Long non-coding RNA HCG18 promotes malignant phenotypes of breast cancer (BC) cells by HCG18/miR-103a-3p/UBE2O/mTORC1/HIF-1α positive feedback loop

Background: In recent years, a growing number of studies have reported that long non-coding RNAs (LncRNAs) play crucial roles in breast cancer (BC) progression and metastasis. Another study group of our research center reported that LncRNA HCG18 was one of the 30 upregulated lncRNAs in BC tissues related to normal tissues in TCGA database. However, the exactly biological roles of HCG18 in BC remains unclear. Method: qRT-PCR was used to detect the expression prole of HCG18 in BC tissues and cell lines. In vitro assays were used to evaluate the pro-tumor function of HCG18 in BC cells. Animal study were used to explore the role of HCG18 in vivo. Bioinformatic analysis, dual-luciferase reporter assay, RNA immunoprecipitation (RIP) assay and Chromatin Immunoprecipitation (ChIP) assays were used to investigate the regulatory relationship of HCG18, miR-103a-3p, UBE2O in BC. Results: HCG18 was upregulated in BC tissues and cells, and BC patients with high HCG18 expression tended to have poor prognosis. HCG18 could promote BC cells proliferation, invasion and provided BC cells with tumor stemness properties (CSPs) in vitro and facilitate tumor growth and lung metastasis in vivo. In terms of mechanism, HCG18 functioned as a miRNA sponge which positively regulated the expression of Ubiquitin-conjugating enzyme E2O (UBE2O) by sponging miR-103a-3p and our previous research achievement have already veried UBE2O could promote malignant phenotypes of BC cells through UBE2O/AMPKα2/mTORC1 axis. Furthermore, as a downstream target of HCG18/miR-103a-3p/UBE2O/mTORC1 axis, HIF-1α transcriptionally promoted HCG18 expression and then formed a positive feedback loop in BC. Conclusion: HCG18 played an oncogenic role in BC and it might serve as a prognostic biomarker and a potential target for BC survival established stable HCG18 KD OE BC cells employed in vitro in vivo assays to evaluate HCG18’s biological function. the certied HCG18 oncogene by promoting proliferation, invasion and providing CSPs in BC cells.


Background
Although early detection, high-quality prevention and advanced therapeutic strategies have been devoted into the clinical practice, breast cancer (BC) is still the most common female cancer diseases (30%) and the rst leading cause of cancer related death (15%) worldwide 1 . Actually, the high recurrence and metastasis characteristics are the main cause of high mortality in BC. It has been reported that the overall 5-year survival rate of patients in early stage of BC is 92.5%-85.9%, however, it is decreased to 25.1% for patients in stage 2 . Therefore, exploring promising detection markers and developing effective therapeutic methods are urgently needed in BC.
Long non-coding RNAs (LncRNAs) are a class of RNA transcripts which are longer than 200 nucleotides in length and have no or limited protein-coding capability 3 . Recent studies have con rmed that LncRNAs are involved in a variety of physiological and pathological processes of human diseases, especially cancers 4,5 . Another group of our research center identi ed LncRNA HLA complex group 18 (HCG18) was one of the 30 upregulated LncRNAs in BC by analyzing data from two cohorts in TCGA 6 . Previously studies have also reported that HCG18 has oncogenic functions in gastric cancer, hepatocellular carcinoma, lung adenocarcinoma and colorectal cancer. However, its biological function in human BC remains unclear [7][8][9][10] .
Ubiquitin-conjugating enzyme E2O (UBE2O) is a large E2-ubiquitin-conjugation enzyme which has both E2 and E3 activities 11 . Abnormal expression of UBE2O existed in many types of human cancers and deregulation of UBE2O plays crucial roles in tumor progression and metastasis [12][13][14] . Our previous research achievement demonstrated that UBE2O could promote BC cells proliferation, epithelialmesenchymal transformation (EMT) and conferred BC cells with cancer stemness properties (CSPs) depending on UBE2O/AMPKα2/mTORC1-MYC positive feedback loop 15 . However, the exactly regulatory mechanism of UBE2O in BC is still needed to be further investigated.
In this study, we con rmed that HCG18 was signi cantly upregulated in BC patients and had closely relationship with poor prognosis in BC patients. HCG18 could promote BC cells proliferation, invasion and conferred BC cells with CSPs in vitro and facilitate tumour growth and metastasis in vivo. In terms of mechanism, we found that HCG18 could competitively adsorb miR-103a-3p, indirectly facilitating UBE2O expression and thus activating the UBE2O/mTORC1 axis in BC cells. Furthermore, as a downstream target of mTORC1, HIF-1α could transcriptionally promote HCG18 expression and constituted a positive feedback loop in BC.

BC cells and specimens
All the BC cells and normal mammary cells (MCF-10A) were purchased from the cell bank of the Chinese Academy of Sciences (Shanghai, China) and were cultured according to the supplier's instructions. Fresh BC tissues, adjacent noncancerous tissues and corresponding para n embedded cancer tissues (n=120) were obtained from Harbin Medical University Cancer Hospital between 2012 and 2013. All the patients rerolled in our study had complete clinicopathological information and patients who received neoadjuvant chemotherapy, radiotherapy, immunotherapy and those with recurrent tumors, bilateral BC or other previous tumors were excluded. Our study was approved by the Research Ethics Committee of Harbin Medical University and all the patients enrolled in our study had signed informed consent.
RNA extraction and quantitative real-time PCR (qRT-PCR) Total RNA was extracted from fresh BC tissues and cells with TRIzol reagent (Invitrogen, USA) according to the manufacturer's instruction. Prime Script RT Reagent Kit with gDNA Eraser (Takara, Kyoto, Japan) was used to synthesize the cDNA and qRT-PCR was conducted using SYBR Premix Ex Taq™ II (Takara, Kyoto, Japan) on a CFX96 Touch Detection System (Bio-Rad, USA). Primers used in our study were listed in the supplementary table S1. GAPDH , U6 and 18S tRNA were used as internal controls, and relative RNA abundances were calculated by the standard 2 -△△CT method.

Western blotting
Western blotting assays were performed as previously study reported 15 . Brie y, indicated cells were lysed with RIPA (Beyotime, China) buffer containing PMSF (Beyotime, China). The lysate was collected and western blotting assays were then performed. Antibodies used were listed in the supplementary table S2.

Cell proliferation assay
For the CCK-8 assays, indicated cells were seeded into 96-well plates at a density of 2×10 3 cells per well. At each pre-set time, 10μl CCK-8 reagent (Beyotime, China) were added into each well containing 90 μl culture medium and the plates were incubated at 37 °C for 2 hours. After that, the cell viability was detected by measuring the absorbance at 570 nm.
For the colony formation assays, indicated cells were seeded into six-well plates (5×10 2 cells/well) and incubated with culture medium containing 10% FBS at 37 °C for 2 weeks. After that, the cells were xed with formalin for 30 minutes and stained with crystal violet. FluorChem M system was used to take the photos.
Edu staining assays were performed as follows: indicated cells were seeded into 96-well plates and cultured with 10% FBS medium at 37 °C. After the con uence reached to 60%-70%, the cells were subjected to Edu staining (Beyotime, China) according to the manufacture's instruction. After xing and permeabilizing, the cells were stained with Edu and DAPI solutions. The results were captured and analyzed under a uorescence microscope.

Scratch assays and cell invasion assays
For wound scratch assays, indicated cells were seeded into six-well plates and cultured at 37°C. After the cells reached to 95% con uence, sterile micropipette tips were used to made the scratch wounds. Then the cells were washed by PBS and cultured with medium containing 10% FBS continuously. Images were taken in schedule under a microscope at preset time and the migration rate were subsequently analyzed.
For cell invasion assays, the upper champers of 24-well transwell plates (Coring, USA) were coated with Matrigel and incubated at 37℃ for 4 hours. Then 200μl serum-free medium containing indicated cells were added into the upper champers of the transwell plates and 600 μl 10% FBS medium were added into the lower champers of the transwell plates. Then the cells were cultured at 37℃ for 24 hours. Afterwards, the non-invaded cells were scraped from the upper side of the champers and invaded cells at the underside were xed with 4% paraformaldehyde and stained with 1% crystal violet. Finally, the images were captured and the invaded cells were counted under a microscope.
Hematoxylin-eosin (HE) staining and immunohistochemistry (IHC) HE staining assays were performed according to a standard HE staining technique as previously study reported 16 . Brie y, mouse lung tissues were xed in 4% paraformaldehyde, embedded in para n and sectioned (5-mm thickness). After that, the lung sections were subjected to HE staining to evaluate pathological changes and photographed under an optical microscope.
IHC staining were performed on para n-embedded specimens from BC patients in a standard streptavidin-peroxidase complex manner. The staining results were evaluated with histochemistry score (H-score) by three senior pathologists independently. The following antibodies were used: UBE2O

Lentiviral production
Recombinant lentivirus for overexpression HCG18 or UBE2O in BC cells and corresponding negative controls were designed and synthesized from GenePharma (Shanghai, China). The recombinant lentivirus was transfected into indicated cells with polybrene (8μg/ml). Twenty-four hours after transfection, the cells were selected with puromycin and qRT-PCR was applied to detect the transfection e ciency.
To establish stable HCG18 or UBE2O knocking-down BC cells, shRNAs targeting HCG18 or UBE2O were designed from Sigma and transfected into indicated cells according to the manufactory's instruction. The shRNA sequences were listed in supplementary table3. qRT-PCR was used to detect the knocking-down e ciency.

Transfection of BC cells
Small interfering RNAs (siRNAs) targeting HIF-1α and its control group were designed from GenePharma. The microRNA mimics and inhibitor were synthesized by Sigma. For transfection, indicated cells were seeded into six-well plates (5×10 5 per well) and cultured at 37℃ overnight. Then the cells were transfected with corresponding victors using Lipofectamine 2000 transfection reagent (Invitrogen, USA) according to the manufacturer's instructions. Forty-eight hours later, the transfected cells were harvested for the follow-up experiments. The siRNA sequences were listed in supplementary information table3.

Subcellular fractionation
Nuclear and cytoplasmic RNA separation was conducted using PARIS™ Kit (Catalogue Number: AM1921, Invitrogen, USA) according to the manufacturer's instruction. qRT-PCR was used to analyze the expression of HCG18 in different fragments of BC cells.
RNA pull-down (RIP) assays RIP assays were performed with Magna RIP™ RNA-Binding Protein Immunoprecipitation Kit (Millipore, USA). Brie y, indicated cells were collected and lysed with RIPA. Then the lysate was subjected to RIP with an Anti-AGO2 antibody (Abcam, USA) according to the manufacturer's instruction. Finally, the samples were disposed with proteinase K (Merck, USA) and qRT-PCR was used to analyze the binding of HCG18 and miR-103a-3p in BC cells.

Dual-luciferase reporter assay
To investigate whether miR-103a-3p could directly bind to the 3'UTR region of HCG18 or UBE2O, dualluciferase report assays were performed using a double luciferase assay system (Promega, USA). In brief, 3'UTR region of HCG18 (wild /mutant type) or UBE2O (wild /mutant type) luciferase reporter plasmids and miR-103a-3p mimics were co-transfected into HEK-293T cells. Then the cells were lysed and luciferase assays were conducted according to the manufacturer's instruction. Fire y luciferase activity normalized to Renilla luciferase activity was used as an internal control.
To analyze whether HIF-1α could bond to the promotor region of HCG18, HCG18 promotor region (wild /mutant type) luciferase reporter plasmids and HIF-1α plasmids were co-transfected into HEK-293T cells. Forty-eight hours later, cells were harvested and dual-luciferase assays were performed. Fire y luciferase activity normalized to Renilla luciferase activity was used as an internal control.

Chromatin immunoprecipitation (ChIP)
ChIP assays were performed as previously study reported 15 . Brie y, indicated cells were treated with formaldehyde to crosslink target protein and chromatin and sonicated to an average length of 200-500bp on ice. Then a DNA depuration kit (Beyotime, China) was used to extract and clean the DNA fragments. After that, ChIP assays were conducted with an anti-HIF-1α antibody (Catalogue Number: 36169S, CST, USA) or IgG control. qRT-PCR was used to detect the promotor fragments of HCG18. The primers used in our experiment were listed in supplementary information.

Animal study
All the animal studies were approved by the Medical Experimental Animal Care Commission of Harbin Medical University and performed in Second A liated Hospital of the Harbin Medical University Laboratory. Animals used in our study were raised in speci c pathogen-free animal facilities (Temperature was maintained at 25℃ and a12 hour light-dark cycle) and provided free access to clean water and food. Animals were anesthetized with 1-3% iso urane and humanely sacri ced by CO 2 inhalation as previous studies reported 17 . For the tumorigenesis assay, six-week-old BALB/c Nude mouse (Vitalriver, Beijing, China) were randomly assigned into two groups (n=6), and then MDA-MB-231 NC or MDA-MB-231 HCG18-KD (5×10 5 ) were respectively injected into the mammary fat pads of mice. The tumor growth was measured every week and seven weeks after injection, the mice were humanely sacri ced and the tumors were extracted and xed in formalin solution. The tumor volumes were evaluated by the following formula: 1/2 (length × width 2 ). For the lung metastasis model, MDA-MB-231 NC or MDA-MB-231 HCG18-KD (5×10 5 ) were respectively injected into the tail vein of the six-week-old BALB/c Nude mouse. Seven weeks after injection, the mice were humanely sacri ced and the lungs were collected to count metastatic nodules and for HE staining.

Statistical analysis
All the data were presented as the means ± SDs from at least three independent experiments. Student's ttest or one-way ANOVA test were used to analyze the differences between groups. The chi-square test was used to analyze the relationship between HCG18 expression and the clinicopathological features of the BC patients. The Kaplan-Meier method and log-rank test were applied to draw the survival curves. P < 0.05 indicated statistical signi cance, which was evaluated by GraphPad Prism 8.0 software.

Results
1. The expression level of HCG18 was commonly upregulated in BC and was associated with poor progression of BC patients.
Previous study reported that HCG18 was upregulated in BC tissues related to normal tissues in TCGA database 6 . However, the biological function of HCG18 in BC is largely unknown. To investigate this, we rstly detected HCG18 expression pro le in BC tissues and corresponding normal breast tissues. The results showed that HCG18 was signi cantly upregulated in BC tissues in comparison with normal mammary tissues ( Fig.1a-b). Fig.1c exhibited that high invasive BC cells MDA-MB-231 hold the highest HCG18 expression level and the expression was relative lower in low-malignancy cells MCF-7. However, the mammary epithelial cell line MCF-10A hold the lowest HCG18 expression compared with the other BC cells. Then, we sought to assess the relevance of HCG18 expression and clinicopathologic features in BC patients. Results in Fig.1d and Table 1 con rmed that HCG18 was positively correlated with the histilogical and clinical stage, tumor size and axillary lymph node metastasis of BC patients and these had been further con rmed by data from lnCAR database (Fig.S1a-S1c). Fig.1e and Fig.1f showed that BC patients with high HCG18 expression had poor distant metastasis-free survival (DMFS) and overall survival (OS), and these had also been supported by survival information in Kaplan-Meier Plotter database ( Fig.1f-1h). Collectively, all these results revealed that HCG18 was commonly over-expressed in BC and patients with high HCG18 expression tended to have high risks of metastasis and worse progression.
2. Aberrant expression of HCG18 was associated with proliferation, metastatic and CSPs of BC cells.
To investigate the biological roles of HCG18 in BC cells, we employed HCG18 highly expressed cell Colony formation assays (Fig.2b) and Edu staining assays (Fig.2c) further con rmed this conclusion.
Then, Ki-67 (Ki-67 was regarded as an important index of tumor growth in BC) expression state was detected in BC tissues by IHC and chi-square test was used to analyze the relationship between HCG18 and Ki-67 expression. Results in table1 showed that there was a closely relationship between HCG18 expression and the status of axillary lymph node metastasis in BC patients, so we focused on the pro-metastasis function of HCG18 in BC cells. Wound healing assays showed that the migration ability was extremely decreased after knocking-down HCG18 in MDA-MB-231, and the migration ability of MCF-7 HCG18 was increased in comparison with the control group (Fig.3a). Matrigel invasion assays also exhibited the similar results ( Fig.3b). Western blotting assays revealed that the epithelial related protein E-cadherin was increased, but the mesenchymal marker vimentin and metastatic markers MMP2 and MMP9 were declined in MDA-MB-231 HCG18-KD cells. The opposite results were observed in MCF-7 HCG18 compared with the control group (Fig.3d).
In recent years, a growing number of studies have reported that CSPs played important roles in tumor survival, proliferation, invasion, immune escape and drug resistance. Then the following experiments were sought to explore the relationship between HCG18 expression and CSPs in BC cells. Sphere formation assays revealed that the sphere formation ability of MDA-MB-231 HCG18-KD was obviously declined compared with the control group, however the capability of CSPs was remarkably enhanced after overexpressing HCG18 in MCF-7 cells (Fig.3c). Western blotting assays showed that CSP markers CD44, ABCG2 and OCT4 were remarkably decreased in MDA-MB-231 HCG18-KD and increased in MCF-7 HCG18 compared with the control group (Fig.3e). Finally, lung metastasis mouse models were employed to investigate the pro-metastasis function of HCG18 in vivo. The results showed that mice injected with MDA-MB-231 HCG18-KD exhibited fewer metastasis nodes compared with the mice in control group (Fig.3e-3f). In conclusion, all these results con rmed that HCG18 could provide BC cells with CSPs in vitro and promoted BC cells metastasis both in vitro and in vivo.

UBE2O contributed to the tumor promoting function of HCG18 in BC cells.
E2s play important roles in regulating cellular post-translational protein modi cations and are involved in a wide range of key physiological or pathologic processes. Multiple studies have reported that aberrant E2s expression is commonly existed in a variety of human cancers. Some cancer related E2s could facilitate DNA repair, cell cycle progression and activate oncogenic signaling pathways, which promote cancer progression and metastasis 18 . Our previous study reported that ubiquitin-conjugating enzyme E2O (UBE2O) could promote the proliferation, EMT and provide BC cells with CSPs by degrading AMPKα2, thus activating mTORC1-MYC axis 15 . For this reason, we hypothesized there may be a potential interaction network between HCG18 and E2s in BC. Then a panel of cancer related E2s were detected in MDA-MB-231 NC /MDA-MB-231 HCG18-KD and MCF-7 NC /MCF-7 HCG18 cells. Coincidentally, we found that there was a closely positive association between HCG18 and UBE2O expression in the indicated cells ( Fig.4a and Fig.S1e-f). To further con rm this, correlation analysis was performed using GEPIA database.
The results revealed that HCG18 expression was positively correlated with UBE2O expression in BC (Fig.4b), and this correlation was further veri ed in clinical samples (Fig.4c-d). Next, we knocked down Then we explored the effect of UBE2O on the protumor function of HCG18 in BC cells in vitro. CCK-8 (Fig.5a), clone formation assays (Fig.5b) and Edu staining assays (Fig.5c) showed that over-expressing UBE2O could signi cantly reverse the declining proliferation ability of MDA-MB-231 HCG18-KD , and the enhanced multiplication capacity was abolished after knocking-down UBE2O in MCF-7 HCG18 cells. Fig.5d-5e and Fig.6a showed that the migration, invasion and sphere formation capabilities of MDA-MB-231 HCG18-KD+UBE2O were improved compared with the MDA-MB-231 HCG18-KD , and knocking down UBE2O in MCF-7 HCG18 remarkably destroyed these capabilities in comparison with the control group. Taken together, all these results above con rmed that UBE2O played a vital role in the cancer-promoting function of HCG18 in BC cells.

HCG18 promoted the expression of UBE2O by sponging miR-103a-3p in BC cells
To further identify the regulatory mechanism of UBE2O by HCG18 in BC cells, we rstly ascertained the cellular location of HCG18 in BC cells. Subcellular fraction assays exhibited that HCG18 were predominantly distributed in the cytoplasm of BC cells (Fig.6b), which was further con rmed by lncLocator database (Fig.S1g) and in accordance with previous studies reported 19,20 . Recently, many studies con rmed that cytoplasmic lncRNAs could affect the biological process of BC cells by modulating the expression of miRNAs in a competing endogenous RNA (ceRNA) dependent manner 15,21 . Therefore, we speculated that HCG18 could in uence the stability of UBE2O in a ceRNA mechanism. To validated this, Targerscan, miRwalk and starbase were used to predicted the potential miRNAs, and the results in Fig.6c revealed that there were three miRNAs (miR-30a-5p, miR-34a-5p and miR-103a-3p) connected with both HCG18 and UBE2O. Then we knocked down HCG18 in MDA-MB-231 and MCF-7 cells and qRT-PCR was applied to detected the expression change of the three miRNAs above. The results showed that compared with the control group, the expression level of miR-103a-3p was remarkably increased after knocking-down HCG18 both in MDA-MB-231 and MCF-7 cells. However, there were no expression changes of the other two miRNAs in the indicated cells (Fig.6d). We rstly explored the interaction between HCG18 and miR-103a-3p in the following experiments. Dual-luciferase assays revealed that transfection of miR-103a-3p mimics could signi cantly reduce the luciferase activity of HCG18-WT group but failed to affect that of the mutant one (Fig.6e). Anti-AGO2 RIP assays further con rmed their binding potential (Fig.7a). Then the expression level of miR-103a-3p was detected in BC tissues and corresponding normal tissues. The results exhibited that miR-103a-3p was obviously decreased in BC tissues (Fig.7b), and there was a negative relationship between miR-103a-3p and HCG18 in BC tissues (Fig.7c). Collectively, all these results revealed that HCG18 acted as a sponge for miR-103a-3p in BC cells.
Then we validated whether miR-103a-3p could mediate the function of HCG18 in promoting UBE2O expression. The results in Fig.7d showed that ectopic expression of miR-103a-3p by miRNA mimics obviously reduced UBE2O mRNA expression in the indicated cells. Dual-luciferase assays further revealed that overexpression of miR-103a-3p could decrease the luciferase activity of the wild-type UBE2O reporter but not the mutant one, which indicated that UBE2O was a directly target of miR-103a-3p (Fig.7e). Correlation analysis in Fig.7f identi ed that UBE2O expression was negatively correlated with miR-103a-3p expression in BC tissues. Taken together, all these results above claimed that HCG18 promoted UBE2O expression by sponging miR-103a-3p in human BC.

HIF-1α transcriptionally promoted HCG18 expression in BC cells.
Since we veri ed that HCG18 was upregulated in BC, the subsequent experiments were carried out to explore the regulatory mechanism resulting in the aberrant expression of HCG18 in BC. JASPAR database was applied to identify the potential HCG18 transcription factors and coincidentally, we found that there were four potential binding sites for HIF-1α in the promotor region of HCG18 (Fig.8a). Our previous study demonstrated that UBE2O could active mTORC1 signaling pathway by meditating AMPKα2 ubiquitination, and mTORC1 was associated with controlling HIF-1α expression, which in turn promoted cell growth and anabolism 15,22,23 . Therefore, HIF-1α was knocked down in MDA-MB-231 and MCF-7 cells. The e ciency of HIF-1α silencing were detected by qRT-PCR, as well as the relative HCG18 expression. The results in Fig.8b revealed that HCG18 was signi cantly declined after knocking down HIF-1α in the indicated cells. Correlation analysis showed that there was a positive relationship between HIF-1α and HCG18 expression in BC tissues (Fig.8c), and this had been further con rmed by GEPIA database (Fig.8d). Then four sets of luciferase reporter plasmids containing the wild-type or mutant type of HCG18 promotor region were constructed and dual-luciferase assays were performed to validate the potential HIF-1α binding sties. The results claimed that for site 2, transfecting HIF-1α plasmids could signi cantly increase the luciferase activity of WT group but not the mutant one. For the other three sites, however, there were no differences of luciferase activity in both WT and mutant groups after transfecting HIF-1α plasmids (Fig.7e). These results indicated that HIF-1α could bind to the promotor region of HCG18 (site2) and facilitated HCG18 transcription. This conclusion had been further con rmed by ChIP assays (Fig. 7f). Taken together, all these results above demonstrated that HIF-1α could promote HCG18 expression, thus forming a positive feedback in BC cells (Fig.9).

Discussion
From the results above, we can draw the following conclusion: (1) HCG18 was upregulated in BC tissues and patients with high HCG18 expression tended to have a poor prognosis.
(2) HCG18 could promote BC cells proliferation, invasion and CSPs in vitro and facilitate breast tumor growth and metastasis in vivo.
(3) HCG18 endowed the malignant phenotype of BC cells by sponging miR-103a-3p, indirectly enhancing UBE2O expression and activating mTORC1 signaling pathway. (4). As a downstream target of mTORC1 axis, HIF-1α could transcriptionally promote HCG18 expression and constitute a positive feedback loop in BC.
HCG18 was identi ed as one of the 30 upregulated LncRNAs in BC by analyzing data from two cohorts in TCGA database 6 . However, its exact expression pro le and biological function in BC remains undetermined. Therefore, we detected HCG18 expression in BC tissues for the rst time and the results were consisted with the result in TCGA database. We found HCG18 hold a high expression pro le in BC tissues and cells. Furthermore, we found that BC patients with high HCG18 expression had worse DMFS and OS, and these results were further con rmed by survival information from Kaplan-Meier Plotter database. Then we established stable HCG18 KD /HCG18 OE BC cells and employed both in vitro and in vivo assays to evaluate HCG18's biological function. All the results certi ed that HCG18 served as an oncogene by promoting proliferation, invasion and providing CSPs in BC cells. Previously studies also reported that HCG18 played tumor promotor roles in many other types of carcinomatosis. HCG18 could increase the expression level of WIPF1, DNAJB12 and activate PI3K/AKT axis in gastric cancer cells, thus promoting gastric cancer progression 7,9,10 . HCG18 could contribute to hepatocellular carcinoma progression by improving CENPM expression 8 . HCG18 could also act as an oncogene in lung adenocarcinoma through enhancing HMMR expression and accelerate nasopharyngeal carcinoma progression by upregulating CCND1 9, 24 . All these achievements above portended that HCG18 may be a potential diagnostic marker and a novel anti-tumor target for cancer therapy. Our research expanded our knowledge on the role of HCG18 in human malignant neoplastic diseases.
UBE2O is a large E2 ubiquitin-conjugation enzyme which acts as a combination of E2 and E3 enzymes and has both E2 and E3 activities 11 . Deregulation of UBE2O has been associated with several human diseases, especially cancers. UBE2O could mediate Mxi1 ubiquitination and then promote lung cancer progression and radioresistance 25 . It has been reported that UBE2O could decrease the stability of MLL in a polyubiquitination-dependent manner, which results in aggressive leukaemia 26  HIF-1α is a crucial regulator of metabolism and a well characterized oncoprotein in cancer cells 28 .
Through modulating a series of glycolytic enzyme gene expression, HIF-1α could stimulate angiogenesis, enhance aerobic glycolysis and mediate metabolic reprogramming, which in turn facilitates BC cells malignant transformation 29 . HIF-1α could also bind to the hypoxia response regions (HREs) in the promotor region of target genes and acts as a transcriptional factor in regulating multiple cancer-related genes expression 30 . Several LncRNAs participating in tumorigenesis and metastasis have been proved to be regulated by HIF-1α mediated transcriptional regulation, such as LncRNA EFNA3, LncRNA BCRT1 and LncRNA HITT 15,31,32 . In our study, we identi ed HIF-1α maybe a potential transcriptional factor of HCG18 and we found there were four potential HREs in the promoter region of HCG18 by JASPAR database.
Knocking down HIF-1α could suppress HCG18 expression in BC cells and there was a positive correlation between HCG18 and HIF-1α expression in both BC tissues and GEPIA database. ChIP and dual-luciferase reporter assay further con rmed that HIF-1α could bind to the speci c HREs in the promoter region of HCG18 and promote HCG18 expression in BC cells. This self-control positive feedback loop further highlights the signi cance of the HCG18/miR-103a-3p/UBE2O/mTORC1-HIF-1α axis in BC.

Conclusion
Our study con rmed that HCG18 functioned as a miRNA sponge to positively regulate UBE2O expression through sponging miR-103a-3p and subsequently mediated the malignant phenotypes of BC cells, thus playing an oncogenic role in BC progression. Our research provided a new theoretical basis for exploring the mechanism of the proliferation, invasion and CSPs in BC cancer. Collectively, our study rstly demonstrated that HCG18/miR-103a-3p/UBE2O/mTORC1-HIF-1α axis constituted a positive feedback loop in promoting proliferation, invasion and providing CSPs in BC. HCG18 could become a promising therapeutic target and prognostic predictor against BC.

Availability of data and material
The data in the present study are available from the corresponding authors upon reasonable request.

Competing interests
The authors declare that they have no competing interests.