The Cancer Genome Atlas (TCGA)¹ provided expression data of lncRNAs, mRNAs and miRNAs related to breast cancer. Clinical information was offered in Supplementary Table 1. Differential expression analysis was performed with edgeR (| logFC| > 2, padj < 0.05). Differentially expressed lncRNAs (DElncRNAs) and miRNAs (DEmiRNAs) that had binding sites were screened through miRcode.² Downstream regulatory target genes of DEmiRNAs were predicted by miRDB,³ TargetScan,⁴ along with miRTarBase.⁵ The predicted mRNAs were intersected with DEmRNAs to obtain target DEmRNAs of DEmiRNAs. Pathway enrichment analysis of unigenes was conducted by using Gene Set Enrichment Analysis (GSEA).

Mammary epithelial cells MCF10A (CBP60419) and breast cancer cells MDA-MB-231 (CBP60382), BT-549 (CBP60357), MDA-MB-468 (CBP60387) and MCF-7 (CBP60380) were cultured in complete mediums. All mediums and cell lines were purchased from Nanjing Cobioer Biotech. Cells were inoculated in 6-well plates at 3 × 10⁵ each well until cell growth density reached 70–90%. Then, cells were subjected to transfection utilizing Lipofectamine 2000 (11668-019, Invitrogen, California, United States). Next, 250 μl Opti-MEM (11668-019, Invitrogen, California, United States) medium was used to release 4 μg target plasmids and 10 μl Lipofectamine 2000, respectively, and mixed well by flicking. After being maintained at room temperature for 5 min, two liquids were mixed evenly, and added into cell culture wells after 20 min. Culture plates were shaken gently and mixed evenly. Then plates were generally cultured in an incubator. After 24 h, medium was changed and cells were collected 36–48 h later.

The si-ADAMTS9-AS1 and si-TGFBR2 were chemically synthesized by Genepharma (Shanghai, China). The siRNA transfection was undertaken with Lipofectamine 2000 (Invitrogen) as the standard plan. The oe-negative control (NC) (NC plasmids of overexpressing ADAMTS9-AS1, the vector used was pcDNA™3.1(+) vector (V79020), Invitrogen), oe-ADAMTS9-AS1 (overexpressing ADAMTS9-AS1 plasmids, the vector used was pcDNA™3.1(+) vector), inhibitor NC (NC of miR-301b-3p inhibitor) and miR-301b-3p inhibitor were procured from Genepharma (Shanghai, China).

Total RNA was detached from cells (1 × 10⁶ per well) with Trizol (No. 16096020, Thermo Fisher Scientific, New York, United States). Next, 5 μg samples were chosen for reverse transcription to synthesis into cDNA following the instruction of cDNA kit (K1622; Fermentas Inc., Ontario, CA, United States). MiR-301b-3p expression was detected by qPCR on TaqMan MicroRNA Assay with cDNA as the template. The reaction of qRT-PCR was as follows: 95^°C for 2 min, then 45 cycles of 95^°C for 15 s, 60^°C for 45 s, 72^°C for 45 s. U6 was taken as the internal reference to normalize results.

qRT-PCR was performed on ADAMTS9-AS1 and TGFBR2 according to the instruction of TaqMan Gene Expression Assays protocol (Applied Biosystems, Foster City, CA, United States). GAPDH was taken as the internal reference. PCR procedures were designed as follows: 95^°C for 10 min, 35 cycle of 95^°C for 15 s, 60^°C for 30 s, 72^°C for 45 s. All qRT-PCR was set with 3 parallel wells. Primers were designed as shown in Table 1 and synthesized by TAKARA Company (Beijing, China). Relative transcription level of target gene was formulated by relative quantitative method (2^–△△CT). △△Ct = △Ct (experimental group)−△Ct (control group), △Ct = Ct (target gene)−t (internal reference), relative transcription level of target mRNA = 2^{–△△ Ct}.

Total proteins were acquired from cells (1 × 10⁶ per well), and protein concentration was determined by bicinchoninic acid (BCA) kit (Thermo, United States). Next, 30 μg total proteins were loaded on sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) at constant 80 V for 35 min and then 120 V for 45 min. The proteins were transferred onto a polyvinylidene fluoride (PVDF) membrane (Amersham, United States). The membrane was blocked with 5% skim milk at normal temperature for 1 h. It was followed by an overnight incubation at 4°C overnight with rabbit anti-TGFBR2, Ki67, PCNA, MMP-2, MMP-9, JAK1, p-JAK1, JAK2, p-JAK2, STAT1, p-STAT1, STAT3, p-STAT3, β-actin. All antibodies were offered by Abcam (Cambridge, United Kingdom). The membrane was washed with Tris-buffered saline with 0.1% Tween^®20 detergent (TBST) three times (10 min each time). Afterward, secondary antibody goat anti-rabbit IgG H&L labeled by horseradish peroxidase (HRP) was added for 1 h of incubation at room temperature. Phosphate buffered saline with Tween 20 (PBST) solution was used for 3 washes of the membrane (each time 10 min). Chemiluminescence apparatus (GE, United States) was applied for scanning and development. Protein bands were gray-scale scanned with Image Pro Plus 6.0 (Media Cybernetics, United States) software to analyze relative protein expression.

The transfected breast cancer cells (1 × 10⁴ per well) were inoculated into 12-well plates and cultured at 37^°C with 5% CO₂ for 2 weeks. Then cells were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet. The number of colonies, whose cells are more than 50, was manually counted.

Cells (3 × 10⁵ per well) to be tested were seeded in 24-well plates with 3 parallel wells. EdU (C10310-1, RiboBio, Guangzhou, China) was added into the culture solution to achieve a concentration of 10 μmol/l for 2 h of incubation in the incubator. After removing medium, cells undergone 15 min of fixation with PBS + 4% paraformaldehyde at common temperature for and twice washes with PBS + 3% bovine serum albumin (BSA). PBS containing 0.5% Triton-100 was added for 20 min of incubation at room temperature and then PBS with 3% BSA was used to wash cells twice. Each well was added with 100 μl Apollo^® 567 (C10310-1, RiboBio, Guangzhou, China) for 30 min of incubation in the dark at room temperature. Thereafter, cells were washed with PBS containing 3% BSA. DIPA (C0065, Solarbio, Beijing, China) was added to stain nucleus for 5 min. Next, cells were washed with PBS in triplicate and sealed. Cells were observed under a fluorescent microscope (Type: FM-600, Shanghai PuDa Company) and positive cell number in each field was recorded. Under the microscope, total cells presented as blue and positive cells presented as red. Percentage of EdU stained positive cells was calculated in 3 random fields of each well. EdU positive rate = (EdU positive nucleus number/total nucleus) × 100%.

Extracellular matrix (ECM) gel was maintained at 4°C overnight and then attenuated with serum-free medium at 1:9 ratio to 1 mg/ml. Each polycarbonate membrane of the 24-well Transwell chamber was added with 40 μl ECM gel and maintained generally in the incubator for 5 h. Polymerized gel was prepared by ECM gel and redundant solution was removed. Then, 70 μl/chamber pure Dulbecco’s Modified Eagle Medium (DMEM) was added. Cells (1 × 10⁴ per well) were incubated in an incubator at 37^°C for 0.5 h to rehydrate Matrigel and redundant culture solution was removed. Cells in each group were starved for 24 h without serum, digested, centrifuged and resuspended with DMEM without FBS to a final concentration of 2.5 × 10⁵ cells/ml. The upper chamber where the basement membrane was hydrated was added with added 0.2 ml suspension, while the lower chamber was added with 700 μl pre-cooled DMEM containing 10% FBS. Then, plates were generally cultured in a moist incubator for 24 h. Thereafter, the chamber was discarded, and cells in the chamber and on the membrane were wiped by a wet cotton swab. The remaining cells were fixed with methyl alcohol for 30 min, stained with 0.1% crystal violet for 20 min and dried in inverted position. Cells were observed and photographed under an inverted microscope. Average number of cells that passed through the membrane was counted in 5 random fields.

Co-transfection was, respectively, undertaken on MCD-7 cells (3 × 10⁵ per well) with mimic NC/miR-301b-3p mimic along with oe-NC/oe-ADAMTS9-AS1. The binding between miR-301b-3p with ADAMTS9-AS1 and TGFBR2 was examined using RIP kit. MCF-7 cells were washed with pre-cooled PBS and then the supernatant was removed. Cells were lysed with equivalent volume of lysis solution (P0013B, Beyotime) in ice-bath for 5 min and centrifuged at 4^°C and 14,000 rpm for 10 min. The cell extract was incubated with antibodies for co-precipitation. Specific steps were: 50 μl magnetic beads were taken from each co-precipitation reaction system, washed and resuspended in 100 μl RIP Wash Buffer. According to experimental groups, 5 μg antibodies were added for binding. Magnetic beads-antibody complex was resuspended in 900 μl RIP Wash Buffer after being washed. Thereafter, 100 μl cell extract was added for an overnight incubation at 4^°C. Samples were placed on the magnetic drill to collect magnetic beads-protein complex. The antibodies used in RIP were: Ago2 (ab32381, 1:50, Abcam, United Kingdom) which was evenly shaken at room temperature for 30 min, and IgG (1:100, ab172730, Abcam, United Kingdom) as the NC. Samples were digested by protease K. Then, RNA was extracted from the samples for succeeding analyses.

The whole length of 3′UTR of TGFBR2 gene was amplified. PCR products were cloned into multiple cloning sites on the downstream of pmirGLO luciferase gene (Promega, WI, United States) by using restriction enzyme sites, to construct TGFBR2-wild type (WT) (CCAGCUAUGA CCACAUUGCACUU) group. Binding sites of miR-301b-3p and its target gene were predicted by target gene databases. TGFBR2-mutant (MUT) vector (CCAGCUAUGACCACAUAC GUGAU) was constructed by PCR site-directed mutation method. Renilla luciferase expression vector pRL-TK (TaKaRa, Dalian, China) was taken as the internal reference. HEK-293T cells (3 × 10⁵ per well) were co-transfected with mimic NC/miR-301b-3p mimic and luciferase reporter vectors, respectively. Dual Luciferase activity was tested by using Dual-Luciferase Reporter Assay System (Promega, Madison, WI, United States).

Results were treated with SPSS 21.0 statistical software (SPSS, Inc., Chicago, IL, United States). Three times of biological repetitions and technical repetitions were required for each assay. Measurement data were shown as mean ± standard deviation. Comparison of two or more groups were, respectively, Student’s t-test and analysis of variance (ANOVA). Repeated-measures ANOVA was adopted to compare results at each time point. Patient’s overall survival curve was calculated by Kaplan-Meier. Patient’s survival differences were analyzed by log-rank. Analysis method of enumeration data was chi-square test. P < 0.05 stands for statistical significance.

There were 1,017 DElncRNAs, 74 DEmiRNAs and 2,161 DEmRNAs, as offered by Figure 1A and detailed in Supplementary Tables 2–4. miRcode database was used to compare the results, finding that 60 lncRNAs and 18 miRNAs had binding sites (Supplementary Table 5). Low and high ADAMTS9-AS1 groups were classified by median ADAMTS9-AS1 level. Survival analysis of the 60 lncRNAs revealed notably lower survival of low ADAMTS9-AS1 expression group. Furthermore, ADAMTS9-AS1 levels noticeably pertained to clinical M stage, and patients in high ADAMTS9-AS1 group were mostly in M0 stage (Figure 1B). Hence the downregulation of ADAMTS9-AS1 remarkably impacted patient’s prognostic conditions. Meanwhile, qRT-PCR uncovered that level of ADAMTS9-AS1 was prominently low in cancer cell lines. MDA-MB-231 cells (p < 0.05) and MCF-7 cells (p < 0.05) showed the lowest and highest ADAMTS9-AS1 levels, respectively (Figure 1C).

After transfection of plasmids oe-NC and oe-ADAMTS9-AS1 was done, the latter exhibited a markedly increase in ADAMTS9-AS1 level, with respect to the former group (Figure 2A). Transfected cells were therefore employed in the succeeding analyses. Cell proliferation was examined through colony formation assay and EdU assay (Figures 2B,C). The results revealed that the cell proliferation was remarkably reduced by overexpressed ADAMTS9-AS1. Meanwhile, the result of Transwell assay uncovered that the overexpressed ADAMTS9-AS1 notably constrained the invasive ability of cells (Figure 2D). Western blot result revealed that oe-ADAMTS9-AS1 markedly decreased the expression of tumor markers for proliferation and invasion, such as Ki67, PCNA, MMP-2, MMP-9 (Figure 2E), suggesting that oe-ADAMTS9-AS1 could reduce tumor proliferative, and invasive abilities.

ADAMTS9-AS1 was silenced in MCF-7 and the expression was examined afterward, finding a prominent reduction in si-ADAMTS9-AS1-1/si-ADAMTS9-AS1-2 groups (Figure 2F). The si-ADAMTS9-AS1-1 with the highest interference efficiency was employed hereinafter and was uniformly named as si-ADAMTS9-AS1. Further, the changes of breast cancer proliferative ability (Figures 2G,H) and invasive ability (Figure 2I) were examined by multiple methods. The results uncovered that cancer cell proliferation and invasion were prominently boosted after the downregulation of ADAMTS9-AS1. Western blot result uncovered that levels of Ki67, MMP-9, MMP-2 and PCNA were notably boosted after the downregulation of ADAMTS9-AS1 (Figure 2J). All the above, ADAMTS9-AS1 repressed the proliferation and invasion of breast cancer cells. Morphology of cells after ADAMTS9-AS1 silence or overexpression manifested no noticeable differences between groups (Supplementary Figure 2).

ADAMTS9-AS1 targeted 6 DEmiRNAs (miR-301b-3p, miR-96-3p, miR-144-3p, miR-145-3p, miR-182-3p, and miR-21-3p), among which miR-301b-3p markedly pertained to patient’s prognosis (Figure 3A). Median miR-301b-3p was utilized to classify groups. As displayed in Supplementary Figure 1, miR-301b-3p levels were irrelevant to TNM stage. 8 DEmRNAs binding with miR-301b-3p were discovered in the overlap of target genes and DEmRNAs (Figure 3B). These 8 genes were subjected to Pearson correlation. Pearson correlation coefficient between TGFBR2 and ADAMTS9-AS1 was the highest (Figure 3C). Median TGFBR2 level was utilized for grouping, followed by survival analysis and expression differences. Level of TGFBR2 was of notably difference in varying pathological stages (Figure 3D). Moreover, TGFBR2 participates in the proliferation and invasion of tumor cells (Liu et al., 2018; Zhou B. et al., 2018). Therefore, we speculated that ADAMTS9-AS1 mediated TGFBR2 levels via sequestering miR-301b-3p, thus regulating cancer proliferation and invasion.

Further mechanism verification was conducted. RIP assay exhibited stimulated enrichment of ADAMTS9-AS1/TGFBR2 upon miR-301b-3p mimic treatment, implying that miR-301b-3p bound to ADAMTS9-AS1 and TGFBR2, respectively (p < 0.05) (Figure 3E). Moreover, similar trend was observed concerning enrichment of miR-301b-3p/TGFBR2 in MCF-7 cell with oe-ADAMTS9-AS1, implying that ADAMTS9-AS1/TGFBR2 competitively bound miR-301b-3p (Figure 3F). Dual-luciferase result uncovered that miR-301b-3p mimic markedly repressed luciferase intensity of TGFBR2-WT (p < 0.05) but did not impact TGFBR2-MUT (Figure 3G). TGFBR2 expression was prominently decreased after ADAMTS9-AS1 was interfered (p < 0.05), while miR-301b-3p inhibitor could reverse TGFBR2 expression (p < 0.05) (Figure 3H). Altogether, ADAMTS9-AS1 competitively bound miR-301b-3p, so as to promote the expression of TGFBR2.

We first divided MCF-7 cells into varying transfecting groups. Further, changes in cell proliferation (Figures 4A,B) and invasion (Figure 4C) of breast cancer cells were detected by in vitro colony formation, EdU and Transwell assays, respectively. The results revealed that the invasion (p < 0.05) as well as proliferation (p < 0.05) of breast cancer were markedly boosted after the downregulation of ADAMTS9-AS1 expression, while these abilities reversed upon co-transfection with si-ADAMTS9-AS1 and miR-301b-3p inhibitor. Western blot result uncovered that Ki67, MMP-9, MMP-2, PCNA and other tumor markers were prominently activated after the downregulation of ADAMTS9-AS1, indicating increased tumor proliferative activity. However, in si-ADAMTS9-AS1 + miR-301b-3p inhibitor group, the expression of those tumor makers was reversed (p < 0.05) (Figure 4D), indicating a reduced tumor proliferative activity. To sum up, downregulated ADAMTS9-AS1 facilitated cancer cell invasion and proliferation through miR-301b-3p.

Two TGFBR2 interference sequences were constructed, and TGFBR2 expression was examined. The silencing sequence with the highest silencing efficiency, si-TGFBR2-2 (p < 0.05), was selected and named as si-TGFBR2 (Figure 4E). Three transfecting groups were generated. qRT-PCR indicated that TGFBR2 level was notably boosted after the overexpression of ADAMTS9-AS1 (p < 0.05). Relative to the oe-ADAMTS9-AS1 + si-NC group, the level of ADAMTS9-AS1 in the oe-ADAMTS9-AS1 + si-TGFBR2 group had no notable change (p < 0.05) while TGFBR2 was prominently constrained (p < 0.05) (Figure 4F). Further, cell proliferation (Figures 4G,H) and invasion (Figure 4I) were detected to be markedly reduced upon ADAMTS9-AS1 overexpression(p < 0.05), while overexpressed ADAMTS9-AS1 and silencing TGFBR2 together could reverse that effect (p < 0.05). Western blot result uncovered that Ki67, PCNA, MMP-2, and MMP-9 were notably repressed upon ADAMTS9-AS1 overexpression (p < 0.05), indicating that the tumor proliferation was decreased. However, overexpressed ADAMTS9-AS1 and silencing TGFBR2 together could reverse this regulatory effect (p < 0.05), indicating increased tumor proliferation (Figure 4J). Altogether, ADAMTS9-AS1 restrained the proliferation and invasion of breast cancer via stimulating the level of TGFBR2 through miR-301b-3p.

GSEA analysis result revealed that TGFBR2 was notably enriched in the JAK STAT signaling (Figure 5A), suggesting that ADAMTS9-AS1/miR-301b-3p/TGFBR2 signaling axis may regulate breast cancer cells through the JAK STAT signaling pathway. For further verification, ADAMTS9-AS1 was firstly overexpressed in cell line MDA-MB-231. Corresponding detection uncovered that the protein levels of p-STAT3, p-JAK2, p-STAT1 and p-JAK1 in the oe-ADAMTS9-AS1 group were notably repressed (p < 0.05) (Figure 5B), implying that ADAMTS9-AS1 constrained the JAK STAT signaling pathway. Further analyses depicted that the protein levels of p-STAT3, p-STAT1, p-JAK2, and p-JAK1 were notably boosted after silencing ADAMTS9-AS1, while the regulatory effects were reversed in si-ADAMTS9-AS1 + miR-301b-3p inhibitor (p < 0.05) (Figure 5C). Generally, silencing ADAMTS9-AS1 activated the JAK STAT signaling pathway, while silencing miR-301b-3p could reverse this impact.

MDA-MB-231 cells divided into the following transfecting groups: oe-NC + si-NC, oe-ADAMTS9-AS1 + si-NC and oe-ADAMTS9-AS1 + si-TGFBR2. Western blot examined that p-JAK2, p-STAT3, p-JAK1 and p-STAT1 protein levels were notably reduced in oe-ADAMTS9-AS1 + si-NC group, while co-transfection of oe-ADAMTS9-AS1 and si-TGFBR2 reversed the regulatory effect caused by overexpressed ADAMTS9-AS1 (p < 0.05) (Figure 5D), suggesting that ADAMTS9-AS1 inhibited the JAK STAT signaling pathway, while silencing TGFBR2 reversed that inhibitory effect. Overall, ADAMTS9-AS1 regulated the JAK STAT signaling pathway by targeting TGFBR2 through miR-301b-3p.

MCF-7 cells were simultaneously transfected with si-NC and si-ADAMTS9-AS1, and treated with JAK STAT signaling pathway inhibitor INCB018424 (HY-50856, MedchemExpress; Punwani et al., 2012). qRT-PCR result indicated that the expression of ADAMTS9-AS1 in si-ADAMTS9-AS1 + DMSO group was markedly lower than that in si-NC + DMSO group (p < 0.05). While there was no significant difference in ADAMTS9-AS1 expression between the si-ADAMTS9-AS1 + DMSO group and the si-ADAMTS9-AS1 + INCB018424 group (Figure 6A). The expression of JAK STAT signaling pathway-related proteins was detected by western blot. The result suggested that compared with si-NC + DMSO group, the protein expression levels of p-JAK1, p-JAK2, p-STAT1 and p-STAT3 in si-ADAMTS9-AS1 + DMSO group were remarkably increased. Compared with si-ADAMTS9-AS1 + DMSO group, the protein expression levels in si-ADAMTS9-AS1 + INCB018424 group were notably decreased (p < 0.05) (Figure 6B). In vitro colony formation (Figure 6C), EdU (Figure 6D), and Transwell (Figure 6E) assays indicated that compared with the si-NC + DMSO group, the proliferative ability and invasive ability of breast cancer cells in the si-ADAMTS9-AS1 + DMSO group were prominently increased. Compared with the si-ADAMTS9-AS1 + DMSO group, the proliferative and invasive abilities of breast cancer cells in si-ADAMTS9-AS1 + INCB018424 group were markedly decreased (p < 0.05). Western blot result exhibited that compared with si-NC + DMSO group, the expression levels of Ki67, PCNA, MMP-2 and MMP-9 in si-ADAMTS9-AS1 + DMSO group were markedly boosted, indicating increased tumor proliferative activity. Compared with si-ADAMTS9-AS1 + DMSO group, the expression levels of Ki67, PCNA, MMP-2 and MMP-9 in si-ADAMTS9-AS1 + INCB018424 group were prominently constrained, indicating decreased tumor proliferative activity (p < 0.05) (Figure 6F). In general, ADAMTS9-AS1 could suppress the proliferation and invasion of breast cancer cells by regulating the JAK STAT signaling pathway. Besides, JAK inhibitor did not impact TGFBR2 levels according to detection results (Supplementary Figure 3).