MDA-MB-231 cells were cultured in L15 medium. T-47D, MCF-10A, HeLa, BT549, A549, H446, H460, SW480, HCT116, HepG2, ZR-75-1, SK-BR-3, MDA-MB-488, and MCF-7 cells were cultured in DMEM. All culture media were supplemented with 10% FBS (Hyclone). All cells except MDA-MB-231 were grown at 37°C in 5% CO₂ incubators, and MDA-MB-231 cells were grown at 37°C in 0% CO₂ incubators. All cells were passaged for less than 3 months before renewal from frozen, early-passage stocks. All cells were purchased from the ATCC, and all were tested to ensure that they were mycoplasma negative.

Total RNAs were purified with the RNeasy Mini Kit (Qiagen), and cDNA was synthesized with the SuperScript III First-Strand Synthesis SuperMix for qRT-PCR (Thermo Fisher Scientific). The expression levels of OSBPL3 and GAPDH mRNA were quantified with the LightCycler 480 Real-Time PCR System with Universal ProbeLibrary Probe #36 (Roche). The primers were as follows: OSBPL3, 5′-TTG​GTG​TGT​CCC​AAA​AAT​TGG​T-3′ (forward) and 5′-TCC​TGG​GTG​TAA​TTC​ATC​TCC​C-3′ (reverse), and GAPDH, 5′-TCA​TCC​CTG​CCT​CTA​CTG-3′ (forward) and 5′-TGC​TTC​ACC​ACC​TTC​TTG-3′ (reverse).

Western blotting (WB) was performed as described previously (Hao et al., 2020) with the following modifications. The primary antibodies were OSBPL3 (sc-514097, Santa Cruz, United States) at a 1:500 dilution and β-actin (sc-47778, Santa Cruz Biotechnology) at a 1:1,200 dilution. Images were acquired using the Bio-Rad ChemiDoc MP Imaging System (Bio-Rad). All western blots were a representative image of three separate experiments.

IHC and IF staining was performed as described previously (Hao et al., 2020) with the following modifications. Antigen retrieval for sections of tissue microarrays (TMA PR803b, US Biomax, Inc.) was performed in a pressure cooker. The antibodies was both anti-OSBPL3 (sc-514097, Santa Cruz, United States).

We searched OSBPL3 on the TIMER2 web (tumor immune estimation resource, version 2) (http://timer.cistrome.org/) with the “Gene_DE” module and obtained the expression pattern of OSBPL3 between tumor and corresponding normal tissues from the TCGA project. For tumors’ lack of normal tissues [e.g., GBM (glioblastoma multiforme), LAML (acute myeloid leukemia)], we searched the GEPIA2 web server (Gene Expression Profiling Interactive Analysis, version 2) (http://gepia2.cancerpku.cn/#analysis) with the “Expression analysis-Box Plots” module from the GTEx (Genotype-Tissue Expression) database and added the expression of OSBPL3 with box plots. The related parameters are as follows: p-value cutoff = 0.01, log₂FC (fold change) cutoff = 1, and “matched TCGA normal and GTEx data.” We also obtained the expression difference of OSBPL3 in pathological stages (stages I–IV) of different tumors with violin plots via the “Pathological Stage Plot” module of GEPIA2. Then, log₂ [TPM (transcripts per million) + 1] transformation was applied.

We searched the Oncomine database (https://www.oncomine.org/resource/main.html) and obtained the expression difference data of the OSBPL3 gene between tumor and normal tissues by setting the threshold of p-value = 0.001, gene rank = 10%, and fold change = 2.

We obtained the overall and disease-free survival data of OSBPL3 across all types of tumors via the “Survival Map” module of GEPIA2. High (50%) and low (50%) cutoff values were split into high and low expression cohorts. The log-rank test was used in the hypothesis test, and the survival plots were also plotted. We also used the interactive operation interface of the Kaplan–Meier plotter (http://kmplot.com/analysis/) to pool the GEO datasets for a series of meta-analyses of progression-free interval (PFI) and disease-free interval (DFI) survival. The hazard ratios (HRs), 95% confidence intervals (95% CIs), and log-rank p-values were computed, and the Kaplan–Meier survival plots were generated. The meta-analysis was statistically mapped by STATA 12.0 software (StataCorp LP, College Station, TX, United States).

We queried the genetic alteration characteristics of OSBPL3 and obtained the alteration frequency, mutation site, and type and CNA (copy number alteration) across all TCGA tumors on the cBioPortal web (https://www.cbioportal.org/) with the “TCGA Pan Cancer Atlas Studies” module. The mutation sites of OSBPL3 were plotted with a schematic diagram of the protein and the 3D (three-dimensional) structure with the “Mutations” module. We also generated the data on the overall, disease-free, progression-free, and disease-free survival curves for the different cancer types with or without OSBPL3 genetic alteration by Kaplan–Meier plots via the log-rank p-value.

We investigated the correlation between OSBPL3 expression and tumor mutational burden (TMB)/microsatellite instability (MSI) in different tumors from the TCGA project on the web of “http://sangerbox.com/Tool.” Spearman’s rank correlation test was performed, and the p-value and partial correlation (cor) value are shown.

We searched the expression level of the total protein and phosphoprotein of OSBPL3 (NP_055205.2) using the CPTAC dataset. The available datasets of six tumors are shown—BRCA (breast cancer), OV (ovarian cancer), colon cancer, RCC (clear cell), UCEC (uterine corpus endometrial carcinoma), and lung adenocarcinoma.

We explored the association between OSBPL3 expression and immune infiltrates on the TIMER2 database with the “Immune-Gene” module and analyzed the immune score and stromal score using R software “Estimations” with Spearman’s analysis. The immune cells of neutrophils, macrophages, dendritic cells, CD8⁺ T cells, CD4⁺ T cells, B cells, myeloid-derived suppressor cells (MDSCs), T regulatory cells (Tregs), and cancer-associated fibroblasts (CAFs) were selected. The algorithms of immune infiltration estimations include TIMER, CIBERSORT, CIBERSORT-ABS, QUANTISEQ, XCELL, MCPCOUNTER, and EPIC. The results were visualized with a heatmap and a scatter plot.

We obtained the top 50 available experimentally determined OSBPL3-binding proteins via searching the STRING website (https://string-db.org/). The main parameters are organism (“Homo sapiens”), minimum required interaction score [“low confidence (0.150)”], meaning of network edges (“evidence”), max. number of interactors to show (“no more than 50 interactors” in the first shell), and active interaction sources (“experiments”). We also obtained the top 100 OSBPL3-correlated targeting genes from GEPIA2 with the “Similar Gene Detection” module and performed a pairwise Pearson correlation analysis of OSBPL3 and selected genes applying the “correlation analysis” module of GEPIA2. The value of TPM was converted with Log₂ for the dot plot, and the p-value and the correlation coefficient (R) were labeled. The heatmap of the selected genes used the “Gene_Corr” module of TIMER2, and we conducted an intersection Venn analysis to access the OSBPL3-binding genes using FUNRICH software.

Moreover, we performed KEGG (Kyoto Encyclopedia of Genes and Genomes) and GO (Gene Ontology) enrichment analyses. Briefly, we uploaded the gene lists (above-mentioned) to DAVID (the Database for Annotation, Visualization, and Integrated Discovery) with the settings of selected identifier (“OFFICIAL_GENE_SYMBOL”) and species (“Homo sapiens”) and obtained the functional annotation chart. The enriched pathways were finally visualized with the “tidyr,” “ggplot2,” and “clusterProfiler” R packages. The analysis of GO enrichment including biological process (BP), cellular component (CC), and molecular function (MF) was visualized as cnetplots, using the cnetplot function (circular = F, colorEdge = T, node_label = T).

All the data of gene expression were normalized by log₂ transformation. The comparison of tumor and normal tissues used two sets of t-tests; p < 0.05 indicates statistical significance. The Kaplan–Meier curve and log-rank test were used for all survival analyses. The correlation analysis between the two variables used Spearman’s or Pearson’s test. All statistical analyses were processed by R software (version 4.0.2).

In this study, we aimed to explore the roles of the human OSBPL3 gene in different cancers. We first explored the gene and protein information of OSBPL3. As shown in Supplementary Figure S1A, the OSBPL3 gene is located on chr7: 24,827,146-24,949,571 and contains four isoforms, and the OSBPL3 protein structure is relatively conserved among different species (e.g., H. sapiens, P. troglodytes, M. mulatta) and generally consists of the oxysterol-binding protein (Oxysterol_BP) (pfam01237) domain and pleckstrin homology (PH) domain (cl17171) (Supplementary Figure S1B). The phylogenetic tree shows the evolutionary relationship of the OSBPL3 protein between different species (Supplementary Figure S1C).

To explore the expression pattern of OSBPL3 under physiological conditions, we detected OSBPL3 expression in all types of normal tissues based on the HPA (Human Protein Atlas), GTEx, and FANTOM5 (Functional ANnoTation Of the Mammalian genome 5) datasets (Supplementary Figure S2A). OSBPL3 was expressed in all detected tissues (all consensus normalized expression values > 1) and highly expressed in the parathyroid gland followed by the appendix (Supplementary Figure S2B) but presented low RNA tissue specificity as well as low RNA cell type specificity in different normal cells (Supplementary Figure S2C).

Next, we evaluated the expression status of OSBPL3 across various cancers from the TCGA project. As shown in Figure 1A, the expression level of OSBPL3 is significantly higher than the corresponding normal tissues in multiple tumor tissues except breast cancer. We also compared the expression of OSBPL3 between the tumor tissues from TCGA and the normal tissues from the GTEx dataset. Figure 1B and Supplementary Figure S3A indicate that the expression of OSBPL3 is significantly upregulated in all types of cancers except only kidney renal clear cell carcinoma (KICH). The results of the CPTAC dataset also showed similar results in cancer types of renal, colon, and lung cancers and uterine corpus endometrial carcinoma (Supplementary Figure S3B, p < 0.001).

Furthermore, we identified the mRNA expression of OSBPL3 across different types of human cancers and the corresponding normal tissues from the Oncomine database. Figure 1C shows that OSBPL3 expression was higher in multiple cancer groups, including bladder, brain, breast, cervical, colorectal, liver, lung, ovarian, and pancreatic cancers as well as leukemia and lymphoma. The pooling analysis results of over 40 reports confirmed that OSBPL3 is highly upregulated in different kinds of cancers than in normal tissues—pancreatic cancer, lung cancer, colorectal cancer, liver cancer, and cervical cancer (all p < 0.05) (Supplementary Figures S4A–E).

To fully appreciate the expression and correlation of OSBPL3 with different cancers, we experimented the expression levels of OSBPL3 in multiple types of cancer tissues (breast, lung, stomach, and colon) using microarrays consisting of 63 biopsies (Figure 2A) and three cancer tissues (prostate, cervical, and liver) from the HPA dataset (Figure 2B) by IHC staining. The results showed that OSBPL3 was expressed at abnormally high levels in these tumor biopsies. The clinical pathological grade of a tumor cancer closely correlates with malignancy and differentiation, and we found a positive correlation between elevated expression levels of OSBPL3 and high-grade tumors in both the cancer biopsies (breast, lung, stomach, and colon cancers) (Figure 2A) and the tumor statistics data from the GEPIA2 database (COAD, LIHC, OV, PAAD, and TGCT) (Figure 1D). We also checked the protein (Figure 2C) and mRNA expression levels of OSBPL3 (Figure 2D) in four cancer cells (breast, lung, colon, and liver) and the protein expression in all types of human cancer cells from the CCLE dataset (Figure 2F), and the results showed that OSBPL3 had high expression levels in cancer cell lines. Additionally, the IF assays localized endogenous OSBPL3 to the cytoplasm (Figure 2E).

Based on the above-mentioned findings, OSBPL3 is generally upregulated in multiple human cancer tissues, in particular in digestive system carcinoma (e.g., gastrointestinal cancer, colorectal cancer, liver cancer, and pancreatic cancer) and female cancers (e.g., breast cancer, cervical cancer, and ovarian cancer), which may act as an oncogene in most cancer types.

To determine whether the expression of OSBPL3 is associated with the outcome of patients with different tumors, we performed survival analysis using progression or deaths cases as endpoints. As shown in Figure 3A, high expression level of OSBPL3 correlates with poor prognosis of OS (overall survival) for LGG (low-grade glioma), MESO, THYM, and UVM (uveal melanoma) cancers and DFS (disease-free survival) for GBM, LGG, LUAD, and UVM cancers (all p < 0.05) (Figure 4A). To further evaluate the relationship between OSBPL3 and tumor progression (relapse/metastasis), we confirmed the PFI (Figure 5A) and DFI (Supplementary Figure S5A), and the data showed highly expressed OSBPL3 was linked to poor PFI and DFI for both LGG and PAAD cancers and poor PFI for LIHC, MESO, PCPG, PRAD, TGCT, and UVM cancers (all p < 0.05).

We also performed meta-analysis of the correlation between OSBPL3 expression and prognosis with different cancers using the Kaplan–Meier plotter (Figures 3B, 4B, 5B, Supplementary Figure S5B). The results presented a significant correlation between highly expressed OSBPL3 and poor OS, DFS, PFI, and DFI prognosis for LGG; poor OS, DFS, and PFI for UVM; and poor PFI and DFI for PAAD. In contrast, a low OSBPL3 expression level was associated with poor OS (P = 0.004) and DFS (P = 0.028) prognosis for TGCT (Figures 3A, 4A).

The above data indicated that OSBPL3 expression is significantly associated with the poor prognosis of patients with LGG, UVM, PAAD, and MESO cancers but differentially associated with other tumors and may act as a detrimental prognostic factor in these tumors, which is worth further exploration.

Next, we investigated the genetic alteration status of OSBPL3 in various tumors of the TCGA cohorts. As shown in Figure 6A, the highest mutation frequency of OSBPL3 (>6%) appears for patients with uterine corpus endometrial carcinoma and the amplification (an alteration frequency of ∼5%) appears for the esophageal adenocarcinoma. Figure 6B shows the mutation types, sites, and case numbers of OSBPL3 and the main type of genetic alteration was the missense mutation. And the X676_splice/V676G alteration in the oxysterol domain induces a frame shift mutation of OSBPL3 with the translation from P (Proline) to L (Leucine) at the 673 site of OSBPL3 protein, and Figure 6C shows subsequent OSBPL3 protein truncation with the 3D structure. We also explored the potential association between the genetic alteration of OSBPL3 and the clinical survival outcome with different types of cancers. Figure 6D indicates that patients with altered OSBPL3 showed poorer prognosis compared with non-alteration in overall survival (P = 1.245e-3) and disease-specific survival (P= 5.047e-3), but not progression-free survival (P = 0.0613) in SKCM (skin cutaneous melanoma).

We also analyzed the correlation between OSBPL3 expression and tumor mutational burden (TMB)/microsatellite instability (MSI) across diverse tumors from TCGA. Figures 6E,F show a positive correlation between high OSBPL3 expression and TMB for DLBC, ESCA, LIHC, LUAD, PRAD, SARC, STAD, THYM, and UVM (all P = 0.05) but a positive correlation with MSI only for LUSC (P = 0.019), SKCM (P = 0.042), and HNSC (P = 0.0028). The meta-analysis showed the details (Supplementary Figures S6A,B). This result suggested that most of the cancers with high expression of OSBPL3 have more tumor mutation burden and less microsatellite instability, which need more sample tests for confirmation and in-depth research.

To evaluate whether phosphorylation of OSBPL3 has an effect on tumors, we compared the differences in OSBPL3 phosphorylation levels between primary tumor and normal tissues. Figure 7A summarizes the OSBPL3 phosphorylation sites, and Figures 7B–G analyze the phosphorylation status of OSBPL3 at different sites in different tumors—breast cancer, uterine corpus endometrial carcinoma, ovarian cancer, renal cancer, colon cancer, and lung cancer. The results showed S34, S251, and S273 loci within the PH domain of OSBPL3 represent a higher phosphorylation level in all primary tumor tissues compared with corresponding normal tissues, followed by S406 and S437. We also confirmed the CPTAC-identified phosphorylation of OSBPL3 used the PhosphoNET database and found that OSBPL3 phosphorylation in the cell cycle was experimentally supported by one publication and our previous data (Table 1).

This observation merits further molecular and cellular experiments to explore the potential role of phosphorylation of OSBPL3 in tumorigenesis.

The tumor microenvironment (TME) plays an important role in prognosis and treatment response (Jiang et al., 2020). To evaluate the effect of OSBPL3 expression level on tumor-infiltrating immune and stromal cells—the two main components of the TME closely associated with the initiation, progression, or metastasis of cancer—we performed the ESTIMATE algorithm to calculate the immune scores and stromal scores from different types of tumors (Supplementary Figures S7–9). As shown in Figure 8, OSBPL3 was significantly associated with infiltrating immune and stromal cells in the top three tumors (BRCA, LGG, PRAD) (all p < 0.001).

Based on the characteristics of the immune cell infiltrates in the TME, the tumors were preliminarily divided into “hot” and “cold” tumors (Galon and Bruni, 2019). The “hot” tumors are infiltrated with more immune active cells (neutrophils, macrophages, CD8⁺ T cells, dendritic cells, CD4⁺ T cells, Th1), which can produce a better response to immunotherapy drugs; in contrast, the “cold” tumors with less immune cell infiltration and more proportion of immune suppressive cells (myeloid-derived suppressor cells (MDSCs), T regulatory cells (Tregs)) have a weak response to immunotherapy (Lui et al., 2020; Yang et al., 2020; Biffi and Tuveson, 2021). Thus, we explored the relationship between the OSBPL3 expression level and the active immune cells in different cancer types of TCGA. Figure 9A shows a statistical positive correlation between OSBPL3 expression and the immune active cells of neutrophils, dendritic cells, and CD8⁺ T cells in the tumors of PRAD and LIHC. The top three tumors with the strongest correlation between OSBPL3 expression and all the immune infiltration cells are BRAC, COAD, and KIRC (Figure 9B) (all p < 0.001). Unfortunately, OSBPL3 appears to be weakly correlated with immune suppressive cells of MDSCs and Tregs in multiple tumors, which suggests that patients with high expression of OSBPL3 are more likely to have the “hot” tumors in the TME and might have a better immunotherapeutic response.

Additionally, cancer-associated fibroblasts (CAFs) play a role in regulating the interaction between immune effector cells and cancer cells and are associated with tumor progression and poor prognosis (Desbois and Wang, 2021). Thus, we explored the relationship of OSBPL3 with CAFs, and the results showed a positive correlation between the infiltration level of CAFs and OSBPL3 expression in the tumors of BRCA-luminal A, BRCA-luminal B, LGG, LIHC, MESO, PRAD, TGCT, and THYM (Figures 10A,B) while a negative correlation in TGCT.

Immune checkpoint inhibitors (ICIs) have affected the therapeutic landscape for a variety of tumors, but biomarkers associated with ICIS efficacy are still lacking (Bagchi et al., 2021). To analyze the relationship between OSBPL3 and immune checkpoint genes, we extracted and calculated more than 40 common immune checkpoint genes in diverse cancer types of TCGA (Figure 11). The results showed over 30 immune checkpoint genes were strongly associated with OSBPL3 expression in UVM, PRAD, and ACC tumors, and then in LIHC, LGG, and BRAC, which are the ones that are positively associated with OSBPL3 prognosis.

These results indicated that tumors with high expression of OSBPL3 are infiltrated with more immune active cells and seem to represent “hot” tumors and more immune checkpoint genes, which may play a potential role in response to the immune microenvironment and may benefit from immune therapeutic interventions.

To further investigate the molecular mechanism and function of the OSBPL3 gene in tumorigenesis, we screened out the OSBPL3-binding proteins and correlated genes by a series of pathway enrichment analyses. Figure 12A shows the interaction network of top 20 OSBPL3-binding proteins, and Figure 12B shows the top six of the top 100 positively correlated genes with the OSBPL3 expression level—ANKLE2 (R = 0.44), BIRC6, LRIG2, TMEM170A, ZNF490, and C1GLAT1 genes (all p < 0.001). Similar results were found in the majority of cancers by a corresponding heatmap (Figure 12C). An intersection analysis of the above two groups reached one common member, namely, CRN1 (Figure 12D).

Next, we performed KEGG and GO enrichment analyses. The KEGG enrichment analysis of Figures 12E, 13A suggests that “SNARE interactions in vesicular transport,” “Endocytosis,” “Synaptic vesicle cycle,” and “Ribosome biogenesis in eukaryotes” might be involved in the molecular mechanism of OSBPL3 on tumor pathogenesis. The GO enrichment analysis—biological process (BP), molecular function (MF), and cellular component (CC)—further showed that most of the related genes are associated with the pathways or cellular biology of mRNA metabolism, synaptic/endocytic/exocytic vesicles, SNARE binding, protein molecular adaptor activity, membrane fusion/tethering/docking, and others (Figures 12F, 13B–D–D; Supplementary Figure S10).

Taken together, these data suggest the biological processes and molecular mechanism in which OSBPL3 may be involved and further investigations are needed for exploration and confirmation.