Apelin-mediated deamidation modication of HMGA1 promotes tumorigenesis by enhancing SRBEP1 activity and Lipid Synthesis

Apelin is a ligand of the G protein-coupled receptor that promotes tumor growth in malignant cancers. However, the molecular mechanisms through which apelin promotes tumorigenesis are unknown. Here, we conrmed that apelin promotes tumorigenesis in lung cancer cells by increasing the synthesis of fatty acids, which induces abnormal lipid metabolism. Apelin interacts with high-mobility group A HMGA1 and mediates sterol-regulatory-element-binding protein 1 (SREBP1) activity, which is required for lung tumorigenesis and lipid metabolism. Deamidation modication of HMGA1 is regulated by apelin enhanced SRBEP1 activity and lipid synthesis. Moreover, deamidated HMGA1 can enhance the formation of the apelin-HMGA1-SREBP1 complexes and increases SREBP1 activity, which induces abnormal lipid metabolism. As an energy regulator, Apelin forms a multi-protein complexes with HMGA1 to increase of lung cancer cells viability. Our results indicate that apelin is important in lipid metabolism and cancer cell proliferation.


Introduction
Apelin is a member of adipokines and is abundantly secreted by adipocytes 1 . Apelin has several functions, including energy metabolism, uid homeostasis and angiogenesis 2 . Apelin greatly increases the likelihood of malignant tumors, such as lung cancer 3 , glioblastoma 4 , ovarian cancer 5 , colon cancer 6 , and hepatocellular carcinoma 7 and is associated with a poor prognosis, which suggests apelin might promote tumorigenesis. Exogenous apelin has as anti-apoptotic effect and stimulates the proliferation of colon cancer cells through the JAG-1/Notch3 signalling pathway 8 . Apelin also increases the invasiveness of breast cancer 9 , gastric cancer 10 , and the colon cancer cells by a matrix metalloproteinase (MMP)dependent manner 11 . In addition, apelin is an important proangiogenic factor and could be a potential target in tumor therapy by interfering with angiogenesis 12 . However, the molecular mechanisms of apelin in tumorigenesis are unknown.
Metabolic reprogramming, including both the Warburg effect (aerobic glycolysis) and abnormal lipid synthesis, is one of the hallmarks of cancer, which maintains cell proliferation in a severe microenvironment and provides energy to support the metastasis of cancer cells 13,14 . Increased glucose consumption is a major source of energy in tumor cells, not only for glycolysis, but for increased lipogenesis 15 . Activated de novo lipogenesis accelerates tumor cell growth, participates in immune evasion, and is closely associated with poor prognosis in many tumors 16 .
The important molecules in lipogenesis include acetyl-CoA carboxylase (ACC) 1, which catalyzes the ATPdependent carboxylation of acetyl-CoA to form malonyl-CoA 17 . The multi-domain enzyme fatty acid synthase (FASN) uses malonyl-CoA and acetyl-CoA to catalyze subsequent successive reactions to form fatty acids 18 .
Sterol regulatory-element-binding protein 1 (SREBP1), a member of the transcription factor family, is also important in fatty acid synthesis. As an inactive precursor, SREBP1 is found in the endoplasmic reticulum and in the proteolytic release of an N-terminal fragment in Golgi-activated SREBP1 and promotes the expression of lipogenic genes 19,20 . Apelin might also regulate glucose metabolism by restoring glucose tolerance and increasing glucose metabolism in insulin-resistant mice 21 . However, apelin inhibits lipid accumulation in human hepatoma cells (Hep3B) and in primary mouse hepatocytes. It also suppresses the expression of sterol regulatory element-binding protein 1c (SREBP-1c) and glycerol-3-phosphate acyltransferase (GPAT) 22 . But, whether apelin promotes the proliferation and metastasis of tumors by regulating abnormal lipogenesis is unknown.
In addition to glucose, glutamine is also a source of carbon for replenishing the tricarboxylic acid (TCA) cycle and produces the precursor (glutathione) to nucleotides and lipid synthesis by reductive carboxylation 23 . The glutamine amidotransferases (GATs) are a family of metabolic enzymes that extract nitrogen from glutamine to synthesize nucleotides, amino acids, glycoproteins, and the enzyme cofactor, nicotinamide adenine dinucleotide (NAD), which are required for cell growth and proliferation 24 . Deamidation modi cation increases the a nity of protein interactions and promotes cell metabolism for cancer-cell proliferation and tumorigenesis. For example, deamidating the RelA subunit of NF-kB in cancer cells promotes aerobic glycolysis and increases tumorigenesis. This post-translational modi cation switches RelA function from increasing the expression of NF-kB-responsive genes to increasing the production of glycolytic enzymes, thus shunting the cell's in ammatory response to aerobic glycolysis 25 . Whether apelin is involved with deamidation modi cation in promoting tumorigenesis is unknown.
In this study, we created lung cancer cells lines to establish apelin-overexpression (A549 and H1975) and knock-down (H460) cells to investigate the mechanism of apelin in regulating lung cancer. We found that apelin promoted proliferation of lung cancer cells by increasing fatty acid de novo synthesis. We identi ed a new mechanism in which apelin formed complexes with HMGA1 and SRBEP1, deamidates HMGA1, and elevates SRBEP1 activity, which results in abnormal lipid metabolism.

Apelin Expression was Increased in Lung Cancer Tissues
In this study, the expression of apelin in tissue samples of patients with lung cancer was detected by immunohistochemistry. Of the 51 patients enrolled in the study (32 men and 19 women), 24 had adenocarcinoma of the lung tissues (LUAD) and 27 had squamous carcinoma of the lung tissues (LUSC; Supplementary Table 1). In addition to 51 resected tumor samples, 39 distant healthy tissues were used as controls. Staining intensity was classi ed as low or high (Fig. 1a). Apelin expression was signi cantly higher in tumor tissues than in normal tissues (Fig. 1a). Kaplan-Meier analysis indicated that increased apelin expression was related to poorer survival of lung cancer patients, although not signi cantly so (P=0.362; Fig. 1b). As in our previous study, Apelin expression was markedly elevated in lung tumor tissues.

Apelin Increased the proliferation of Lung Cancer Cells
To further explore the functions of apelin in lung cancer, we rst analyzed the expression of apelin in cell lines from both tumor (A549, H1299, H446, H1975, and H460) and normal lung tissue (HBE). The expression of apelin was higher in H1299 and H460 cells in both mRNA and protein concentrations ( Fig.   1d and Supplementary Fig. 1a) than in A549, H1975, and H446 (Fig. 1d).
In addition, we found that knock-down expression of apelin in A549-OE substantially reduced cell viability and colony formation after apelin-siRNA treatment ( Supplementary Fig. 1f-h). These results con rmed that apelin enhances cellular proliferation and colony formation in lung cancer cells. Finally, A549-OE and -NC cells implanted into nude mice resulted in subcutaneous tumor formation and increased tumor formations (Fig. 1i). Tumor volume (Fig. 1j), and tumor weight ( Supplementary Fig. 1i) in A549-NC and -OE group were detected from 6 weeks after implantation. All these results indicate that apelin promoted tumor growth in lung cancer cells.

Apelin Induced Metabolic Reprogramming in Lung Cancer Cells by Enhancing Lipid Metabolism
To investigate how apelin promotes tumor formation, A549-OE and -NC cells detected by RNA-Seq revealed enrich variation were metabolic-related genes in A549-OE cells (Fig. 2a), indicating changes in apelin expression are associated with changes in metabolism. Quantitative PCR showed that fatty acid synthesis-related genes were upregulated in both A549-OE (OE, over-expression) and H1975-OE cells, including GLUT4, FASN, ACC1, and SREBF1 (Fig. 2b), which were also observed in protein concentrations ( Fig. 2c).
Glucose was the most important material in both the tricarboxylic acid cycle (TCA) and lipid synthesis ( Fig. 2d). To test whether apelin uses glucose as an energy source to increase fatty acids synthesis, both A549-OE and -NC (NC, negative controls) cells were labelled with [U-13 C6] glucose to measure carbon incorporation into newly synthesized fatty acids over 24 hours. The results indicated that apelin promoted the synthesis of both saturated fatty acids (palmitate, stearate, and myristate; Fig. 2e) and unsaturated fatty acids (palmitoleate, oleate, and arachidonate ( Supplementary Fig. 2a). The intermediaries of the TCA cycle were mostly decreased in A549-OE cells except citrate (Fig. 2f). Glucose uptake was also elevated in A549-OE and H1975-OE cells (Fig. 2g), and apelin-siRNAs reduced the elevated glucose uptake ( Supplementary Fig. 2b), which indicated that apelin preferentially used glucose as source for lipogenesis. We next used nile red staining to detect the lipid concentration on apelin-overexpression or knock-down cells. Lipid components markedly were greatly increased in A549-OE and H1975-OE cells ( Fig. 2h and Supplementary Fig. 2c). Instead, glucose uptake declined, and lipid components accumulation was restrained in H460-KD cells ( Fig. 2i and Supplementary Fig. 2d).
To verify the effect of lipids on cell proliferation, H460-negative control (control, H460) and apelin knockdown cells (KD, H460) cells were grown in palmitate and then plated into media containing either regular fetal bovine serum (FBS) or delipidated FBS (FCS) in the presence or absence of exogenous palmitate.
Cellular growth in apelin knock-down cells (KD, H460) was markedly lower than in controls(control, H460) in medium containing regular FBS, a result that was even more pronounced in delipidated FBS, but apelin-KD cells (KD, H460) were rescued by exogenous palmitate (Fig. 2j).
Finally, nile red staining in H460-control and -KD cells, as well as those treated and untreated with palmitic acid in a medium containing FBS or FCS, also con rmed that apelin knock-down greatly decreased lipid content (Fig. 2k). Taken together, these results indicate that apelin can induce metabolic reprograming by promoting lipid synthesis.

SREBP1 was Important in Apelin-Mediated Lipid Metabolism
Our study indicated that apelin markedly promoted the expression of sterol-regulatory-element-binding proteins 1 (SREBP1; Fig. 2b), a transcription factor that induces the expression of genes synthesizing cholesterol and fatty acids 26 . High glucose stimulation enhances SREBP1 expression and prevents SREBP1 binding to ER-anchored, insulin-induced gene protein (Insig1) and activates the expression genes involved in lipid metabolism 14 . Therefore, we further investigated whether SREBP1 was involved in apelininduced abnormal lipid metabolism.
Overexpression of apelin promoted mature (SREBP1-N) and precursor-SREBP1 (SREBP1-P) expression in A549-OE and H1975-OE cells (Fig. 3b). Cellular immuno uorescence experiments con rmed that apelin promoted the expression of endogenous SREBP1 and the localization of SREBP1 to the Golgi apparatus and to the nucleus in A549-OE cells (Fig. 3b), a nding further con rmed by nuclear protein detection (Fig.   3c). To further con rm the effect of apelin on SREBP1, knock-down apelin in H460 cell(H460-KD), SREBP1 expression level and activity were reduced, but were rescued with reconstituted of apelin in H460-KE cell (Fig. 3d).
Next, SREBP1-siRNA was introduced to con rm the function of SREBP1 in apelin-mediated lipid metabolism. We found that inhibiting SREBP1 decreased both mature and precursor-SREBP1, as well as the lipid synthesis-related molecules, ACC1 and FASN, in both protein (Fig. 3e) and mRNA concentrations (Fig. 3f). Nile red staining revealed that lipid synthesis was also inhibited in A549-OE and H1975-OE cells in the knockdown of SREBP1 (Fig. 3g). Finally, si-SREBP1 also inhibited cell viability in A549-OE and H1975-OE cells (Fig. 3h). These results strongly suggest the importance of SREBP1 in abnormal lipogenesis and cell viability mediated by apelin.
Apelin Improved SREBP1 Activity by Forming Apelin-SREBP1 Complexes and Inhibiting SREBP1-Insig1 Binding Th above results con rmed that apelin promotes SREBP1 activity, but the molecular mechanism is still unknown. Insulin-induced gene protein 1 (Insig1) could bind to SREBP1 on the endoplasmic reticulum, restrain the movement of SREBP1 to the Golgi apparatus, inhibit the N-terminal cleavage of SREBP1 into the nucleus, and regulate the expression of lipid metabolism genes. Therefore, a co-immunoprecipitation (Co-IP) assay was performed to verify whether apelin regulated the binding of SREBP1 and Insig1. Overexpression of apelin markedly inhibited SREBP1 from combining with Insig1 in A549-OE and H1975-OE cells (Fig. 4a,4b). To the contrary, knockdown of apelin greatly increased the binding of SREBP1 to Insig1, and re-expression of apelin in H460-KD cells re-inhibited SREBP1-Insig1 binding (Fig. 4c).
Next, we investigated whether apelin inhibited SREBP1 from binding to Insig1 by direct interaction with SREBP1. Using apelin as a bait protein, the Co-IP result con rmed that binding of apelin to SREBP1 was increased in A549-OE and H1975-OE cells (Fig. 4d,e). Our results indicated that apelin promoted SREBP1 activity by dissociating SREBP1 with Insig1 to forming an apelin-SREBP1 complexes.

Apelin-HMGA1 Complexes Facilitates Lipid Metabolism and Lung Cancer Cell Proliferation
Several molecules contribute to metabolic reprograming in malignant tumors. High Mobility Group A1 (HMGA1) is an architectural transcription factor involved in regulating lipid metabolism by promoting GLUT3 expression 27 . Cancer Genome Atlas (TCGA) data analysis showed that expression of HMGA1 was substantially increased in tissues of patients with lung adenocarcinoma (LUAD) and squamous carcinoma (LUSC; Supplementary Fig. 3a,b). Furthermore, elevated HMGA1 concentrations were associated with poor prognosis in patients with LUAD (P=0.0027; Supplementary Fig. 3c) but not with survival in patients with LUSC (data not shown). In addition, Pearson correlation analysis revealed that HMGA1 concentration was related to multiple lipid synthesis-related genes ( Supplementary Fig. 3d), which suggested the importance of HMGA1 in lipid metabolism. Therefore, we speculated that HMGA1 might be involved in apelin-mediated abnormal lipid synthesis.
We rst observed that the expression of HMGA1 was increased in A549-OE and H1975-OE cells ( Supplementary Fig. 3e and Fig. 5a) and decreased in H460-KO cells ( Supplementary Fig. 3f,g). A Co-IP assay revealed increased binding of apelin to HMGA1 in A549-OE and H1975-OE cells (Fig. 5a). In addition, apelin and HMGA1 were co-localized in the nucleus of A549 -OE cells (Fig. 5b).
When HMGA1-siRNA was introduced to con rm this observation, we found that knockdown HMGA1 greatly inhibited cell viability (Fig. 5c) and colony formation in A549-OE and H1975-OE cells (Fig. 5d). Moreover, glucose uptake capacity and acetyl-CoA activity were also restrained in A549-OE and H1975-OE cell after si-HMGA1 treatment (Fig. 5e,f), as with the expression of lipid synthesis genes, including ACC1, SREBP1, and FASN (Fig. 5g).
The most interesting observation was that si-HMGA1 inhibited SREBP1 concentrations in both the mature (SREBP1-N) and precursor (SREBP1-P) forms in A549 -OE and H1975-OE cells (Fig. 5h), which suggested a close relationship between HMGA1 and SREBP1 in abnormal lipid metabolism induced by apelin. Finally, cholesterol -modi ed si-HMGA1 in the treatment of subcutaneous tumors in nude mice indicated that si-HMGA1 slightly reduced the tumor volume and weight of A549-OE engrafted mice (Fig. 5i,j). Oil red staining con rmed that lipid components also decreased more in si-HMGA1 tumor tissues than in A549-OE tissues (Fig. 5k). These results con rm that HMGA1 is important in apelin-mediated abnormal lipid metabolism in lung cancer cells.
Deamidated HMGA1 Increased the Binding of Apelin to HMGA1 and Increased the Activity of SREBP1 Our study con rmed that HMGA1 can increase abnormal lipid synthesis by forming an HMGA1-apelin complexes (Fig. 5a); however, the binding regions in HMGA1 and the regulation mechanism are unknown.
Immunoprecipitation and mass spectrometry revealed that two fragments of the HMGA1 protein directly bind to apelin. We constructed three de cient mutant plasmids of HMGA1: delete 1 and delete 2 were de cient in the binding region of 30-47 and 92-107 respectively, and delete 3 was co-de cient in two binding fragments (Fig. 6a). The three described above de cient mutant plasmids of HMGA1(HMGA1-Del1, Del2, Del3), as well as HMGA1 full length control (HMGA1-FL/WT), were co-transfected into 293T cells. The results indicated that apelin interacted with wild-type HMGA1 (Fig. 6a and Supplementary Fig.  4a ), but this interaction was sharply reduced in HMGA1 deletes 1 and 3, but not in delete 2 (Fig. 6a), which indicated that fragment 30-47 was the region important in binding HMGA1 to apelin.
In this region, Glu 32 was an important site that had undergone deamidation of glutamine into glutamate (Fig. 6a). In addition, expressions of many glutamine amidotransferases increased in A549-OE and H1975-OE cells ( Supplementary Fig. 4b,c), which suggested that apelin might increase the expression of glutamine amidotransferases, induce deamidation in HMGA1, and increase the binding of apelin to HMGA1.
To con rm this hypothesis, glutamine amidotransferase inhibitor (DON) was used to investigate whether deamidation modi cation regulated the binding of HMGA1 to apelin. We found that DON reduced binding of apelin to HMGA1 (Fig. 6c). To con rm the importance of deamidation modi cation in apelin-HMGA1 interaction, two HMGA1 mutants, Q32E (glutamic acid, Glu) and Q32A (alanine, Ala), at a deamidation site resistant to deamidated were generated. Substituting Gln-32 with Glu, not Ala, increased binding of apelin to HMGA1 in both 293T (Fig. 6d) and A549-OE cells (Fig. 6e), showing that HMGA1 deamidation increased apelin-HMGA1 binding. Furthermore, we also found that HMGA1-WT and Q32E increased the viability of A549-OE cells compared with Q32A (Fig. 6f), as well as promoted mRNA expression of lipid metabolism genes (Fig. 6g). SREBP1 activity also was elevated in A549-OE and 293T cells with transfected HMGA1-WT and Q32E (Fig. 6h).
When we studied the effect of endogenous deamidation modi cation in lung cancer cells, we found that DON inhibited glutamine amidotransferases expression (Fig. 6i, Supplementary Fig. 4d) in A549-OE and H1975-OE cells and restrained apelin-HMGA1 interactions (Fig. 6j). DON treatment also reduced the expression of lipid-related genes in A549-OE and H1975-OE cells (Fig. 6k and Supplementary Fig. 4e,f). Finally, adding DON markedly restrained the viability of A549-OE cells (Fig. 6l). All these results indicated that deamidation modi cation of HMGA1 increased the binding a nity of apelin and HMGA1, subsequently activated SRBEP1, and promoted SREBP1-responsive lipogenesis.
Deamidated HMGA1 Promoted the Formation of Apelin-SREBP1-HMGA1 Complexes We found that apelin can bind to SREBP1 and HMGA1 separately and that HMGA1 deamidation modi cation promoted the formation of an apelin-HMGA1 complexes and improved SREBP1 activity.
However, whether HMGA1 directly interacts with SRBEP1 is unknown. Immunoprecipitation revealed that HMGA1 could interact with SREBP1 in A549-OE (Fig. 7a) and H1975-OE cells ( Supplementary Fig. 5a), which indicated the formation of an apelin-SREBP1-HMGA1 complexes. Deamidated HMGA1 increased the activity of SREBP1, but whether it affected the binding of HMGA1 to SREBP1 was still unclear.
We found that si-HMGA1 and DON treatment increased the binding of SREBP1 to Insig1 (Fig. 7b). Furthermore, si-HMGA1 and DON treatment both inhibited the formation of an apelin-SREBP1-HMGA1 complexes, then suppressed SREBP1 activity and the expression of lipogenic-related molecules (Fig.  7c,d). It illustrated that HMGA1 was essential for formation of multi-proteins complexes.
To further explore the relationship between apelin, HMGA1, and SREBP1, the 293T co-transfection experiment con rmed that apelin and HMGA1 both directly interact with SREBP1, while apelin can enhance the combination with HMGA1 and SREBP1 (Fig. 7e). It was showed that apelin was key regulator to form of multiprotein complexes. To further con rmed whether deamidation modi cation promoted the formation of an apelin-HMGA1 -SREBP1complexes, it was showed multiprotein complexes was restrained in the presence of DON (Fig. 7f), like effect on A549-OE cell (Fig. 7c). The HMGA1-Q32E mutation increased the binding of SREBP1 to HMGA1( Fig.7g and Supplementary Fig. 5b) and dissociated SREBP1 from Insig1 in the presence or absence of apelin, whereas HMGA1-Q32A had no similar effect ( Supplementary Fig. 5c,d), indicating that deamidation modi cation of HMGA1 competitively combined with SREBP1 and reduced the binding of Insig1 with SREBP1. We further con rmed that the HMGA1-Q32E mutation promoted the formation of an apelin-HMGA1-SREBP1 complexes both in lung cancer A549-NC and -OE cells, whereas Q32A had no effect (Fig. 7h). This phenomenon further illustrated that deamidated HMGA1 was essential for regulating the activity of SREBP1 and the interaction SREBP1 with Insig1. Deamidated HMGA1 acted independently to regulate the SREBP1 activity. Deamidation modi cation of HMGA1 also reduced Insig1-SREBP1 interaction in A549 cell (Supplementary Fig. 5e). Taken together, deamidation modi cation strongly enhanced the formation of an apelin-SREBP1-HMGA1 complexes and increased of SREBP1 activity.

Discussion
Development and progression of cancer are frequently associated with increased de novo lipogenesis in tumor cells. In this study, we con rmed that apelin had greatly increased concentrations in tissues from lung cancer tumors and promoted tumor growth in lung cancer cell lines. We then determined that apelin promoted lung cancer proliferation by enhancing lipid synthesis. We also identi ed two important lipidregulating factors, HMGA1 and SREBP1, that, when combined with apelin, formed an apelin-HMGA1-SREBP1 complexes that increased lipid metabolism in lung cancer cells. We identi ed a new mechanism through which apelin promoted tumorigenesis. The theoretical model (Fig.8) illustrated that overexpressed apelin in lung cancer cells resulted in increasing of glucose uptake and weakened oxidative phosphorylation for enhancing lipogenesis. This occured, at least partially, due to apelinmediated deamidation modi cation of HMGA1. Apelin interacted with HMGA1 and regulated the deamidation modi cation of HMGA1. Deamidation of HMGA1 increased the binding of HMGA1 to apelin, subsequently bound to SREBP1 and inhibited the interaction SREBP1 with Insig1 on endoplasmic reticulum (ER). SREBP1 complexes dissociated from Insig1, transported to the Golgi where SREBP1 released the transcriptionally active N-terminal fragment and entered the nucleus with HMGA1 and apelin to promote the expression of lipid synthesis genes.
Although apelin had been reported to enhance proliferation of multiple tumor cells, the mechanism was still unknown. Apelin is secreted by adipocytes and affects several biological functions by regulating glucose metabolism and fatty acid oxidation 28,29 . Apelin stimulates glucose uptake by regulating AMPactivated protein kinase-(AMPK), MAPK/Erk-, and PI3K-Akt-dependent signaling pathways in normal tissues 21,30 . Here, we established that apelin induced abnormal lipogenesis by increasing the synthesis of both saturated (palmitate and stearate) and unsaturated fatty acids, as well as by up-regulating lipidrelated genes, such as GLUT4, FASN, ACC1, and SREBP1.
High-mobility-group A (HMGA1) protein is a nonhistone chromatin remodeling protein and is a transcription enhancer in elevating oncogene expression by interacting with DNA and protein in cancer cells 31 . The HMGA1 protein regulates, and is critically important to, tumor progression in diverse malignancies 32 . HMGA1 increases glucose uptake and ATP production by stimulating the transcription of the glucose transporter GLUT3 by binding GLUT3 promoter in colon cancer 25 . Sterol-regulatory element binding protein 1 (SREBP1) is an important transcription factor in lipid synthesis that stimulates the expression of lipogenic genes 33 . Inactive SREBP1 binds to insulin-induced gene (Insig1) and is located in endoplasmic reticulum, whereas activated SREBP1 translocate to the nucleus and increase lipid synthesis in malignant tumor cells 34,35 . The results of the current study demonstrated that apelin interacted with HMGA1, reduced binding of SREBP1 to Insig1, and promoted SREBP1 tra cking from the endoplasmic reticulum to the Golgi for activating SREBP1.
Our study revealed that apelin induced deamidation modi cation of HMGA1 and increased the binding of apelin to HMGA1, which activated SREBP1 and induced the translocation of SREBP1, leading to increasing lipogenesis.
Glutamine amidotransferases (GATs), an enzyme important in regulating deamidation modi cation, extracts nitrogen for sustaining cell growth and proliferation 36 . Tumor-associated deamidation promotes the proliferation of cancer cells by inducing metabolic reprogramming 22 . We found that deaminated HMGA1 acted as a new substrate of apelin in enhancing lipogenesis. The SREBP1 activity can be regulated by deamidated HMGA1 independently. Deamidated HMGA1 was su cient for SREBP1 tra cking and activation, and si-HMGA1 and glutamine amidotransferase inhibitor (DON) both restrained SREBP1-responsive lipid gene expression.
We also determined that HMGA1 Glu 32 was the most important deamidated-related region of HMGA1. The glutamate mutant of HMGA1-Q32D can enhance the binding to apelin and SREBP1 than wild-type HMGA1 and HMGA1-Q32A, increasing the expression of the lipid genes, including GLUT4, ACC1, and FASN, but the alanine mutant of HMGA1-Q32A had the opposite effect.
In conclusion, we identi ed an important mechanism of apelin in promoting the lung cancer cells proliferation. We also found that apelin, as well as the apelin-HMGA1-SREBP1 complexes, might be new candidate metabolic target for lung cancer therapy. Committee, Chinese Academy of Sciences (Shanghai, China). HBE and HEK293T were gifts from the transthoracic surgery department of West China Hospital of Sichuan University. HBE and HEK293T were maintained in Dulbecco's modi ed Eagle's medium (DMEM, HyClone) supplemented with 10% fetal bovine serum (FBS, HyClone). H1975, A549, and other lung cancer cells were cultured in RPMI 1640. All cell cultures were supplemented with 1% penicillin/streptomycin, and cells were incubated at 5% CO 2 at 37 ºC. Plasmids were transfected with a lip3000 Transfection reagent (Thermo Fisher) following the manufacturer's instructions.

Lentivirus-Mediated Stable Cell Line Construction
To establish an apelin-knockdown cell line, H460 cells were transfected with the apelin shRNA Lentiviral Particles (Santa Cruz). Controls were also transfected with shRNA Lentiviral Particles (Santa Cruz). Apelin was restored by infecting apelin-knock down H460 cell with lentivirus expressing full-length human apelin cDNA (Santa Cruz). H1975 and A549 were infected with puromycin-resistant lentiviral apelin particles. Stable integrants were selected and maintained with puromycin (2 μg/ml).

Plasmid Construction and RNAi Transfection
All plasmids were constructed by Gene Inc. Both siRNAs targeting apelin (sc-44741) and HMGA1(sc-37115) were purchased from Santa Cruz. Transfection assays for plasmids or si-RNA were performed using Lipofectamine 3000 (Life Technologies) according to the manufacturer's protocol.

Cell Viability and Colony-Formation Assays
Cell viability was assessed with a CCK8 kit in 96-well culture plates, according to the manufacturer's instructions. For delipidated media, cells were rst seeded in media containing regular 10% FBS, and the following day were switched into media containing 20% delipidated FBS (Gemini, Cat. 900-123) on treatment. Palmitate (Sigma Aldrich, Cat. P5585) rescue experiments were performed in delipidated FBS by cotreating cells, in 96-well plates, with palmitate conjugated to BSA. For colony-formation assays, cells were seeded at 800 per/well in 6-well culture plates. After 14 days, cells were xed with ice methanol for 30 min and washed with PBS three times, combined with crystal violet staining solution buffer, and incubated for another 30 min. Finally, the cells were rinsed with PBS, air dried, and photographed.

RNA Isolation and Real-Time PCR Analysis
Cells were lysed in TRIzol Reagent (Life Technologies) to extract total RNA according to the manufacturer's instructions. Total RNA was then reverse-transcribed to cDNA using the iScript cDNA Synthesis kit (Bio-Rad). The resulting cDNA underwent quantitative PCR with gene-speci c primers in the presence of SYBR Green PCR master mix (Bio-Rad) using a Step-One-Plus Real-Time PCR System (Bio-Rad). Some metabolism-relative genes primers were cited from reference 14 . All primers are listed in Table  S2.
Immunohistochemistry Staining and Immuno uorescence The Department of Pathology, West China Hospital, Sichuan University, provided specimens of primary lung tumor tissues and normal pulmonary tissues surgically resected from patients treated without radiotherapy and chemotherapy. Para n embedding, tissue sectioning (5-μm) for tissues and immunohistochemical staining were performed as previously described 3 . Staining intensity was classi ed as low or high. Then, 8-mm-thick, fresh frozen tissues were mounted on glass slides for lipid Oil Red O staining to assess tumor lipid contents according to the manufacturer's instructions (Baso Diagnostics).
For nile red staining, cells were added to 100 ul of Hoechst (0.5 mmol/l) and stained for 10 min in a 96well plates, then nile red buffer was added for 10 min. Cells were washed with PBS for 10 min and their uorescence intensity was measured (Bio-Tek). To measure immuo uorescence, cells (20 ´ 10 4 ) were placed on glass coverslips in media containing 24-well plates. After transfection, cells were xed with icemethanol for 30 min and then blocked with fetal calf serum for 1 hr. A secondary antibody was added and the cells were incubated for 1 hr after adding a primary antibody. Finally, cells were counterstained with DAPI.

Nuclear and Cytoplasmic Extraction
Nuclear and cytosolic extraction was performed using the NE-PER nuclear and cytoplasmic extraction kit (Abcam). Brie y, 1 ´ 10 6 cells were harvested with trypsin-EDTA and then centrifuged at 500×g for 5 min. Next, the cell pellets were washed with PBS and centrifuged at 500×g for 5 min. To get the cytoplasmic extracts, the supernatant was removed and discarded, and 100 μl of ice-cold 1X pre-extraction buffer was added to the cell pellet. The tube was vortexed vigorously for 10 s, incubated on ice for 10 minutes, and centrifuged at 16,000×g for 2 min. The supernatant containing the cytoplasmic fraction was transferred to a new tube. To get the nuclear extract, the pellet fraction was suspended in 100 μl of ice-cold NER buffer, the tube was vortexed vigorously for 15 s, and the sample was placed on ice and vortexed for 5 s every 3 min during a period of 10 minutes. After a nal centrifugation at 16,000×g for 5 min, the supernatant containing the nuclear fraction was transferred to a new tube. The cytoplasmic and nuclear extracts were collected and used for western blotting.

Immunoprecipitation and Western Blotting
Western blotting was carried out as described in reference 4 . In brief, cells were homogenized and lysed at 4°C in lysis buffer with protease and phosphatase inhibitors (Thermo Fisher #A32961). Cell lysates were sonicated and centrifuged at 12,000 ×g for 10 min at 4 °C, and the resulting supernatant was used for immunoblotting analysis. Protein concentrations in cell lysates were measured using Beyotime Protein Assay Dye Reagent. For co-immunoprecipitation, cells were transfected with the indicated expression plasmids for 24 to 48 hr. Whole cell lysates were prepared with NP40 buffer (Beyotime) supplemented with protease and phosphatase inhibitors. Cell lysates were sonicated, centrifuged, and pre-cleared with normal IgG and protein A/G beads for 3 hr to reduce non-speci c binding. Then, pre-cleared cell lysates were incubated with their respective antibodies overnight and with protein A/G beads for another 3 hr at 4°C. The agarose beads were washed extensively and eluted by boiling at 100°C for 5 min. The precipitates were then analyzed by immunoblotting.

Glucose Uptake
Glucose uptake was measured using the Glucose Uptake-Glo™ Assay kit (Promega #tm467) according to according to the manufacturer's instructions. Brie y, cells were seeded 8000 per/well in 96-well plates. and cultured in normal medium overnight.
The medium was removed, and cells were washed with 100 µl PBS to most e ciently remove glucose from the cell culture. Cells were added to 50 µl of the prepared 1mM 2DG per well and incubated for 10 min at room temperature. After adding the 2DG6P detection reagent, luminescence was recorded using a 0.3-1 second integration on a luminometer (BioTek).

Acetyl CoA Activity Assay
Acetyl CoA activity assay was performed using Acetyl CoA uorometric assay kit (BioVision, cat. k312-100) according to the manufacturer's instructions. Brie y, cells were collected and deproteinized using a perchloric acid/KOH (BioVision, cat. k808-200). Samples were added to 10 µl CoA quencher and quench remover to correct background. Finally, cells were added to a 50 µl reaction mix buffer to measure uorescence using Ex/Em=535/587 nm.

Metabolic Pro ling and Isotope Tracing
Isotope tracing experiments were performed as previously described 35  10 OD cholesterol -modi ed HMGA1siRNA and control siRNA was administered at the bottom of the tumor every 3 days when the tumor volume was about 500 mm3. One weeks after the injection, the mice were euthanized, and the tumors were removed. All animals were cared for according to the Association for the Assessment and Accreditation of Laboratory Animal Care. All animal procedures were approved by the Subcommittee on Research Animal Care at West China Hospital of Sichuan University.

Statistical Methods
The statistical methods are described in each gure. All analyses were performed using Graph Pad Prism 7. Data are given as means and standard errors. Groups which 2 groups, experimental vs control were compared with two-tailed, unpaired Student's t-tests. Asterisks in the gures indicate statistical signi cance (*p < 0.05; **p < 0.01; ***p < 0.001).

Data Availability
All data supporting the ndings of this study are available in the article and its supplementary material or from the corresponding authors on reasonable request. Competing Interests: The authors declare no competing interests.