge to promote the development of cancers (30,36), we presumed that HULC might contribute to the elevation of HMGA2 by sequestering miRNAs. If the miRNA can bind to HMGA2 mRNA and it simultaneously interacts with HULC, then HMGA2 mRNA and HULC will have partially the same sequences. Based on this hypothesis, we compared the sequences of HULC and those of HMGA2 mRNA 3Ј-UTR. Interestingly, bioinformatics analysis showed that three sites in HULC RNA were completely the same as those in the HMGA2 mRNA 3Ј-UTR. Furthermore, we predicted that miR-186 could potentially bind with one of those sites, suggesting that miR-186 may be involved in the regulation of HMGA2 mediated by HULC. Increasing evidence has highlighted the significance and molecular mechanisms of miR-186 in many cancers (37). It has been reported that miR-186 suppresses the migration and invasion of human non-small cell lung cancer cells by targeting pituitary tumor transforming gene (PTTG1) mRNA 3Ј-UTR (38). Overexpression of miR-186 also can decrease the proliferation, invasion, and migration of human gastric cancer through the inhibition of Twist1 expression (39). In our study, we found that the expression levels of miR-186 were negatively associated with those of HMGA2 in clinical HCC tissues, which is consistent with the suppressive function of miR-186 in other cancers. In addition, miR-186 was capable of directly binding to the 3Ј-UTR of HMGA2 mRNA in HCC cells, resulting in the inhibition of HMGA2 expression. It suggests that miR-186 is able to suppress the expression of HMGA2 in HCC cells. Next, we examined whether HULC could block the decrease of HMGA2 mediated by miR-186. We observed that HULC rescued the inhibition of HMGA2 mediated by miR-186. This suggests that HULC can elevate the expression of HMGA2 through the sponge of miR-186 at the post-transcriptional level. It has been reported that miR-186 can down-regulate some important oncogenes, such as YAP, Twist1, PTTG1, and HIF-1␣ (15,17,38,40). This suggests that HULC may widely modulate oncogenes through sequestering miR-186 to promote hepatocarcinogenesis. Our group has reported that HULC can sequester miR-107 through a different site from miR-186, leading to an increase in the expression of E2F1. In addition, we have also reported that miR-107 inhibits the expression of HMGA2 in HCC cells (10,22), implying that HULC may also up-regulate HMGA2 expression by sequestering miR-107. However, our data failed to show that HULC could increase HMGA2 expres-

HULC up-regulates HMGA2 in liver cancer
sion through the sponge of miR-107 (data not shown). Functionally, we found that HULC was able to facilitate the growth of HCC cells in vitro and in vivo through HMGA2, in which miR-186 is involved in the event. Therapeutically, HULC and HMGA2 may serve as targets in liver cancer.
Taken together, we illustrate a novel mechanism by which HULC modulates HMGA2 in hepatocarcinogenesis in this study (Fig. 8). Our findings indicate that HULC is able to upregulate HMGA2 in HCC cells. Mechanistically, miR-186 is capable of decreasing the expression of HMGA2 through targeting 3Ј-UTR of HMGA2 mRNA. Strikingly, HULC can sponge miR-186, leading to the increase of HMGA2 expression in hepatocarcinogenesis. Thus, we conclude that HULC increases HMGA2 expression by sequestering miR-186 posttranscriptionally and thereby promotes liver cancer growth. Our finding provides new insights into the mechanism of hepatocarcinogenesis mediated by HULC.

Patient samples
HCC tissue samples and their corresponding peritumor liver tissues (n ϭ 30) were obtained during surgery at Tianjin First Center Hospital and Tianjin Tumor Hospital (Tianjin, China).  (n ϭ 3). B, the proliferation of HepG2 cells transfected with control, HULC, or HMGA2 or co-transfected with HULC and miR-186 or si-HMGA2 was detected by MTT assays (n ϭ 3). C, the proliferation of HepG2 cells transfected with control, HULC, or HMGA2 or co-transfected with HULC and miR-186 or si-HMGA2 was measured by EdU incorporation assays (n ϭ 3). D, the proliferation of HepG2 cells transfected with control, HULC, or HMGA2 or co-transfected with HULC and miR-186 or si-HMGA2 was examined by colony formation assays (n ϭ 3). E, the proliferation of HepG2 cells transfected with control, HULC, or HMGA2 or co-transfected with HULC and miR-186 or si-HMGA2 was assessed by flow cytometry analysis. Error bars, S.D.

Figure 7. HULC promotes the growth of HCC cells through HMGA2 in vivo.
A, dissected tumors from nude mice. B, growth curves of tumors from nude mice. C, average tumor weight of each group. D, the expression levels of HULC were measured by quantitative real-time PCR analysis in tumor tissues from nude mice. E, the expression levels of HMGA2 were examined by Western blot analysis in tumor tissues from nude mice. F, the expression levels of miR-186 were evaluated by quantitative real-time PCR analysis in tumor tissues from nude mice. G, the expression of Ki67 in tumor tissues from nude mice was detected by IHC assays. Error bars, S.D.

HULC up-regulates HMGA2 in liver cancer
The detailed information on the patients with HCC, obtained from patient records, is presented in supplemental Table S2. Informed consent approving the use of tissue samples for study purposes was obtained from each patient. The study was approved by the Institute Research Ethics Committee at Nankai University and was performed in accordance with relevant guidelines and regulations.

Cell transfection
Cells were cultured in 6-well or 24-well plates for 12 h and then were transfected with plasmids or siRNAs. All transfections were performed using Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol.

Western blot analysis
HepG2 (or Huh7) cells were seeded into a 6-well plate and incubated in a humidified 5% CO 2 atmosphere at 37°C in antibiotic-free DMEM supplemented with 10% FBS for 12 h to about 50% confluence. Then cells were transfected with pcDNA3.1-HULC, pcDNA3.1-HMGA2, miR-186, or their corresponding controls using TRIzol reagent (Invitrogen) according to the manufacturer's protocol. 48 h after transfection, Western blot analysis was carried out as described previously (44). Primary antibodies used were rabbit anti-HMGA2 (Genetex, Irvine, CA), mouse anti-␤-actin (Sigma). All experiments were repeated three times.

Luciferase reporter gene assays
Cells were transferred into 24-well plates at 3 ϫ 10 4 cells/ well. After 12 h, the cells were co-transfected with the pRL-TK plasmid containing the Renilla luciferase gene (Promega, Madison, WI) used for internal normalization and various constructs containing HMGA2 3Ј-UTR, pCDNA3.1-HULC, or their corresponding mutants. 36 h after transfection, the luciferase activities were measured using the Dual-Luciferase reporter assay system (Promega, Madison, WI) according to the manufacturer's protocol. All experiments were performed at least three times.

Analysis of cell proliferation
HepG2 or Huh7 cells were seeded into 96-well plate at 1000 cells/well for 12 h before transfection. The MTT assay was performed as described previously (44). Briefly, cells were cultured after different days (day 0, 1, 2, and 3), and 15 l of MTT was added to each well, followed by incubation for 4 h. The supernatant was discarded, and 100 l of dimethyl sulfoxide was added to stop the reaction. Absorbance at 490 nm was assessed using an ELISA reader system (Multiskan Ascent, LabSystems). The EdU incorporation assay was applied by using the Cell-Light TM EdU imaging detecting kit following the manufacturer's instructions (RiboBio). For the colony formation assay, 12 h after transfection, 500 transfected cells were placed into a 6-well plate and maintained in complete medium for 2 weeks. Colonies were fixed with methanol and stained with crystal violet (45). Flow cytometry analysis was processed as described earlier (46).

IHC staining assays
The microarrays including HCC tissues were obtained from Xi'an Aomei Biotechnology Co., Ltd. (Xi'an, China). These microarrays were composed of 198 HCC tissues, which included duplicate core biopsies (1 mm in diameter) from fixed, paraffin-embedded tumors. The medical records of the patients are listed in supplemental Table S1. The IHC staining assay was performed as described previously (47), and the primary antibodies of rabbit anti-HMGA2 (Genetex) and rabbit anti-Ki67 (Thermo Fisher Scientific) were used. The results of IHC were evaluated as described earlier (48). In brief, the staining level of HMGA2 was classified into three groups using a modified scoring method based on the intensity of staining (0, negative; 1, low; 2, high) and the percentage of stained cells (0, 0% stained; 1, 1-49% stained; 2, 50 -100% stained). A multiplied score (intensity score ϫ percentage score) of 0 was considered to be negative staining (Ϫ), 1 was considered to be low staining (ϩ), 2 was considered to be moderate staining (ϩϩ), and 4 was considered to be high staining (ϩϩϩ). Categorization of immunostaining intensity was performed by three independent observers.

In vivo tumorigenicity assays
Nude mice were housed and treated in accordance with the guidelines established by the National Institutes of Health Guide for the Care and Use of Laboratory Animals. We conducted animal transplantation according to the Declaration of Helsinki principles. All experiments were approved by the

HULC up-regulates HMGA2 in liver cancer
Institute Research Ethics Committee at Nankai University. Tumor transplantation in nude mice was performed as described previously (49 -51). Briefly, HepG2 cells were pretreated with control, pCDNA3.1-HULC, pCDNA-3.1-HULC, and si-HMGA2 for 24 h. Then the cells were harvested and resuspended at 2 ϫ 10 7 cells/ml with sterile PBS. Groups of 4-week-old male BALB/c athymic nude mice (Experimental Animal Center of Peking, China; each group, n ϭ 6) were subcutaneously injected at the shoulder with 0.2 ml of the cell suspensions. Tumor growth was measured after 5 days from injection and then every 5 days. Tumor volume was monitored by measuring the length (L) and width (W) with calipers and calculated with the formula, (L ϫ W 2 )/2. After 30 days, the mice were sacrificed, and the tumors were excised and measured.

Statistical analysis
Each experiment was repeated at least three times. Statistical significance was assessed by comparing mean values Ϯ S.D. using Student's t test for independent groups and was assumed for p Ͻ 0.05 (*), p Ͻ 0.01 (**), p Ͻ 0.001 (***), and not significant (NS). Association between expression levels of HULC and HMGA2 in tumor tissues was assessed using Pearson's correlation coefficient. HMGA2 expression in tumor tissues and their corresponding peritumor tissues was compared using the Wilcoxon signed-rank test.
Author contributions-X. Z. and L. Y. conceived the project, designed the experiments, and drafted the manuscript. X. Z. and Y. W. designed the experiments, drafted the manuscript, and performed the experiments. F. C., M. Z., Z. Y., J. L., S. Z., and W. Z. performed the experiments. All authors reviewed the results and approved the final version of the manuscript.