Breast cancer is the most common cancer type in women. Long non-coding RNAs (lncRNAs) have been reported as potential new diagnostic markers, prognostic factors, and therapeutic targets in cancer. However, the specific roles and mechanisms of lncRNAs in breast cancer remain to be elucidated. Here we demonstrated the downregulation of lncRNA SNORD3A in breast cancer cells and tissues and verified its non-protein-coding property. SNORD3A overexpression had no effect on cell proliferation but specifically sensitized breast cancer cells to 5-fluorouracil (5-FU) in vitro and in vivo. Mechanistically, SNORD3A exerts its effect via enhancing uridine monophosphate synthetase (UMPS) protein expression. SNORD3A acts as a competing endogenous RNA for miR-185-5p, leading to UMPS protein upregulation. miR-185-5p overexpression disrupted the effect of SNORD3A on chemosensitization to 5-FU in vitro and in vivo. Moreover, Meis1 overexpression transcriptionally promotes SNORD3A expression, and Meis1 is downregulated in breast cancer cells and tissues. In breast cancer tissues, SNORD3A level positively correlates with Meis1 and UMPS protein levels, whereas miR-185-5p level negatively correlates with UMPS protein level. High SNORD3A transcript and Meis1 and UMPS protein levels predicts a better outcome, but high miR-185-5p level predicts a worse outcome in breast cancer patients receiving 5-FU-based chemotherapy. Our findings indicate that Meis1-regulated SNORD3A specifically sensitizes breast cancer cells to 5-FU via enhancing UMPS expression. The SNORD3A–UMPS axis may serve as a potential biomarker and therapeutic target to improve the efficacy of 5-FU-based chemotherapy for breast cancer patients.
Breast cancer is the most common cancer type in women, accounting for approximately 25% of new cancer cases1. Despite great advances in diagnosis and treatment strategies over the past decades, approximately 15% of the cancer-related deaths in women are due to breast cancer2. Breast cancer is a heterogeneous cancer type that exhibits a variety of histopathological characteristics and genetic alterations, with varying clinical outcomes. Based on intrinsic gene expression profiling, breast cancer can be divided into five major subtypes: luminal A, luminal B, HER2-positive breast cancer, triple-negative breast cancer (TN-BC), and normal breast-like subtype. TN-BC includes the basal-like and claudin-low subtypes3–5. The therapeutic strategy depends on the subtype and can include endocrine therapy, anti-HER2 targeting therapy, and chemotherapy. However, chemotherapy resistance remains a major obstacle for successful cancer treatment of all breast cancer types, resulting in recurrence, metastasis, and poor outcome6. Therefore, identifying strategies to overcome chemoresistance and increase chemosensitivity in breast cancer is critical to improve treatment efficacy.
Emerging evidence has revealed long non-coding RNAs (lncRNAs) as major regulators in both physiological and pathophysiological processes. LncRNAs are endogenous transcripts that are >200 nucleotides (nt) and lack protein-coding potential7. LncRNAs have been reported to be involved in a variety of biological activities, including proliferation, differentiation, migration, invasion, cell cycle, stem cell pluripotency, and lineage differentiation as well as cancer development and progression8–10. LncRNAs exert their roles at multiple mechanistic levels, such as regulating gene transcription, mRNA stability and translation, protein abundance and location, and chromatin and protein conformation11,12. Recent studies have revealed lncRNAs deregulation in a variety of cancer types and showed that lncRNAs can exhibit oncogenic function, cancer-suppressor function, or both13–16. Several lncRNAs have been shown to be involved in breast cancer. LncRNA NKILA binds nuclear factor (NF)-kB/inhibitor of kB (IkB) and masks the phosphorylation motif of IkB to inhibit IkB kinase-induced IkB phosphorylation and NF-kB activation, thus preventing breast cancer metastasis17. We previously reported that the lncRNA LINC01638 was highly expressed in the TN-BC subtype and LINC01638 activates MTDH-Twist1 signaling by preventing SPOP-mediated c-Myc degradation to maintain the epithelial–mesenchymal transition and cancer stem cell-like state of TN-BC cells12. However, the specific roles and mechanisms of lncRNAs in breast cancer are still not fully understood.
In this study, we investigated the expression and potential role of the lncRNA SNORD3A in breast cancer. We confirmed the downregulation of SNORD3A in breast cancer cells and tissues and demonstrated a novel mechanism by which SNORD3A regulates chemosensitivity to 5-fluorouracil (5-FU) in breast cancer.
MCF10A, MCF-7, MDA-MB-231, T47D, SKBR3, ZR7530, BT549, HCC1937, BT474, and HEK293T cell lines were obtained from ATCC (Rockville, MD, USA) and cultured under standard conditions in media containing 10% fetal bovine serum (Gibco, Carlsbad, CA, USA).
To establish stable transfectants, cell lines were transfected with pReceiver-Lv201 lentiviral vectors containing SNORD3A or miR-185-5p, EX-T1651-Lv217 lentiviral vector containing Meis1, and psi-LVRU6GP vector with UMPS short hairpin RNAs (target sequence for sh-1#: 5’- CCAAUCAAAUUCCAAUGCU-3’, sh-2#: 5’-GAGUUGAUAACUCUGGCAA-3’) using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) following the manufacturer’s instructions.
To inhibit miR-185-5p function, cells were transfected with miR-185-5p inhibitor (miRCURY LNA™ microRNA inhibitor for miR-185-5p; Exiqon, Vedbaek, Denmark).
Frozen fresh and paraffin-embedded breast cancer and non-cancerous tissues were collected from patients at the Affiliated Cancer Hospital of Guangzhou Medical University. All samples were collected with informed consent from patients, and all procedures were performed after the internal review and approval of the Ethics Committees of Guangzhou Medical University and the Affiliated Cancer Hospital.
HEK293T cells were co-transfected with various cloned MS2bs vectors (MS2bs, MS2bs-SNORD3A-WT or MS2bs-SNORD3A-Mut) and MS2bp-GFP overexpression vector (Addgene, Watertown, MA, USA). After 48 h, RIP was performed with the EZ-Magna RIP Kit (Millipore, Burlington, MA, USA) using anti-green fluorescent protein (anti-GFP) according to the manufacturer’s instructions. After RNA extraction, miR-185-5p level was examined by quantitative reverse transcriptase polymerase chain reaction (qRT-PCR).
Female BALB/c athymic nude mice were obtained from Guangdong Medical Laboratory Animal Center, China. Cell lines were injected subcutaneously into the armpit of female BALB/c athymic nude mice to generate xenograft tumors (n = 5/group). Ten days after cancer cell implantation, mice were injected intraperitoneally with 5-FU (30 mg/kg) every 3 days for six cycles. At the experimental endpoint, animals were sacrificed, and then tumors were harvested and weighed. The animal studies were approved by the Institutional Animal Care and Use Committee (IACUC) of Guangzhou Medical University. Standard animal care and laboratory guidelines were conducted according to the IACUC protocol.
qRT-PCR, proliferation assay, colony-formation assay, chemosensitivity assay, western blot, luciferase reporter assay, chromatin immunoprecipitation (ChIP)-qPCR assay, in situ hybridization (ISH), immunohistochemistry (IHC) methods and primers are described in Supplemental Experimental Procedures.
Data are presented as mean ± s.d. Student’s t test and χ2 test were used to compare the differences among different groups and correlation analysis. Survival curves were plotted using the Kaplan–Meier method and compared using log-rank test. Statistical analyses were performed using GraphPad Prism 6. p < 0.05 was considered statistically significant.
We first examined the expression level of the annotated potential lncRNA SNORD3A in a series of breast cancer cell lines and the normal mammary epithelial cell line MCF10A. The results demonstrated that SNORD3A expression was downregulated in breast cancer cells compared with MCF10A cells (Fig. 1a). We then examined the expression pattern of SNORD3A in paired breast cancer tissues and adjacent non-cancerous tissues. SNORD3A expression was also downregulated in breast cancer tissues compared with levels in paired non-cancerous mammary tissues (Fig. 1b).
SNORD3A is a small nucleolar RNA located on human chromosome 17. SNORD3A is composed of three exons, with a transcript length of 699 nt. We analyzed the SNORD3A transcript using the online software (RegRNA 2.0, LNCipedia 5.0, Coding Potential Calculator), which predicted no protein-coding potentiality of SNORD3A (Fig. 1c). CPAT online software analysis also predicted no protein-coding ability but predicted an open reading frame (ORF) in the SNORD3A transcript (Fig. 1c).
To determine whether the SNORD3A ORF has coding potentiality, we constructed a series of vectors with a mutation in the initiation codon of GFP (the start codon ATGGTG was mutated to ATTGTT) that was fused downstream of the full SNORD3A transcript or potential ORF (Fig. 1d). While GFP expression was detected in HEK293T cells transfected with the wild-type GFP vector, no substantial expression of GFP was observed in HEK293T cells transfected with full-length-GFPmut or ORF-GFPmut construct (Fig. 1d). Western blot analysis using the anti-GFP antibody further confirmed that SNORD3A lacks protein-coding ability (Fig. 1e). We also examined the subcellular location of SNORD3A and found that SNORD3A resides in both the nucleus and the cytoplasm of breast cancer cells (Fig. 1f). These data indicated that SNORD3A as an lncRNA is downregulated in breast cancer cells and tissues.
To explore the biological function of SNORD3A in breast cancer, SNORD3A was stably overexpressed in MCF-7 and MDA-MB-231 cells (Fig. S1a). Cell proliferation assays showed that SNORD3A overexpression had no significant effect on proliferation (Fig. S1b). Colony-formation assay also showed that SNORD3A overexpression had no significant effect on colony-formation capacity of breast cancer cells (Fig. S1c).
We then examined whether SNORD3A was involved in the chemosensitivity of breast cancer cells. SNORD3A-overexpressing MCF-7 and MDA-MB-231 cells and controls were treated with 5-FU, cisplatin (cDDP), or paclitaxel (PTX) at different concentrations. We found that SNORD3A overexpression specifically enhanced the chemosensitivity of breast cancer cells to 5-FU but not to cDDP and PTX (Fig. 2a and Fig. S2a). Both plate colony-formation and soft agar colony-formation assays also confirmed that SNORD3A overexpression specifically promoted the chemosensitivity of breast cancer cells to 5-FU (Fig. 2b, c and Fig. S2b, c).
The 5-FU antimetabolite drug has been widely used for the treatment of different types of cancer. To investigate whether SNORD3A enhanced the chemosensitivity of breast cancer cells to 5-FU via modulating a metabolic pathway for 5-FU, we examined the expression of 5-FU metabolic-related genes, including uridine monophosphate synthetase (UMPS), thymidylate synthase (TS), dihydropyrimidine dehydrogenase (DPD), and methylene tetrahydrofolate reductase (MTHFR). SNORD3A overexpression had no effect on the expression of TS, DPD, and MTHFR in MCF7 and MDA-MB-231 cells, whereas SNORD3A overexpression markedly increased the UMPS protein level but not the mRNA level (Fig. 2d and Fig. S2d). To determine whether SNORD3A enhancement of the sensitivity to 5-FU in breast cancer cells is dependent on UMPS, we stably knocked down UMPS in breast cancer cells with ectopic SNORD3A overexpression (Fig. 2e). Knockdown of UMPS significantly diminished the promoting effect of SNORD3A overexpression on 5-FU sensitivity (Fig. 2f–h and Fig. S3a–c). Together, these data suggest that SNORD3A may promote 5-FU sensitivity by activating UMPS expression in breast cancer cells.
Our results demonstrated that SNORD3A overexpression increased UMPS protein level but did not alter UMPS mRNA level in breast cancer cells. We next investigated the mechanism by which SNORD3A regulates UMPS. Recent studies showed that lncRNAs can function as competing endogenous RNAs (ceRNAs) by acting as molecular sponges for microRNAs (miRNAs), thus regulating gene expression at the posttranscriptional level. We used the online DIANA tool (http://diana.imis.athena-innovation.gr) and TargetScan (http://www.targetscan.org) to screen potential miRNAs that may bind with the SNORD3A transcript and the 3’-untranslated region (UTR) of UMPS mRNA. We found that miR-185-5p may bind with SNORD3A transcript and UMPS mRNA 3’-UTR (Fig. 3a). The interaction between SNORD3A and miR-185-5p was further predicted, and the minimum free energy at the binding site was calculated using RNAhybrid (Fig. 3b).
To determine whether SNORD3A interacts with miR-185-5p via the predicted binding sites, wild-type SNORD3A or SNORD3A with deletion mutations in the miR-185-5p targeting site were cloned into the MS2b plasmid to transcribe RNA combined with MS2-binding sequences and then co-transfected with the MS2bp-GFP expression plasmid and miR-185-5p-expressing construct into HEK293T cells. We then performed RIP assays to pull-down miRNAs associated with SNORD3A via GFP antibody, followed by qRT-PCR. The results showed that miR-185-5p associated with SNORD3A via the targeting site (Fig. 3c). In contrast, miR-191-5p, a negative control, did not associate with SNORD3A. To further confirm the interaction of SNORD3A with miR-185-5p, we constructed luciferase reporters containing wild-type SNORD3A or SNORD3A with deletion mutation of the miR-185-5p targeting site. Luciferase constructs were co-transfected with miR-185-5p plasmid into HEK293T cells. We found that miR-185-5p reduced the reporter activity of the construct with wild-type SNORD3A (Fig. 3d).
We next examined whether UMPS is a direct target of miR-185-5p. miR-185-5p inhibitor upregulated UMPS protein levels in MCF-7 and MDA-MB-231 cells (Fig. 3e). We also constructed luciferase reporters containing the wild-type 3’-UTR UMPS mRNA or construct with mutated miR-185-5p targeting sites. miR-185-5p reduced the reporter activity of the construct with the wild-type 3’-UTR of UMPS mRNA (Fig. 3f), suggesting that miR-185-5p negatively regulates UMPS mRNA.
We further found that miR-185-5p expression was upregulated in breast cancer cell lines compared with the normal mammary epithelial cell line MCF10A (Fig. 3g). We also determined the expression of miR-185-5p in the cohort of the paired breast cancer and benign adjacent tissues and found that miR-185-5p levels were higher in breast cancer tissues than that in the paired non-cancerous mammary tissues (Fig. 3h). Collectively, these data suggested that miR-185-5p is highly expressed in breast cancer and that SNORD3A acts as a sponge to miR-185-5p to promote UMPS expression in breast cancer cells.
Given that SNORD3A enhanced chemosensitivity to 5-FU depending on UMPS and that SNORD3A acts as miR-185-5p sponge to promote UMPS expression in breast cancer cells, we examined the involvement of miR-185-5p in SNORD3A-mediated chemosensitization to 5-FU. miR-185-5p was stably overexpressed in MDA-MB-231 and MCF-7 cells with SNORD3A overexpression. MTS assays demonstrated that miR-185-5p overexpression diminished SNORD3A -induced sensitization to 5-FU (Fig. 4a and Fig. S4a). Moreover, both plate colony-formation and soft agar colony-formation assays confirmed that miR-185-5p overexpression reversed SNORD3A -induced sensitization to 5-FU (Fig. 4b, c and Fig. S4b, c).
We further evaluated the biological role of SNORD3A and miR-185-5p in vivo. We subcutaneously injected MDA-MB-231 cells with SNORD3A overexpression combined with or without miR-185-5p overexpression or control MDA-MB-231 cells into nude mice. Mice were then treated with 5-FU as described in “Methods.” SNORD3A overexpression significantly enhanced chemosensitivity to 5-FU compared with control mice treated with 5-FU. The volume and weight of tumors derived from SNORD3A -overexpressing cells were significantly decreased compared with those of control xenografts in response to 5-FU (Fig. 4d). However, miR-185-5p overexpression disrupted the effect of SNORD3A overexpression on chemosensitivity to 5-FU (Fig. 4d). Together these in vitro and in vivo results indicated that miR-185-5p is involved in SNORD3A-mediated sensitization to 5-FU in breast cancer cells.
Given that SNORD3A expression was downregulated in breast cancer cells and tissues, we next explored the mechanisms underlying SNORD3A downregulation. To identify the transcription factors that regulate SNORD3A transcription, we used the JASPAR database (http://jaspar.binf.ku.dk) to predict potential transcription factor-binding sites in the 2-kb region upstream of the transcription start site in the SNORD3A gene. The analysis identified four potential Meis1-binding sites in the SNORD3A promoter (Fig. 5a). We further found that Meis1 expression was downregulated in breast cancer cells compared with the normal mammary cell line MCF10A (Fig. 5b). Meis1 level was also decreased in breast cancer tissues from the cohort compared with non-cancerous tissues (Fig. 5c). Correlation analysis indicated that Meis1 level was positively correlated with SNORD3A level in breast cancer tissues (Fig. 5d). To investigate the potential role of Meis1 in regulating SNORD3A expression, Meis1 was overexpressed in MCF-7 and MDA-MB-231 cell lines. Meis1 overexpression increased SNORD3A expression (Fig. 5e) as well as UMPS, the downstream target of SNORD3A (Fig. 5f). We next performed ChIP assays to confirm Meis1 occupancy on the SNORD3A promoter and found that Meis1 overexpression resulted in enrichment of Meis1 at the predicted Site C and Site D in the SNORD3A promoter in MCF-7 and MDA-MB-231 cells (Fig. 5g). Together these results suggested that Meis1 transcriptionally regulates SNORD3A expression in breast cancer cells.
To further define the role of SNORD3A and verify its correlation with the newly identified upstream regulator and downstream effectors in breast cancer clinical samples, we first examined SNORD3A and miR-185-5p levels via ISH in breast cancer tissues (N = 72). SNORD3A transcript was expressed at low levels in approximately 68.06% of the specimens, while miR-185-5p transcript was highly expressed in approximately 73.61% of the specimens (Fig. 6a). We performed IHC staining of Meis1 and UMPS proteins in breast cancer tissues that were examined by ISH. As expected, Meis1 and UMPS protein were expressed at low levels in approximately 77.78% and 65.28% breast cancer specimens, respectively (Fig. 6a). Further analysis showed that Meis1 level was positively correlated with SNORD3A level. UMPS level was positively correlated with SNORD3A level but negatively correlated with miR-185-5p level (Fig. 6b). Importantly, high SNORD3A transcript and Meis1 and UMPS protein levels predicts a better outcome, but high miR-185-5p level predicts a worse outcome in breast cancer patients receiving 5-FU-based chemotherapy (Fig. 6c).
We then used online tools to analyze the expression pattern and prognosis in breast cancer. Online analysis using UALCAN (http://ualcan.path.uab.edu) based on 1097 breast cancer tissues and 114 non-tumor mammary tissues indicated that Meis1 was downregulated in breast cancer tissues (Fig. 6d). We used a registration-free online service to generate Kaplan–Meier plots and found that high Meis1 expression in breast cancer predicted a better outcome in breast cancer (Fig. 6e), while high miR-185-5p predicted a worse outcome (Fig. 6f). There were no available records for SNORD3A in the online database. Together these data demonstrated the clinical significance of SNORD3A-mediated signaling in breast cancer patients.
In this study, we for the first time delineated the expression pattern, critical role, and mechanisms of the lncRNA SNORD3A in breast cancer. Our findings provide several insights into the underlying mechanisms and potential strategy for enhancing 5-FU chemosensitivity in breast cancer: (1) SNORD3A as an lncRNA is downregulated in breast cancer (2) resulting from downregulation of the Meis1 transcription factor and (3) SNORD3A acts as a ceRNA of miR-185-5p (4) to upregulate UMPS protein expression and specifically enhance chemosensitivity to 5-FU.
Emerging evidence has revealed the aberrant expression of lncRNAs in cancer and their critical roles in cancer development, progression, and prognosis18. Some lncRNAs have been reported to regulate important biological processes in cancer, such as proliferation19, metabolism20, cancer stem cells15, and metastasis21. LncRNAs deregulation has also been shown to be involved in the chemotherapeutic response and chemoresistance that is correlated with eventual cancer mortality22. Here we demonstrated low expression of lncRNA SNORD3A in breast cancer cells and tissues. SNORD3A overexpression had no significant effects on breast cancer cell proliferation and growth but specifically enhanced the chemosensitivity to 5-FU, validated by in vitro and in vivo studies. Importantly, higher SNORD3A level was positively correlated with clinical outcome in breast cancer patients receiving 5-FU-based chemotherapy. 5-FU is a pyrimidine analog that disrupts nucleoside metabolism and a widely used chemotherapy agent for some cancer types including breast cancer and colon cancer23. 5-FU is converted into its active metabolite fluorouridine monophosphate (FUMP), either directly by the orotate phosphoribosyltransferase domain of UMPS combining with phosphoribosyl pyrophosphate or indirectly via the activity of fluorouridine. FUMP is further phosphorylated and converted to fluorouridine triphosphate (FUTP). 5-FU induces cytotoxic effects and cell death through the incorporation of its active metabolite FUTP into RNA24. Abnormal alteration in the expression of 5-FU pathway genes directly correlates with chemosensitivity to 5-FU25. Here we demonstrate that SNORD3A enhanced chemosensitization via inducing UMPS expression at the protein level and showed that UMPS knockdown impeded SNORD3A.
Given that SNORD3A increased UMPS expression at the protein level but not at the mRNA level, we next explored the molecular mechanism by which SNORD3A regulates UMPS expression. Some lncRNAs function as ceRNAs to regulate the functions of miRNAs through miRNA response elements26. The lncRNA H19 promotes 5-FU resistance in colorectal cancer (CRC) by sponging miR-194-5p and regulating SIRT1-mediated autophagy27. The lncRNA HOTAIR promotes gastric cancer progression by sponging miR-331-3p to upregulate HER2 expression28. Exosome-transmitted lncARSR functions as a sponge of miR-34/miR-449 to induce c-MET and AXL expression to mediate sunitinib resistance in renal cell carcinoma29. In this study, we found that SNORD3A shared miR-185-5p response elements with UMPS and facilitated UMPS expression by sponging miR-185-5p. miR-185-5p was highly expressed in breast cancer, which predicted a poor outcome in breast cancer patients. UMPS was experimentally validated to be a bona fide target of miR-185-5p. Ectopic miR-185-5p expression overcame the chemosensitization induced by SNORD3A in in vitro and in vivo models, indicating that SNORD3A acts as a ceRNA for miR-185-5p in breast cancer cells.
Recent studies reported the tumor-suppressive roles of miR-185-5p in cancer progression. miR-185 function is blocked by miR155HG, resulting in ANXA2 upregulation and glioblastoma growth and progression30. In CRC, miR-185 downregulation by TCF1/LEF1 contributed to DC-SIGN-induced cancer metastasis31. Interestingly, in non-small cell lung cancer, miR-185 inhibited cancer growth and invasion but increased chemosensitivity32. Here we showed that SNORD3A enhanced the sensitivity of breast cancer cells to 5-FU via sponging miR-185-5p and that miR-185-5p acts as a suppressive miRNA of UMPS to mediate the insensitivity to 5-FU. Thus miR-185-5p may exhibit different effects in cancer progression and the therapeutic response. The precise effects of miR-185-5p and detailed mechanisms in cancer should be further validated by independent cohorts and prospective studies.
We also explored preliminary the mechanism for SNORD3A downregulation in breast cancer. Bioinformatics analysis combined with experimental validation indicated that Meis1 was downregulated in breast cancer and positively regulated SNORD3A expression at the transcriptional level. Clinically, low Meis1 level predicts a poor prognosis in breast cancer patients. Meis1, a 3-amino acid loop extension class transcription factor, activates its target genes via interacting with Hox transcription factors33. Meis1 plays important roles in tumorigenesis, such as in the development of neuroblastomas34, ovarian carcinomas35, and leukemia36. Sebastian et al. reported that Meis1 enhances Syk signaling via downregulating miR-146a expression in Hoxa9-driven acute myeloid leukemia37. The detailed molecular mechanisms for Meis1 transcriptionally regulating SNORD3A will be investigated in our next studies.
In conclusion, our study delineates the expression pattern, role, and mechanism of the lncRNA SNORD3A in breast cancer. SNORD3A is downregulated in breast cancer resulting from Meis1 downregulation. SNORD3A enhances UMPS expression by sponging miR-185-5p to specifically promote chemosensitivity to 5-FU in breast cancer. Our findings provide insight into the SNORD3A–UMPS axis as a promising therapeutic target against breast cancer, with important translational implications for improving the efficacy of 5-FU for breast cancer patients. Further studies should be performed to develop precise strategies targeting SNORD3A–UMPS signaling.
Supplementary Information accompanies this paper at (10.1038/s41419-020-2557-2).
This study was supported by grants from the National Natural Science Foundation of China (No.81672616 and No.81872197); grants from Guangdong Natural Science Funds for Distinguished Young Scholars (No.2016A030306003); Guangdong Special Support Program (No.2017TQ04R809); grants from Guangdong Natural Science Funds (2017A030313867), supported by Guangzhou key medical discipline construction project fund; Medical Scientific Research Foundation of Guangdong Province of China (No.B2018274); and grants from Science and Technology Program of Guangzhou, China (No.201710010100).