3.1 Single-cell analysis of melanoma tumor tissues

Figure 1 illustrated the overall workflow of this study. At the beginning, we obtained single-cell data from 9 melanoma samples (GSM5708993-GSM5709001) representing tumors from 8 patients, including 5 primary and 4 metastatic tumor tissues. Employing rigorous data filtration and dimensionality reduction clustering techniques, we subdivided the melanoma tumor tissue into 34 subpopulations (Fig. 2A). Investigation into the origins and metastatic attributions of these cells revealed heterogeneous distribution patterns across the subpopulations, indicating diverse cellular compositions within the TME. The study mainly focused on primary acral melanoma tissues (AM2, AM3, AM4, AM6, AM8-Primary) and metastatic acral melanoma tissues (AM1, AM5, AM7, AM8-Mets, Mets indicated Metastatic). Notably, two types of samples were extracted from a single patient (Fig. 2B). Subsequent annotation and classification of the 34 clusters, delineated 9 distinct cellular populations within the melanoma tissue (Fig. 2C, D). These populations comprised Melanoma cells (18,083), NK-T cells (10,548), Endothelial cells (ECs) (1418), B-Plasma cells (1295), Myeloid cells (1535), Fibroblasts (1651), Pericytes (552), Epithelial cells (EPCs) (225), and Mast cells (79). Notably, melanoma cells constituted the largest cluster, serving as the predominant cellular entity within the melanoma TME, followed by NK-T cells, indicative of their immunological relevance. To investigate cellular migratory capabilities, we analyzed the distribution and proportions of cellular origins in primary and metastatic tumor tissues across the major cell populations (Fig. 3A–C). Cell cycle distribution (Fig. 3D–F) revealed a prominent G2/M phase predominance in EPCs, indicative of heightened proliferative activity. These findings underscored the necessity for further subtyping of melanoma cells to elucidate mechanisms underlying tumor progression and metastasis.

Fig. 1

Workflow of single cell sequencing analysis for melanoma. We used single-cell sequencing technology, combined with bulk sequencing, in vitro animal experiments, and a series of analyses to comprehensively investigate melanoma. The aim was to identify novel therapeutic targets and improve patient prognosis

Fig. 2

Clustering of melanoma tumor tissues. A Following single-cell clustering, a total of 34 clusters were identified from tumors and surrounding tissues of 8 melanoma patients. B These samples originated from 9 specimens of 8 patients. It could be divided into 5 primary acral melanoma tissue and 4 metastatic acral melanoma tissues. C, D Annotation of different cell populations based on distinct cell surface markers, including Melanoma cells, NK-T cells, ECs, B-Plasma cells, Myeloid cells, Fibroblasts, Pericytes, EPCs, and Mast cells

Fig. 3

The distribution of melanoma cell populations and the cell cycle phases. A–C The UMAP plots illustrated the spatial distribution of different cells within primary and metastatic melanoma tissues. Pie charts were used to depict the proportions of primary and metastatic melanoma tissues from different groups within each cell type. D–F The UMAP plots illustrated the spatial distribution of various cells within primary and metastatic melanoma tissues. Pie charts were used to describe the proportions of primary and metastatic melanoma tissues from different cell types across different cell cycle phases (G1, G2/M, S)

3.2 Subpopulation and primary-metastasis analysis of melanoma cells

Further exploration was done to understand the relationship between heterogeneity and subpopulations in melanoma patients. Through further clustering of melanoma cells, we subdivided them into 8 subpopulations (Fig. 4A). Investigation into the cellular origins of these subpopulations revealed their derivation (Fig. 4B). According to the expression levels of cell-specific genes, we reclassified them into 8 subpopulations, including C0 SOD3 + Melanoma cells, C1 MTRNR2L12 + Melanoma cells, C2 PMEL + Melanoma cells, C3 IGF2 + Melanoma cells, C4 COMP + Melanoma cells, C5 CXCL9 + Melanoma cells, C6 THBS2 + Melanoma cells, C7 PECAM1 + Melanoma cells (Fig. 4C). To delve deeper into distribution of subpopulations and metastatic relationships of each subpopulation, we evaluated the cell cycle phases and primary metastatic status of each subpopulation. Remarkably, the majority of cells in the C0 SOD3 + Melanoma cells subpopulation was found to be in the G1 phase (Fig. 4D–F). In addition, the proportion of metastatic tumor tissues in the C0 SOD3 + Melanoma cells subpopulation was relative higher (Fig. 4G–I).

Fig. 4

Comprehensive distribution landscape of melanoma cell subpopulations. A The melanoma cells underwent dimensionality reduction and clustering, resulting in the identification of 8 distinct subpopulations. B The 8 subpopulations were derived from different melanoma tissues, primarily categorized into primary and metastatic tissues. C The melanoma cells were distinguished into 8 subpopulations based on the different characteristic genes. D–F The UMAP plots illustrated the distribution characteristics of the 8 subpopulations across different cell cycle phases. Pie charts depicted the proportions of each phase within the various subpopulations. G–I The UMAP plots showed the distribution characteristics of the 8 subpopulations based on their tissue origins. Pie charts displayed the proportions of primary and metastatic melanoma tissues within the different subpopulations

To further investigate the interplay between cell cycle phases and primary metastasis among different subpopulations, we visualized the expression of key genes across all subpopulations (Fig. 5A). Strikingly, the key genes expression profile of the C0 SOD3 + Melanoma cells subpopulation closely mirrored that of metastatic tumor tissues, suggesting an association between the C0 SOD3 + Melanoma cells subpopulation and tumor metastasis. Analysis of the cell cycle in each subpopulations revealed that the C0 SOD3 + Melanoma cells subpopulation showed distinct differences compared to other subpopulations, exhibiting a considerably higher proportion of cells in the G1 phase, reaching 50.7%, while other subpopulations showed more uniform distributions across G2/M and S phases (Fig. 5B, C). Furthermore, our analysis of the cell proportions involved in primary and metastatic tissues across different subpopulations indicated that within the C0 SOD3 + Melanoma cells subpopulation, metastatic tumor tissue cells represented the higher proportion, accounting for 78.1% (Fig. 5D, E). Further analysis of Ro/e preference analysis across subpopulations revealed that within all subpopulations, the C0 SOD3 + Melanoma cells subpopulation exhibited the highest Ro/e value of metastatic tumor tissue cells and G1 phase cells (Fig. 5F, G). Hence, we hypothesized that certain life activities or metabolic pathways of the C0 SOD3 + Melanoma cells subpopulation during the G1 phase could be associated with tumor metastasis, potentially influencing other subpopulations through specific metabolic pathways to promote tumor proliferation and metastasis.

Fig. 5

Phase and distribution of subpopulations within primary and metastatic tumor tissues. A The bubble diagram showed the expression levels of different marker genes across the 8 subpopulations and various tissues. Pie charts illustrated the cell cycle proportions for different cell subpopulations and tissues, while violin plots presented the G2/M Score, S Score, and nCount-RNA levels for different cell subpopulations and tissues. B, C The bar graphs showed the proportion of each subpopulations in the different phases. D, E The bar graphs showed the proportion of each subpopulations in the primary and metastatic groups. The proportion of primary and metastatic melanoma tissues within subpopulations C0 to C7. F Ro/e values of 8 melanoma cell subpopulations in metastatic and primary tumor tissues G Ro/e values of 8 melanoma cell subpopulations were assessed across different cell cycle phases (G1, G2/M, and S phases)

3.3 Metabolic pathways and activity analysis of subpopulations

To better understand the mechanisms of melanoma cell subpopulations in the context of staging and metastasis, we assessed the metabolic pathways across distinct subpopulations, cell cycle phases, as well as primary and metastatic tumor tissues (Fig. 6A–C). We observed that the pathways of oxidative phosphorylation, glutathione metabolism, and glycolysis/gluconeogenesis consistently ranked high across different groups. To further explore and compare the scores of these pathways across different subpopulations, cell cycle phases, and primary and metastatic tumor tissue cells, we visualized them using UMAP and violin plots (Fig. 6D–O). Notably, there was a significant upregulation in the expression of these three metabolic pathways in the C0 SOD3 + Melanoma cells subpopulation. In addition, we found that the above three metabolic pathways exhibited higher expression levels in the G1 phase and metastatic tumor tissues. This suggested a potential correlation between these pathways and the activity and metastasis of melanoma tumor cells. Interestingly, the expression levels nFeature-RNA and nCount-RNA were high in the C0 SOD3 + Melanoma cells subpopulation. Metastatic tumor tissue cells and those in the G1 phase also exhibited significantly higher expression compared to other cell types and cell cycle phases (Fig. 6P–W). We hypothesized that the C0 SOD3 + Melanoma cells subpopulation might influence the activity and metastasis of tumor cells through oxidative-related pathways and potentially promote tumor proliferation by affecting other subpopulations' cells via differentiation or signaling pathways.

Fig. 6

Cell metabolic pathway analysis landscape. A–C Top 20 metabolic pathways analyzed across different subpopulations, phases, as well as primary and metastatic tumor tissues. D–O UMAP and violin plots depicted the distribution and scores of oxidative phosphorylation, glutathione metabolism, glycolysis/gluconeogenesis pathways in different subpopulations, phases, as well as primary and metastatic tumor tissues. P–S UMAP and violin plots showed nCount-RNA distribution and scores across different subpopulations, phases, as well as primary and metastatic tumor tissues. T–W UMAP and Violin plots displayed nFeature-RNA distribution and scores across different subpopulations, phases, as well as primary and metastatic tumor tissues. (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, "ns" was used to say that there was no significant difference)

3.4 Enrichment analysis of subpopulations

Then, we computed the DEGs between each subpopulation of melanoma cells and showcased the top 5 upregulated and downregulated genes for each comparison (Fig. 7A–H). Subsequently, we proceeded with GO enrichment analysis on these DEGs (Fig. 7I–P). We discovered that the enriched pathways associated with the C0 SOD3 + Melanoma cells subpopulation included “extracellular matrix organization”, “extracellular structure organization”, “external encapsulating structure”, and “cell-substrate adhesion organization”. Existing studies had shown that the “extracellular matrix organization” pathway was closely related to cancer cell metastasis [51]. Therefore, we inferred that the C0 SOD3 + Melanoma cells subpopulation was intricately linked to melanoma metastasis and might achieve this through pathways involving extracellular matrix and tissue adhesion.

Fig. 7

Enrichment analysis of melanoma cell subpopulations. A–H DEGs within each melanoma cell subpopulations. I–P Functional enrichment analysis of GO-BP terms for differential genes within each melanoma cell subpopulation

3.5 Cell trajectory analysis of the subpopulations

For a clearer insight into the developmental trajectories of melanoma cell subpopulations, we used pseudotime analysis to infer their progression. We utilized CytoTRACE to predict the cellular stemness of each subpopulation of melanoma cells (Fig. 8A, B). We observed that the C0 SOD3 + Melanoma cells subpopulation exhibited the strongest cellular stemness, followed by the C2 PMEL + Melanoma cells subpopulations. Additionally, we utilized slingshot to elucidate the temporal ordering of subpopulations (Fig. 8C, D), which revealed four distinct trajectories corresponding to various cellular differentiation pathways. Notably, all four trajectories originated from the C0 SOD3 + Melanoma cells subpopulation, suggesting that this subpopulation exhibited higher level of cellular stemness and functions as the primary source for the differentiation of the other subpopulations. Subsequently, we validated these results through monocle analysis and examined the temporal expression order of stemness-associated genes (Fig. 8E). Additionally, we analyzed the subpopulation trajectories calculated by monocle (Fig. 8F, G), revealing that C0 SOD3 + Melanoma cells was likely positioned at the initial phase of the temporal order, consistent with the results obtained through slingshot and CytoTRACE. Through our investigations, we identified the C0 SOD3 + Melanoma cells subpopulation as the key subpopulation, characterized by its strong cellular stemness, metabolic activity, and high correlation with metastasis.

Fig. 8

Pseudotime analysis of subpopulation cells. A Using CytoTRACE, predicted ordering was performed on the 8 subpopulations, with higher scores indicating greater differentiation potential. B Distribution characteristics of predicted order and corresponding cell subpopulations of melanoma cells. C, D Prediction of time trajectories for subpopulations using slingshot, revealing 4 distinct trajectories, with UMAP plots illustrating each trajectory. (Lineage1:C0 → C1 → C2 → C3; Lineage2:C0 → C1 → C4; Lineage3:C0 → C1 → C5; Lineage4:C0 → C1 → C7. “ → ” represented the direction of pseudotemporal trajectory differentiation. E Dynamic expression trend of different stemness genes over time based on monocle analysis. F, G Violin and UMAP plots illustrated the pseudotime expression levels of different melanoma cell subpopulations

3.6 Stemness and TFs analysis

To further investigate the crucial role played by tumor development, cell proliferation, metastatic behavior, and drug resistance, we investigated the stemness-associated genes of each cell subpopulation, different cell cycle phases, as well as primary and metastatic tumor tissues. (Fig. 9A–D). We observed significantly higher cellular stemness in the C0 SOD3 + Melanoma cells subpopulation, G1 phase, and metastatic tumor tissues. Moreover, we examined the expression patterns of stemness-associated genes across subpopulations, cell cycle phases, as well as primary and metastatic tumor tissues (Fig. 9E–H). Notably, TWIST1, MYC, HLFA, CD44, exhibited significant expression across C0 SOD3 + Melanoma cells subpopulation, and their expression levels were visualized using UMAP plots (Fig. 9I–L). Furthermore, stemness-associated genes showed pronounced expression in the C0 SOD3 + Melanoma cells subpopulation, G1 phase, and metastatic tumor tissues, suggesting their involvement in the activity of the C0 SOD3 + Melanoma cells subpopulation and potentially in the metastatic process.

Fig. 9

Stemness analysis of cells. A–C The AUC value of cell stemness within each subpopulation, phase, as well as in the primary and metastatic tumor tissues. D The UMAP plot demonstrated the AUC density distribution characteristics of stemness across the cell subpopulations. E The expression levels of stemness genes in the 8 subpopulations, as well as in primary and metastatic melanoma tissues, were analyzed. F–H Expression levels of stemness genes within each subpopulation of cells, across different phases, as well as within primary and metastatic tumor tissues. I–L Expression distribution characteristic of stemness genes TWIST1, MYC, CD44, and HIF1A within UMAP plots

To delve deeper into the specific mechanisms underlying tumor metastasis, we investigated the TFs associated with primary and metastatic tumor tissues, computed and selected the top 5 TFs, and visualized them using UMAP plots (Fig. 10A–N). We found that TFs related to metastatic tumor tissues were predominantly expressed in the region associated with the C0 SOD3 + Melanoma cells subpopulation, while TFs related to primary tissues were distributed among various subpopulations. Additionally, we observed that TCF7L2 ranked highly in specificity scores within melanoma metastasis samples. Consequently, we conducted an in-depth analysis of this TF, focusing on its visualization across different melanoma subpopulations, cell cycle phases, and various classifications (Fig. 10O–R). The results indicated that TCF7L2 exhibited expression levels in the C0 SOD3 + Melanoma cells subpopulation and metastasis samples. Notably, TCF7L2 was linked to the advancement of various cancers, particularly gastric cancer, where it played a key role in regulating upstream and downstream genes during transcription, enhancing tumor cell drug resistance, and significantly promoting apoptosis and metastasis. Moreover, the presence of TCF7L2 was often observed in cases where gastric cancer patients had poor prognoses [52].Therefore, we might hypothesize that TCF7L2 played an integral part in the metastatic process of melanoma. This TF could serve as a potential therapeutic target.

Fig. 10

TFs analysis. A, H TFs distribution characteristics of primary and metastatic melanoma tissues. B, I The scatter diagrams displayed the specificity scores of TFs within regulons in primary and metastatic melanoma tissues. C–G, J–N UMAP plots showed distribution of top 5 TFs in primary and metastatic tumor tissues. O UMAP plot displayed the distribution of TCF7L2 within subpopulations. P–R Violin plots depicted expression levels of TCF7L2 within each different subpopulations, cell cycle phases, as well as primary and metastatic tumor tissues. (****P < 0.0001. "ns" was used to say that there was no significant difference)

3.7 Construction and analysis of the prognostic model

To further explore the patients' prognostic status, we performed an intersection analysis between the DEGs from the TCGA database and the signature genes of the C0 SOD3 + Melanoma cells subpopulation in melanoma. Using univariate Cox regression analysis, we identified 38 risk genes associated with prognosis (Fig. 11A). The stability and reliability of these genes were confirmed through LASSO Cox regression analysis (Fig. 11B, C). Furthermore, through multivariable Cox regression analysis, we ultimately identified 15 prognostic genes and performed a coef analysis. The figure presented the coef values of these prognostic genes (Fig. 11D). Given that IGF1 had relatively higher coef values, we hypothesized its important role in melanoma metastasis and progression. Consequently, we selected this gene for further experimental validation. Among them, IGF1, CD81, FBLN1, GPC3, and APCDD1 were classified as high-risk genes, while APOD, PLSCR4, MGP, MRPS6, GADD45A, SERPING1, SPRY1, MT2A, RGS16, and IFITM1 were categorized as low risk genes. Risk scores were used to categorize patients into high-risk and low-risk groups (Fig. 11E), and their survival status over time was assessed (Fig. 11F). We observed a concentration of deceased patients in the high-risk group, with a significant decrease in the number of survivors over time. Additionally, we displayed the expression levels of prognostic genes in the high-risk and low-risk groups using a heatmap (Fig. 11G). The ROC curve's area under the curve values at 1-year, 3-years and 5-years were 0.69, 0.69, and 0.74, respectively (Fig. 11H), indicating the stable and reliable predictive capability of our model. Kaplan–Meier survival curves (Fig. 11I) revealed a substantial difference in survival rates between the high-risk and low-risk groups, with a statistically significant p-value of less than 0.0001. We also explored the correlation between genes and risk scores, highlighting genes with strong correlations (Fig. 11J, K). The expression levels of APCDD1, IGF1, CD81, and FBLN1 were positively correlated with risk, while MRPS6, PLSCR4, SERPING1, and IFITM1 were negatively correlated with risk. In other words, higher expression levels of APCDD1, IGF1, CD81, and FBLN1 were associated with an poor prognosis.

Fig. 11

Independent prognostic analysis. A Univariate regression analysis uncovered 38 genes connected to prognosis. (HR < 1: protective factors, HR > 1: risk factors). B The coef spectrum distribution of 15 prognostic genes as determined by LASSO regression analysis. C Parameter selection in optimal cross-validation LASSO regression. D The coef values of genes associated with prognosis. E Sorting patients into high-risk and low-risk groups according to risk scores. F Scatter diagram showed the temporal trends in the occurrence of death and alive events. G The heatmap visualized the expression levels of prognostic genes in high-risk and low-risk groups. H Time-dependent ROC curves with AUC values of 0.69, 0.69, and 0.74 for 1-year, 3-years, and 5-years survival, respectively. I Kaplan–Meier survival analysis curves for high-risk and low-risk groups. J Correlation analysis between genes and risk scores. K Scatter diagrams illustrated the correlation of 8 prognostic genes with risk scores

Furthermore, we investigated the association between clinical variables like age, tumor stages, and prognosis (Fig. 12A–D). In addition, we performed risk assessments based on patients' risk groups, race, age, and tumor stages. Forest map revealed some independent risk factors (P < 0.05) (Fig. 12E). Based on the prognostic genes and clinical factors, we constructed a nomogram chart to predict the survival rates of melanoma patients at 1-year, 3-years, and 5-years (Fig. 12F). The accuracy of our nomogram chart model was validated using ROC curve, showing AUC values of 0.72, 0.79, and 0.79 at 1-year, 3-years, and 5-years, respectively (Fig. 12G), indicating its high specificity and sensitivity.

Fig. 12

Clinical relevance analysis. A–D Comparative analysis of risk scores between different prognostic factor groups. E The forest plot demonstrated the results of multivariate cox regression analysis integrating risk scores and clinical factors (age, race and tumor clinical stage T, M and N). F Nomogram showed the prediction of 1-year, 3-years, and 5-year of OS based on race, tumor clinical stage (T, M, and N), age, and risk score. G Time-dependent ROC curves with AUC values of 0.72, 0.79, and 0.79 for 1-year, 3-years, and 5-years survival, respectively

3.8 Immuno-related and enriched functional analysis

Additionally, we carried out an in-depth analysis to explore the role of immune infiltration in disease progression. We conducted an analysis of immune infiltration in both high-risk and low-risk groups (Fig. 13A, B) and identified immune cell populations with higher proportions (Fig. 13C), where macrophages constituted a significant portion. Further comparison of immune cell compositions between the two groups (Fig. 13D) revealed a higher proportion of M1 macrophages in the low-risk group and a higher proportion of M2 macrophages in the high-risk group. Additionally, M1 macrophages exhibited a negative correlation with the risk score, while M2 macrophages showed a positive correlation (Fig. 13E). In summary, we hypothesized that macrophages might be involved in the proliferation and metastasis of melanoma and thus affect the prognosis of patients, but the roles played by M1 and M2 macrophages were different.

Fig. 13

Immune-related analysis. A The heatmap displayed the differences in immune infiltration expression of various factors between the high-risk and low-risk groups. B Differential proportions of immune cells between high-risk and low-risk groups. C Comparison of immune cell infiltration and estimated proportions. D Differential estimated proportions of immune cells between high-risk and low-risk groups. E Correlation analysis between risk scores and different immune cells. F Heatmap analysis of the correlation between gene expression, risk score, OS, and immune cell infiltration. G–I Differential immune scores, stromal scores, and ESTIMATE scores of TME in high-risk and low-risk groups. J Violin plot displayed TIDE expression levels in high-risk and low-risk groups. (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.)

We also explored the interplay between prognostic genes, risk, OS and immune cells (Fig. 13F) and found an association between the OS and both M1 and M2 macrophages, displaying consistent with previous findings, M1 macrophages were positively correlated with OS, while M2 macrophages were negatively correlated with OS. M2 macrophages might have been associated with tumor progression. Regarding TME-related scores, the low-risk group exhibited higher scores for stromal, immune, and estimate scores (Fig. 13G–I). While TIDE scores were also higher in the low-risk group, suggesting that patients in the high-risk group might have a poorer response to treatment with immune checkpoint inhibitors (Fig. 13J).

Furthermore, GSEA revealed notable enrichment of pathways connected to "epidermis development", “skin development” and "keratinization" in the positive expression trend group, which were associated with skin metabolism. Conversely, pathways related to "positive regulation of leukocyte cell–cell adhesion", "regulation of lymphocyte mediated immunity," and "adaptive immune response" showed a negative expression trend (Fig. 14A–F).

Fig. 14

GSEA depicted enriched functional pathways. A–C Enrichment analysis of pathways associated with positive trend expression. D–F Pathways enrichment analysis related to pathways with negative expression trends

3.9 In vitro experiment of IGF1

Based on analysis of prognostic genes, we selected IGF1 as the target gene for our experiments. We conducted experiments using two melanoma cell lineages, employing methods to compare negative control and knockdown infection groups. In the cell viability assay (Fig. 15A, B), CCK-8 tests indicated significantly reduced cell viability after knockdown. In colony formation assays (Fig. 15C, D), the colony numbers formed by IGF1 knockdown cells was notably lower in contrast to the negative control group. Cell proliferation assays (Fig. 15E, F) showed a decreased cell count in both melanoma cell lines following IGF1 knockdown compared to the negative control group. In the wound healing assays (Fig. 15G, H), the width of scratches made on IGF1 knockdown cells was significantly wider than those on the negative control group after 48 h. In the Transwell assays, the migration of both cell lines treated with IGF1 knockdown showed significantly smaller stained areas against the negative control group (Fig. 15I, J). Similarly, in the invasion assay (Fig. 15K, L), the two melanoma cell lines with IGF1 knockdown exhibited markedly reduced invasion compared to the negative control group. Therefore, through these experiments, we found that knockdown of IGF1 in melanoma cells led to decreased proliferation, migration, and invasion capabilities, indicating that IGF1 probably played a positive role in the proliferation, migration, and invasion of melanoma, thereby promoting its progression.

Fig. 15

In vitro assays of prognostic gene IGF1 on proliferation and invasion metastasis. A, B CCK-8 assays revealed significantly reduced cell viability after IGF1 knockout. C, D Colony formation assays demonstrated a notable decrease in colony numbers in cells with IGF1 knockdown against the negative control group. E, F EdU staining results indicated suppressed proliferation of SK-MEL2 and A375 cells upon IGF1 knockdown. G, H Wound healing assays revealed markedly decelerated migration of SK-MEL2 and A375 cells following IGF1 knockdown. I, J Transwell assays showed significantly decreased migration of SK-MEL2 and A375 cells upon IGF1 knockdown. K, L Transwell assays demonstrated a significant reduction in invasion of SK-MEL2 and A375 cells following IGF1 knockdown