Exosomal miR-224 was screened to be associated with LNM of GC
Two non-coding RNA serum expression profiles were selected and screened from the GEO database, which were GSE112264 and GSE85589. We downloaded the expression profile data and clinical information of the two datasets and screened gastric cancer (GC) serum and health control (HC) serum. The GSE112264 datasets contained 50 GC serum and 41 HC serum, while the GSE85589 datasets contained 7 GC serum and 19 HC serum. The differential analysis was performed between GC serum and HC serum extracted from GSE112264 and GSE85589, respectively. In GSE112264, 1790 differential serum miRNAs were screened, and 108 differential serum miRNAs were screened in GSE85589. In addition, 18 human-derived differential serum miRNAs (DE-miRNAs) were screened out from the serum miRNA dataset analyzed by sequencing (LNM vs nonLNM). We found that miR-224-3p was shown in Fig. 1A. The Venn diagram of the three datasets was intersected to obtain an up-regulated miRNA (miR-224-3p) and a down-regulated miRNA (miR-873-5p) related to lymph node metastasis (LNM) of GC (Fig. 1B). In this study, we focused on miR-224-3p.
We extracted and analyzed the serum expression profile data of miR-224-3p. As shown in Fig. 1C, compared with the HC group, the relative expression of serum miR-224-3p in the GC group was increased, suggesting that miR-224-3p may be involved in GC progression. Then, RNA was extracted from the collected serum of 65 patients with newly diagnosed GC, and the relative expression of miR-224-3p was detected by quantitative real-time PCR (QRT-PCR). It was found in Fig. 1D that the expression of miR-224-3p in serum of LNM patients (n = 38) was higher than that of nonLNM patients (n = 27), which demonstrated that serum miR-224-3p was probably associated with LNM in GC.
Vascular endothelial growth factor C (VEGFC) is the most important cytokine to promote lymphangiogenesis. Therefore, we detected the content of serum VEGFC in GC and explored the correlation between miR-224-3p and VEGFC in serum. As shown in Fig. 1E, there was almost no correlation between serum miR-224-3p and serum VEGFC. We extracted RNA from the serum exosomes of patients and the serum after ultracentrifugation and measured the relative expression of miR-224-3p in each component by QRT-PCR. It was found that the relative expression level of exo-miR-224-3p in the serum of patients was significantly higher than that in the serum which excluded exosomes by ultracentrifugation (Fig. 1F). The results suggested that miR-224-3p in serum mainly existed in exosomes.
To further explore the relationship between serum miR-224-3p and LNM of GC, tumor tissues of 32 GC patients were collected for immunohistochemistry (IHC), and the lymphatic endothelial marker Padoplanin was used to evaluate the lymphatic vessel density (LVD) of tumor tissues. Meanwhile, the expression level of serum miR-224-3 in 32 patients was detected by QRT-PCR to explore the correlation between them. As shown in Fig. 1G, the amount of LVD in the LNM group was significantly higher than that in the nonLNM group. There was a positive correlation between GC serum miR-224-3p and LVD (Fig. 1H). The above results indicated that miR-224-3p was associated with LNM in GC, while it may regulate LNM of GC in a VEGFC-independent manner.
Exosomes derived from GC cells could deliver miR-224-3p into HLECs
We have found that the expression of miR-224-3p was increased in the serum of GC patients with LNM. Then, we intend to investigate the expression of miR-224-3p in GC cell lines and their exosomes. As shown in Fig. 2A, compared with the normal gastric mucosal epithelial cell line GES-1, the expression of miR-224-3p was increased in the GC cell lines HGC27, AGS, and MKN45. Meanwhile, we extracted exosomes from each cell culture medium to detect the expression of miR-224-3p. The trend of exosomal miR-224-3p expression level secreted by each cell line was roughly the same as that secreted by cell lines (Fig. 2B). The expression of miR-224-3p was the highest in HGC27, and the expression of exo-miR-224-3p was also the highest in the culture medium. The expression of miR-224-3p in MKN45 and its exosomes was also higher than that of GES-1. While the expression of exo-miR-224-3p derived from the AGS culture medium was not significantly different from that of GES-1. Therefore, HGC27 and MKN45 were selected as exosome donors in the subsequent study.
We collected the medium of HGC27 and MKN45 excluded serum exosomes, then used differential gradient centrifugation to obtain a batch of purified HGC27/MKN45-derived particles. To prove that these extracted particles were exosomes, we detected the characteristic proteins CD9, CD63 and TSG101 expressed by exosomes. In Fig. 2C, we found that three proteins are highly expressed in the particles obtained by this method. Under the electron microscope, these particles appeared as round or oval discs with a diameter of about 100nm (Fig. 2D and 2E). Nanoparticle tracer analysis (NTA) displayed that the size of HGC-derived particles was 112.7 ± 25.8nm, and the maximum size distribution was about 113nm. The size of MKN45-derived particles was 109.2 ± 32.3.8nm, and the maximum size distribution was about 110nm. These results demonstrated that we have successfully obtained purified exosomes for the subsequent experiments.
Exosomes obtained by ultracentrifugation were labeled with fluorescent dye PKH26, and then co-cultured with human lymphatic endothelial cells (HLECs), as shown in Fig. 2F. After 4–6 hours, we could observe a large number of co-cultured exosomes endocytosed by HLECs or bound around HLECs under the fluorescence microscope (Fig. 2G). In addition, we extracted RNA from HLECs co-cultured with exosomes after 0, 6, 12, 24, and 36 hours, respectively, and detected the expression level of miR-224-3p. In Fig. 2H, it was revealed that the relative expression of miR-224-3p was increased gradually with the prolongation of co-culture time.
Exo-miR-224-3p from GC cells promoted tube formation and migration of HLECs
MKN45 and HGC27 were transfected with miR-224-3p mimic (MKN45/miR-224-3p-exo) or miR-224-3p inhibitor (HGC27/anti-miR-224-3p-exo) and corresponding normal control (NC), and were cultured with the medium excluded exosomes for 48–72 hours. Exosomes were extracted from the culture medium to detect the relative expression of miR-224-3p. Compared with the NC group, the expression level of miR-224-3p in the MKN45/miR-224-3p-exo group was significantly increased (Fig. 3A). However, the expression level of miR-224-3p was observably decreased in the HGC27/anti-miR-224-3p-exo group (Fig. 3C). The above results indicated that exosomes derived from GC could function as mediators to regulate the expression level of miR-224-3p in HLECs and facilitate the communication between HLECs and GC cells in the microenvironment.
Next, we wonder about the effect of exo-miR-224-3p on the biological behavior of HLECs. The number of lymphatic branch tubes formed by HLECs cells in matric gel represented the lymphangiogenesis ability, and the Transwell assay was used to detect the change in the migration ability. As shown in Fig. 3B and 3D, compared with the NC group, HLECs in the MKN45/miR-224-3p-exo group significantly enhanced the lymphangiogenesis ability and migration ability. However, these abilities of HLECs in the HGC27/anti-miR-224-3p-exo group were certain impaired. To further confirm that the stronger lymphangiogenic ability of HLECs is caused by the change of miR-224-3p expression in exosomes, we performed the rescue experiment. The results were shown in Fig. 3E, which proved that exosomal miR-224-3p indeed affected the biological functions of HLECs.
Exo-miR-224-3p derived from GC cells could promote LNM in vivo
We infected MKN45 with lentivirus overexpressing luciferase. The aminoglycoside antibiotic G418 was used in the drug screening test for the stable line with high expression of luciferase (MKN45-Luci). Then, MKN45-Luci cells were infected with lentivirus overexpressing miR-224-3p or sponge miR-224-3p and corresponding NC. Finally, they were screened by puromycin to obtain MKN45-Luci-OE, MKN45-Luci-SP, and MKN45-Luci-NC.
The constructed MKN45-Luci-OE, MKN45-Luci-SP, and MKN45-Luci-NC cells were implanted into the foot pads of nude mice (n = 5 in each group). The basic status of nude mice and the growth of tumors on foot pads were recorded every three days. After 6–8 weeks, luciferase substrate was intraperitoneally injected into nude mice, and in vivo imaging was performed to observe tumor metastasis. When the size of tumors on foot pads grew to more than 200 mm3, the nude mice were sacrificed (Fig. 4A). Meanwhile, the in situ and popliteal lymph nodes of nude mice were removed, as shown in Fig. 4B and 4C. The results of in vivo imaging displayed that four nude mice in the MKN45-Luci-OE group developed popliteal LNM, with a metastasis rate of 80%. No metastases were found in the MKN45-Luci-SP group, whereas two metastases occurred in the MKN45-Luci-NC group (Fig. 4D and 4I). To more accurately observe the LNM status of xenografts, we used luciferase antibody to immunostain the popliteal lymph nodes. It was found that compared with the MKN45-Luci-NC group, the MKN45-Luci-OE group had a higher positive rate of luciferase cells, while the MKN45-Luci-SP group had almost no luciferase expression (Fig. 4E). The volume and weight of lymph nodes in the three groups were exhibited in Fig. 4F and 4G. In addition, we measured the expression of miR-224-3p in the serum of nude mice in each group. The expression of serum miR-224-3p in the MKN45-Luci-OE group was higher than that in the MKN45-Luci-NC group, and the expression of serum miR-224-3p in the MKN45-Luci-SP group was lower than that in MKN45-Luci-NC group (Fig. 4H). Based on these results, it was demonstrated that overexpression of miR-224-3p could promote LNM of GC in vivo.
Exo-miR-224-3p derived from GC cells could promote lymphangiogenesis in vivo
To explore the effect of miR-224-3p on lymphangiogenesis in vivo, the constructed MKN45-Luci cells were implanted subcutaneously in nude mice (Fig. 5A). One week later, the basic condition and tumor growth of nude mice were observed. When the volume of subcutaneous tumors reached about 20 mm3, nude mice were randomly divided into two groups (n = 6 in each group). In the first group, exosomes with high expression of miR-224-3p (MKN45/ miR-224-3p-exo) were injected around the transplanted tumors, and in the second group, normal control exosomes were injected around the transplanted tumors (MKN45/NC-exo). Tumor growth was measured and recorded every three days. When the tumor in situ grew to more than 250 mm3, in vivo imaging of small animals was performed to detect these tumors. In Fig. 5B, we could see that the fluorescence area and intensity of the MKN45/miR-224-3p group were larger than those of the MKN45/NC-exo group. After that, nude mice were sacrificed and transplanted tumors were removed. The tumor tissues in each group were shown in Fig. 5C. The curve of tumor growth volume change suggested that the MKN45/miR-224-2p-exo group had faster tumor growth and larger tumor volume (Fig. 5D). The volume and weight of tumor tissues in each group were shown in Fig. 5E and 5F. Ki67 antibody, a tumor cell proliferation marker, was used to detect the proliferation of tumor cells in the transplanted tumor tissues by IHC[27]. It was found that the proliferation of tumor cells in the MKN45/miR-224-3p-exo group became vigorous, which was consistent with the trend of tumor volume and weight. Furthermore, to verify the specific effects of exo-miR-224-3p on lymphangiogenesis in GC, we used the podoplanin antibody to perform IHC on the removed tumor tissues. Podoplanin is a specific marker of HLECs that can reveal the status of lymphangiogenesis in and around tumors by IHC. The results indicated that the density of lymphatic vessels around tumor tissues was higher in the MKN45/miR-224-3p-exo group than in the MKN45/NC-exo group (Fig. 5H). In conclusion, we confirmed that exo-miR-224-3p could promote tumor lymphangiogenesis in vivo.
Exo-miR-224-3p targeted GSK3B to promote lymphangiogenesis
We combined an online database, miWalk (http://mirwalk.umm.uni-heidelberg.de/), DIANA (http://diana.imis.athenainnovation.gr/DianaTools/index.php?r=microT_CDS/index), and TargetScan (http://www.targetscan.org/vert_72/) to explore miR-224-3p potential targets, which were downstream targets controlled by miR-224-3p. We found a binding site between miR-224-3p and the 3'UTR region of GSK3B. HEK293FT and HLECs were used to verify the targeting regulatory relationship between miR-224-3p and GSK3B by the dual luciferase reporter assay. There was divided into two groups, wild-type (WT) and mutant type (MUT), which was mutated plasmid targeting the binding site. The binding site of miR-224-3p and GSK3B mRNA 3'UTR predicted by the website was shown in Fig. 6A. In HEK293FT and HLECs, the fluorescence value of the WT group transfected with miR-224-3p mimic was significantly decreased compared with that of the NC group, while the fluorescence value of the WT group transfected with miR-224-3p inhibitor was increased. However, there was no significant difference in fluorescence value of the MUT group, whether transfected with miR-224-3p mimic or miR-224-3p inhibitor (Fig. 6B).
We continued to explore the specific regulatory relationship between miR-224-3p and GSK3B. We collected exosomes derived from different groups (MKN45/miR-224-2p-exo, MKN45/NC-exo, HGC27/anti-miR-224-3p-exo, and HGC27/NC-exo) and co-cultured them with HLECs, respectively. The expression of GSK3B was detected by Western blot (WB). It was exhibited in Fig. 6C that the protein level of GSK3B in the MKN45/miR-224-3p-exo group was significantly decreased compared with the MKN45/NC-exo group, while the protein level of GSK3B in the HGC27/anti-miR-224-3p-exo group was higher than that in the HGC27/NC-exo group. There was no significant difference in GSK3B mRNA level among the above co-cultured groups by QRT-PCR (Fig. 6D). The results suggested that the GSK3B protein level may be regulated by post-transcription, and the expression of GSK3B in HLECs could be suppressed by exo-miR-224-3p derived from GC.
We further sought to explore the possible mechanism of altered GSK3B expression affecting lymphangiogenesis. In the classical Wnt-β-catenin signaling pathway, β-catenin in the cytoplasm is phosphorylated and degraded by GSK3B. The inhibition of the GSK3B protein expression level leads to the accumulation of β-catenin in the cytoplasm, and the activation of β-catenin into the nucleus as a transcription factor to promote gene transcription[24, 25]. It has been reported that the entry of β-catenin into the nucleus could enhance the expression of Prospero homeobox 1 (PROX1), thereby promoting lymphangiogenesis[26]. Next, we designed to verify whether the change of miR-224-3p expression level would affect the stable entry of β-catenin into the nucleus and promote the expression of PROX1 after affecting the expression of GSK3B. As shown in Fig. 6C, inhibiting the expression of GSK3B led to the phosphorylation level of β-catenin being decreased. The degradation of β-catenin was reduced, and this caused β-catenin stabilization.
To verify whether β-catenin aggregates in cytoplasm and increases in nuclear entry, we used the MKN45/miR-224-3p-exo group and the MKN45/NC-exo group, as well as the HGC27/anti-miR-224-3p-exo group and the HGC27/NC-exo group for immunofluorescence to detect the expression and distribution of β-catenin. After co-culture with MKN45/miR-224-3p-exo and HLECs, β-catenin accumulated in the cytoplasm and increased into the nucleus of HLECs compared with the MKN45/NC-exo group. However, compared with the HGC27/NC-exo group, the accumulation and nuclear localization of β-catenin in HLECs of the HGC27/anti-miR-224-3p-exo group was decreased (Fig. 6E). Meanwhile, we detected the expression of PROX1 in the lysates of each group. It was found that compared with the MKN45/NC-exo group, the expression level of PROX1 in the MKN45/miR-224-3p-exo group was up-regulated, while the expression level of PROX1 in the HGC27/ anti-miR-224-3p-exo group was down-regulated.
To clarify that GSK3B is the direct target of exo-miR-224-3p affecting HLECs, we transfected GSK3B overexpression plasmid into HLECs (GSK3B OE), and then co-cultured it with exosomes derived from GC cells transfected with miR-224-3p mimic (GSK3B OE + MKN45/miR-224-3p-exo). The results of WB were shown in Fig. 6F. After overexpression of GSK3B, the protein level of p-β-catenin was significantly increased, while that of β-catenin was slightly decreased, indicating that GSK3B promoted the phosphorylation and degradation of β-catenin by inhibiting the β-catenin signaling pathway. However, compared with the GSK3B OE group, the expression level of p-β-catenin was partially suppressed and β-catenin expression was slightly increased in the GSK3B OE + MKN45/miR-224-3p-exo group. As shown in Fig. 6G, the tube formation ability of HLECs in the GSK3B group was significantly decreased compared with that in the NC group. While we observed that the tube formation ability of HLECs in the MKN45/miR-224-3p-exo group was somewhat improved, this ability was still worse than that in the NC group. In addition, immunofluorescence experiments were performed on the above treatment groups (Fig. 6H). The intracellular nuclear localization of β-catenin in the GSK3B OE group decreased significantly. And the intracellular nuclear localization of β-catenin in the GSK3B OE + MKN45/miR-224-3p-exo group increased compared with that of the GSK3B OE group. We also found that the expression level of PROX1 decreased in the GSK3B OE group, and increased slightly in the GSK3B OE + MKN45/miR-224-3p-exo group. In conclusion, the exo-miR-224-3p derived from GC directly targets GSK3B in HLECs, then inhibits its phosphorylation and degradation of β-catenin, and promotes the nuclear activation of β-catenin, leading to the transcription of PROX1, and thus affects lymphangiogenesis.
hnRNPA1 mediated the sorting of miR-224-3p into exosomes
To investigate how miR-224-3p is specifically sorted into exosomes in GC cells, we analyzed the specific interactions between miR-224-3p sequences and RNA-binding protein (RBP) motifs via the RBPDB database (database of RBP specificities, http://rbpdb.ccbr). The analysis exhibited that there were three RBPS with scores greater than 5 and binding bases greater than 4 to miR-224-3p (Fig. 7A). They are YTHDC1 (YTH Domain Containing 1), ACO1 (Aconitase 1), and hnRNPA1 (heterogeneous nuclear ribonucleoprotein A1). We designed specific siRNA for YTHDC1, ACO1, and hnRNPA1 to knock down their expression in HGC27. These siRNAs could effectively reduce the expression levels of YTHDC1, ACO1, and hnRNPA1 (Fig. 7B-7D). We found that the expression level of miR-224-3p in exosomes transfected with YTHDC1 siRNA or ACO1 siRNA did not change significantly, while the expression level of miR-224-3p in exosomes transfected with hnRNPA1 siRNA decreased (Fig. 7E-7G). These results suggested that hnRNPA1 could affect the expression level of exo-miR224-3p, but YTHDC1 and ACO1 could not.
We hypothesized that hnRNPA1 might affect the abundance of exo-miR-224-3p by affecting the expression of total miR-224-3p in cells. To verify this hypothesis, HGC27 was transfected with hnRNPA1 siRNA. The results showed that the knockdown of hnRNPA1 had no significant effect on the expression of miR-224-3p (Fig. 7H). In recent years, it has been reported that the heterogeneous nuclear ribonucleoprotein (hnRNPs) family interacted with miRNA motifs to mediate the sorting of miRNAs into exosomes, such as hnRNPA2B1[28]. Combined with the pieces of literature and our results, it is suggested that hnRNPA1 may be involved in the sorting of miR-224-3p into exosomes. It was shown in Fig. 2I that the motif of miR-224-3p binding to hnRNPA1 was predicted by the RPBDP database.
To more fully explain that hnRNPA1 may participate in the sorting of miR-224-3p into exosomes, we designed a plasmid with overexpressing hnRNPA1, and the effect was shown in Fig. 7J. HGC27 was transfected with the plasmid, and the expression levels of miR-224-3p in exosomes and cells were detected by QRT-PCR. We found that after overexpression of hnRNPA1, the level of exosomal miR-224-3p increased, but the increase was not obvious in HGC27 (Fig. 7K and 7L). In Fig. 7M, it was proved that hnRNPA1 was effectively knockdown or overexpressed in HGC27. Moreover, targeting the predicted binding site of hnRNPA1 and miR-224-3p (UAGUGA sequence), we synthesized miR-224-3p with biotin (bio-Wnt), mutated miRNA with biotin (bio-Mut) and bio-negative control sequence with biotin (bio-Negative) as shown in Fig. 7N. Figure 7O displayed the results of miRNA pull-down. The interaction between hnRNPA1 and miR-224-3p was observed in whole cell lysates, and miR-224-3p could pull hnRNPA1 out. However, when the UAGUGA sequence of miR-224-3p was mutated, its binding ability to hnRNPA1 was certain impaired. The enrichment of hnRNPA1 in the bio-Wnt group was also found in exosome lysates. We found that hnRNPA1 played a regulatory role in the level of exo-miR-224-3p and mediated the sorting of miR-224-3p into exosomes.
Finally, we investigated the effect of PKM2 on the secretion of exo-miR-224-3p. We transfected HGC27 with siRNA to reduce the expression of PKM2 and detected the expression of miR-224-3p in cells and exosomes (Fig. 7P-7S). Compared with the NC group, the exo-miR-224-3p from cell lines with down-expression of PKM2 was significantly reduced, while the miR-224-3p in the cell lines was not significantly changed, which suggested that PKM2 could promote the secretion of exo-miR-224-3p derived from GC cells. However, the specific mechanism of how it promotes secretion needs to be further explored. Furthermore, we note that hnRNPA1 as a splicing protein has been reported to splice intracellular PKM in the direction of PKM2. Therefore, we transfected hnRNPA1 overexpressing plasmid or siRNA and corresponding NC into HGC27. Overexpression of hnRNPA1 could increase the mRNA and protein expression of PKM2, while knockdown of hnRNPA1 down-regulated both mRNA and protein expression of PKM2 (Fig. 7T-7V). The above results indicated that hnRNPA1 promoted the secretion of exo-miR-224-3p by enhancing the expression of PKM2, together with PKM2.