Identification of a robust biomarker LAPTM4A for glioma based on comprehensive computational biology and experimental verification

Data acquisition and processing

The RNAseq data from the TCGA-GBM and TCGA-LGG projects, processed using the STAR pipeline, along with clinical data, were downloaded and organized from the TCGA database (https://www.cancer.gov/ccg/research/genome-sequencing/tcga) [15] using R packages “stats (4.2.1)” and “car (3.1–0)”. The extracted data were in FPKM format. In total, there were 706 samples, consisting of 701 neuroglioma patients and 5 normal tissue samples. Among the neuroglioma patients, there were 532 samples classified as low-grade glioma and 174 samples as glioblastoma multiforme. Since the TCGA database only contained 5 normal samples, we supplemented the analysis by including an additional 1152 normal brain tissue samples from the GTEx database (https://www.gtexportal.org/) [16]. The CGGA mRNAseq_693 data set, obtained from the CGGA database (http://www.cgga.org.cn/), was used as a validation cohort. Additionally, in order to obtain the LAPTM4A expression of distinct subtypes, we downloaded and analyzed the Bao, Phillips, and Rembrandt data set of the Gliovis website (http://gliovis.bioinfo.cnio.es/) [17].

Weighted gene co-expression network analysis (WGCNA)

Weighted Gene Co-expression Network Analysis (WGCNA) was used to identify clusters (modules) of highly correlated genes, establish module-to-module relationships (using the feature gene network approach), and calculate module membership metrics for the identification of candidate biomarkers or therapeutic targets [18]. First, sample clustering of all DEmRNAs was applied to ensure the inclusion of high-quality samples with reliable RNA. Subsequently, a soft-thresholding power value of 12 (mRNA) (scale-free R² = 0.8) was chosen to determine the scale-free topology model. The adjacency matrix was transformed into a topological overlap matrix (TOM). Based on the dissimilarity measurements from TOM, the mRNAs were grouped into different modules. A minimum module size of 30 and a cut height of 0.45 were applied to define the key modules. Module Eigengenes (MEs) were considered the major principal components of a given RNA module, while module membership (MM) was used to assess their correlation with gliomas. Furthermore, the turquoise module was determined as the most significant module correlation. Pearson algorithm was employed to calculate the connectivity of genes and genes with high connectivity were identified as hub genes for the module.

DNA methylation and genetic alterations of LAPTM4A

MethSurv (https://biit.cs.ut.ee/methsurv/) [19] can combine DNA methylation data for multivariate survival analysis. To evaluate the prognostic values (OS) of CpG methylation of LAPTM4A in LGG and GBM, we entered “LAPTM4A” in the “Query” module, and selected an accessible CpG methylation site to analyze. Moreover, the methylation heatmap of LAPTM4A can be found in the “Gene visualization” module.

What’s more, the genomic profiles of LAPTM4A in cBioPortal (https://www.cbioportal.org) [20] were analyzed utilizing three datasets: Brain Lower Grade Glioma (TCGA, Firehose Legacy), Glioblastoma Multiforme (TCGA, Firehose Legacy), and Merged Cohort of LGG and GBM (TCGA, Cell 2016). We entered “LAPTM4A” in the “Query” module, and the LAPTM4A loci, type, and number of variants can be found in the “Cancer Type Summary” and “Mutation” modules.

Protein interaction, structure and docking analysis

In the present study, GeneMANIA (http://genemania.org/) [21] was initially utilized to construct an interactive functional network for LAPTM4A.

Through the utilization of the cBioPortal portal website (https://www.cbioportal.org/), we investigated the secondary structures of LAPTM4A, MCOLN1, and IGF2BP3 by leveraging samples from a merged queue consisting of LGG and GBM entities (study ID, Merged Cohort of LGG and GBM).

Moreover, we employed the Protein Data Bank (PDB; https://www.rcsb.org/) to identify the advanced structures of MCOLN1 and IGF2BP3 (with PDB IDs: 5TJA and 6FQ1, respectively). Furthermore, the AlphaFold Protein Structure Database (https://swissmodel.expasy.org/) [22] was used to predict the advanced structure of LAPTM4A (ID: AF-Q15012-F1).

Finally, HDOCK (http://hdock.phys.hust.edu.cn/) [23] was employed to predict the interaction docking patterns among LAPTM4A, MCOLN1, and IGF2BP3 and visualized them using PyMOL software.

Functional relevance analysis

CancerSea is a database for examining the biological role of target genes in tumor cells in the single-cell pattern (http://biocc.hrbmu.edu.cn/CancerSEA/) [24]. We entered “LAPTM4A” in the “Query” module. Subsequently, the single-cell sequence data in the Correlation Graph module were analyzed for correlations between LATPM4A and 14 cancer functional states. LATPM4A and function were screened for correlations at a p-value of 0.05.

LinkedOmics is a general online website covering 32 TCGA carcinoma-related datasets (http://www.linkedomics.org/login.php) [25]. The “HiSeq RNA” platform and the “TCGA_ GBMLGG” cohort were selected for analysis. The correlation between LAPTM4A and co-expressed genes was detected by Spearman’s test. Our team utilized the LinkFinder module in Variomics to study differentially expressed genes associated with LAPTM4A in TCGA GBMLGG.

Single-cell sequencing analysis and immunoassay

TISCH (http://tisch.comp-genomics.org/home/) is a collection of scRNA-seq data for multiple cancer types that allows for cell type-specific analysis of target gene expression based on single-cell TME expression [26]. The following were our analysis parameters: LAPTM4A, major lineage, and all cancers. The heatmaps, scatter plots, and violin plots quantified and visualized the expression levels of LAPTM4A in each cell type. The TIMER2 database (http://timer.cistrome.org/) [27] studies immune infiltration of multiple cancer types systematically. We investigated the relationships between the expression of LAPTM4A and 12 immune cell subgroups, including cancer-associated fibroblasts (CAF), regulatory T cells (Tregs), macrophages, monocytes, and so on. Additionally, TISIDB, an integrated knowledge base portal (http://cis.hku.hk/TISIDB/) [28], is crucial in identifying how cancer and the immune system interact. To investigate the association between LAPTM4A and the expression of MHC, chemokines, chemokine receptors, immunostimulators, and immune inhibitors, we evaluated the expression levels of chemokines/chemokine receptors in TIIC by the corresponding module. Furthermore, we utilized the Wilcoxon test to investigate the correlation between LAPTM4A expression and immune checkpoint genes to guide clinical immunotherapy. The Tumor Immune Dysfunction and Exclusion (TIDE) algorithm was utilized to forecast possible immune checkpoint inhibitor (ICI) reactions.

Drug sensitivity analysis

First of all, we utilized the Comparative Toxicogenomics Database (CTD, http://ctdbase.org/) [29] to query chemical agents that affected the expression of LAPTM4A. Subsequently, GSCALite (http://bioinfo.life.hust.edu.cn/web/GSCALite/) [30], a general site for the study of immune infiltration and drug sensitivity was applied to seek the drugs relevant to LAPTM4A. Additionally, we predicted chemotherapy response based on the Genomics of Drug Sensitivity in Cancer (GDSC, https://www.cancerrxgene.org/) [31, 32] and conducted the procedure through the R package ‘pRRophetic’. Ridge regression was employed to estimate the half-inhibitory concentration. Eventually, we calculated the association between LAPTM4A expression and the IC₅₀ of anticancer drugs.

Prediction and construction of ceRNA networks

The TargetScanHuman 8.0 (http://www.targetscan.org) [33], DIANA-microT (http://diana.imis.athena-innovation.gr/DianaTools/index) [34], and RNAinter (http://www.rnainter.org) [35] online websites are employed to forecast and analyze the miRNA potentially bound to the LAPTM4A. Next, we use miRNet 2.0 (https://www.mirnet.ca/miRNet/home) [36] and starBase 3.0 (https://rnasysu.com/encori/) [37] for prediction and analysis of the target lncRNA of the relevant miRNA to establish a possible ceRNA network.

Cell culture

The U251 glioblastoma cell line was obtained from the Chinese Academy of Sciences in Shanghai, China. U251 cells were cultivated in RPMI-1640 medium, while the remaining cell lines were cultured in DMEM supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin. All cell cultures were maintained at 37°C in a CO₂ incubator with 5% carbon dioxide.

The small interfering RNA (siRNA) and plasmid transfection

The siRNA was purchased from Hanbio (Shanghai, China) and transfected with liposomes 3000 (Shanghai, China). Cells were cultured in a 6-well plate until they reached a density of 50–60% during the logarithmic growth phase. 5 μL of liposomes 3000 and 125 μL Opti-MEM solutions were mixed. DNA (2 μg) and P3000 reagent (10 μl) were diluted to a final volume of 125 μl using Opti-MEM solution. The DNA-lipid complex was incubated at room temperature for 10–15 minutes before being added to the cells. The cells were then incubated at 37°C for 2–4 days [38].

Real-time PCR (qRT-PCR)

Total RNA extraction was carried out using the Trizolrogen (USA) protocol. The extracted RNA was subsequently converted into complementary DNA (cDNA) using an RT-PCR kit. To assess the mRNA levels of the target genes, a Real-Time qPCR kit was employed.

Western blotting

The RIPA protein extraction method (Beyotime, Shanghai, China) was used to extract total protein from glioblastoma cells. Following centrifugation, the protein concentration was determined with the BCA protein assay kit (Thermo Fisher Scientific, Waltham, MA, USA). For protein separation, SDS-polyacrylamide gel electrophoresis (SDS-PAGE) was employed, and PVDF membrane was utilized for protein transfer. After incubating the membrane overnight at 4°C with the specified antibodies, it was washed three times with TBST. Subsequently, the membrane was incubated with LAPTM4A-conjugated secondary antibodies at room temperature for the next hour. Finally, protein expression was detected using the Amersham™ Image Quant 800 system (GE Healthcare, Chicago, IL, USA).

Transwell assay

To evaluate cell invasion, Transwell chambers (Shanghai, China) were utilized. In a nutshell, cells were seeded in the upper chamber of Transwell coated with a matrix gel in serum-free DMEM medium at a concentration of 1 mg/ml. Once cell confluence reached around 80%, DMEM medium supplemented with 10% fetal bovine serum (FBS) was added to the lower chamber, and cells were cultured for 24 hours. Following this, invasive cells were fixed with 4% paraformaldehyde and stained with 0.5% crystal violet for 30 minutes. The invaded cell count was assessed using an Olympus microscope to determine the cell’s invasion capability.

Dual-luciferase reporter gene system

The sections encompassing the binding sites of miR-103a-3p or the respective mutants from FGD5-AS1 and LAPTM4A were cloned into a pmiRGLO Vector (Shanghai, China). The vectors mentioned above, together with the miR-103a-3p mimic and inhibitor, were co-transfected into U251 cells using Lipofectamine 2000 (Shanghai, China). The dual-luciferase reporter gene system (Shanghai, China) was used to assess luciferase activities.

Statistical analysis

We utilized R (version 4.1.2) for the ensuing analysis. We extracted the expression data of LAPTM4A from the normal and tumor tissues in LGG, GBM, and GBMLGG of TCGA and GTEx. The RNAseq data and clinical data were downloaded in level 3 HTSeq-FPKM format in the TCGA GBMLGG project, and three data points, WHO grade, IDH mutation status, and 1p/19q deletion, were obtained from the study by Ceccarelli et al. The visualization was performed using the ggplot2 package. We utilized the R package “survival” to construct COX proportional hazards regression models investigating the relationship between LAPTM4A expression and prognosis in glioma and performed statistical tests using the log-rank test to obtain prognostic significance [39]. P-values < 0.05 were considered significant (^*P < 0.05, ^**P < 0.01, ^***P < 0.001).

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Constructing WGCNA and identifying LAPTM4A

Initially, we extracted a total of 30,029 genes from the TCGA-GBM, TCGA-LGG, and GTEx-brain datasets for our analysis, as depicted in Figure 1A. Through rigorous screening, we identified 8,762 genes (meeting the criteria of positive logFC and p-value < 0.05) for subsequent WGCNA analysis. To establish a scale-free topological network, we applied a soft threshold power (β) of 12, as shown in Figure 1B. The resulting hierarchical clustering tree revealed four distinct co-expression modules, as depicted in Figure 1C. Notably, the turquoise module demonstrated the strongest positive correlation (R = 0.98, p < 0.001) with the tumor proportion, as illustrated in Figure 1D. This module encompassed a total of 5,679 genes (Supplementary Table 1).

Figure 1. Identification of the key gene modules in WGCNA. (A) The volcano map showed differentially expressed genes. (B) Determination of the soft-thresholding power. (C) Dendrogram of differentially expressed genes clustered based on a dissimilarity measure (1-TOM). (D) The correlation of gene modules with clinical traits. (E) Gene correlation scatter plot of the turquoise module. (F) The 1-, 3-, and 5-years ROC for the top ten genes. (G) Comparison of the clinical significance of LAPTM4A and RP2.

We further selected the top 5% hub genes (Supplementary Table 2) and identified 50 genes associated with glioma patient survival using HR >1 and p < 0.05 as criteria. Through univariate and multivariate COX regression analysis, as well as ROC survival curves, we determined that 12 out of the 50 genes exhibited superior predictive capability for 1-year and 5-year survival in glioma patients, as presented in Figure 1E. Notably, of these rigorously selected genes, RP2 and LAPTM4A demonstrated not only independent prognostic values for glioma patients but also displayed high Area Under ROC Curve (AUC) values. Importantly, these two genes have not been previously reported or studied extensively in the context of glioma (refer to Figure 1F).

Furthermore, we investigated the correlation between the expression of RP2 and LAPTM4A and the clinicopathological features of glioma patients using the CGGA database. Interestingly, we observed that LAPTM4A exhibited greater clinical significance (Figure 1G). Ultimately, LAPTM4A was identified as the gene of primary interest in our study.

Expression and clinical parameters of LAPTM4A in glioma

To determine the expression level of LAPTM4A, we conducted a comprehensive analysis of LAPTM4A mRNA levels using data from the TCGA and GTEx databases. Our results, as depicted in Figure 2A, demonstrated a significant increase in LAPTM4A expression in 16 different types of carcinomas, with particularly high expression observed in LGG, GBM, and GBMLGG. This upregulation of LAPTM4A in LGG, GBM, and GBMLGG was further confirmed by box diagrams displayed in Figure 2B–2D. Moreover, in order to establish the association between LAPTM4A expression in GBMLGG and various clinic parameters, we assessed LAPTM4A expression levels in different cohorts stratified by WHO grade, histological type, IDH status, 1p19q codeletion, age, gender, OS event, and primary therapy outcome, utilizing the TCGA and GTEx databases (Figure 2E–2L). Our analysis revealed a significant correlation between LAPTM4A expression and clinicopathological features, with the exception of gender (p < 0.01). Notably, LAPTM4A expression exhibited an increasing trend as the WHO grade advanced. Furthermore, LAPTM4A was found to be overexpressed in IDH wild-type cases when compared to IDH mutant cases, and higher expression levels were observed in 1p19q non-co-deletion as opposed to co-deletion. Collectively, these findings suggest that LAPTM4A expression is elevated in gliomas and holds promise as a potential biomarker for assessing glioma progression.

Figure 2. Expression of LAPTM4A in glioma. (A) The expression level of LAPTM4A in different types of tumor tissues and normal tissues in the TIMER database. (p < 0.05) (B–D) Expression levels of LAPTM4A were higher than corresponding normal tissues in LGG, GBM, and GBMLGG samples. The box plot showed the association of LAPTM4A expression with clinicopathological characteristics. (E) WHO grade, (F) Histological type, (G) IDH status, (H) 1p/19q codeletion, (I) Age, (J) Gender, (K) OS event, (L) Primary therapy outcome.

Over-expression of LAPTM4A predicted an unfavorable prognosis in glioma

Considering its significant expression in LGG, GBM, and GBMLGG, the prognostic and diagnostic value of LAPTM4A was assessed. Utilizing the TCGA and GTEx databases, we calculated the correlation between LAPTM4A expression in LGG, GBM, and GBMLGG, and patient prognosis, including overall survival (OS), disease-specific survival (DSS), and progression-free interval (PFI). Our findings revealed a significant association between overexpression of LAPTM4A and poor OS (p = 2.8 e-27) and DSS (p = 6.9 e-27) as well as PFI (p = 4.7 e-24) in GBMLGG (Figure 3A–3C). Similarly, overexpression of LAPTM4A was associated with unfavorable OS (p = 1.1 e-6) and DSS (p = 1.5 e-6), and PFI (p = 4.8 e-6) in LGG (Figure 3D–3F), and correlated with poor OS (p = 0.03) and DSS (p = 0.03) and PFI (p = 0.02) in GBM (Figure 3G–3I). The validation of survival across all WHO grades of primary (p = 0.028) and recurrent glioma (p = 0.04) was performed using the CGGA database (Supplementary Figure 1A), yielding consistent results. Furthermore, the area under the ROC curve was 0.982, 0.992, and 0.984 in LGG, GBM, and GBMLGG, respectively, indicating superior diagnostic accuracy for LAPTM4A (Supplementary Figure 1B–1D). Subsequently, we developed a nomogram integrating the significant clinic parameters mentioned above to estimate the survivability of GBMLGG patients (Supplementary Figure 1E). The predictive performance of the nomogram was evaluated using calibration curves, which demonstrated close agreement between predicted and actual 1, 3, and 5-year survival durations (Supplementary Figure 1F). In conclusion, our findings support the conclusion that overexpression of LAPTM4A is predictive of an unfavorable prognosis in gliomas.

Figure 3. Relationship between LAPTM4A and prognosis of glioma patients. LGG patients with higher expression levels of LAPTM4A had unfavorable (A) OS, (B) DSS, and (C) PFS. GBMLGG patients with higher expression levels of LAPTM4A had awful (D) OS, (E) DSS, and (F) PFS. GBM patients with higher expression levels of LAPTM4A had undesirable (G) OS, (H) DSS, and (I) PFS.

Association of LAPTM4A with DNA methylation and genetic alterations in glioma

To elucidate the underlying cause of LAPTM4A dysregulation, we conducted investigations into DNA methylation and genetic variation of LAPTM4A. Our study aimed to determine the methylation status of LAPTM4A across different subtypes and WHO grades of glioma. Boxplots were employed to visualize the variations in LAPTM4A methylation levels among the subtypes and WHO grades (Supplementary Figure 2A, 2B). Intriguingly, high-grade gliomas exhibited lower levels of methylation compared to low-grade gliomas. Furthermore, we utilized the MethSurv network tool to delve deeper into the methylation patterns of LAPTM4A in GBM and LGG. The resulting heatmap illustrated a predominance of hypomethylation at most sites within LAPTM4A (Supplementary Figure 2C, 2D). Notably, hypomethylated sites, such as cg11645081, were associated with poor survival outcomes in LGG (Supplementary Figure 2E). However, this phenomenon was more prevalent in GBM, as exemplified by sites cg04515480, cg10383839, and cg17989428 (Supplementary Figure 2F–2H). In conjunction with previous findings, we observed an increase in LAPTM4A expression with increasing grade, while the level of LAPTM4A methylation decreased with glioma grade.

Moreover, genetic variation plays a crucial role in gene expression and tumorigenesis. Through analysis using the cBioPortal website, we determined that the mutation rate of LAPTM4A in gliomas was generally low, with amplification being the most common alteration (Supplementary Figure 3A). However, a detailed examination of copy number variation revealed significant differences in LAPTM4A expression among neutral, gain, and loss groups in glioma (Supplementary Figure 3B). Notably, we discovered an association between LAPTM4A expression and neoplasm-related mutant genes in glioma. In GBM, the LAPTM4A high expression group exhibited a higher frequency of SSPO mutations and a lower frequency of TP53 and ATRX mutations (Supplementary Figure 3C). Similarly, in LGG, the LAPTM4A high expression group displayed a higher frequency of TP53, ATRX, and EGFR mutations, while having a lower frequency of IDH1, CIC, FUBP1, NOTCH1, and ZBTB20 mutations (Supplementary Figure 3D). Furthermore, in GBMLGG, the LAPTM4A high expression group demonstrated higher frequencies of TTN, PTEN, and EGFR mutations, and lower frequencies of IDH1, CIC, and FUBP1 mutations (Supplementary Figure 3E).

LAPTM4A is associated with multiple immune-related and cancer-relevant pathways in glioma

To gain further insights into the functional significance of LAPTM4A, a series of comprehensive analyses were conducted. Leveraging the LinkedOmics web resource, we performed GO pathway enrichment analysis, which revealed significant enrichment of LAPTM4A in various terms, including neutrophil-mediated immunity, acute inflammatory response, interferon-gamma production, control of humoral immune response, and immunological effector processes (Figure 4A). Additionally, KEGG pathway enrichment analysis highlighted the enrichment of LAPTM4A in specific pathways such as complement and coagulation cascades, the phagosome, antigen processing and presentation, cytokine-cytokine receptor interaction, and the lysosome (Figure 4B). Moreover, utilizing the TCGA database, we demonstrated the association between immune and cancer-associated pathways and LAPTM4A expression. Notably, LAPTM4A expression showed positive associations with cellular responses to hypoxia, apoptosis, the inflammatory response, angiogenesis, and epithelial-mesenchymal transition (EMT) (Figure 4C, 4D). To validate these findings, we further employed the CancerSea database to explore the comprehensive enrichment pathways in various glioma subtypes, which corroborated our previous results (Figure 4E). Notably, high-grade glioma (HGG) exhibited the strongest association with the aforementioned pathways (Figure 4F). In conclusion, LAPTM4A plays a significant role in multiple immune-related and cancer-relevant pathways in glioma.

Figure 4. Pathway enrichment analysis of LAPTM4A. (A, B) Significantly enriched GO and KEGG pathways of LAPTM4A. GO: Gene Ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes. (C, D) Correlation of LAPTM4A expression and cancer-related pathways. (E) Functional relevance of LAPTM4A in pan-cancers from cancerSEA. (F) Functional relevance of LAPTM4A in GBMLGG from cancerSEA red plots suggested a positive correlation, while blue plots suggested a negative correlation.

LAPTM4A may affect the invasion and migration through the EMT pathway in glioma

Previously, the Cancer Genome Atlas Network classified glioblastoma multiforme into four distinct molecular subtypes: proneural, neural, mesenchymal, and classical [40]. Mesenchymal GBM cells are known to exhibit enhanced motility and invasion, in addition to higher expression levels of proteins related to cell movement, as compared to epithelial tumor cells. To explore the expression discrepancies of LAPTM4A among these subtypes, we assessed LAPTM4A expression levels in the Bao, Phillips, and glioma Rembrandt datasets. Our findings showed that mesenchymal subtype gliomas express LAPTM4A more frequently than other subtypes (Supplementary Figure 4A–4C). We further conducted a ROC analysis to evaluate the diagnostic power of LAPTM4A on the mesenchymal phenotype, which yielded an AUC of 0.859, 0.815, and 0.790 for LAPTM4A in Bao [41], Phillips [42], and Rembrandt [43], respectively, indicating its strong diagnostic potential (Supplementary Figure 4D–4F). To corroborate these results, we performed plasmid-mediated knockdown of LAPTM4A in the U251 cell line and assessed the expression of key molecules involved in epithelial-mesenchymal transition (EMT), such as N-cadherin, E-cadherin, and MMP9, using qRT-PCR and Western blotting. Interestingly, LAPTM4A knockdown resulted in higher levels of E-cadherin protein expression, while simultaneously lowering the protein levels of N-cadherin and MMP9 (Figure 5A, 5B). As EMT programming has been implicated in the metastasis of malignant tumor cells originating from epithelial cells [44], we further explored the relationship between LAPTM4A expression and tumor metastasis using the Transwell assay. The results demonstrated that the knockdown of LAPTM4A inhibited the invasion and migration of glioblastoma cells (Figure 5C, 5D). Collectively, our findings suggest that suppression of LAPTM4A expression may hinder glioblastoma invasion and migration through the EMT pathway.

Figure 5. LAPTM4A may affect the invasion and migration through the EMT pathway in glioma. Changes in the expression of n-cadherin, e-cadherin, and MMP 9 after LAPTM4A knockdown (A) with mRNA aspect, (B) with protein aspect. (C, D) Effect of LAPTM4A knockdown on glioma cell invasion and migration.

PPI network analysis and molecular docking model of LAPTM4A

Significant scientific advancements have shed light on the pivotal role of protein-protein interactions in various fundamental biological processes within living cells [45]. To explore this further, we utilized the GeneMANIA website to construct an interaction network involving LAPTM4A and other proteins (Supplementary Figure 5A). This analysis revealed that LAPTM4A physically interacts with six proteins. Of interest was the notable protein-protein interaction observed between LAPTM4A, MCOLN1, and IGF2BP3. Furthermore, we investigated the secondary structures of these proteins, including valuable information about chemical remodeling sites such as glycosylation, acetylation, ubiquitination, and phosphorylation, using the cBioPortal database (Supplementary Figure 5B). To gain insights into the three-dimensional structures of LAPTM4A, MCOLN1, and IGF2BP3, we employed the AlphaFold Protein Structure and PDB databases for tertiary structure prediction. Subsequently, employing the H-DOCK server, we predicted potential binding sites between LAPTM4A, IGF2BP3, and MCOLN1, as visually represented in Supplementary Figure 5C, 5D.

Location and infiltration analysis of LAPTM4A in the tumor microenvironment integrating of single-cell sequencing analysis and the ESTIMATE algorithm

The intricate molecular compositions of both the internal and external tumor microenvironment, primarily comprised of immune and stromal cells alongside other associated cell types, play a crucial role in nurturing tumor growth. In order to identify the key cell types expressing LAPTM4A within the cancer microenvironment, we conducted a comprehensive single-cell analysis of LAPTM4A across 80 cancer sample datasets. By utilizing the TISCH online tool, we examined the expression levels of LAPTM4A in 34 distinct cell types, including immune cells, stromal cells, malignant cells, and functional cells, as depicted in the heatmap presented in Supplementary Figure 6A.

Notably, our findings highlighted a predominant expression of LAPTM4A in immune cells, particularly monocytes and macrophages, across various cancer types. This observation was consistent across multiple databases, such as GliomaGSE102130 (Supplementary Figure 6B) and Glioma_GSE131928_10X (Supplementary Figure 6C), where LAPTM4A was primarily expressed in monocytes/macrophages and AC-like malignant cell clusters.

Furthermore, to elucidate the relationship between LAPTM4A expression and the degree of infiltration within LGG, GBM, and GBMLGG, we performed correlation studies using stromal scores, immune scores, and ESTIMATE scores obtained from the ESTIMATE method (Figure 6). The results demonstrated a positive association between LAPTM4A expression and immunological, stromal, and estimation scores across LGG, GBM, and GBMLGG, with a particularly noteworthy correlation observed in GBMLGG (correlation coefficient >0.5). These findings further support the notion that LAPTM4A is predominantly expressed in monocytes/macrophages and AC-like malignant cells, and is intricately linked to the tumor microenvironment in gliomas.

Figure 6. The association between LAPTM4A with the tumor microenvironment. (A) The relevance between LAPTM4A expression and the stromal score in GBM, LGG, and GBMLGG. (B) The relevance between LAPTM4A expression and the immune score in GBM, LGG, and GBMLGG. (C) The relevance between LAPTM4A expression and the ESTIMATE score in GBM, LGG, and GBMLGG.

Relationship between LAPTM4A expression and the tumor immune cell infiltration and immune-related molecules

Previous pathway enrichment analyses have established a strong association between LAPTM4A and immune-related pathways in gliomas. First, we utilized the TIMER method to investigate the relationship between LAPTM4A and immune infiltration in GBMLGG. Our results, depicted in scatter plots, revealed a significant association of LAPTM4A with neutrophils, macrophages, and dendritic cells in GBMLGG, and a mild correlation with B cells, CD4+ T cells, and CD8+ T cells (Figure 7A).

Figure 7. Analysis of the correlation between LAPTM4A expression and immune cell infiltration. (A) The relevance between LAPTM4A expression and the infiltration of five immune cells. (B) The association between LAPTM4A expression and the infiltration of various immune cells in pan-cancers. (C) The connection between LAPTM4A expression and several notable biomarkers of Macrophage/Monocyte and cancer-associated fibroblast.

To further explore this relationship between LAPTM4A expression and immune cell subtypes in LGG and GBM, we utilized the TIMER2 website. Our results, as illustrated in the heatmap, demonstrated a positive correlation between LAPTM4A expression and the infiltration levels of common lymphoid progenitor, cancer-associated fibroblast, macrophage, monocyte, and neutrophil in LGG, and macrophage, common lymphoid progenitor, and cancer-associated fibroblast in GBM. Conversely, LAPTM4A was negatively correlated with NK cells and T cells CD4+ Th1 in LGG and GBM (Figure 7B).

Given the strongly positive association between LAPTM4A and monocytes/macrophages, as well as cancer-associated fibroblasts, we further examined the relationship between LAPTM4A and their biomarkers at the mRNA level (Figure 7C). To comprehensively investigate the link between LAPTM4A and immune system function, we performed gene co-expression analyses of LAPTM4A and immune-involved genes, including MHC, immunological activation, immunosuppression, chemokine, and chemokine receptor genes. Our outcomes revealed that the majority of immune-related molecules were co-expressed with LAPTM4A and positively correlated with LAPTM4A in gliomas (Supplementary Figure 7). Taken together, our findings suggest that LAPTM4A plays an essential role in immune infiltration and immune system function in gliomas.

Association between LAPTM4A expression and immunotherapy

Immunotherapy has emerged as a pivotal approach in the management of carcinoma. To delve into the association between LAPTM4A and immunotherapy, we focused on the expression patterns of eight immune checkpoint (ICP) genes in relation to LAPTM4A expression levels. Our findings demonstrated that when LAPTM4A was highly expressed, there was a corresponding upregulation of immune checkpoint genes (Figure 8A, 8B). Notably, LAPTM4A exhibited a strong correlation with PDCD1LG2, CD274, and HAVCR2.

Figure 8. The relationship between LAPTM4A expression and immunotherapy. (A) LAPTM4A differential expression status of the immune checkpoint genes under the high and low expression groups. (B) The correlation between LAPTM4A and the immune checkpoint genes. (C) The TIDE score of the LAPTM4A.

To gain a deep understanding of the impact of LAPTM4A expression on the response to immune checkpoint blockade (ICB) treatment, we determined the TIDE scores for GBMLGG using the TIDE algorithm. It is worth noting that high TIDE scores are indicative of a diminished response to ICB therapy and shorter survival following such therapy. Figure 8C showed that the LAPTM4A high expression group exhibited higher TIDE scores in GBMLGG, thus classifying LAPTM4A as a risk factor (Figure 8C). In summary, the downregulation of LAPTM4A expression may hold promise for enhancing the efficacy of immunotherapy in glioma patients.

Drug sensitivity analysis of LAPTM4A

To evaluate the influence of LAPTM4A on drug resistance in patients, we conducted an in-depth analysis of the correlation between LAPTM4A expression and multidrug sensitivity using the CTD and GSCALite platforms. Our study revealed that the expression of LAPTM4A was under the regulation of 35 small-molecule drugs (Figure 9A). Leveraging the CGP2016 database, we further examined the IC50 values of drugs based on LAPTM4A expression. Venn diagrams created with CTD and CGP2016 data showed that a specific medication has the capacity to downregulate LAPTM4A expression (Figure 9B), making patients more receptive to doxorubicin in cases where LAPTM4A expression was high (Figure 9C).

Figure 9. Prediction of LAPTM4A expression-related drugs. (A) An advanced network diagram shows 37 cancer-related drugs that can modulate LAPTM4A expression. (B) A Venn diagram demonstrates drugs related to LAPTM4A expression in CTD and cgp2016. (C) Relationship between LAPTM4A expression and IC50 of doxorubicin. (D) LAPTM4A is resistant to 73 drugs and sensitive to 9 drugs.

We also harnessed the power of the GSCA website to investigate the drug sensitivity associated with LAPTM4A expression. Remarkably, patients with high levels of LAPTM4A expression were found to be resistant to 41 small-molecule drugs (Figure 9D). An additional Venn diagram, integrating data from GSCA and CGP2016, revealed a total of eight drugs, seven of which were consistent with our previous results and exhibited sensitivity in patients with high LAPTM4A expression (Supplementary Figure 8).

These findings not only enhance our understanding of the relationship between LAPTM4A expression and drug responses but also present novel therapeutic avenues for glioma patients with elevated LAPTM4A expression.

An FGD5-AS1-hsa-miR-103a-3p-LAPTM4A axis may regulate the progression of glioma

Emerging evidence has consistently highlighted the pivotal role of competing endogenous RNA (ceRNA) networks in the context of cancer [46]. In this study, we endeavored to construct a comprehensive ceRNA regulatory network centered around LAPTM4A in gliomas. Through meticulous screening of databases such as TargetScan, DIANAmicroT, and RNAinter, we identified a total of 67 miRNAs that exhibit binding potential with LAPTM4A (Supplementary Figure 9A). Notably, 15 of these miRNAs displayed a negative correlation with LAPTM4A (Supplementary Figure 9B). Given the significance attributed to hsa-miR-103a-3p and hsa-miR-107 in glioma, as elucidated by pertinent literature, we proceeded to investigate the relationship between LAPTM4A expression and these two miRNAs in glioma (Supplementary Figure 9C). Interestingly, a stronger correlation was observed between LAPTM4A and hsa-miR-103a-3p, prompting us to delve deeper into the underlying regulatory mechanism between hsa-miR-103a-3p and LAPTM4A. Intriguingly, putative binding sites were identified, leading to subsequent mutations in LAPTM4A mRNA (Supplementary Figure 9D). Experimental results subsequently revealed that the introduction of the hsa-miR-103a-3p mimic led to a reduction in luciferase activity in U251 cells, while the hsa-miR-103a-3p inhibitor elicited an increase in luciferase activity (Figure 10A, 10B). Moreover, the hsa-miR-103a-3p mimic effectively downregulated LAPTM4A expression in U251 cells, while the hsa-miR-103a-3p inhibitor exerted the opposite effect, elevating LAPTM4A levels (Figure 10C, 10D). Clinically, abnormal down-expression of miR-103a-3p in glioma tissue was observed in comparison to normal adjacent tissue (Supplementary Figure 9E). Furthermore, the expression levels of LAPTM4A in glioma tissue demonstrated a negative correlation with miR-103a-3p (Supplementary Figure 9F).

Figure 10. Prediction of the ceRNA network in glioma. (A, B) The regulating relationship of has-miR-103a-3p and LAPTM4A was investigated by a dual-luciferase reporter gene system. (C, D) Real-time qPCR was used to determine LAPTM4A mRNA levels in U251 cells. (E) The regulating relationship of has-miR-103a-3p and FGD5-AS1 was investigated by a dual-luciferase reporter gene system. (F) RNA-pull down assay was performed to detect has-miR-103a-3p enrichment in FGD5-AS1. (G) Western blotting assay was performed to detect LAPTM4A expression levels in the control and FGD5-AS1 group. (H) Western blotting assay was performed to detect LAPTM4A expression levels in the control, FGD5-AS1, and has-miR-103a-3p group. (I) Transwell assay was used to detect U251 cell metastasis in the control, FGD5-AS1, inh-NC, and inh-has-miR-103a-3p group. ^*p < 0.05, ^**p < 0.01.

Subsequently, leveraging the miRNet and starBase databases, we employed predictive analysis to identify potential regulatory relationships between hsa-miR-103a-3p and long non-coding RNAs (lncRNAs) (Supplementary Figure 9G). A scatter plot analysis revealed LAPTM4A exhibited a statistically significant positive correlation with FGD5-AS1 (r = 0.381) (Supplementary Figure 9H). Similarly, FGD5-AS1 as the most significantly correlated lncRNA with miR-103a-3p (r = −0.318) (Supplementary Figure 9I). Encouragingly, potential binding sites between FGD5-AS1 and miR-103a-3p were predicted, further leading to mutations in FGD5-AS1 mRNA (Supplementary Figure 9J). Notably, the dual-luciferase reporter gene system demonstrated that transfection of the FGD5-AS1 mimic successfully reduced relative luciferase activity in U251 cells transfected with wild-type hsa-miR-103a-3p, while no such reduction was observed in cells transfected with the corresponding mutant (Figure 10E). Additionally, over-expressed FGD5-AS1 decreased hsa-miR-103a-3p levels in U251 cells (Figure 10F). Furthermore, overexpression of FGD5-AS1 demonstrated an elevation in LAPTM4A levels in U251 cells (Figure 10G). Strikingly, the co-overexpression of FGD5-AS1 and hsa-miR-103a-3p effectively reversed the heightened expression levels of LAPTM4A (Figure 10H). Notably, Transwell assay results revealed a substantial inhibition in the invasion and migration capacities of U251 cells upon FGD5-AS1 knockdown, which could be rescued by the hsa-miR-103a-3p inhibitor (Figure 10I). Collectively, our findings illuminate the intricate regulatory mechanism governing LAPTM4A, underscoring the significance of the FGD5-AS1-hsa-miR-103a-3p-LAPTM4A axis in the regulation of glioma progression.