Introduction

The biological functions of miR-152-3p are contradictory to the current understanding of the carcinogenesis of multiple malignant tumours [1,2,3]. miR-152-3p is recognized as an oncosuppressor in breast cancer [1], prostate cancer [3], colorectal cancer [4, 5] and glioma [6, 7]. Conversely, up-regulation of miR-152-3p is revealed in plasma from prostate cancer patients compared with healthy control subjects [2]. In our previous study [8], miR-152-3p repressed cyclin-dependent kinase 8 to restrain hepatic carcinogenesis. Herein, we further explored the roles of miR-152-3p and its gene target in the prognosis and immune infiltration of HCC.

ROBO1 is a member of the roundabout transmembrane protein receptor family and contributes to axonal guidance in neurogenesis [9, 10]. Recently, ROBO proteins have been implicated in tumour angiogenesis, endothelial cell migration and immune cell recruitment, interacting with Slit2 as a corresponding ligand [11, 12]. The SLIT/ROBO signalling pathway exhibits Janus-faced properties in cancer progression [9]. In several studies, the Slit2/ROBO1 axis restrains the malignant phenotypes, such as migration, invasion and epithelial–mesenchymal transition, of cancer cells [13,14,15]. However, up-regulation of ROBO1 is correlated with poor prognosis and accelerates osteosarcoma cell growth [16]. Moreover, ROBO1 expression is elevated in nasopharyngeal cancer and is associated with worse overall survival [17]. In HCC, ROBO1 is up-regulated in tumour tissues and is one of the poor-prognosis-related and immune-related genes that may contribute to hepatic carcinogenesis [18, 19]. ROBO1 is also substantiated as a serologic marker for the diagnosis of HCC [19].

In our study, miR-152-3p was up-regulated in nine cancer types and down-regulated in five cancer types in a pan-cancer analysis of the TCGA database. Compared with non-tumour tissues, the elevation of ROBO1 and the reduction of miR-152-3p were observed in HCC tissues. Bioinformatics prediction and experimental measurements validated that ROBO1 is a direct gene target of miR-152-3p that can repress the protein expression of ROBO1 in HCC cells. We further investigated the antineoplastic activity, prognosis and immune infiltration of the miR-152-3p/ROBO1 axis in HCC.

Materials and methods

Prediction of miR-152-3p-related gene targets

Three miRNA prediction databases, TargetScan, miRDB and RNA22, were used to predict miR-152-3p-related gene targets. A Venn diagram was utilized to visualize the potential gene targets with R software and the ggplot2 package (version 3.3.3).

TCGA data analysis

The expression profiles of miR-152-3p and ROBO1 in pan-cancer or HCC were evaluated using the TCGA database with the ggplot2 package (version 3.3.3). The prognostic analysis was carried out using the TCGA database with the survminer package (version 0.4.9) and the survival package (version 3.2–10) as described previously [20]. In addition, the HCCDB database and the HPA database were used to evaluate ROBO1 gene expression and immunostaining in HCC tissues, respectively.

Prognostic analysis of ROBO1

The Kaplan–Meier Plotter database (https://kmplot.com/analysis/index.php?p=background) was used to analyse the prognostic significance of ROBO1 in HCC. In addition, to investigate whether ROBO1 is an unfavourable factor for OS by mediating the enrichment of immune cells, the Kaplan–Meier Plotter database was used to evaluate OS in the subgroups with enriched and decreased immune cells.

Immune infiltration

The association between the enrichment of immune cells and ROBO1 expression in HCC was analysed using the GSVA package (version 1.34.0) with the ssGSEA algorithm [21, 22]. In addition, the correlation of ROBO1 with immune checkpoint molecules [cytotoxic T-lymphocyte associated protein 4 (CTLA-4), programmed cell death 1 (PDCD-1), CD274, T-cell immunoreceptor with Ig and ITIM domains (TIGIT), butyrophilin subfamily 2 member A1 (BTN2A1) and BTN2A2] was assessed in the TCGA database using the ggplot2 package (version 3.3.3).

GSEA, GO and KEGG pathway enrichment

An analysis of single-gene differences of ROBO1 in the TCGA database was prepared for GSEA using the DESeq2 package (version 1.26.0) as described previously [23]. In addition, GO and KEGG pathway enrichment analyses were predicted by the DAVID online database (https://david.ncifcrf.gov/).

Cell experiments

Cell experiments, including cell culture, cell transfection, luciferase reporter assays (Promega, USA), western blotting (anti-ROBO1: ab7279; dilution: 1:500; Abcam) and CCK8 assays (Beyotime), were carried out as described previously [8]. miR-Com-, miR-152-3p-, sh-Con-, sh-ROBO1- and ROBO1-overexpressing plasmids were obtained from GenePharma (Shanghai, China). A TUNEL kit (Roche) was utilized to analyse cell apoptosis.

Statistical analysis

The data were analysed using Mann–Whitney U test, Wilcoxon signed rank test, one-way analysis of variance log-rank test, univariate Cox regression analysis and Spearman’s correlation analysis.

Results

Prediction of gene targets of miR-152-3p

Based on three miRNA prediction databases, TargetScan, miRDB and RNA22, a total of 101 gene targets of miR-152-3p were collectively identified in those three databases (Fig. 1A). The heatmap shown in Fig. 1B presents the differential expression profiles of the 101 gene targets in the TCGA database, which contained 160 non-tumour tissues and 371 HCC specimens. According to log2(fold change) > 2 and p < 0.05, ROBO1 [log2(fold change) = 2.21; p < 0.001] and COL4A1 [log2(fold change) = 2.07; p < 0.001] expression levels were elevated in HCC tissues and filtered out for further investigations (Fig. 1C). In 50 paired tissues, both ROBO1 and COL4A1 were expressed at significantly higher levels in cancerous tissues than in paracarcinoma tissues (Fig. 1D).

Fig. 1
figure 1

Prediction of gene targets of miR-152-3p. TargetScan, miRDB and RNA22 were implemented to predict gene targets of miR-152-3p (A). Heatmap represents the differential expression of 101 gene targets (B). ROBO1 and COL4A1 expression in non-paired HCC tissues and non-tumour tissues (C). ROBO1 and COL4A1 expression in paired HCC tissues and adjacent tissues (D). ***p < 0.001

Full size image

Prognostic significance of ROBO1 and COL4A1 in HCC

The Kaplan–Meier Plotter database was used to evaluate the association of ROBO1 and COL4A1 with prognosis in HCC. Worse OS, PFS and DSS were observed in patients with high ROBO1 expression than in those with low ROBO1 expression. However, COL4A1 expression had no significant correlation with OS, RFS, PFS or DSS in HCC patients (Fig. 2B).

Fig. 2
figure 2

Prognostic significance of ROBO1 and COL4A1 in HCC. Kaplan–Meier Plotter database was used to evaluate the prognostic significance of ROBO1 (A) and COL4A1 (B) in HCC

Full size image

Further validation of ROBO1 expression in HCC

The HCCDB database was utilized to analyse ROBO1 gene expression in 12 datasets. In 11 of the 12 datasets, up-regulation of ROBO1 gene expression in HCC tissues was validated by the HCCDB database (Fig. 3A). Intriguingly, the HPA database revealed prominent positive staining of ROBO1 expression in HCC tissues (Fig. 3B). As shown in Fig. 3C and Table 1, ROBO1 expression was not significantly different in the T, N, M and pathologic stage subgroups. However, up-regulation of ROBO1 showed significant correlations with high histologic grade and AFP levels in HCC patients.

Fig. 3
figure 3

Further validation of ROBO1 expression in HCC. HCCDB database was utilized to analyse ROBO1 gene expression (A). HPA database revealed prominent positive staining of ROBO1 (B). The association between ROBO1 expression and clinical parameters in HCC (C)

Full size image
Table 1 The association of ROBO1 expression with clinical parameters in HCC patients
Full size table

Relationship of ROBO1 expression with immune infiltration and immune checkpoint molecules

To evaluate whether ROBO1 expression was associated with tumour immunity, the ssGSEA algorithm was implemented to investigate whether ROBO1 expression mediates immune cell enrichment in HCC tissues. Spearman correlation analysis revealed that ROBO1 was positively correlated with five immune cell enrichments and negatively correlated with six immune cell enrichments in HCC tissues (Fig. 4A). The top three correlated immune cells, including T-helper cells (r = 0.329; p < 0.001), DCs (r = −0.203; p < 0.001) and cytotoxic cells (r = −0.190; p < 0.001), are listed in Fig. 4B and C. In addition, ROBO1 was positively correlated with five immune checkpoint molecules (Fig. 4D), including PDCD1 (r = 0.170; p = 0.001), CTLA4 (r = 0.220; p < 0.001), TIGIT (r = 0.190; p < 0.001), BTN2A1 (r = 0.320; p < 0.001) and BTN2A2 (r = 0.260; p < 0.001).

Fig. 4
figure 4

The relationship of ROBO1 expression with tumour immunology. ssGSEA algorithm was implemented to explore the correlation between ROBO1 expression and immune cell enrichment in HCC tissues (A). Spearman correlation analysis evaluated the correlation of ROBO1 with T helper cells (B), DC and cytotoxic cells (C). Spearman correlation analysis evaluated the correlation of ROBO1 with 6 immune checkpoint molecules (D)

Full size image

Prognostic significance of ROBO1 expression based on immune cell enrichment in HCC patients

The TIMER database was also used to validate the association between ROBO1 expression and immune cell enrichment in HCC. As shown in Fig. 5A, ROBO1 expression was positively correlated with B cells (r = 0.154; p = 4.31 × 10−3), CD4+ T cells (r = 0.244; p = 4.55 × 10−6), macrophages (r = 0.176; p = 1.11 × 10−3), neutrophils (r = 0.202; p = 1.56 × 10−4) and dendritic cells (r = 0.116; p = 3.25 × 10−2). Based on ROBO1, in the above-mentioned results, high ROBO1 expression was correlated with poor prognosis and immune cell enrichment. Therefore, we hypothesized that the ROBO1-regulated poor prognosis was associated with immune cell enrichment. The association between ROBO1 and OS was analysed in the subgroups with enriched and decreased immune cells. In both the enriched and decreased subgroups of B cells (Fig. 5B, I), CD4+ T cells (Fig. 5C, I) macrophages (Fig. 5D, I), natural killer T cells (Fig. 5E, I), regulatory T cells (Fig. 5F, I) and type-1 T-helper cells (Fig. 5G, I), high ROBO1 expression was significantly correlated with poor OS in HCC patients. As shown in Fig. 5H and I, high ROBO1 expression correlated with poor OS in HCC patients with enriched type-2 T-helper cells, suggesting that poor ROBO1-related OS may be partially mediated by the enrichment of type-2 T-helper cells. To further validate the association of ROBO1 with immune cell enrichment in HCC, the correlation between ROBO1 and multiple biomarkers of immune cells was evaluated using TCGA database. As shown in Table 2, ROBO1 was significantly positively correlated with T-helper cell biomarkers (CXCR3, CCR4, CCR6 and CCR10), Tcm biomarkers (CD62L, CCR5, CD58 and TCF7), Th2 cell biomarkers (PTGDR2, IL5 and IL10) and macrophage biomarkers (PPARG, IRF5 and CD68) in HCC.

Fig. 5
figure 5

Prognostic significance of ROBO1 expression based on immune cells enrichment in HCC patients. TIMER database was used to validate the association between ROBO1 expression and immune cell enrichment in HCC (A). The association between ROBO1 and OS was analysed in subgroups of enriched and decreased immune cells of B cells (B), CD4+ T cells (C), macrophages (D), natural killer T cells (E), regulatory T cells (F) and type-1 T-helper cells (G) and type-2 T-helper cells (H). Univariate Cox regression was used to analyse the association between ROBO1 and OS in enriched and decreased immune cell subgroups (I)

Full size image
Table 2 The association of ROBO1 expression with the biomarkers of immune cells
Full size table

Differentially expressed genes in HCC based on ROBO1 expression

According to |log2(fold change)|> 2 and adjusted p < 0.05, 132 up-regulated and 68 down-regulated genes were filtered based on the ROBO1 high- and low-expression subgroups (Fig. 6A). A heatmap was constructed to visualize the expression profiles of the top five up-regulated and down-regulated genes in HCC tissues (Fig. 6B). GSEA suggested that ROBO1-related genes were enriched in the Biocarta intrinsic pathway, Reactome CD22 mediator BCR regulation, the PID/PLK1 pathway, Reactome mitotic prometaphase, the PID/MYC activity pathway and the PID/beta-catenin Nuc pathway (Fig. 6C).

Fig. 6
figure 6

Differentially expressed genes in HCC based on ROBO1 expression. Differentially expressed genes were categorized on the basis of ROBO1 high- and low-expression subgroups (A). Heatmap presents the expression of the top five up-regulated and down-regulated genes in HCC tissues (B). GSEA analysis based on ROBO1-related differentially expressed genes in HCC (C)

Full size image

GO and KEGG analyses

The ROBO1-related top 100 up-regulated and down-regulated genes were used to perform GO and KEGG analyses. Seven BP terms (Fig. 7A), 15 MF terms (Fig. 7B) and 7 KEGG pathways (Fig. 7C) were enriched in HCC.

Fig. 7
figure 7

GO and KEGG analysis. ROBO1-related top 100 up-regulated and down-regulated genes were used to perform GO [BP terms (A) and MF terms (B)] and KEGG analysis (C)

Full size image

Validation of ROBO1 as a gene target of miR-152-3p

The analyses performed in the online databases revealed that ROBO1 might be a potential gene target of miR-152-3p, which was predicted to bind with the 3′-UTR of ROBO1 (Fig. 8A). In vitro luciferase activity was significantly diminished in HepG2 and Huh7 cells after transfection with miR-152-3p mimics compared with the control group (Fig. 8B), indicating that miR-152-3p can directly bind with ROBO1. Western blot analysis revealed significant decreases in ROBO1 protein expression in HepG2 and Huh7 cells transfected with miR-152-3p mimics (Fig. 8C). To investigate the roles of miR-152-3p and ROBO1 in the progression of HCC, miR-152-3p mimics, sh-ROBO1 or ROBO1 overexpression plasmids were delivered into HepG2 and Huh7 cells. As shown in Fig. 8D and E, transfection of miR-152-3p mimics or sh-ROBO1 inhibited cell proliferation and induced cell apoptosis of HepG2 and Huh7 cells. However, the inhibition of cell proliferation and the enhancement of cell apoptosis by miR-152-3p mimics were counteracted by overexpression of ROBO1 in HepG2 and Huh7 cells (Fig. 8D, E).

Fig. 8
figure 8

Validation of ROBO1 as a gene target of miR-152-3p. miR-152-3p binds with the 3′-UTR of ROBO1 (A). In vitro luciferase activity in HepG2 and Huh7 cells (B). ROBO1 protein expression in HepG2 and Huh7 cells with miR-152-3p mimics transfection (C). Cell proliferation (D) and apoptosis (E) of HepG2 and Huh7 cells were analysed using CCK8 and TUNEL. *p < 0.05 versus miR-Con group; #p < 0.05 versus sh-Con group; $p < 0.05 versus mimics group

Full size image

The association between miR-152-3p and HCC

As shown in Fig. 9A, miR-152-3p was up-regulated in nine cancer types and down-regulated in five cancer types in pan-cancer analysis of the TCGA database. Compared with non-tumour tissues, miR-152-3p was significantly down-regulated in HCC tissues (Fig. 9B), while miR-152-3p expression had no significant correlation with OS (Fig. 9C), T stage (Fig. 9D), N stage (Fig. 9E), M stage (Fig. 9F), pathologic stage (Fig. 9G) or histologic grade (Fig. 9H). As shown in Fig. 9I, low miR-152-3p expression was significantly correlated with high AFP levels in HCC patients.

Fig. 9
figure 9

Clinical significance of miR-152-3p in HCC. miR-152-3p expression in pan-cancer analysis of TCGA database (A). miR-152-3p expression in TCGA database (B). The association between miR-152-3p and OS (C), T (D), N (E), M (F) and pathologic stage (G), and histologic grade (H). miR-152-3p low expression was significantly correlated with high AFP levels in HCC patients (I). *p < 0.05; ***p < 0.001

Full size image

The association between miR-152-3p and immune infiltration

To determine whether miR-152-3p expression was associated with tumour immunity, the ssGSEA algorithm was implemented to explore the correlation between miR-152-3p expression and immune cell enrichment in HCC tissues. Spearman correlation analysis revealed that ROBO1 was positively correlated with six immune cell enrichments and negatively correlated with two immune cell enrichments in HCC tissues (Fig. 10A). The top six correlated immune cells, including Th17 cells (r = 0.180; p < 0.001), NK cells (r = 0.180; p = 0.001), DCs (r = 0.170; p = 0.001), mast cells (r = 0.160; p = 0.002), neutrophils (r = 0.140; p = 0.007) and Tgd (r = 0.130; p = 0.012), are shown in Fig. 10B.

Fig. 10
figure 10

The association between miR-152-3p and immune infiltration. ssGSEA algorithm was implemented to evaluate miR-152-3p mediated immune cell enrichment (A). The top six correlated immune cells, including Th17 cells, NK cells, DC, mast cells, neutrophils and Tgd (B)

Full size image

Discussion

Our findings revealed a novel signalling axis, miR-152-3p/ROBO1, that contributes to hepatic tumorigenesis by regulating cell proliferation and apoptosis in vitro and mediates immune cell enrichment in HCC. ROBO1 was identified as an oncogene to accelerate the enrichment of type-2 T-helper cells that may be correlated with poor prognosis in patients with HCC.

The convertible role of ROBO1 as a tumour suppressor or an oncogene has been observed in different cancer types [15, 24,25,26]. At present, the transcription level and functions of ROBO1 are consistent in the progression of HCC [19, 27, 28]. For example, the enhancement of ROBO1 expression triggers tumour growth, invasion and metastasis in HCC Sk-hep-1 cells [27]. Down-regulation of ROBO1 by miR-490-5p contributes to the induction of apoptosis and inhibits malignant phenotypes in HCC Hep3B cells [28]. Hirotaka et al. substantiated that ROBO1 is overexpressed in the serum of HCC patients, HCC tissues and cell lines and may be a serologic marker for the diagnosis of HCC [19]. In our study, marked elevations in ROBO1 gene and protein expression were corroborated by three databases, TCGA, HCCDB and HPA, in HCC tissues, and ROBO1, as an unfavourable prognostic marker, was correlated with worse OS, PFS and DSS. We also found that ROBO1 loss of function augmented the proportion of apoptotic cells and restrained the proliferation of HCC cell lines.

Diversiform immune cells are an important component of the tumour microenvironment and correlate with prognosis, metastasis and immunotherapy in cancer [29, 30]. HCC is associated with inflammatory processes, such as viral infection and cirrhosis, which drive the enrichment of immune cells, especially lymphocytes, contributing to poor prognosis [30, 31]. In our study, both the TCGA and TIMER databases indicated that ROBO1 was significantly and positively correlated with macrophages in HCC tissues. Tumour-associated macrophages (TAMs) contribute to the initiation and progression of HCC by secreting pro-inflammatory cytokines and triggering the expansion of cancer stem cells [32]. In addition, TAMs facilitate metastasis and predict poor prognosis in patients with HCC [32, 33].

Our findings also suggested that poor OS in ROBO1HCC patients with high ROBO1 expression was associated with the enrichment of Th2 cells. The TCGA database also validated a significant positive correlation (r = 0.182; p < 0.001) between ROBO1 and Th2 cells in HCC tissues. Enrichment of Th2 cells has frequently been reported in HCC patients and is implicated in tumour invasion and metastasis [34, 35]. Duan et al. indicated that inhibition of Th2 cell activity by immune checkpoint blockades (ICBs) may be associated with increased survival time and decreased tumour recurrence in a mouse model of HCC [36]. These findings suggested that enrichment of Th2 cells represented an unfavourable biomarker of HCC prognosis.

ICBs are a class of antitumour immunotherapeutic drugs that suppress immune checkpoint molecules, such as PD-1, PD-L1, CTLA-4 and TIGIT, to restore immune recognition and immunogenicity in HCC [37,38,39]. Our findings suggested that ROBO1 was positively correlated with five immune checkpoint molecules, PDCD1, CTLA4, TIGIT, BTN2A1 and BTN2A, in HCC, reflecting that ROBO1 inhibitors may have a synergistic effect and enhance the potency of ICBs to improve therapeutic efficiency in HCC patients.

In this study, both predictions and experiments corroborated that ROBO1 is a direct gene target of miR-152-3p. In our previous study [8], miR-152-3p expression was decreased in HCC tissues, and overexpression of miR-152-3p targeted cyclin-dependent kinase 8 to mediate antineoplastic activity in HCC. Other studies have also shown that miR-152-3p possesses outstanding anticancer properties in colorectal cancer, prostate cancer and lymphoma [5, 40, 41]. Based on our previous findings [8] and the present results, miR-152-3p may function as a tumour suppressor by mediating multiple gene targets to prevent hepatic tumorigenesis.

In conclusion, ROBO1 was identified as an unfavourable prognostic marker and was correlated with the enrichment of Th2 cells in HCC. ROBO1 expression was also positively correlated with multiple immune checkpoint molecules, suggesting that ROBO1 may be a potential drug target to enhance the potency of immunotherapy. Further, ROBO1 was identified as a direct target of miR-152-3p, indicating that the miR-152-3p/ROBO1 signalling axis may be involved in the pathogenesis of hepatic tumorigenesis.