The TCGA-SKCM dataset was selected from the Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov) and contains 469 melanoma tissue samples. Normal skin tissue samples (n = 813) were selected from the Genotype-Tissue Expression Project (GTEX, https://www.gtexportal.org/). GSE24996 was selected from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) and contains 15 melanoma tissue samples and 8 benign nevi samples.

The Agilent platform was leveraged to generate raw data from the GEO database. Background correction and normalization were achieved using a robust multi-chip averaging (RMA) algorithm. The TCGA and GTEX databases provided RNA sequencing data. Fragment per kilobase (FPKM) values were converted to transcript per kilobase (TPM) values with signal intensities similar to RMA-treated values. Following normalization, data were analyzed using the limma package in R language. The screening standards for miRNA were $|$logFC$|$$>$log${}_2$1.5 and p 0.05. The gene screening standards were $|$logFC$|$$>$1 and p 0.05. Volcano plots were used to show all miRNAs and genes. The heatmap showed the top 45 differentially expressed miRNAs (DEMs) and the top 89 EMT-related DEGs. FPKM values were converted to TPM values with signal intensities similar to RMA-treated values.

Correlation analyses were performed between DEGs and EMT. EMT hallmark gene sets were downloaded from the MSigDB database. For the TCGA-SKCM gene expression profile, the enrichment score for the HALLMARK_EPITHELIAL_MESENCHYMAL_TRANSITION pathway was calculated using Genomic Variation Analysis (GSVA) and EMT activity was assessed using the GSVA R package. DEGs that were significantly associated with this pathway were then identified. Correlation coefficients of $>$0.3 were considered to show a correlation between DEGs and EMT. These are detailed in Supplementary Table 1.

Survival analysis for 89 EMT-related genes was performed using the Kaplan-Meier method and the R language survival package. Cox regression was used to test significance, with p 0.05 considered statistically significant.

The R multiMiR package was used to identify miRNAs that target galectin 1 (LGALS1) in the pita, diana, and targetscan databases. Venn diagrams were used to display common miRNAs present in the three databases.

The multiMiR package was used to search for genes targeted by miR-22-3p. The R language clusterProfiler package was then used for GO and KEGG enrichment pathway analysis.

Normal human skin melanocytes (PIG1) and human melanoma cells (WM-266-4) originated from Shanghai Zeye Biotechnology Co., Ltd. MSCs originated from iCell Bioscience Inc. Both PIG1 and WM-266-4 cells were cultured in DMEM (D5796-500ML, SIGMA) with 10% fetal bovine serum and 1% penicillin-streptomycin. MSCs were cultured in a special medium (human bone marrow MSC complete medium, CM-H166, Procell) for primary MSCs. WM-266-4 cells were separated into three groups: cells in the control group did not undergo any treatment; cells in the mimics NC group were transfected with mimics NC plasmid; cells in the miR-22-3p mimics group were transfected with miR-22-3p mimics plasmid. MSCs were divided into two groups. In the MSCs group, MSCs were cultured normally and the cells collected for evaluation. In the MSC-exos group, MSCs were cultured in serum-free medium, the cell supernatant was then collected and the exosomes extracted. In addition, MSCs were first transfected with mimics NC plasmids and miR-22-3p mimics plasmids, and the cell supernatant subsequently collected for the extraction of exosomes. WM-266-4 cells were divided into four groups. In the NC group, WM-266-4 cells were cultured normally. In the MSC-exos group, MSC-exos were added to the WM-266-4 cells for 12 h. In the MSC-exos${}^{\text{mimics NC}}$ group, WM-266-4 cells were treated with exosomes secreted from MSCs that had previously been transfected with mimics NCs for 12 h. In the MSC-exos${}^{\text{miR-22-3p mimics}}$ group, WM-266-4 cells were treated with exosomes secreted from MSCs that had previously been transfected with miR-22-3p mimics for 12 h. The plasmid transfection steps were: 5 $\mu$L mimics NC and 5 $\mu$L miR-22-3p mimics plasmids were added to 95 $\mu$L of basal medium, mixed gently, and allowed to stand for 5 min. Five $\mu$L of Lipofectamine 2000 (11668019, ThermoFisher) was added to 95 $\mu$L of basal medium. The plasmid and Lipofectamine 2000 were mixed separately and after standing for 20 min, 200 $\mu$L of mixture and 800 $\mu$L of basal medium were added to the cells. The medium was replaced by fresh complete medium after 6 h. Cells and cell supernatant were collected after 48 h for subsequent evaluation. All plasmids were obtained from Genechem Co., Ltd.

MSCs were cultured with serum-free medium for 2 days. The cell supernatant was then collected, centrifuged at 3500 rpm for 15 min, and the pellet discarded. The cell supernatant was mixed with ExoQuick exosome precipitation solution (EXOQ20A-1, System Biosciences Inc., USA) at a ratio of 5:1 and incubated at 4 ${}^{\circ }$C overnight. The mixture was then centrifuged at 2500 rpm for 30 min, with the exosomes contained in the resulting pellet. Two uL of PKH67 reagent (Fluorescence, China) was then added to the resuspended exosomes and incubated at 37 ${}^{\circ }$C for 20 min. Ten uL of serum was added to stop marking. WM-266-4 cells were treated with 100 uL of labeled exosomes and incubated at 37 ${}^{\circ }$C for 1 h. After discarding the medium, cells were washed 3 times with Phosphate Buffered Saline (PBS) and fixed with paraformaldehyde. Nuclei were stained by adding 1 mg/mL of 4’,6-diamidino-2-phenylindole (DAPI) working solution (AWC0291a, Abiowell) for 10 min at 37 °C. Confocal fluorescence microscopy was used to observe the cells after washing with PBS.

TargetScan v.7.0 was used to predict the binding site between LGALS1 and miR-22-3p (http://www.targetscan.org/vert_70/). Plasmid constructs were purchased from HonorGene. The pHG-MirTarget-LGALS1 WT-3U and pHG-MirTarget-LGALS1 MUT-3U plasmids were transfected into 293A cells seeded in 12-well plates. Subsequently, the miR-22-3p mimics plasmid and its empty vector mimics-NC were transfected into 293A cells. Each set of experiments was repeated three times. The activities of firefly luciferase and Renilla luciferase in each group were detected sequentially according to the instructions for the dual luciferase detection kit.

Cells were lysed on ice using Radio Immunoprecipitation Assay (RIPA) lysis buffer (AWB0136, Abiowell). The cell and lysate mixture was then centrifuged at 12,000 rpm for 15 min. The bicinchoninic acid (BCA) protein kit (AWB0104a, Abiowell) was used to measure protein concentrations. The marker and the denatured protein were mixed with loading buffer and subjected to agarose gel electrophoresis for 2 h. After cutting the gel according to molecular weight, the proteins were transferred to PVDF membranes. The membranes were blocked with 5% skimmed milk powder for 90 min at room temperature. Next, the membranes were incubated overnight at 4 ${}^{\circ }$C with primary antibodies to CD63 protein (CD63) (25682-1-AP, proteintech, US), Heat Shock Protein Family A (Hsp70) Member 4 (HSPA4) (25405-1-AP, proteintech, US), galectin 1 (LGALS1) (11858-1-AP, proteintech, US), carboxypeptidase X, M14 family member 1 (CPXM1) (orb467709, biorbyt, UK), apelin receptor (APLNR) (20341-1-AP, proteintech, US), Vimentin (VIM) (10366-1-AP, proteintech, US), snail family transcriptional repressor 2 (SNAI2) (12129-1-AP, proteintech, US), cadherin 1 (CDH1) (20874-1-AP, proteintech, US), actin beta (ACTB) (66009-1-Ig, proteintech, US). ACTB was used as an internal control parameter. The membranes were then incubated with the corresponding secondary antibodies HRP goat anti-mouse IgG (SA00001-1, proteintech, US) and HRP goat anti-rabbit IgG (SA00001-2, proteintech, US) for 90 min at room temperature. Enhanced chemiluminescence (ECL) solution (AWB0005a, Abiowell) was used to reveal the bands, which were then photographed using a gel imaging system.

Total RNA was extracted from cells according to instructions for the Trizol kit (15596026, Thermofisher). miRNA and mRNA were then transcribed into cDNA according to instructions for the miRNA cDNA Synthesis Kit (CW2141, Cowin bio.) and the HiFscript cDNA Synthesis Kit (CW2569, Cowin bio.), respectively. Primer 5.0 software (PREMIER Biosoft, San Francisco, California, USA) was used to design primers, which were then synthesized by Beijing Qingke Biotechnology. The primer sequences are shown in Table 1. Next, the primers were added to cDNA, ddH${}_2$O, and the UltraSYBR Mixture (CW2601, Cowin bio.). qRT-PCR was performed using ACTB as the internal reference for mRNA and RNU6V as the internal reference for miRNA. The qRT-PCR amplification program was: 95 °C for 10 min (pre-denaturation), 95 °C for 15 s, and 60 °C for 30 s (denaturation, annealing, and extension; 40 cycles in total). Preliminary experiments confirmed that the efficiency of all primers was between 90–110%. Samples were tested in triplicate and the 2${}^{-△△ct}$ method was used to analyze experimental data.

The cells in different groups were digested, counted, and seeded into 96 well plates at a density of 1 $\times$ 10${}^4$ cells/well. Three wells were used for each group. Following specific treatment of the different groups, the culture medium was discarded and 10 $\mu$L of CCK8 working solution (NU679, Dojindo) and 90 $\mu$L of complete medium were added to each well. The cells were incubated at 37 °C for 4 h and a Bio Tek microplate reader was then used to measure the absorbance at 450 nm.

Cells were grouped and then digested with trypsin, with PBS used as a washing agent. Experiments were performed strictly according to instructions for the Annexin V-APC/PI Apoptosis Detection Kit (KeyGEN BioTECH, KGA1030). Briefly, 500 uL of binding buffer was used to suspend the cells prior to testing, followed by the addition of 5 uL Annexin V-APC and 5 uL of propidium iodide. The mixture was left to stand for 10 min at room temperature and protected from light. Flow cytometry was then used for the measurement of fluorescence.

All cell experiments were conducted in triplicate. Data for each group are shown as the mean $\pm$ standard deviation (SD) and were analyzed using Graphpad Prism 9.0 (GraphPad Software LLC, San Diego, California, USA). Student’s t-test was used to compare two data groups, while ordinary one-way ANOVA was used to compare three data groups. p 0.05 was considered statistically significant.

The TCGA-SKCM and GTEX datasets were screened to identify DEGs between the melanoma and control groups, while data from the GSE24996 dataset was screened to identify DEMs between these groups. All DEMs (Fig. 1A) and DEGs (Fig. 1B) are shown in the volcano plot, while the heatmap shows the top 45 DEMs (Fig. 1C). DEMs included miR-211-5p, miR-205-5p, miR-125b-5p, miR-200c-3p, miR-768-3p, and miR-22-3p.

Fig. 1.

Screening for DEMs and DEGs. (A) Volcano plot of all miRNAs that were differently expressed between the melanoma and control groups in the GEO database. (B) Volcano plot of all mRNAs that were differently expressed between the melanoma and control groups in the TCGA and GTEX databases. Red dots indicate significant up-regulation and down-regulation of DEMs and DEGs between the two groups ($|$logFC$|$$>$log${}_2$1.5, p-value 0.05). (C) Heatmap showing the top 45 DEMs between the melanoma and control groups. Darker red indicates higher expression, while darker blue indicates lower expression.

The 89 EMT-related DEGs are shown in the heatmap (Fig. 2A), and their correlation coefficients shown in Supplementary Table 1. EMT-related DEGs included ADAM19, ITGAM, DCSTAMP, PLAUR, CSF3R, LGALS1, CPXM1, and APLNR. Survival analysis was performed according to the expression of each of these 89 genes. Kaplan-Meier analysis showed that expression of three genes (LGALS1, CPXM1, APLNR) was significantly associated with patient survival. Low gene expression in the tumor tissue correlated with better patient survival (Fig. 2B). Next, Boxplots were drawn based on the expression level of these three genes in the melanoma and control groups. These showed that the expression of all three genes was higher in the melanoma group compared to the control group (Fig. 2C).

Fig. 2.

Survival analyses for the expression of three DEGs associated with EMT. (A) Heatmap of 89 DEGs identified from the TCGA and the GTEX databases and associated with EMT. Darker red corresponds to higher expression and darker blue corresponds to lower expression. (B) Survival curves according to expression of the DEGs LGALS1, CPXM1, and APLNR. The red line represents the survival probability of patients with high gene expression, while the blue line represents the survival probability of patients with low gene expression. (C) Boxplots show the expression level of the three DEGs LGALS1, CPXM1, and APLNR in the melanoma and control groups. *p 0.05 Melanoma group vs. Control group.

To confirm the observed differences in expression of the above three DEGs, we compared the expression levels of the LGALS1, CPXM1, and APLNR genes (Fig. 3A) and the LGALS1, CPXM1, and APLNR proteins (Fig. 3B) between PIG1 and WM-266-4 cells. Both the mRNA and protein levels for all three DEGs were significantly higher in WM-266-4 cells than in PIG1 cells, with the difference being greatest for LGALS1.

Fig. 3.

LGALS1 is highly expressed in melanoma cells. (A) qRT-PCR was performed to evaluate LGALS1, CPXM1 and APLNR mRNA expression in PIG1 and WM-266-4 cells. (B) Western blot was used to evaluate the expression of LGALS1, CPXM1 and APLNR protein in PIG1 and WM-266-4 cells. *p 0.05 WM-266-4 cells vs. PIG1 cells.

LGALS1 was selected for the follow-up research because it showed the most difference in expression between melanoma and control. Three databases (pita, diana, and targetscan) were used to identify miRNAs predicted to target LGALS1. Only one miRNA, miR-22-3p, was found in all three databases to target LGALS1, as shown by Venn diagram (Fig. 4A). Boxplots showed that the expression of miR-22-3p in the melanoma group was lower than in the control group (Fig. 4B). Next, GO analysis and KEGG analysis were performed (Fig. 4C). GO analysis showed that in biological processes (BP), LGALS1-targeted miR-22-3p was significantly associated with endocytosis, platinum drug resistance, the sphingolipid signaling pathway, the AMPK signaling pathway, and proteoglycans in cancer. In cellular components (CC), LGALS1-targeted miR-22-3p bound to proximal promoter sequence-specific DNA binding, protein serine/threonine kinase activity, transcription cofactor binding, RNA polymerase II proximal promoter sequence-specific DNA binding, and enzyme activator activity. For molecular function (MF), LGALS1-targeted miR-22-3p was related to the dendritic spine, neuron spine, spindle, main axon, and cell leading edge. The KEGG signaling pathway was mainly enriched in learning or memory, cognition, regulation of GTPase activity, cell morphogenesis involved in neuron differentiation, regulation of cell morphogenesis, and regulation of neuron death. Next, qRT-PCR was used to confirm the expression of miR-22-3p in cells. The miR-22-3p level in PIG1 cells was higher than in WM-266-4 cells (Fig. 4D). The targeted binding of miR-22-3p to LGALS1 was predicted by LncBase Predicted v.2 (Fig. 4E) and confirmed using a dual luciferase experiment (Fig. 4F). The results demonstrated that miR-22-3p targeted LGALS1.

Fig. 4.

Screening for LGALS1-interacting miRNAs. (A) Venn diagram of miRNAs identified to target LGALS1 in the pita, diana and targetscan databases. Only one DEM, miR-22-3p, was identified in all three databases. (B) Confirmation of miR-22-3p expression in the GEO database. (C) GO and KEGG functional enrichment analysis of miR-22-3p. Blue represents biological process (BP)-related functions in GO analysis. Yellow represents functions associated with cellular components (CC) in GO analysis. Red represents functions related to molecular function (MF) in GO analysis. Green represents KEGG functional analysis. (D) qRT-PCR showing miR-22-3p expression in cells. (E) Bioinformatics prediction of the LGALS1 binding site for miR-22-3p. (F) Dual luciferase experiments were performed to verify the binding of LGALS1 to miR-22-3p. *p 0.05 vs. Control group, ${}^{\mathrm{\#}}$p 0.05 vs. PIG1 group, p 0.05 vs. mimics NC group.

The effect of miR-22-3p on EMT was investigated by overexpressing miR-22-3p in WM-266-4 cells. The qRT-PCR result showed that the expression of miR-22-3p in the miR-22-3p mimics group was significantly higher than in the mimics NC group, indicating successful overexpression of miR-22-3p in WM-266-4 cells (Fig. 5A). Cell viability was found to decrease significantly following the overexpression of miR-22-3p (Fig. 5B). In addition, flow cytometry showed that cell apoptosis increased after transfection with miR-22-3p mimics (Fig. 5D). The mRNA expression level of LGALS1 and other EMT-related genes (CDH1, VIM, and SNAI2), together with their protein expression level, was examined by qRT-PCR and western blot, respectively (Fig. 5C,E). The expression levels of LGALS1 and LGALS1 were significantly lower in the miR-22-3p mimics group than in the mimics NC group. Furthermore, the expression of LGALS1 and LGALS1 was negatively correlated with miR-22-3p. The expression of CDH1 and CDH1 was significantly higher in the miR-22-3p mimics group than in the mimics NC group, while the mRNA expression levels for VIM, VIM, SNAI2, and SNAI2 were significantly lower in the miR-22-3p mimics group. These results indicate that miR-22-3p can inhibit EMT.

Fig. 5.

miR-22-3p regulates EMT. (A) The expression level of miR-22-3p in WM-266-4 cells was evaluated by qRT-PCR. (B) CCK8 was used to determine WM-266-4 cell viability, with absorbance measured at 450 nm. (C) qRT-PCR was used to determine the cellular mRNA levels for LGALS1, CDH1, VIM, and SNAI2. (D) Detection of apoptosis by flow cytometry. (E) Cellular protein levels for LGALS1, CDH1, VIM, and SNAI2 were determined by western blot. *p 0.05 vs. mimics NC group.

The exosomes secreted by MSCs were isolated and observed by transmission electron microscopy (TEM) to be round vesicles with a diameter of about 50 nm (Fig. 6A). Western blot assay was used to detect the exosome-associated proteins CD63 and HSPA4. Both proteins were expressed in the MSC-exos group (Fig. 6B). Next, PCR revealed that miR-22-3p expression was higher in the MSC-exos group than in the MSC group (Fig. 6C). These results indicate that exosomes secreted by MSCs contain a large amount of miR-22-3p. The uptake of exosomes by WM-266-4 cells was observed by confocal fluorescence microscopy (Fig. 6D).

Fig. 6.

Identification of MSC-EXOs and uptake of exosomes by melanoma cells. (A) Exosome morphology examined by TEM. The diameter of exosomes was about 50 nm (scale bar = 100 nm). (B) Western blot technique was used to determine protein levels for the exosome markers CD63 and HSPA4. (C) qRT-PCR was used to determine miR-22-3p levels in cells and exosomes. (D) PKH67 labeling was used to observe exosome uptake. Green represents exosomes, while blue represents cell nuclei. *p 0.05 vs. MSCs.

The amount of miR-22-3p in each cell group was evaluated. The expression of miR-22-3p in the MSC-exos${}^{\text{miR-22-3p mimics}}$ group was significantly higher than in the MSC-exos${}^{\text{mimics NC}}$ group (Fig. 7A), demonstrating that miR-22-3p was successfully overexpressed in exosomes. Next, the CCK8 assay was used to measure the proliferative activity of each cell group. The proliferation of cells in the MSC-exos group was slower than that of cells in the NC group (Fig. 7B). In contrast, the proliferation of cells in the MSC-exos${}^{\text{miR-22-3p mimics}}$ group was slower than that of cells in the MSC-exos${}^{\text{mimics NC}}$ group. This results indicates that miR-22-3p in exosomes secreted by MSC-exos can inhibit the proliferation of WM-266-4 cells. Furthermore, flow cytometry showed that exosomes carrying miR-22-3p can promote cell apoptosis (Fig. 7C). The expression levels of LGALS1, LGALS1, VIM, VIM, SNAI2, and SNAI2 in cells from the MSC-exos group were significantly lower than in the NC group, whereas the expression level of CDH1 and CDH1 in MSC-exos${}^{\text{miR-22-3p mimics}}$ cells was significantly increased (Fig. 7D,E). The expression levels of LGALS1, LGALS1, VIM, VIM, SNAI2, and SNAI2 in MSC-exos${}^{\text{miR-22-3p mimics}}$ cells were significantly lower compared to the MSC-exos${}^{\text{mimics NC}}$ group, whereas the expression of CDH1 and CDH1 in MSC-exos${}^{\text{miR-22-3p mimics}}$ cells was significantly increased.

Fig. 7.

miR-22-3p in MSC-exos inhibited EMT of melanoma cells by regulating LGALS1. (A) The expression level of miR-22-3p in WM-266-4 cells was evaluated by qRT-PCR. (B) The cck8 assay was used to evaluate the viability of WM-266-4 cells by measuring absorbance at 450 nM. (C) Cell apoptosis was detected by flow cytometry. (D) qRT-PCR was used to measure mRNA levels for LGALS1, CDH1, VIM, and SNAI2 in WM-266-4 cells. (E) Western blot was used to determine the protein levels of LGALS1, CDH1, VIM, and SNAI2 in WM-266-4 cells. *p 0.05 vs. NC. #p 0.05 vs. MSC-exos ${}^{\text{mimics NC}}$ group.