PIWIL1/piRNA-DQ593109 Regulates the Permeability of the Blood-Tumor Barrier via the MEG3/miR-330-5p/RUNX3 Axis.

The blood-tumor barrier (BTB) restricts the efficient delivery of anti-glioma drugs to cranial glioma tissues. Increased BTB permeability may allow greater delivery of the therapeutic agents. Increasing evidence has revealed that PIWI proteins and PIWI-interacting RNAs (piRNAs) play an important role in tumor progression. However, whether PIWI proteins and piRNAs regulate BTB permeability remains unclear. In the present study, we demonstrated that the PIWIL1/piRNA-DQ593109 (piR-DQ593109) complex was the predominant regulator of BTB permeability. Briefly, PIWIL1 was upregulated in glioma endothelial cells (GECs). Furthermore, piR-DQ593109 was also overexpressed in GECs, as revealed via a piRNA microarray. Downregulation of PIWIL1 or piR-DQ593109 increased the permeability of the BTB. Moreover, PIWIL1 and piR-DQ593109, which formed a piRNA-induced silencing complex, degraded the long non-coding RNA maternally expressed 3 (MEG3) in a sequenced-dependent manner. Furthermore, restoring MEG3 released post-transcriptional inhibition of Runt related transcription factor 3 (RUNX3) by sponging miR-330-5p. In addition, RUNX3 bounded to the promoter regions and reduced the promoter activities of ZO-1, occludin, and claudin-5, which significantly impaired the expression levels of ZO-1, occludin, and claudin-5. In conclusion, downregulating PIWIL1 and piR-DQ593109 increased BTB permeability through the MEG3/miR-330-5p/RUNX3 axis. These data may provide insight into glioma treatment.

Introduction

Glioma is the most common primary malignant tumor of the central nervous system. Patients suffering from glioblastoma bear a median survival of 15 months. Chemotherapy plays a crucial role in clinical glioma management. However, the blood-tumor barrier (BTB) limits the delivery of therapeutics into the glioma microenvironment, which dramatically attenuates any chemotherapeutic effect. Properties of specialized brain microvessel endothelial cells in glioma tissues (GECs) provide the structural basis of BTB permeability. Accordingly, strategies targeting these GECs may be conducive to glioma chemotherapy.2, 3

PIWI proteins are a subfamily of Argonaute proteins that are most abundantly expressed in the germline and are involved in stem cell self-renewal and gametogenesis in previous studies. However, growing evidence has indicated that PIWI proteins are also expressed in somatic cells and are linked to tumor cell behaviors, such as cell proliferation, apoptosis, and invasion. PIWIL1 is a member of the PIWI protein subfamily. Deregulation of PIWIL1 can be observed in several types of tumors, such as seminoma and soft-tissue sarcoma, and may predict colorectal cancer and renal cell carcinoma patients’ survival. However, no report has assessed the expression level of PIWIL1 in GECs or its potential role in BTB permeability.

PIWI-interacting RNAs (piRNAs) constitute one of the major classes of small RNAs. Unlike microRNAs (miRNAs) and small interfering RNAs (siRNAs), piRNAs are longer (∼25–30 nt) and do not require Dicer (Dcr)-1 or Dcr-2 for their production. As their name suggests, they bind to PIWI proteins to exert their biological function. Recent evidence revealed that the PIWI/piRNA complex modulates transposon repression, epigenetic states, and RNA cleavage.11, 12, 13 Based on these studies, we hypothesized that PIWIL1 might regulate BTB permeability by forming a complex with piRNAs. piRNA microarray results showed that piR-DQ569979, piR-DQ582921, piR-DQ-585094, and piR-DQ593109 are four of the most elevated piRNAs in GECs. Subsequent assays demonstrated that only piR-DQ593109 could regulate BTB permeability. Accordingly, we hypothesize that piR-DQ593109 may synergistically regulate BTB permeability with PIWIL1 by forming the PIWI/piRNA complex.

Recent studies reported that long non-coding RNA (lncRNA) may play a critical role in regulating multiple steps of tumor progression. The lncRNA maternally expressed 3 (MEG3) has been demonstrated to act as a tumor suppressor in human glioma.15, 16 In addition, MEG3 is involved in endothelial cell dysfunction.17, 18 However, it is unknown whether MEG3 affects BTB permeability. It is well-accepted that mature small RNAs bind to Argonaute proteins and guide them to cleave target RNAs.19, 20 Similarly, PIWI and piRNA complex were shown to exert a gene-silencing function.21, 22 Bioinformatics tools (RepeatMasker, http://www.repeatmasker.org/ and piRNABank, http://pirnabank.ibab.ac.in/) show that MEG3 is a target of piR-DQ593109. Therefore, we hypothesize that PIWIL1/piR-DQ593109 degrades MEG3 with the guidance of piR-DQ593109.

miRNAs are another group of small RNAs and usually regulate gene expression at the post-transcriptional level. It is well established that lncRNAs sponge miRNAs to reduce mRNA degradation or inhibit translation of target mRNAs. The online tool miRanda (http://www.microrna.org/microrna/home.do) shows that MEG3 harbors a putative miR-330-5p binding site, which indicates that MEG3 may regulate BTB permeability by mediating miR-330-5p.

Runt-related transcription factor 3 (RUNX3) belongs to the Runt-related gene family and functions as a tumor suppressor in glioma.24, 25 The expression level of RUNX3 in GECs and its potential role in BTB permeability also remain unknown. By employing bioinformatics tools (targetscan, http://www.targetscan.org/vert_71/ and miRanda), we hypothesize that RUNX3 is a putative target of miR-330-5p, which suggests that miR-330 may regulate BTB permeability by affecting RUNX3 expression and function.

Tight junction-related proteins that localize to the boundary of GECs are an important molecular basis for BTB function. Occludin, claudins, junction adhesion molecules, and cell-selective adhesion molecules are all linked to zonula occludens (ZO) proteins, which restrict the absorption of hydrophilic drugs through the paracellular pathway.28, 29 ZO-1 serves as a signaling molecule or scaffolding protein that recruits other signaling molecules. Meanwhile, claudin-5 is abundant in brain endothelial cells and is an important regulator of endothelial permeability. Thus, targeting ZO-1, occludin, and claudin-5 may contribute to increasing BTB permeability. In silico analysis using JASPAR (http://jaspar.genereg.net/) reveals putative associations between RUNX3 and the promoter regions of ZO-1, occludin, and claudin-5, which indicates that RUNX3 may affect BTB permeability by regulating ZO-1, occludin, and claudin-5 at the transcription level.

In the present study, cell culture models of the blood-brain barrier (BBB), in which human brain microvessel endothelial cells are co-cultured with normal astrocytes, and the BTB, in which human brain microvessel endothelial cells are co-cultured with U87 glioblastoma cells, were introduced to mimic the endothelial cell phenotypes in the normal brain or glioblastoma tissues, respectively. Furthermore, the expression patterns of PIWIL1, piR-DQ593109, MEG3, miR-330-5p, and RUNX3 in GECs were investigated. Because the BTB model in vitro simulates the microenvironment of endothelial cells in vivo, we used this model to further explore the potential roles of the above factors in BTB as well as to elucidate their underlying mechanisms.

Results

PIWILI Knockdown Increased BTB Permeability and Decreased the Expression Levels of Tight Junction-Related Proteins

Western blot and qRT-PCR analyses were performed to detect the protein and mRNA levels, respectively, of PIWIL1 in endothelial cells (ECs) and GECs. As shown in Figures 1A and 1B, PIWIL1 was detected in ECs and GECs, and its protein and mRNA levels were enhanced in GECs. Next, PIWIL1 stable knockdown GECs were constructed, and transendothelial electric resistance (TEER) values and horseradish peroxidase (HRP) flux were measured. As shown in Figures 1C and 1D, no significant differences were detected in TEER values and HRP flux between the control groups and the shPIWIL1 negative control (shPIWIL1 NC) groups. However, as compared to shPIWIL1 NC groups, significant decreases in TEER value and increases in HRP flux were observed in shPIWIL1 groups. Furthermore, knockdown of PIWIL1 reduced the expression levels of ZO-1, occludin, and claudin-5 (Figure 1E). Consistent with the western blot results, immunofluorescence staining showed that expression levels of ZO-1, occludin, and claudin-5 were decreased in shPIWIL1 groups. Meanwhile, the continuous distributions of ZO-1, claudin-5, and occludin in the GEC border were disrupted in the shPIWIL1 groups (Figure 1F).

Figure 1

Knockdown of PIWIL1 Increased BTB Permeability In Vitro as well as Reduced the Expression Levels of ZO-1, Occludin, and Claudin-5 in GECs

(A) Relative PIWIL1 protein levels in ECs and GECs determined by western blot. Data represent mean ± SD (n = 4, each). *p < 0.05 versus EC group. (B) Relative PIWIL1 mRNA levels in ECs and GECs determined by qRT-PCR. Data represent mean ± SD (n = 5, each). *p < 0.05 versus EC group. (C and D) Effects of PIWIL1 knockdown on TEER values (C) and HRP flux (D). (E) Effects of PIWIL1 knockdown on ZO-1, occludin, and claudin-5 expression levels determined by western blot. Data represent mean ± SD (n = 4, each). *p < 0.05 versus shNC group. (F) Effects of PIWIL1 knockdown on ZO-1, occludin, and claudin-5 expression levels and distribution determined by immunofluorescence staining (n = 3, each). ZO-1, occludin, and claudin-5 (red) were labeled with secondary antibody against anti-ZO-1, anti-occludin, and anti-claudin-5 antibody, respectively, and nuclei (blue) were labeled with DAPI. Scale bar represents 30 μm.

Downregulation of piR-DQ593109 Increased BTB Permeability and Inhibited the Expression Levels of ZO-1, Occludin, and Claudin-5

We next performed a piRNA microarray to assess the expression patterns of piRNAs in GECs. As shown in Figure 2A, piR-DQ569979, piR-DQ585094, piR-DQ582921, and piR-DQ593109 are four of the most abundant piRNAs in GECs. Meanwhile, qRT-PCR analysis showed that expression levels of piR-DQ569979, piR-DQ585094, piR-DQ582921, and piR-DQ593109 were significantly elevated in GECs (Figure 2B). Subsequently, we performed loss-of-function tests to clarify the potential roles of these piRNAs in BTB permeability. Remarkably, downregulation of piR-DQ593109 decreased TEER values and increased HRP flux in BTB models in vitro (Figures 1C and 1D). However, piR-DQ569979, piR-DQ585094, and piR-DQ582921 showed few effects on TEER values or HRP flux (Figure S2). Accordingly, piR-DQ593109 was selected to perform the subsequent analyses. As shown in Figures 2E and 2F, compared with piR-DQ593109(-)NC groups, expression levels of ZO-1, occludin, and claudin-5 were decreased in piR-DQ593109(-) groups in western blot and immunofluorescence assays. Meanwhile, ZO-1, occludin, and claudin-5 presented discontinuous distributions at the GEC border. Moreover, the combination of PIWIL1 and piR-DQ593109 knockdown strengthened increases in TEER values (Figure 2G) and decreases in HRP flux (Figure 2H) as well as inhibitory effects on ZO-1, occludin, and claudin-5 expression (Figure 2I).

Figure 2

Downregulation of piR-DQ593109 Increased BTB Permeability as well as Inhibited the Expression Levels of ZO-1, Occludin, and Claudin-5

(A) piRNA microarray analysis of total RNAs isolated from EC and GEC cells. Red indicates high relative expression and green indicates low relative expression. (B) Relative expression levels of piR-DQ569979, piR-DQ585094, piR-DQ-582921, and piR-DQ593109 determined by qRT-PCR. Data represent mean ± SD (n = 5, each). *p < 0.05 versus EC group. (C and D) Effects of piR-DQ593109 knockdown on TEER values (C) and HRP flux (D). (E) Effects of piR-DQ593109 knockdown on ZO-1, occludin, and claudin-5 expression levels determined by western blot. Data represent mean ± SD (n = 4, each). *p < 0.05 versus piR-DQ593109(-)NC group. (F) Effects of piR-DQ593109 knockdown on ZO-1, occludin, and claudin-5 expression levels and distribution determined by immunofluorescence staining (n = 3, each). ZO-1, occludin, and claudin-5 (red) were labeled with secondary antibody against anti-ZO-1, anti-occludin, and anti-claudin-5 antibody, respectively, and nuclei (blue) were labeled with DAPI. Scale bar represents 30 μm. (G) PIWIL1 knockdown combined with piR-DQ593109 decreased TEER values in BTB models in vitro. (H) PIWIL1 knockdown combined with piR-DQ593109 increased HRP flux in BTB models in vitro. (I) PIWIL1 knockdown combined with piR-DQ593109 decreased the expression levels of ZO-1, occludin, and claudin-5. Data represent mean ± SD (n = 5, each). *p < 0.05 versus shNC+piR-DQ593109(-)NC group; #p < 0.05 versus shNC+piR-DQ593109(-) group; &p < 0.05 versus shPIWIL1+piR-DQ593109(-)NC group.

MEG3 Was Downregulated in GECs, and Increased BTB Permeability Followed MEG3 Overexpression

As shown in Figure 3A, as compared to EC groups, MEG3 was downregulated in GECs. Next, we performed gain-of-function tests to verify the potential roles of MEG3 in BTB permeability. As shown in Figures 3B and 3C, overexpression of MEG3 produced a decrease in TEER values and an increase in HRP flux. Moreover, as compared to the shPIWIL1+piR-DQ593109+MEG3(+) NC groups, TEER values (Figure 3D) were decreased and the HRP flux (Figure 3E) was increased significantly in shPIWIL1+piR-DQ593109+MEG3(+) groups. The above results indicated PIWIL1 knockdown, piR-DQ593109 knockdown, and MEG3 overexpression induce synergistic effects on BTB permeability.

Figure 3

MEG3 Was Involved in PIWIL1/piR-DQ593109 Complex-Regulated BTB Permeability

(A) Relative MEG3 expression levels determined by qRT-PCR in ECs and GECs. Data represent mean ± SD (n = 5, each). *p < 0.05 versus MEG3(+)NC group. (B) TEER values were detected to evaluate the effects of MEG3 overexpression on BTB integrity. (C) HRP fluxes were detected to evaluate the effects of MEG3 overexpression on BTB permeability. (D and E) The combination PIWIL1 knockdown, piR-DQ593109 knockdown, and MEG3 overexpression on TEER values (D) and HRP flux (E). Data represent mean ± SD (n = 4, each). *p and #p < 0.05 versus control group; &p < 0.05 versus shPIWIL1+piR-DQ593109(-)+MEG3(+)NC group. (F) Effects of PIWIL1 and piR-DQ593109 knockdown on MEG3 expression levels in GECs. Data represent mean ± SD (n = 3, each). *p < 0.05 versus shNC group; #p < 0.05 versus piR-DQ593109(−)NC group; &p < 0.05 versus shNC+piR-DQ593109(-)NC group. (G) RNA immunoprecipitation assays were performed with anti-IgG or anti-PIWIL1 in GECs. Relative enrichment of piR-DQ593109 and MEG3 were determined by qRT-PCR in GECs. Data represent mean ± SD (n = 5, each). **p < 0.01 versus anti-IgG group. (H) Dual luciferase reporter assays were performed to determine the binding sites of MEG3 and piR-DQ593109 in HEK293T cells. Data represent mean ± SD (n = 5, each). *p < 0.05 versus piR-DQ593109(+) NC group.

The PIWIL1/piR-DQ593109 Complex Regulated MEG3 in a Sequence-Specific Manner

PIWI/piRNA complexes are reported to degrade RNAs. To clarify whether MEG3 is regulated by PIWIL1 and piR-DQ593109, we performed qRT-PCR to measure MEG3 levels in the PIWIL1- and piR-DQ593109- deficient GECs. As shown in Figure 3F, MEG3 was upregulated in the shPIWIL1, piR-DQ593109(-), and shPIWIL1+piR-DQ593109(-) groups, and the shPIWIL1+piR-DQ593109(-) group had the highest MEG3 expression level. Subsequently, we performed an RNA immunoprecipitation assay to further explore the interaction between PIWIL1, piR-DQ593109, and MEG3. As shown in Figure 3G, piR-DQ593109 and MEG3 were enriched in the PIWIL1 immunoprecipitates as compared to the immunoglobulin G (IgG) immunoprecipitates.

Bioinformatic tools (RepeatMasker and piRNABank) show that MEG3 harbors three putative piR-DQ593109 binding sites. We next performed a dual-luciferase reporter assay to confirm the above binding sites. As shown in Figure 3H, luciferase activity was decreased in HEK293T cells co-transfected with piR-DQ593109(+) and MEG3-Wt, piR-DQ593109(+) and MEG3-Mut1, piR-DQ593109 and MEG3-Mut2, as well as piR-DQ593109 and MEG3-Mut3. However, luciferase activity in HEK293T cells co-transfected with piR-DQ593109(+) and MEG3-Mut4 returned to control levels, indicating that piR-DQ593109 bound to MEG3 in a sequence-specific manner and all of the three putative binding sites were functional. The above data revealed that MEG3 may be degraded by the PIWIL1/piR-DQ593109-induced silencing complex.

MEG3 Bound to and Negatively Regulated miR-330-5p

lncRNAs are shown to sponge miRNAs to release post-transcriptional inhibition. We used a bioinformatics tool (miRanda) that revealed that MEG3 has one putative miR-330-5p binding site, which indicated that MEG3 may regulate BTB permeability by sponging miR-330-5p. To test this hypothesis, we measured the expression level of miR-330-5p in GECs stably overexpressing MEG3. As shown in Figure 4A, as compared to the MEG3(+)NC group, miR-330-5p was downregulated in the MEG3(+) group. We then performed RNA immunoprecipitation and dual-luciferase reporter assays to define the interaction between MEG3 and miR-330-5p. As shown in Figure 4B, MEG3 and miR-330-5p were enriched in the anti-Ago2 group as compared to the anti-normal IgG group. Moreover, miR-330-5p knockdown decreased the enrichment of MEG3 and miR-330-5p immunoprecipitated with Ago2. A subsequent dual-luciferase reporter assay showed that luciferase activity was lower in the MEG3-Wt+pre-miR-330-5p group than in the MEG3-Wt+pre-NC group. However, luciferase activity remained similar to control levels in HEK293T cells co-transfected with pre-miR-330-5p and a reporter vector containing an MEG3 fragment, in which the miR-330-5p binding site was mutated (Figure 4C). Collectively, these results suggested that MEG3 negatively regulated miR-330-5p by binding it in a sequence-dependent manner.

Figure 4

miR-330-5p Was a Mediator of MEG3-Regulated BTB Permeability

(A) Effect of MEG3 overexpression on miR-330-5p levels in GECs by qRT-PCR. Data represent mean ± SD (n = 5, each). *p < 0.05 versus MEG3(+)NC group. (B) RNA immunoprecipitation assay was performed with normal mouse IgG or anti-Ago2 in GECs. Relative enrichment of MEG3 and miR-330-5p were determined by qRT-PCR. Data represent mean ± SD (n = 5, each). *p < 0.05 versus IgG group; **p < 0.01 versus IgG group; #p < 0.05 versus control group. (C) Relative luciferase activity in HEK293T cells co-transfected with MEG3-miR-Wt and pre-miR-330-5p. Data represent mean ± SD (n = 5, each). *p < 0.05 versus MEG3-miR-Wt+ pre-NC group. (D) Relative miR-330-5p levels in ECs and GECs were detected by qRT-PCR. Data represent mean ± SD (n = 4, each).*p < 0.05 versus EC group. (E) Effect of miR-330-5p on TEER values. (F) Effect of miR-330-5p on HRP flux. Data represent mean ± SD (n = 5, each). *p < 0.05 versus pre-NC group; #p < 0.05 versus anti-NC group. (G) TEER values were detected to evaluate the effects of MEG3 and miR-330-5p on BTB integrity. (H) HRP fluxes were detected to evaluate the effects of MEG3 and miR-330-5p on BTB permeability. Data represent mean ± SD (n = 5, each). *p < 0.05 versus MEG3(+)NC+pre-NC group; #p < 0.05 versus MEG3(+)+pre-NC group.

miR-330-5p Was Upregulated in GECs and Increased BTB Permeability

We next interrogated the expression pattern of miR-330-5p in ECs and GECs, and investigated its potential role in BTB permeability. qRT-PCR results showed that miR-330-5p expression level was enhanced in GECs as compared to ECs (Figure 4D). To further define the potential function of miR-330-5p in BTB, we knocked down and overexpressed miR-330-5p in GECs. As shown in Figure 4E, TEER value was increased in the pre-miR-330-5p group and decreased in the anti-miR-330-5p group. Meanwhile, HRP flux was reduced in the pre-miR-330-5p group and increased in the anti-miR-330-5p group (Figure 4F). The above results revealed that miR-330-5p inhibition enhanced BTB permeability.

miR-330-5p Mediates MEG3-Regulated BTB Permeability

To further investigate the role of miR-330-5p in MEG3-regulated BTB permeability, pre-miR-330-5p was transfected into GECs that stably overexpressed MEG3 before the TEER value and HRP flux tests. As shown in Figures 4G and 4H, overexpression of miR-330-5p significantly reversed the MEG3 overexpression-induced decrease in TEER and increase in HRP flux. These data suggested that an miR-330-5p was involved in MEG3-regulated BTB permeability.

RUNX3 Was a Target of miR-330-5p

miRNAs are generally considered to negatively regulate gene expression by targeting the 3′ UTR of a target mRNA. By searching bioinformatics databases, the transcription factor RUNX3 was selected as a candidate. As shown in Figure 5A, miR-330-5p upregulation reduced RUNX3 protein expression in GECs, and miR-330-5p inhibition increased RUNX3 protein expression. Next, a dual-luciferase reporter assay was performed to confirm the putative binding sites of miR-330-5p in the RUNX3 mRNA 3′ UTR. As shown in Figure 5B, luciferase activity was decreased in HEK293T cells that were co-transfected with pre-miR-330-5p and RUNX3 3′-UTR-Wt, pre-miR-330-5p and RUNX3 3′-UTR-Mut1, as well as pre-miR-330-5p and RUNX3 3′-UTR-Mut2. However, luciferase activity remained at control levels in HEK293T cells co-transfected with reporter vector containing the combined mutated miR-330-5p binding sites and pre-miR-330-5p. The above data suggested that miR-330-5p post-transcriptionally inhibited RUNX3 by targeting the aforementioned two sites in the 3′ UTR of RUNX3 mRNA.

Figure 5

RUNX3 Was Involved in miR-330-5p-Regulated BTB Permeability

(A) Effects of miR-330-5p on RUNX3 expression levels by western blot. Data represent mean ± SD (n = 4, each). *p < 0.05 versus pre-NC group; #p < 0.05 versus anti-NC group. (B) Dual luciferase reporter assay was performed to determine the binding site of miR-330-5p and RUNX3 3′ UTR in HEK293T cells. Data represent mean ± SD (n = 5, each). *p < 0.05 versus pre-NC group. (C) RUNX3 expression levels in ECs and GECs were determined by western blot. Data represent mean ± SD (n = 4, each). *p < 0.05 versus EC group. (D) Relative RUNX3 mRNA levels in ECs and GECs were detected by qRT-PCR. Data represent mean ± SD (n = 4, each). *p < 0.05 versus ECs group. (E) Effects of RUNX3 on TEER values. (F) Effects of RUNX3 on HRP flux. (G) Effects of RUNX3 on ZO-1, occludin, and claudin-5 mRNA levels by qRT-PCR. (H) Effects of RUNX3 on ZO-1, occludin, and claudin-5 protein levels by western blot. Data represent mean ± SD (n = 5, each). *p < 0.05 versus RUNX3(+)NC group; **p < 0.01 versus RUNX3(+)NC group; #p < 0.05 versus shNC group; ##p < 0.01 versus shNC group. (I) Effects of RUNX3 on ZO-1, occludin, and claudin-5 protein levels and distribution were determined by immunofluorescence staining (n = 3, each). ZO-1, occludin, and claudin-5 (red) were labeled with secondary antibody against anti-ZO-1, anti-occludin, and anti-claudin-5 antibody, respectively, and nuclei (blue) were labeled with DAPI. Scale bar represents 30 μm. (J) Effects of miR-330-5p and RUNX3 on TEER values. (K) Effects of miR-330-5p and RUNX3 on HRP flux.

RUNX3 Was Downregulated in GECs, and RUNX3 Restoration Increased BTB Permeability and Inhibited ZO-1, Occludin, and Claudin-5 Expression at the Transcriptional Level

To better understand the role of RUNX3 in BTB permeability, we first detected RUNX3 expression in ECs and GECs. As shown in Figures 5C and 5D, protein and mRNA levels of RUNX3 were decreased in GECs as compared to ECs. We next tested changes in BTB permeability after RUNX3 knockdown or overexpression. As shown in Figures 5E and 5F, RUNX3 overexpression reduced the TEER values and enhanced the HRP flux. In addition, impaired mRNA and protein levels of ZO-1, occludin, and claudin-5 were observed in the RUNX3(+) group via qRT-PCR (Figure 5G), western blot assay (Figure 5H), and immunofluorescence staining (Figure 5I). The above data indicated that the enhanced BTB permeability induced restoring RUNX3 in GECs may occur as a consequence of ZO-1, occludin, and claudin-5 downregulation.

RUNX3 Was Involved in miR-330-5p-Mediated BTB Permeability

Co-transfection of miR-330-5p and RUNX3 was performed to determine the effects of RUNX3 on miR-330-5p-mediated BTB permeability. As shown in Figures 5J and 5K, RUNX3 overexpression substantially attenuated the increase in TEER value and decrease in HRP flux induced by miR-330-5p overexpression. The above data revealed that RUNX3 participated in miR-330-5p-induced increases in BTB permeability and decreases in ZO-1, occludin, and claudin-5 expression.

RUNX3 Decreased the Promoter Activity of ZO-1, Occludin, and Claudin-5

RUNX3 overexpression decreased both the mRNA and protein levels of ZO-1, occludin, and claudin-5; because of this, we infer that RUNX3 might regulate the expression of the above proteins at the transcriptional level. To confirm this hypothesis, we first screened for putative RUNX3 binding sites in the promoter region of ZO-1, occludin, and claudin-5. Subsequent dual-luciferase reporter and chromatin immunoprecipitation (ChIP) assays were performed to identify any association between RUNX3 and the above promoters. As shown in Figures 6A–6C, following co-transfection with pEX3-RUNX3, ZO-1, occludin, and claudin-5 promoter activities were substantially reduced. Subsequently, putative RUNX3 binding sites in the promoter reporter constructs were deleted one by one. As shown in Figure 6A, deletion of the −1731 site did not induce a significant change in ZO-1 promoter activity as compared to the full-length ZO-1 promoter reporter construct. However, as compared to the construct containing the −1731 site deletion, the construct containing a −309 site deletion produced reversed promoter activity. Similarly, deletion of the +15 and −917 site regions significantly reversed the promoter activity decrease in occludin and claudin-5 promoter constructs, respectively (Figures 6B and 6C). In addition, the DNA binding pattern of RUNX3, as determined by ChIP assay, was consistent with the locations of the binding sites on the ZO-1, occludin, and claudin-5 promoters determined in luciferase reporter assays. As shown in Figures 6D–6F, no PCR construct was observed in IgG immunoprecipitates. However, in RUNX3 immunoprecipitates, PCR constructs were observed using primers specific for the −309, +15, and −917 site regions of the ZO-1, occludin, and claudin-5 promoters, respectively. Collectively, our data indicated that RUNX3 exhibited transcriptional repression of ZO-1, occludin, and claudin-5 by binding to specific sites in these gene’s promoters. (Figure S3 shows the effects of MEG3, miR-330-5p, and RUNX3 on ZO-1, occludin, and claudin-5 expression and distribution.)

Figure 6

RUNX3 Decreased the Promoter Activity of ZO-1, Occludin, and Claudin-5

(A–C) Schematic depiction of the different reporter plasmids and relative luciferase activity: ZO-1 (A), occludin (B), and claudin-5 (C) are shown. The Y-bar shows the deletion positions on the promoter fragments. The X-bar shows the reporter vector activity after normalization with the cotransfected reference vector (pRL-TK), and relative to the activity of the pEX3 empty vector, and the activity was set to 1. Data represent mean ± SD (n = 5, each). *p < 0.05. (D–F) Schematic representation of the human ZO-1 (D), occludin (E), and claudin-5 (F) promoter region 3,000 bp upstream of the transcription start sites (TSSs), which were designated as +1. ChIP PCR products for putative RUNX3 binding sites and an upstream region not expected to associate with RUNX3 were depicted with bold lines. Dashed arrows represent the primers used for each PCR. GECs were used to conduct ChIP assays. PCR was conducted with the resulting precipitated DNA. Images are representative of independent experiments (n = 4).

Discussion

In the present study, we found that PIWIL1 and piR-DQ593109 were upregulated and MEG3 was downregulated in GECs. PIWIL1 knockdown, piR-DQ593109 knockdown, and MEG3 overexpression increased BTB permeability. The combination of the above treatments produced the strongest effects. PIWIL1/piR-DQ593109 led to MEG3 degradation via a piRNA-induced silencing complex. miR-330-5p was upregulated in GECs and sponged by MEG3. miR-330 inhibition permeabilized the BTB by promoting RUNX expression. Moreover, RUNX3 mRNA and protein levels were decreased in GECs, and restoration of RUNX3 increased BTB permeability via transcriptional repression of ZO-1, occludin, and claudin-5.

Because the cell co-culture model of BTB in vitro simulates high transendothelial electric resistance, restricted permeability across the monolayer, and changes in the morphological characteristics of endothelial cells in vivo, it has been utilized to investigate the nature of BTB cellular interactions occurring in vivo.32, 33 TEER values and HRP flux, which are based on the BTB model in vitro, are important indices used to evaluate BTB permeability according to previous studies.32, 34 Therefore, we detected these two indices to determine the effects of loss- or gain-of function tests on BTB permeability. However, more results should be achieved from primary human cell samples.

PIWIL1 was detected in tissues and cell lines beyond the germline. Several studies reported that PIWIL1 acts as an oncogene in several types of human tumors. Li et al. demonstrated that PIWIL1 is upregulated in uterine cervical cancer, breast carcinoma, ovarian cancer, and endometrial cancer. In addition, the expression level of PIWIL1 was positively correlated with that of Ang-2 in the above cancers. PIWIL1 protein expression is also correlated with tumor clinicopathological factors and may be a tumor prognostic marker. Chen et al. found that PIWIL1 is significantly correlated with Federation International of Gynecology and Obstetrics (FIGO) stage, lymphovascular space involvement, lymph node metastasis, and level of myometrial invasion in endometrial cancer. Furthermore, PIWIL1 mRNA expression is associated with the depth of tumor invasion and the stage of tumorigenesis progression in colorectal cancer. Moreover, PIWIL1 has been shown to be an independent prognostic marker of non-small cell lung cancer. An in vivo xenograft mouse model confirmed that downregulation of PIWIL1 inhibits lung cancer growth. PIWIL1 is also detected in human gliomas and is correlated with tumor grade. Silencing PIWIL1 suppresses the malignant behaviors of glioma cells. Consistent with the above findings, we detected the expression of PIWIL1 in GECs and demonstrated its functional role in BTB permeability.

Similar to PIWI proteins, piRNAs were originally described in the germline. Repression of transposable elements, epigenetic activation, and mRNA decay have been widely studied in germline tissue. Recently, the effects of piRNAs on tumorigenesis have attracted a great deal of attention. Aberrant expression of piRNAs is correlated with clinical features in malignant tumors. piR-651 was shown to be downregulated in classical Hodgkin lymphoma serum samples, and it returns to normal levels when patients achieve complete remission. Furthermore, piR-823 is deregulated in tumor tissues, blood serum, and urine of patients with renal cell carcinoma. In addition, it has been confirmed that piR-598 has a tumor-suppressive function in glioma, and a single nucleotide polymorphism in piR-598 increases glioma risk. In the present study, we found that piR-DQ593109 was upregulated in GECs, and that piR-DQ593109 knockdown increased BTB permeability.

It is well-acknowledged that piRNAs bind to PIWI proteins to exert their functions. De Fazio et al. showed that piRNAs can be loaded onto MIWI2 (mouse PIWI protein), and that the MIWI2/piRNA complex directly silences long interspersed nuclear element 1 (LINE1) elements. Recent evidence indicates that PIWI/piRNA pathways exert additional functions to transposon repression. Huang et al. demonstrated that the PIWI/piRNA complex regulates chromatin architecture by promoting de novo heterochromatin formation. Post et al. suggested that the PIWI transcriptional silencing mechanism triggers robust chromatin changes on targets with sufficient piRNA binding. Moreover, PIWI/piRNA complexes have been shown to regulate RNA degradation. Watanabe et al. found that mRNAs and lncRNAs are upregulated in PIWI-deficient or piRNA-biogenesis-inhibited mice. These researchers attribute the increased levels of mRNAs and lncRNAs to the impaired degradation of the PIWIL1/piRNA complex. In the mouse central nervous system, MIWI and piRNAs colocalize in neuronal dendrites. In addition, piR-DQ541777 inhibition decreases the dendrite spine area, presumably due to direct targeting of Cdk5rap1 and Mark1/2 mRNA. In the present study, we found that PIWIL1 and piR-DQ593109 knockdown synergistically increased BTB permeability. Furthermore, the PIWIL1/piR-DQ593109 complex regulated MEG3 in a piR-DQ541777-sequence-dependent manner.

As an increasing number of studies have elucidated that lncRNAs play crucial roles in tumor malignancy, there has been a great interest in defining the role of lncRNAs in EC function. For instance, lncRNA GAS5 and ANRIL have been shown to be involved in hypertension-induced ECs dysfunction. Additionally, during diabetes-mediated vascular destruction, myocardial infarction-associated transcript (MIAT) knockdown ameliorates endothelial cell proliferation, migration, and tube formation induced by high glucose. It has been reported that MEG3 is one of the conserved lncRNAs, and is abundantly expressed in human endothelial cells. Moreover, MEG3 was demonstrated to negatively regulate angiogenesis after ischemic brain injury. Furthermore, Boon et al. confirmed that MEG3 controls endothelial cell aging and function. Besides EC regulation, MEG3 is downregulated in glioma and negatively regulates glioma biology.15, 55 In the present study, we found that MEG3 was downregulated in GECs, and enhanced BTB permeability was observed following MEG3 overexpression.

lncRNAs have been intensively investigated as miRNA regulators. Recent evidence showed that MEG3 functions as an miRNA sponge to regulate miRNA functions. Liu et al. demonstrated that MEG3 functions as a competing endogenous RNA for miR-181b to regulate hypoxia-induced apoptosis of neuronal cells. In addition, MEG3 overexpression impaired the proliferation of hepatocellular carcinoma cells by sponging miRNA-664. In glioma, MEG3 also functions as an miRNA sponge. For instance, Qin et al. showed that MEG3 inhibits glioma malignancy by acting as an miR-19a sponge. Consistent with the above studies, we found that MEG3 acted as an miR-330-5p sponge in GECs.

miRNAs play a key role in gene post-transcriptional regulation. miRNAs recruit Ago2 and Dicer to the 3′ UTR of target mRNAs, inhibiting mRNA translation or promoting mRNA degradation. Previous publications have demonstrated that a single miRNA can have totally different expression patterns in different types of tumor cells. As shown in previous studies, miR-330-5p is downregulated in pancreatic cancer cells and malignant melanoma. In the present study, we found that miR-330-5p was upregulated in GECs. Moreover, miR-330-5p inhibition increased BTB permeability.

RUNX3 has been characterized as a tumor suppressor in several malignant tumors. Loss of RUNX3 expression is highly associated with poor prognosis in gastric cancer. Moreover, RUNX3 interacts with TEA domain (TEAD), negatively regulating the TEAD-Yes-associated protein 1 (YAP) oncogenic complex in gastric carcinogenesis. In hepatocellular cancer, RUNX3 deficiency leads to tumor development and progression. Additionally, knockdown of RUNX3 in endothelial progenitor cells leads to increased cell proliferation, colony formation, and migration. Consistent with the above findings, we demonstrated herein that RUNX3 was lowly expressed in GECs and was negatively regulated by miR-330-5p. Impaired expression of RUNX3 has been linked to hypermethylation of the RUNX3 promoter region in several studies.25, 65 Whether epigenetic regulation is involved in the decreased expression of RUNX3 in GECs deserves to be further explored.

In the present study, ChIP and dual-luciferase reporter assays revealed that RUNX3 transcriptionally repressed ZO-1, occludin, and claudin-5. Consistent with our results, Inoue et al. found that RUNX3 downregulates neurotrophic receptor tyrosine kinase 2 (NTRK2) through transcriptional repression. In addition, ChIP assays and electrophoretic mobility shift assays (EMSAs) have demonstrated that RUNX3 interacted with the transcriptional regulatory region of JAG1 and inhibits JAG1 transcription.

In conclusion, PIWIL1/piR-DQ593109 knockdown in GECs increased the BTB permeability by reducing MEG3 degradation. Restoration of MEG3 in GECs upregulated the miR-330-5p target gene RUNX3 through sponging and negative regulation of miR-330-5p. Overexpression of RUNX3 enhanced BTB permeability via transcriptional repression of ZO-1, occludin, and claudin-5. We described for the first time that downregulation of PIWIL1/piR-DQ593109 increased BTB permeability and inhibited ZO-1, occludin, and claudin-5 expression levels by modulating the MEG3/miR-330-5p/RUNX3 pathway. This research broadens the mechanistic understanding of BTB permeability and provides a basis for a novel strategy for combined glioma treatment.

Materials and Methods

Cell Lines and Cell Culture

The immortalized human brain endothelial cell line hCMEC/D3 was kindly provided by Dr. Couraud from the Institut Cochin, Paris, France. Endothelial cells were cultured on culture inserts (0.4-mm pore size; Corning, Lowell, MA, USA) coated with 150 μg/mL Cultrex Rat Collagen I (R&D Systems, Minneapolis, MN, USA). Cells were maintained in endothelial basal medium (EBM-2) (Lonza, Walkersville, MD, USA), containing 5% fetal bovine serum (FBS) “Gold” (PAA Laboratories, Pasching, Austria), 1% penicillin-streptomycin (Life Technologies, Paisley, UK), 1.4 μmol/L hydrocortisone (Sigma-Aldrich, St Louis, MO, USA), 1% chemically defined lipid concentrate (Life Technologies, Paisley, UK), 5 μg/mL ascorbic acid (Sigma-Aldrich), 10 mmol/L HEPES (PAA Laboratories), and 1 ng/mL human basic fibroblast growth factor (bFGF) (Sigma-Aldrich). Endothelial cells were limited from 30 to 40 passages. Human glioblastoma U87 cell line and HEK293T cell line were purchased from the Shanghai Institutes for Biological Sciences Cell Resource Center and maintained in DMEM with high glucose with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin (Life Technologies, Paisley, UK). Normal human astrocytes (NHAs) were purchased from Sciencell Research Laboratories (Carlsbad, CA, USA) and maintained in astrocyte medium (Carlsbad, CA, USA). All cells were maintained at 37°C, 5% CO2, in a humidified atmosphere.

Establishment of In Vitro BTB and BBB Model

The in vitro BTB and BBB model were established by co-culturing endothelial cells and U87 glioblastoma cells, as described previously. For the BTB model in vitro, U87 cells were seeded in six-well plates at 2 × 104 per well and cultured for 2 days prior to the addition of the endothelial cell inserts. Endothelial cells were seeded at 2 × 105 per well on the upper side of inserts coated with Cultrex Rat Collagen I (R&D Systems, Minneapolis, MN, USA). Both endothelial cells and U87 cells were cultured with prepared EBM-2 medium, and medium was refreshed every 2 days. After co-culturing for 4 days, the GEC modeling endothelial cell phenotype of glioma in vivo was obtained. For the BBB model in vitro, endothelial cells were co-cultured with NHA with the same approach described above. The normal endothelial cells modeling endothelial cell phenotype of the normal brain in vivo were obtained.

qRT-PCR Assay

Total RNAs were separated with Trizol reagent (Life Technologies, Carlsbad, CA, USA), following the manufacturer’s description. The RNA concentration and quality were determined using a Nanodrop Spectrophotometer (ND-100, Thermo Scientific, Waltham, MA). For measuring the levels of PIWIL1 (NM_004764.4), MEG3 (NR_033359.1), and RUNX3 (NM_001031680.2), One-Step SYBR PrimeScript RT-PCR Kit (Perfect Real Time) (Takara Bio) was used. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as endogenous control. The levels of piR-DQ593109 (hsa_piR_016792) and miR-330-5p (NR_029886.1) were detected by TaqMan MicroRNA Reverse Transcription kit and TaqMan Universal Master Mix II (Applied Biosystems). U6 was used as endogenous control. All qRT-PCR reactions were performed by the 7500 Fast RT-PCR System (Applied Biosystems, Foster City, CA, USA). Relative expression values were calculated using the relative quantification (2−ΔΔCt) method. Primers and probes used in this study are shown in Table S1.

Cell Transfection

Short-hairpin RNA directed against human PIWIL1 gene or RUNX3 was ligated into the pGPU6/Neo vector (GenePharma, Shanghai, China) to construct the shPIWIL1 or shRUNX3 plasmid, respectively. The human MEG3 gene or RUNX3 gene coding sequence was ligated into the pGCMV/Blasticidin vector (GenePharma, Shanghai, China) to construct the MEG3 and RUNX3 overexpression plasmid, respectively. The pGPU6/GFP/Neo and pIRES/EGFP empty vectors were used as NCs. Endothelial cells were seeded in a 24-well plate and transfected at approximately 80% confluence by using Opti-MEM I and Lipofectamine LTX Reagent (Life Technologies, Carlsbad, CA, USA) according to the manufacturer’s instructions. The stable expressing cell lines were created through the selection by means of Geneticin (G418) or Blasticidin (Sigma-Aldrich, St. Louis, MO, USA). G418- or Blasticidin-resistant clones were obtained after 4 weeks of application.

Anti-piR-DQ593109 and anti-piR-NC (GenePharma, Shanghai, China) were transiently transfected into cells by using Lipofectamine 3000 Reagents (Life Technologies, Carlsbad, CA, USA) according to the protocols of the manufacturer. For co-transfection of shPIWIL1, MEG3(+), and anti-piR-DQ593109, cells that stably knocked down PIWIL1 were first transfected with pGCMV/MEG3/Blasticidin. G418 and Blasticidin dual-resistant clone was selected. qRT-PCR and western blot were performed to measure the transfection efficacy. Cells that stably knocked down PIWIL1 and overexpressed MEG3 were then transiently transfected with anti-piR-DQ593109. Sequences of the small hairpin RNA (shRNA) template are shown in Table S2.

Agomir-330-5p, antagomir-330-5p, and their respective NC were transfected into GECs. For co-transfection of MEG3(+) and agomir-330-5p or RUNX(+) and agomir-330-5p, cells that stably overexpressed MEG3, RUNX3, or their respective NC were transiently transfected with agomir-330-5p or its NC. All the cells were harvested 48 hr after transfection. The transfection efficacy is shown in Figure S.

TEER Assays

TEER assay was performed after the in vitro BTB model was established by using the millicell-ERS instrument (Millipore, Billerica, MA, USA). Before each measurement, to ensure temperature equilibration and uniformity of culture environment, TEER values were measured after 30 min at room temperature directly after mediums were exchanged. Background electrical resistances were subtracted before the final resistances were calculated. TEER values were expressed as Ω·cm2 using the surface area of the transwell insert.

HRP Flux Assays

After in vitro BTB models were established, 1 mL culture medium containing 10 μg/mL HRP (0.5 μM, Sigma-Aldrich, USA) was added into the upper system and 2 mL of culture medium was added to the well. 1 hr later, 5 μL culture medium in the lower chamber was collected from each well and the HRP content was analyzed with a spectrophotometer at 370 nm by using the tetramethylbenzidine colorimetry approach. The final HRP permeability was calculated from the standard curve and expressed as picomole passed per square centimeter surface area per hour (pmol/cm2/hr).

Western Blot Assays

Total proteins were extracted with RIPA buffer (Beyotime Institute of Biotechnology, Jiangsu, China) supplemented with protease inhibitors (10 mg/mL aprotinin, 10 mg/mL phenyl-methylsulfonyl fluoride [PMSF], and 50 mM sodium orthovanadate) and centrifuged at 17,000 × g for 30 minutes at 4°C. Then, the BCA protein assay kit (Beyotime Institute of Biotechnology, Jiangsu, China) was used to determine the protein concentration of the supernatant. Equal amounts of these total proteins (50 μg) were separated by SDS-PAGE and electrically transferred onto a polyvinylidenedifluoride (PVDF) membrane (Millipore, Shanghai, China). After non-specific binding was blocked with 5% nonfat milk in Tris-buffered saline/Tween 20 (TBST) at room temperature for 2 hr, membranes were incubated with primary antibodies as follows: PIWIL1 (1:1,000; Proteintech, USA), RUNX3 (1:800; Abcam, USA), GAPDH (1:10,000; Proteintech, USA), ZO-1 (1:300; Life Technologies, Frederick, MD, USA), occludin (1:1,000; Proteintech, USA), and claudin-5 (1:300; Life Technologies, Frederick, MD, USA) at 4°C overnight. Then, membranes were incubated with HRP-conjugated secondary antibody diluted at 1:5,000 at room temperature for 2 hr. After washing three times with TBST, these protein blots were visualized by an enhanced chemiluminescence kit (ECL) (Santa Cruz Biotechnology, Dallas, TX) and detected by ECL Detection Systems (Thermo Scientific, Beijing, China). Then, the protein bands were scanned using Chemi Imager 5500 V2.03 software, and the integrated light density values (IDVs) were calculated by Fluor Chen 2.0 software, with GAPDH as an internal control.

Immunofluorescence Assays

Cells were fixed with 4% paraformaldehyde for 20 min at room temperature and permeated in PBS containing 0.2% Triton X-100 for 10 min (ZO-1 and claudin-5) or fixed with methanol for 10 min at −20°C (occludin), followed by incubating in 5% BSA blocking buffer for 2 hr at room temperature. Then, cells were incubated with primary antibodies against ZO-1 (1:50; Life Technologies), occludin (1:50; Abcam), and claudin-5 (1:50; Life Technologies) overnight at 4°C. After washing three times with PBS/Tween 20 (PBST), cells were incubated with Alexa-Fluor-555-labeled goat anti-mouse IgG or anti-rabbit IgG secondary antibody (1:500; Beyotime Institute of Biotechnology, Jiangsu, China) for 2 hr at room temperature. Then, the nuclei were counterstained with 0.5 μg/mL DAPI for 5 min. The staining was visualized using confocal microscopy (the confocal microscopy parameters used were gain value of 2, gamma value of 1, and DAPI laser strength of 79%; Alexa Fluor is 68%). The method of piRNA Microarray analysis is shown in the Supplemental Materials and Methods.

Reporter Vector Construction and Dual Luciferase Reporter Assays

The putative piR-DQ593109 binding regions in the MEG3 gene were amplified and cloned downstream of pmirGLO dual-luciferase vector (Promega, Madison, WI, USA), yielding the wide-type plasmid (MEG3-piR-Wt). Meanwhile, the piR-DQ593109 potential binding sequences in MEG3 gene were mutated as indicated and cloned downstream of pmirGLO dual-luciferase vector too, yielding the mutant-type plasmid (MEG3-piR-Mut1, MEG3-piR-Mut2, MEG3-piR-Mut3, and MEG3-piR-Mut4). Similarly, pmirGLO dual-luciferase vectors that contain a putative miR-330-5p binding sequence in the MEG3 gene (MEG3-miR-Wt) or its respective mutant sequence (MEG3-miR-Mut), putative miR-330-5p binding sequence in the RUNX3 3′ UTR (RUNX3-miR-Wt) or its respective mutant sequences (RUNX3-miR-Mut1 and RUNX3-miR-Mut2) were constructed. HEK293T cells were co-transfected with the pmirGLO dual-luciferase vectors with either Wt fragments (or mutation fragments) and pre-piR-DQ593109 (or pre-piR-NC) using Lipofectamine 3000. Dual luciferase was performed 48 hr after transfection. Dual Luciferase Reporter System (Promega) was used to analyze luciferase activity, and the relative luciferase activity was expressed as the ratio of firefly luciferase activity to renilla luciferase activity.

For the reporter constructs, the ZO-1, occludin, and claudin-5 promoter regions (−2,000 to +200 bp) were amplified from human genomic DNA by PCR. In addition, putative RUNX3 binding sites in the PCR conducts were deleted one by one. The PCR products were subcloned into the pGL3 vector (Promega) upstream of a luciferase gene. Human full-length RUNX3 gene was constructed in pEX3 (pGCMV/MCS/Neo) plasmid vector (GenePharma). HEK293T cells were co-transfected with the pGL3 vector with either full-length promoter regions (or deleted promoter regions) and pEX3-RUNX3 (or pEX3 empty vector) using Lipofectamine 3000. Relative luciferase activity was analyzed as described previously.

RNA Immunoprecipitation Assays

Magna RNA-binding protein immunoprecipitation kit (Millipore) was applied in this study to perform RNA immunoprecipitation assay according to the manufacturer’s instructions. Briefly, GECs in different groups were lysed using RNA lysis buffer. Different cell lysate were incubated with anti-human argonaute 2 (Ago2) antibodies (Millipore, USA) and anti-human PIWIL1 (Proteintech, China). IgG (Millipore, USA) was used as a NC. Co-precipitated RNAs were isolated. Next, RNAs were purified and applied to qPCR with reverse transcription analysis.

ChIP Assays

The ChIP assays were performed using the Simple Chip Enzymatic Chromatin IP kit (Cell Signaling Technology, Danvers, MA, USA) according to the manufacturer’s instructions. Briefly, GECs were crosslinked with EBM-2 containing 1% formaldehyde for 10 min and then glycine was added, incubating for 5 min at room temperature to quenched crosslink. Cells were collected in lysis buffer containing PMSF. Further, micrococcal nuclease was used to digest chromatin and incubated for 20 min at 37°C with frequent mixing. Immunoprecipitation was incubated with 3 μg anti-RUNX3 antibody (Abcam, USA) followed by immunoprecipitation with Protein G Agarose Beads in each sample during an overnight incubation at 4°C with gentle shaking. Normal rabbit IgG was used as a NC. Meanwhile, 2% input reference was removed and stored at −20°C before adding antibody. The ChIP DNA crosslink was reversed by 5 mol/L NaCl and Proteinase K at 65°C for 2 hr and then DNA was purified. Immunoprecipitated DNA was amplified by PCR using their specific primers. In each PCR reaction, the corresponding inputs were taken in parallel for PCR validation. PCR products were resolved on a 3% agarose gel. Primers used for ChIP PCR are shown in Table S3.

Statistical Analysis

GraphPad Prism v5.01 (GraphPad, La Jolla, CA) software was used for statistical analysis. All data were presented as the mean ± SD. Statistical analysis of data was performed using the Student’s t test in significant differences between two groups. Moreover, for three or more groups, statistical analysis was performed using one-way ANOVA followed by Dunnett’s post hoc test. p < 0.05 was considered as statistically significant.

Author Contributions

Y.X. and Y.L. conceived and designed the project. S.S. and W.G. performed most of the experiments. J.Z. and J.C. did the statistics analysis. H.Y. and X.L. wrote the manuscript. L.Z. did western blotting experiments. Y.X. and Y.L. contributed to the manuscript revision.

Conflicts of Interest

The authors declare that they have no competing interests.