We firstly examined the mRNA expression patterns of RETSAT in different types of human cancers using the Tumor Immune Estimation Resource (TIMER) online database (Li et al., 2017). The consistent lower expression of RETSAT was observed in BRCA (Breast invasive carcinoma), CHOL (Cholangiocarcinoma), COAD (Colon adenocarcinoma), HNSC (Head and Neck squamous cell carcinoma), KICH (Kidney Chromophobe), KIRC (Kidney renal clear cell carcinoma), LUAD (Lung adenocarcinoma), LUSC (Lung squamous cell carcinoma), PCPG (Pheochromocytoma and Paraganglioma), PRAD (Prostate adenocarcinoma), READ (Rectum adenocarcinoma) and THCA (Thyroid carcinoma) compared with the corresponding normal tissues (Figure 1A). The above results showed that RETSAT expression is commonly decreased in most human cancers, indicating that RETSAT may function as a tumor suppressor.

Based on the fact that mammals living in high altitude are exposed to both hypoxic and intensive ultra violet conditions, we decided to validate the expression pattern of RETSAT in skin cutaneous melanoma (SKCM), and identified lower RETSAT transcripts in cancerous tissues than that in normal skin tissues, by using the gene expression profile from TCGA and GETx dataset (Figure 1B). Consistently, we also verified the downregulation of RETSAT in tumor tissues in the GEO database (Figure 1C). The correlation between RETSAT and SKCM tumor stage was further tested using the GEPIA online tool (Tang et al., 2017). We found that lower expressions of RETSAT transcripts correlate with higher tumor stages (Figure 1D). In line with this finding, we revealed that patients with SKCM containing higher RETSAT expression exhibited longer overall survival (OS) and disease-free survival (DFS) time, compared to SKCM patients with lower RETSAT expression levels (Figures 1E,F). In addition, we identified many RETSAT mutations in SKCM by applying the cBioPortal web resource (Table 1) (Gao et al., 2013). The above results suggest that RETSAT is decreased in SKCM, indicating its inhibitory role during SKCM tumor progression.

To further elucidate the mechanism by which RETSAT is commonly downregulated in tumors, we firstly examined the methylation status of RETSAT promoter region, which is major cause for repressing gene expression in tumors. By using the methSurv analysis (Modhukur et al., 2018), we found that there are many methylation sites in the promoter region of RETSAT, and the differential methylation regions were indicated in the heatmaps (Figure 2A). Importantly, by using the shiny methylation analysis resource tool (SMART) analysis (Li et al., 2019) we uncovered that the methylation of RETSAT was significantly higher in SKCM cancerous tissues compared to that in normal tissues (Figure 2B). Consistently, we found that the methylation levels on the specific methylation site (cg05514932) within RETSAT promoter region negatively correlated with its expression in SKCM (Figures 2C,D). Furthermore, we showed that the elevated methylation levels on cg05514932 site correlates with worse OS in the TCGA-SKCM cohorts, using the methSurv dataset (Figure 2E). Therefore, we hypothesized that the highly methylation in RETSAT promoter DNA fragment might cause its decreased expressions in multiple types of human cancers, including SKCM (Figure 2F). In addition, we treated B16 cells with 5-Azacytidine, the specific inhibitor for DNA methylases (Christman, 2002), and revealed that RETSAT mRNA expression level was dramatically increased (Figure 2G).

Since RETSAT might function as a tumor suppressor and immune infiltration was considered as a promising independent prognostic factor in cancers, we next used the TIMER database to investigate the correlation between RETSAT expression and immune infiltration (Li et al., 2017). Specifically, RETSAT CNV significantly correlates with infiltrating levels of B cells, CD8⁺ T cells, macrophages, neutrophils and dendritic cells (Figure 3A). Next, we analyzed the correlation between RETSAT expression and six types of infiltrating immune cells, including B cells, CD8⁺ T cells, CD4⁺ T cells, macrophages, neutrophils and dendritic cells. The results demonstrated that the expression level of RETSAT was positively correlated with the infiltration level of B cells (r = 0.597, p = 3.63e-48), CD8⁺ T cells (r = 0.422, p = 2.09e-22), CD4⁺ T cells (r = 0.671, p = 1.25e-64), macrophages (r = 0.478, p = 4.55e-29), neutrophils (r = 0.744, p = 2.67e-86), and dendritic cells (r = 0.837, p = 3.66e-129) in SKCM (Figure 3B). Next, we used TISIDB database to further explore the relationship between RETSAT expression level and immunostimulators and immunoinhibitors, respectively in SKCM (Ru et al., 2019). We found that RETSAT was mostly positively, but negatively, associated with the expression of immunostimulators and the immunoinhibitors, respectively (Tables 2, 3). These results strongly implicate that RETSAT could serve as a key regulator for tumor immune infiltration in SKCM.

To examine the functional role of RETSAT in SKCM cells, RETSAT was inhibited by two independent lenti-viral shRNAs in B16 cells and mouse embryonic fibroblasts (MEFs), and the knockdown and overexpression efficiencies were verified by Real-time RT-PCR, cell line expressing scramble shRNA was used as control. As expected, RETSAT knockdown promoted the cell proliferation ability of B16 and MEFs (Figures 4A–D,I,J,L), while RETSAT overexpression led to the opposite effect (Figures 4E–H,K,M). Taken together, these data suggest that RETSAT functions as a tumor suppressor in SKCM.

To verify the in vivo functional role of RETSAT during tumorigenesis, we performed both the xenograft tumor formation and DMBA/TPA induced cutaneous keratinocyte carcinoma formation assays. Five-weeks old male mice of RETSAT^R/R mutant and wild-type littermates were randomly divided into indicated groups, and B16 cells were injected subcutaneously (2 × 10⁵ cells/point). As expected, the subcutaneous tumors in the wild-type littermates (RETSAT^Q/Q) were detected quickly, whereas the xenograft tumors were markedly retarded in RETSAT^R/R mice, which was visualized by the reduced tumor mass and volume compared to the control group (Figures 5A–C). Consistently, significant lower proliferation as measured by Ki67 immunohistochemistry (IHC) staining in the xenograft tumor sections from RETSAT^R/R group was detected, compared to control group (Figures 5D,E). In addition, by using the DMBA/TPA induced cutaneous keratinocyte carcinoma formation assay, better overall survival rate was observed in RETSAT^R/R mice compared to the littermate wild-type RETSAT^Q/Q mice (Figures 5F,G).

To further explore the molecular events affected by RETSAT^R/R mutation in SKCM, a protein-protein interaction (PPI) network for RETSAT was created using the STRING database (Szklarczyk et al., 2017) (Figure 6A). Among the interacting proteins, Pin1 and Akt1 caught our attention. The oncogenic factor PIN1 has been well demonstrated to promote the occurrence and development of various cancers (Min et al., 2016; Rustighi et al., 2017; Chen et al., 2018; Nakatsu et al., 2020; Yu et al., 2020). Previous study also documented that Pin1 inhibition using small molecule inhibitor such as ATRA or short hairpin RNA, reduces tumor growth via inhibiting PI3K/AKT signaling pathways (Sun et al., 2019). In addition, we revealed that RETSAT is involved in regulating PI3K/AKT signaling pathway by GSEA dataset analysis (Subramanian et al., 2005), suggesting that RETSAT could inhibit the oncogenic effect mediated by Pin1 in SKCM (Figure 6B).

Therefore, we decided to validate the correlation between RETSAT expression and Pin1 related signaling (Sun et al., 2019). We showed that RETSAT expression was significantly negatively correlated with Pin1 in SKCM (r = −0.29, p = 2.9e-10) (Figure 6C). To validate the signaling axis, we used various adult mouse tissues, including brain, heart, liver and lung, and uncovered that the markedly reduced Pin1 and phosphorylated Akt1 proteins in RETSAT^R/R background could only be detected in MEFs, liver and lung, but not in brain and heart, suggesting that the inhibitory effect could be cellular context or developmental stage dependent (Figures 6D–H). Furthermore, we revealed that RETSAT^R/R mutation inhibits Pin1 expressions in B16 xenograft tumors, compared to the RETSAT^Q/Q wild-type control groups, detected by immunohistochemistry (IHC) staining (Figures 6I,J).

We employed the TIMER to analyze the correlation between RETSAT expression and immune infiltration (Li et al., 2017). The TISIDB database was employed analyzing the association between RETSAT and immune regulators (immunostimulators or immunoinhibitors) (Ru et al., 2019). GEPIA and UALcan databases were employed to analyze the expression and prognosis of RETSAT in TCGA SKCM (Chandrashekar et al., 2017; Tang et al., 2017).

Mouse embryonic fibroblasts (MEFs) were generated from RETSAT^R/R (mutant) or RETSAT^Q/Q (wild-type) mice, and B16 cells were cultured in DMEM medium (Corning) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. Cells were all incubated in a humidified atmosphere with 5% CO₂ at 37°C.

As described before (Xu et al., 2021), RETSAT CDS (or mutant CDS DNA fragment) was sub-cloned into pCDH-MSCV-E2F-eGFP lenti-viral vector with a 3×Flag tag at the C-terminus. Independent shRNAs targeting RETSAT mRNA were synthesized and sub-cloned into the lenti-viral vector pLKO.1 (Addgene, Cambridge, United States). Cells were transfected with indicated shRNAs or control scramble shRNA using Lipofectamine 2000 (Invitrogen), and then collected for various experiments, mouse mRETSAT-shRNA#1: GTG​GTG​TCC​TCC​CTC​CTA​CAG, mRETSAT-shRNA#2: AGC​AAT​TCC​TTG​CAC​ATA​TAA.

For Real-time RT-PCR assay, indicated cells were lysed by RNAiso Plus (Takara Bio, Beijing, China, Cat. 108-95-2). Total RNAs were extracted according to the manufacturer’s protocol, and then reverse transcribed using RT reagent Kit (Takara Bio, Beijing, China, Cat. RR047A; TIANGEN Biotech, Beijing, China, Cat. KR211-02). Real-time PCR was performed by FastStart Universal SYBR Green Master Mix (Roche, Cat. 04194194001; TIANGEN Biotech, Beijing, China, Cat. FP411-02) using an Applied Biosystems 7500 machine. The primers used in this study are: mRETSAT-qPCR-F: CCC​ATC​AAG​CAA​GGA​TCC​AA, mRETSAT-qPCR-R: ATG​GGT​ACC​AGC​GCA​GTC​A.

The cell proliferation assay was performed as previously described (Xu et al., 2020). For cell growth assay, indicated cells were plated into 12-well plates and the cell numbers were subsequently counted each day. For BrdU incorporation assay, indicated cells were cultured in 8-well plates for 24 h, pulsed with 10 μM BrdU (Abcam, Cat# ab142567) for 20 min, and fixed with 4% PFA (paraformaldehyde). Cells were then incubated with BrdU (Cell Signaling Technology, Cat# 5292s, dilution 1:1,000) primary antibody followed by secondary antibody detection (Abclonal, Cat# 61303, dilution 1:500). Cell nuclei were stained with DAPI (4′, 6-diamidino-2-phenylindole).

The transgenic mice (RETSAT^R/R) and their wild-type littermates (RETSAT^Q/Q) were derived from C57BL/6 mouse. About 4–6 weeks old age mice were subcutaneously injected with B16 cell lines (2 × 105 cells), 3 weeks later, all mice were sacrificed. The xenograft tumors from indicated groups were harvested and weighted. The mice were monitored every other day, xenograft tumor weights and volumes were measured with a sliding caliper, and tumor volumes were calculated using the formula (L×W2)/2. All animals were kept in a SPF environment and the protocols were pre-approved and conducted under the policy of Animal care and Use Committee at the Kunming Institute of Zoology, CAS.

The assay was performed as previously described (Darido et al., 2011; Jiang et al., 2017). Briefly, 25 μg DMBA (Sigma Aldrich) in 200 µl acetone were applied to the dorsal skin after shaving. After 2 weeks, TPA (10 nmol) in 200 μl was applied to the same area twice weekly for up to 30 weeks. Skin specimens were collected 5 and 8 weeks after DMBA treatment, and when papilloma and SCC formed. The number of tumors per mouse was counted each week as palpable mass >1 mm in size.

To detect the protein expressions of Pin1, Akt1 and P-Akt1, cells were lysed in IP lysis buffer, supplemented with complete protease inhibitor cocktail (Complete Mini, Roche). Indicated proteins were detected with indicated antibodies by western blot. Proteins were resolved on SDS polyacrylamide gels, and then transferred to a polyvinylidene difluoride membrane. After blocking with 5% (w/v) milk, the membrane was stained with indicated primary antibodies as follows: Pin1(Catalog number, R25374, Dilution, 1:1,000, Supplier, ZENBIO), (Akt, Catalog number, 9272, Dilution, 1:1,000, Supplier, Cell Signaling Technology), (p-S473-Akt, Catalog number, 9271, Dilution, 1:1,000, Supplier, Cell Signaling Technology), β-actin (Catalog number, 60008-1-1g, Dilution, 1:20,000, Supplier Proteintech).

For immunohistochemical staining, the sections were deparaffinized in xylene and rehydrated through graded ethanol. Antigen retrieval was performed for 20 min at 95°C with sodium citrate buffer (pH 6.0). After quenching endogenous peroxidase activity with 3% H₂O₂ and blocking non-specific binding with 1% bovine serum albumin buffer, sections were incubated overnight at 4°C with indicated primary antibodies: Pin1(Catalog number, R25374, Dilution, 1:1,000, Supplier, ZENBIO), Ki67 (Catalog number, 170, Dilution, 1:400, Supplier, NOVUS). Following several washes, the sections were treated with HRP conjugated secondary antibody for 40 min at room temperature, and stained with 3, 3-diaminobenzidine tetrahydrochloride (DAB). Slides were photographed with microscope (Olympus BX43F, Japan). The photographs were analyzed with the Image-Pro Plus 7.0 software (Media Cybernetics, Inc., Silver Spring, MD, United States).

All data are presented as the mean ± SEM. All experiments were performed at least three times. All analyses were performed using GraphPad Prism 7 (GraphPad Software). Two-tailed Student’s t-test was used for statistical analysis for experiments with two comparisons. p-values less than 0.05 were considered statistically significant. For all figures, *p < 0.05, **p < 0.01, ***p < 0.001.