Exosomes deep sequencing identi es miR-363-5p as a non-invasive biomarker of axillary lymph node metastasis and prognosis in breast cancer


 Background: Breast cancer is the most common malignant tumor in females. Axillary lymph node (ALN) metastasis is an independent risk factor of prognosis and correlates with distant metastasis. However, the false-negative rate and complications of the standard procedure are not neglectable hitherto. It is necessary to develop an accurate non-invasive approach for lymph node staging.Methods: In this study, circulating exosomal miRNA profiles in peripheral blood from 10 patients with breast cancer and 10 age-matched healthy women were obtained and analyzed using small RNA deep sequencing. Integrative profiles analysis and multiple cross-validation using in-house and external independent datasets were performed to evaluate the diagnostic performance. Functional assays were used to analyze cellular phenotypes and downstream targets of miR-363-5p.Results: We found that the aberrant expression of exosomal miR-363-5p is significantly associated with tumorigenesis (p=0.047) and ALN metastasis (p=0.019). Serum exosomal miR-363-5p demonstrated the consistent down-regulated expression pattern in LN positive patients compared with LN negative patients across multiple independent datasets, which achieved high diagnostic performance in predicting lymph node metastasis (AUC=0.621-0.958). Furthermore, patients with a high expression level of miR-363-5p had significantly improved overall survival (HR=0.63, ±95% CI=0.45-0.89, p=0.0075). Functional experiments discovered that miR-363-5p acts by inhibiting the migration ability of breast cancer cells. Here we further identified Platelet Derived Growth Factor Subunit B (PDGFB) as a downstream target of miR-363-5p.Conclusions: The miR-363-5p deficiency promoted metastasis via facilitating PDGFB expression, leading to overactivity of PDGF signaling in cancer cells. These results demonstrated that tumor metastasis is mediated by tumor-derived exosomes that affect cell growth signal regulation. Therefore, exosomal miR-363-5p may serve as a marker for ALN metastasis diagnosis in a non-invasive manner.


Introduction
Breast cancer is the most common malignant tumor in females with a global annual incidence of 266,120 (30%) and 40,920 (14%) deaths [1]. Axillary lymph node (ALN) metastasis is one of the most important independent risk factors of the prognosis of early breast cancer [2]. Sentinel lymph node biopsy (SLNB) and ALN dissection are two major procedures for the ALN status evaluation and treatment for ALN metastasis. Currently, SLNB is recommended as the standard approach for ALN evaluation in clinically node-negative breast cancer patients [3]. Although unnecessary axillary clearance procedures might be spared, sentinel lymph nodes could have a false negative rate of 7.3%, which leads to patient under-treatment and causes an increased risk of recurrence [4,5]. Additionally, the presence of SLNB complications especially lymphedema is still inevitable. The incidence rate is 3.5% in breast cancer patients who received SLNB alone without ALN dissection [6]. And thus, it is highly necessary to develop an accurate and non-invasive method to identify patients at low risk of ALN metastasis before surgery.
Patients without ALN metastasis would bene t vastly if SLNB could be avoided safely using pre-surgery ALN status evaluation.
Exosomes are 30 to 100 nm microvesicles formed in multivesicular bodies and released into the extracellular environment by most cell types [7]. Abundant studies have shown that exosomes can serve as mediators of cell-to-cell communication by delivering cargo molecules especially nucleic acids that regulate the tumor microenvironment and promote cancer metastasis and progression [8][9][10]. Meanwhile, serum exosomes of cancer patients were proved to have a higher concentration than healthy individuals and considered as reliable markers in cancer diagnosis [11,12]. Within the serum exosomes, there are stuffed with a selection of mRNAs, microRNAs (miRNAs), long non-coding RNAs (lncRNAs), proteins, and lipids [13]. The miRNAs are small non-coding RNAs that regulate the cellular process by suppressing target mRNA translation and present with a high level in an exosome [14]. Recent studies have shown that exosomal miRNAs exhibit essential biological effects on tumor metastasis [15,16]. Besides, there is growing evidence suggesting the potential role of exosomal miRNAs in early detection of cancer metastasis [13,17,18]. However, the relation between tumor-derived exosomal miRNAs and the lymph node metastases in breast cancer is still unclear. Meanwhile, biomarkers based on serum exosomes for clinical applications were underdeveloped.
Here, we conducted a prior study to investigate the potential use of circulating exosomal miRNAs in the detection of lymph node metastasis (LNM). In this study, small RNA deep sequencing (RNAseq) analysis was used aiming to characterize the miRNA expression landscape in the serum exosomes from breast cancer patients. The candidate miRNAs responsible for breast cancer LNM were generated by comparing the miRNA expression difference between patients with or without LNM. Additionally, candidate miRNAs were veri ed in multiple independent patient datasets. Furthermore, in silico and experimental studies were performed to identify potentially relevant target genes of the candidate miRNA for improving our understanding of mechanisms underlying the LNM in breast cancer.

Materials And Methods
Patient enrollment and sample preparation All participants were enrolled through the Genetic investigation of Inherited and Familial Tumor Syndrome (GIFTS) study between January 2018 and June 2018 from the Cancer Hospital, Chinese Academy of Medical Sciences (CHCAMS). Patients were eligible for enrollment if they had an evident histologic diagnosis of breast cancer and without distant metastasis. Age-matched healthy women were also recruited as a control group. Peripheral blood samples of 10ml from these breast cancer patients and ten age-matched healthy women were collected at CHCAMS. Blood samples were collected in vacuum tubes with EDTA and centrifugated at 3,000 × g for 15 minutes at 4℃, the collected supernatant (5 ml plasma) was preserved at -80 °C before use. All participants signed written informed consent. Ethics approval for the study was obtained from the Research Ethics Committee of CHCAMS.

Exosome isolation
The collected plasma was thawed at 37℃ and then centrifugated at 3,000 ×g for 15 minutes to remove cell debris. Aspirated supernatant was diluted 7-fold with PBS and centrifuged at 13,000 ×g for 30 minutes [19]. Large particles were removed using 0.22 μm lters. Then the collected supernatant was ultra-centrifuged at 100,000 ×g, 4℃ for 2 h (CP100NX; Hitachi, Brea, CA, USA). The pellet containing exosomes was re-suspended in PBS and ultra-centrifuged again at 100,000 ×g 4℃ for 2 h. The isolated exosomes were re-suspended in 100 µl PBS after PBS washing for further analysis.

Exosomes Characterization
The nanoparticle tracking analysis (NTA), transmission electron microscopy (TEM), and western blot analysis using rabbit polyclonal antibody CD63, TSG101, and calnexin were conducted following the previously reported protocols [20].

Exosomal RNA isolation and RNA analyses
The RNAs were extracted from plasma-isolated exosomes using the miRNeasy® Mini kit (Qiagen, cat. No. 217004). RNA yields as well as DNA contamination were monitored on a 1.50% agarose gel. The NanoDrop 2000 spectrophotometer (Thermo Fisher Scienti c, Wilmington, Germany) was used to assess RNA concentration and purity. The integrity and distribution of RNAs were analyzed using the Agilent Bioanalyzer 2100 system with RNA Nano 6000 Assay Kit (Agilent Technologies, CA, USA).

Library preparation and sequencing
First, a total amount of 5 ng RNA per sample was depleted of rRNA using the RiboZero magnetic kit (Epicentre, Madison, Wisconsin, USA). Then, the sequencing libraries were generated using the Ovation® RNA-Seq System (NuGEN, CA, USA). A total amount of 2.5 μg RNA per sample was used as input material for sample preparation of small RNA libraries. Then, the libraries were generated using the NEB Next Multiplex Small RNA Library Prep Set for Illumina (NEB, USA). The index codes were added to attribute sequences to each sample. Finally, the polymerase chain reaction (PCR) products were puri ed using the Agencourt Ampure XP system (Beckman Coulter, Brea, CA, USA). The the library quality was evaluated on an Agilent Bioanalyzer 2100 (Agilent Technologies, Palo Alto, CA, USA) and quantitative PCR. The cluster of the index-coded samples was generated by the acBot Cluster Generation System using TruSeq PE Cluster Kitv3-cBot-HS (Illumina, San Diego, CA). At last, the sequencing was performed on the Illumina Hiseq platform using the library preparations and paired-end reads were generated.
Subsequently, the rRNA, transfer RNA, small nuclear RNA, small nucleolar RNA, and other non-coding RNA were ltered. Then, the miRNA including known miRNAs and novel miRNAs were detected using the remaining reads, in which the novel miRNAs were predicted according to the miRbase database and Human Genome (GRCh38), respectively. Read counts of the miRNAs were generated from the mapping results and have been standardized as the total mapped reads (TPM) per million. Circulating exosomal miRNA pro les of samples with two conditions were compared by the Student's t test (two-tailed), and each miRNAs with a log 2 |fold change|>0.58 and p <0.05 was considered as differential expression.
Hierarchical clustering was performed with R package 'pheatmap' using the ward.D2 method.

Cell proliferation assay
The MCF-7 cells were planked in 96-well plates with a density of 5 × 10 3 cells/well. The proliferation of cells at 0h, 24h, 48h, and 72 h after transfection was examined using the CCK-8 proliferation assay kit (MA2018-L, Meilunbio, China). Ten microliters of CCK-8 reagent was added to the medium at every time phase. Absorbance at 450nm was measured after 3 h incubation using microplate spectrometer reader (Molecular Devices, CA, USA).

Transwell migration assay
After transfection with miR-363-5p or NC for 48 h, MCF-7 cell was washed twice with FBS-free medium, and then re-suspended in FBS-free medium at a density of 1×10 5 cells/ml. Transwell chamber (pore size 8.0 μm, 3422, Corning Costar, MA, USA) pre-treated with the FBS-free medium was placed in a 24-well plate. After removing the pre-treatment medium, the lower chamber was added with 600 μL 10% FBS containing medium, while the upper chamber was added with 100 μL cell suspension. After incubating for 48 h, the chambers were xed and stained with methanol and 0.2% crystal violet. After staining, cells on the chamber surface were removed carefully with water and cotton swabs. The number of perforated cells in the outer layer of the basement membrane of each chamber (migrating cells) was counted in ve random high power elds with phase-contrast microscope (NIB-100F, Novel, China).

Statistical analysis
Analyses were performed with R Statistical Software (version 3.5.3). Pre-set p<0.05 was de ned as statistically signi cant. Quantitative data were measured as mean ± standard deviation. The comparison of mean values between two groups was analyzed using Student's t test and Mann-Whitney U-test. Pearson's test was used to evaluate the exosome-tissue miRNA correlation and miRNA-target mRNA correlation. Receiver Operating Characteristic (ROC) curve analysis was used to determine the diagnostic performance, and the area under the curve (AUC) were calculated with R package 'ROCit' [24]. Kaplan Meier method and log-rank test were applied to compare survival differences using R package 'Survival'.

Result
Characterization of exosomes from the serum of breast cancer patients In this study, 10 breast cancer patients and 10 age-matched healthy women were enrolled. Clinical information of the patients was listed in Table 1. 10 breast cancer patients were further divided into two groups according to their lymph node status, namely four patients with LNM and six patients without LNM. Blood samples were collected from both breast cancer patients and healthy controls. The integrity of exosome preparation was con rmed with TEM followed by western blot. The exosomes isolated from the serum exhibited the classic cup-shaped morphology under TEM ( Supplementary Fig. 1A). Exosome markers TSG101 and CD63 expression were detected from the exosome isolated from the serum ( Supplementary Fig. 1B). The NTA indicated that the average size of the vesicles was 105.7nm, and the main peak of particle diameter was at 85.5nm (Supplementary Fig. 1C). The above-mentioned results demonstrated that the extracellular vesicles isolated from serum samples are purified exosomes.

RNA-seq identi ed dysregulated exo-miRNAs in breast cancer patients
To identify exo-miRNAs that play a pivotal role in inducing breast cancer LNM, circulating exosomal miRNAs were isolated and pro led using small RNA deep sequencing analysis. A total of 1631 miRNAs were mapped in exosomes isolated from serum samples. To minimize noise and improve accuracy, the miRNAs with TPM values less than 5 were removed, leaving 367 miRNAs for further analysis. Through differential expression analysis, 43 signi cantly differentially expressed (DE) miRNAs were identi ed in breast tumor exosomes and 7 signi cantly DE miRNAs in breast tumor exosomes with LN positive status. Fig. 1A shows a different expression pattern of the 43 DE miRNAs between healthy and breast cancer patients. Fig. 1B shows a different expression pattern of the 7 DE miRNAs between breast cancer patients with and without LNM.
Identi cation of serum exosomal miR-363-5p as a potential biomarker of axillary lymph node metastasis and prognosis Integrative pro les analysis indicated that the aberrant expression of exosomal miR-363-5p is signi cantly associated with both tumorigenesis (p=0.047) and ALN metastasis (p=0.019) (Fig. 2A).
Exosomal miR-363-5p expression was signi cantly higher in breast cancer patients compared with healthy controls and was signi cantly lower in LN positive patients compared with LN negative patients (Fig. 2B). Since the miRNA concentration in exosomes is distinctively related to its cellular abundance [25], we also hypothesized that reliable circulating markers should coordinate with their expression alterations in tumor tissues. To verify the reliability of miR-363-5p, the expression levels of miR-363-5p in tumor tissues with LN positive and LN negative status were compared in four external independent datasets (i.e. three GEO datasets GSE38167, GSE31309, GSE42072, and The Cancer Genome Atlas [TCGA] dataset). The expression pattern of miR-363-5p in tumor tissues was identical to that in exosomes in in-house and four external independent datasets (Fig. 2C). Patients with LN positive status had a signi cantly lower miR-363-5p expression in GSE38167 (p=0.013), GSE31309 (p=0.136), GSE42072 (p=0.123) and TCGA (p=0.014) datasets. We subsequently pro led the expression levels of the miRNAs in tumor tissue of the 10 breast cancer patients (in-house data) using qRT-PCR. MiR-363-5p expression in breast cancer tissue was signi cantly lower in LN positive patients (Fig. 2D). Additionally, our in-house data showed the exosomal concentrations of miR-363-5p correlated with its expression in tumor tissue (Pearson's r=-0.679 and p=0.0307, Fig. 2E). These validation analyses indicated the potential of miR-363-5p as a potential and stable non-invasive biomarker for further investigation Performance evaluation and validation of miR-363-5p in the in-house set and multiple independent datasets In order to retrospectively evaluate the predictive power of exosomal miR-363-5p on detecting LN metastasis, we performed ROC analysis and found that the miR-363-5p achieved high diagnostic performance with AUC of 0.958 for in-house data and 0.733, 0.621 and 0.639 in GSE38167, GSE31309, and GSE42072 (Fig. 3A), respectively. These results indicated that low miR-363-5p expression levels could be a new diagnostic marker for breast cancer LN metastasis. Furthermore, we assessed the association between miR-363-5p expression level and survival of breast cancer patients and found that patients with the low expression level of miR-363 had signi cantly worse overall survival (p=0.0075, Fig.  3B). Moreover, in patients with negative lymph nodes upon the rst diagnosis, low expression of miR-363 in primary tumors correlates with the signi cantly worse outcome (p=0.00094, Fig. 3C). Survival analysis indicated that the downregulation of miR-363 could serve as a prognostic marker of poor survival in breast cancer. miR-363-5p inhibits metastatic properties of breast cancer cell The miR-363-3p and miR-363-5p (miR-363*) are both the mature forms of miR-363. Previous studies have focused on the biological function and pathophysiological signi cance of miR-363-3p, but little explored the role of miR-363-5p, possibly because of its relatively low abundance compared with miR-363-3p. To investigate the role of circulating exosomal miR-363-5p in breast cancer progression, we hypothesized that miR-363-5p in uence breast cancer cell mobility. We selected an ER-positive breast cancer cell line MCF-7, and transfected breast cancer cells with liposomal vectors overexpressing miR-363-5p or negative control (NC). Migration assays showed that overexpressing miR-363-5p signi cantly represses migration ability and silencing miR-363-5p results in intensi ed cell migration (Fig. 4A & Fig. 4B). The cell viability showed no signi cance in overexpression or knock-down of miR-363-5p (Fig. 4C).
Exosomal miR-363-5p modulates platelet-derived growth factor (PDGF) signaling activity by targeting PDGFB To identify reliable targets of miR-363-5p, we utilized both experimentally validated miRNA-target interaction databases and co-expression analysis (Fig. 5A). We analyzed miRNA and mRNA expression pro les of the TCGA breast cancer dataset and yielded four mRNAs co-expressed with miR-363-5p based on the negative regulation of target gene expression and miRNA level. We also retrieved gene lists of experimentally established targets of mir-363-5p from two databases (mirTarbase and Tarbase). We merged the results from two databases and yielded a list of 234 target genes. Among them, PDGFB was the only one exhibiting a signi cant negative correlation (Pearson's r=-0.208, p<0.001) with the miR-363-5p level in breast cancer tissues from TCGA (Fig. 5B). The target location (Fig. 5C) was provided by the previous study, validated using PAR-CLIP [26]. Thus, it was selected as a potential target for further functional target characterization. We next performed the qRT-PCR for validation. Consistent with the bioinformatics analysis, qRT-PCR analysis also showed that PDGFB expression is signi cantly downregulated by miR-363-5p overexpression, which is subsequently rescued by miR-363-5p knockdown as well (Fig. 5D). These findings indicated that miR-363-5p regulates PDGFB expression in breast cancer. miR-363-5p de ciency promoted metastasis via facilitating PDGFB expression, leading to overactivity of PDGF signaling in cancer cells.

Discussion
Assessment of miRNA expression signatures in exosomes is a promising tool for cancer research and clinical diagnosis, here in this study we report the different miRNA signature and identi ed several deregulated miRNAs in breast cancer patients with ALN metastasis compared with those without ALN metastasis. We identi ed that the level of exosomal miR-363-5p in ALN positive breast cancer patients was signi cantly lower than that in ALN negative patients.
Evaluation of miRNA expression in tumor tissues is necessary since the parallel down-regulation acts as the logical foundation of a tumor-derived diagnostic marker, and indispensable for mechanism interpretation as well. We investigated miR-363-5p level in breast cancer tissue of both in-house patients and external datasets. The results were consistent with that of serum exosomes. Those patients who were diagnosed with LN positive breast cancer have a signi cantly lower level of miR-363-5p. These results indicated that deregulated exosomal miR-363-5p level is associated with transcriptional changes in primary tumor tissue. And these changes substantially contribute to LNM in breast cancer.
We performed in silico diagnosis test and veri ed that miR-363-5p alone has an AUC of 0.621-0.958 in predicting LN metastasis in multiple independent datasets. Previous studies have shown that imaging approaches namely axillary ultrasound and MRI have similar performance in ALN staging, and the AUC of MRI alone was 0.665 [27,28]. We deem that exosomal miR-363-5p can help elevate the accuracy of clinical prediction models if taken into consideration. Furthermore, survival analysis revealed that patients with lower miR-363-5p have a signi cantly worse prognosis, especially in node-negative patients at their initial diagnosis. This suggests that patient strati cation using miR-363-5p can help distinguish individuals with a high risk of breast cancer death. Node-negative patients with low miR-363-5p level might consider adjuvant endocrine therapy.
In this study, we also investigated the functional signi cance of miR-363-5p. We found that restoration of miR-363-5p using mimics signi cantly inhibited breast cancer cell migration, while it seemed proliferation was not affected. Studies have revealed that low miR-363 expression is associated with carcinogenesis and metastasis [29,30]. Overexpression of miR-106a-363 cluster (miR20b, miR-363-3p and miR-363-5p) exhibited an anti-proliferative effect on cancer cells [31]. Our data indicated that among them miR-363-5p does not cause proliferation changes, but only migration ability. The previous study constructed a prognosis model of node-negative patients mainly based on receptor status and tumor size [32]. MiRNA signature can provide distinct tumor information on its cellular and molecular characteristics and would increase the accuracy of clinical prediction models. Our functional study de ned the basic rationale and supported miR-363-5p as a speci c complementary predictor for LN metastasis as well as patient prognosis.
It has been reported that miR-363-5p modulates endothelial cell-speci c genes including angiocrine factors [33], which is consistent with our result. We found that miR-363-5p regulates PDGFB by binding to its 3'-UTR, which inhibits the activation of PDGF/PDGFR-related pathways. It is reported that the metastatic potential of mammary epithelial cells depends on the PDGF-PDGFR loop [34]. PDGF autocrine activates STAT1 and other pathways, contributing to the induction and maintenance of the EMT in breast cancer. PDGFB and dimer protein PDGF-BB is an important lymphangiogenic factor and contributes to cancer lymphatic metastasis by stimulating MAP kinase activity [35,36]. PDGFB exhibited both proliferative and chemotactic effects on lymphatic endothelial cells and directly cause lymphatic metastasis in breast cancer bearing mice [37]. In summary, miR-363-5p/PDGFB might play a pivotal role in breast cancer carcinogenesis and progression, especially related to lymph node staging. Furthermore, miR-363-5p might represent a relatively downstream element in a complicated regulation network, different pathways involved in this process require further exploration.
Nevertheless, this study has several limitations as follows: 1) The present study only included ER+ HER2patients and further veri cation is necessary for other molecular types; 2) The sample size of our discovery cohort is relatively small;3) Accuracy of exosomal miR-363-5p alone is not enough and thus construction of an integrated model is required.

Conclusion
In conclusion, our study identi ed exosome miRNA markers that help evaluate LN status in a noninvasive manner. Exosomal miR-363-5p showed good accuracy and was con rmed with a functional and molecular basis. These results indicate that exosomal miR-363-5p may be applicable in developing liquid biopsy strategies to effectively diagnose LN metastasis in breast cancer.    miR-363-5p inhibits metastatic properties of breast cancer cell (A and B) Migration assay of MCF-7 transfected miR-363-5p-mimic or normal control. Migrated cells were counted using Image J and representative images were shown. Scale bar, 20 μm. Results are shown as mean ± SE. Student's t test was used to analyze the data (*p < 0.05; **p < 0.01). (C) Proliferation ability of MCF-7 transfected with miR-363-5p-mimic or normal control was determined by CCK-8 assay.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.