The brain imaging feature-related gene NRP2 drives the malignant progression of glioblastoma through the FAK pathway: a Mendelian randomization study

Highlights

• •255 brain imaging features show causal links to glioblastoma risk through MR analysis.
• •NRP2 is identified as a key driver gene connecting imaging traits to tumor progression via FAK signaling.
• •NRP2 knockdown inhibits glioblastoma growth in vitro and in vivo, reversible by FAK activation.
• •Dactinomycin is nominated as a potential therapeutic agent targeting the NRP2-FAK axis.

Abstract

Background

Glioblastoma (GBM) is an aggressive primary brain tumor with poor prognosis. Although brain imaging features are related to biological behaviors of GBM, the causal relationship between them remain unclear.

Objectives

To explore the causal relationship between brain imaging features and GBM, identify key pathogenic genes, and provide a perspective for GBM therapy.

Methods

Two-sample MR analysis was employed. Causal relationships were evaluated based on brain imaging features, eQTL, and GWAS data. Differentially expressed brain imaging-related genes were screened through gene mapping and differential expression analysis. MR analysis on eQTL data identified key genes, and GSEA was performed. Given its robust genetic association, high expression in GBM, and enrichment association with tumor malignancy, NRP2 was determined as the core gene, with its function verified by in vitro/in vivo experiments.

Results

MR analysis identified 255 GBM-associated brain imaging features, with 9 key genes selected. NRP2 was identified as a risk gene. NRP2 knockdown significantly inhibited GBM proliferation, migration, and invasion and promoted apoptosis. The inhibitory effects were reversed by activated FAK-signaling pathway. Mechanistically, NRP2 regulated FAK phosphorylation through direct binding, thereby activating the Focal-adhesion pathway and promoting tumor malignancy. In animal experiments, inhibiting NRP2 slowed tumor growth, which was weakened by FAK agonists.

Conclusion

This study establishes the causal relationship between brain imaging features and GBM from a genetic perspective. NRP2 activates Focal-adhesion pathway through FAK signaling to drive GBM progression. NRP2 is a key molecule connecting imaging phenotypes and GBM malignant behaviors, serving as a potential therapeutic target.

Graphical abstract

Introduction

Glioma is the most common primary intracranial malignant tumor in adults, accounting for approximately 80 % of all malignant brain cancer cases [1]. Among them, glioblastoma (GBM) is the most common subtype, accounting for 60.2 % of gliomas and 50.9 % of all intracranial malignant tumors [2]. The median diagnosis age of GBM patients is 66 years, and the median survival period is usually <8 months, with a poor prognosis. Currently, the standard treatment for GBM includes surgical resection, followed by adjuvant fractionated radiotherapy, and concurrent or adjuvant chemotherapy with temozolomide [3]. However, GBM often develops resistance to radiotherapy and chemotherapy, and due to its aggressive nature and its location deep within the brain tissue, complete surgical resection is extremely difficult [[4], [5]]. In recent years, advances in brain imaging technology have provided significant support for the surgical treatment of GBM. For instance, magnetic resonance imaging (MRI) can present the structural and functional information of the tumor non-invasively and without radiation [[6], [7]], which is of great significance for guiding surgical resection.

In the surgical treatment of GBM, brain imaging technology can be used for preoperative tumor grading, localization, and intraoperative navigation; its imaging features are also closely related to the biological behavior of GBM [[8], [9]]. For example, before bevacizumab treatment, diffusion MRI features can serve as an independent biomarker for the overall survival of patients with recurrent GBM with high tumor burden, but the causal nature of this association is not yet clear [10]. Genome-wide association studies (GWAS) have successfully identified single nucleotide polymorphisms (SNPs) corresponding to brain imaging features associated with the risk and progression of GBM [11]. However, GWAS results are difficult to use to directly determine the pathogenic genes, and thus have limited translational value in drug development. Incorporating genetic information into the drug development process is considered an important strategy to improve success rates. Evidence has shown that drugs with genetic support have greater potential in clinical trials [[12], [13]]. By mapping brain imaging feature-related SNPs to corresponding genes using tools such as SNPnexus, it is expected to explain the intrinsic relationship between brain imaging features and GBM at the mechanistic level, thereby providing a theoretical basis for the development of new therapeutic approaches [14].

Therefore, this study explores the causal association between brain imaging features and GBM through a two-sample Mendelian Randomization (MR) analysis. By integrating multi-omics data, key genes related to brain imaging features are screened out, and pathway enrichment analysis is conducted to analyze their potential mechanisms. Subsequently, functional experimentsin vitro andin vivo are combined to verify the impact of key genes on GBM and explore the underlying molecular mechanisms.

Method

Data sources

The three types of genetic data involved in this study were all derived from public databases. The GWAS data for brain imaging features were sourced from the UK Biobank (UK Biobank), encompassing 39,691 samples with 3935 imaging features [15]. The eQTL data for the brain came from a study of 2865 European individuals’ cerebral cortex samples, covering 16,704 genes and approximately 11.6 million cis-acting SNPs (minor allele frequency MAF > 0.01) [16]. The GWAS data for GBM were obtained from the FinnGen database (https://www.finngen.fi/en), including 345,496 samples (378 cases and 345,118 controls) [17]. To avoid bias due to population heterogeneity, all analyses were based on data from the European population. Detailed information for each dataset can be found in Table 1.

Table 1.

Detailed information on the genome-wide association studies in our analysis.

Traits	Year	Consortium/ID	Sample size	Case	Control
Brain imaging	2021	GCST90002426–6360	39,691	/	/
Brain cis‐eQTL	2022	/	2865	/	/
Brain glioblastoma	2024	FinnGen	345,496	378	345,118

The gene expression data for GBM were downloaded from the Gene Expression Omnibus (GEO) database, including the following two independent datasets: GSE68848 (228 GBM samples and 28 non-tumor samples) and GSE4290 (77 GBM samples and 23 non-tumor samples) [[18], [19]]. Both datasets were generated using the GPL570 (Affymetrix Human Genome U-133 plus 2.0 GeneChip) platform.

Selection of instrumental variables (IVs)

All the MR analyses in this study were conducted based on the following three core hypotheses: (1) IV is significantly correlated with the exposure factor; (2) IV is not correlated with any confounding factors between the exposure and the outcome; IV only affects the outcome through the exposure factor and has no other direct paths.

The selection criteria for IV are as follows. (1) SNPs that are significantly correlated with the exposure factor at the genome-wide level were selected as IV (significance threshold set at P < 5 × 10–6). (2) To exclude the influence of linkage disequilibrium (LD), clumping was conducted on the SNPs (parameter set as r² = 0.001, region length = 10,000 kb). (3) The F statistic was calculated to evaluate the strength of IV, and only SNPs with an F statistic > 10 were retained to ensure the validity of IV. (4) The allele direction of the SNP in the exposure and outcome datasets was adjusted, with ambiguous or conflicting palindrome structures and incompatible SNPs eliminated. (5) The MR-Pleiotropy Residual Sum and Outlier methods (MR-PRESSO) were used to detect and eliminate outliers in the SNPs.

For the brain tissue eQTL data, we extracted the cis IVs closely related to the genes associated with differential expression brain imaging (DEBIRGs), that is, SNPs located within 100 kb upstream and downstream of the gene coding region, and we set the LD coefficient r² = 0.001. The remaining screening criteria are consistent with the above (1), (3), (4), and (5). The IV finally included in all the MR analyses of this study are detailed in Table 2.

Table 2.

IVs in MR analysis.

No.	Exposure	Outcome	IVs
1	Brain image features	GBM	Supplementary Table 1
4	DEBIRGs in brain tissues	GBM	Supplementary Table 2

GBM: Glioblastoma; DEBIRGs: Differentially expressed Brain imaging-related genes.

MR analysis

This study employed five MR analysis methods for estimation, including the random effects inverse variance weighted method (IVW), MR-Egger regression, weighted median method, weighted mode method, and simple mode method. Among them, the results of the IVW method were taken as the main analytical basis. When there was only one IV (SNP) for the exposure factor, the Wald ratio method was used to calculate the effect value [20]. In all MR analyses, a P value < 0.05 was used as the statistical significance criterion.

Sensitivity analysis

To verify the robustness of the MR analysis results, we conducted a comprehensive sensitivity analysis [21]. Firstly, Cochran’s Q test was used to determine the heterogeneity of SNPs. When P < 0.05, it indicated the presence of heterogeneity, and then the random-effects IVW model was used. Secondly, MR-Egger regression and MR-PRESSO global tests were employed to evaluate the level of pleiotropy. A P value < 0.05 indicated the possible existence of significant pleiotropic bias. The “Leave-one-out” test was conducted for sensitivity analysis to ensure that our MR results were not overly influenced by a single SNP. In addition, the Steiger analysis method was applied [22] to determine the directionality of the causal association to ensure that the outcome was not influenced by reverse causation, with P > 0.05 indicating the presence of a reverse causal association.

Gene mapping

By using the SNPnexus online platform [[14], [23]], the SNPs corresponding to the brain imaging features that had a significant causal association with GBM were mapped to the overlapping genes in their genomic locations, thereby converting the genetic signals into candidate genes.

Screening of key DEBIRGs and key genes

The “limma” R package, we separately screened the differentially expressed genes (DEGs) between GBM samples and non-tumor samples in two GEO datasets (screening conditions: |logFC| > 1, FDR < 0.05) [24]. We took the intersection of these DEGs with the genes related to brain imaging features to obtain the differentially expressed brain imaging-related genes (DEBIRGs). Finally, using the above key DEBIRGs as instrumental variables in the eQTL data of brain tissues, we conducted MR analysis with GBM again to determine the final key genes.

Gene set enrichment analysis (GSEA)

The GSEA software was used to conduct GSEA on the key genes. During the analysis, all genes were sorted based on their log₂FC values in the differential expression analysis, and the enrichment significance was determined by setting the false discovery rate (FDR) to be <0.25.

Cell culture

The human GBM cell lines U251 and T98G used for functional experiments were purchased from Mingzhou Bio (MINGZHOUBIO, MZ-0186, China) and American Type Culture Collection (ATCC, CRL-1690, USA), respectively. U251 cells were cultured in DMEM medium containing 10 % FBS (ThermoFisher, 11,965,092, USA). T98G cells were cultured in MEM medium containing 10 % FBS (ATCC, 30–2003, USA). The 293T cells used for lentivirus packaging and production were purchased from ATCC (CRL-3216, USA) and cultured in DMEM medium containing 10 % FBS. 1 % penicillin/streptomycin (Corning, New York, USA) was added to all the cells and they were cultured in a 37 °C, 5 % CO₂ incubator (ThermoFisher, 51,032,873, USA).

Cell transfection and drug treatment

To obtain a cell line with stable knockdown of NRP2 and simultaneous expression of luciferase, this study employed the lentiviral system for operation. The designed sh-NRP2 sequence was inserted into the pLKO.1-puro plasmid. Subsequently, the corresponding recombinant plasmids were co-transfected with the packaging plasmid (psPAX2) and the envelope plasmid (pMD2.G) into 293T cells for virus packaging. Virus supernatants were collected 48–72 h after transfection and used for infecting target cells. After 48 h of infection, the drug screening was performed using puromycin. Finally, a stable NRP2 knockdown cell line was obtained. For co-transfection, the pLKO.1-puro-sh-NRP2 plasmid was mixed with the FAK overexpression plasmid (oe-FAK) and transfected together into 293T cells according to the above method for virus packaging, and then the virus supernatant was used to infect target cells. The transfection effect was verified by qRT-PCR.

GBM cells were treated with 10 μM focal adhesion kinase (FAK) activator Adhesamine (HY-122,672, MCE, USA) for 24 h, and the cells were divided into sh-NC group, sh-NRP2 group and sh-NRP2+Adhesamine group.

qRT-PCR

Total cellular RNA was extracted by utilizing TRIzol reagent (Invitrogen, USA), with RNA concentration and purity determined using NanoDrop One micro-Ultraviolet-Vis spectrophotometer (ThermoFisher, 840–317,500, USA). SuperScript II reverse transcriptase (ThermoFisher, 15596018CN, USA) was utilized to reverse transcribe RNA into cDNA. qRT-PCR was carried out using TB Green ® Premix Ex Taq ™ II (Takara, RR820A, Japan) on an ABI 7500 PCR system (ThermoFisher, 4351,105, USA). GAPDH was chosen as a reference gene, with the relative mRNA expression levels calculated by using the 2^−ΔΔCt method. The primer sequence can be found in Table 3.

Table 3.

qPCR primers.

Gene	Forward Primer (5′−3′)	Reverse Primer (5′−3′)
NRP2	CTACATCAAGTTCACCTCCGAC	GGATACTTCTCAGGAAACCCAG
FAK	CAGGGTCCGATTGGAAACCA	CTGCAGGATTTCTTTCCGCC
GAPDH	CGGGAAGCTTGTCATCAAT	TCTCCATGGTGGTGAAGA

CCK-8 assay

Cell viability was measured using the Cell Counting Kit-8 (MCE, HY-K0301, USA). The transfected cells were collected and the cells were inoculated into 96-well plates, with 2000 cells per well. After 0, 24, 48, 72, and 96 h of culturing of cells, 10 μL of CCK-8 test reagent was added. Cells were then incubated in the incubator for 2 h. Absorbance was measured at a wavelength of 450 nm using a microplate reader. The biological experiments were repeated 3 times.

Colony formation assay

The transfected cells were digested with 0.25 % trypsin and inoculated in a 12-well plate at a density of 800 cells/well. Subsequently, the cells were incubated in a DMEM or MEM medium containing 10 % FBS and 5 % CO₂ at room temperature for two weeks. The culture medium was discarded when visible colonies appeared. Colonies were fixed with 4 % paraformaldehyde, stained with 0.1 % crystal violet for 30 min, and rinsed with PBS. The cell colonies were then enumerated.

Scratch healing assay

In six-well plates, GBM cells were inoculated (each well with 1 × 10⁵ cells). When the cells reached 80 %, they were scratched with a 10 μL pipette tip, followed by a wash in phosphate-buffered saline (PBS) to remove cell debris. They were then further cultivated for another 24 h. The imaging situation was observed on a microscope and analyzed using ImageJ software. The cell migration ability was represented by the relative migration rate of cells before and after healing.

Transwell assay

In the Transwell experimental apparatus (Corning, 3460, USA), 100 μL of matrix gel (BD Biosciences, USA) was coated onto the upper chamber. GBM cells (1 × 10⁵ per well) were resuspended in serum-free culture medium and inoculated into the upper chamber, while 10 % FBS (Thermo Fisher Scientific, USA) cell culture medium was added to the lower chamber. After 24 h of culture, the cells were fixed with 4 % paraformaldehyde and stained with 0.1 % crystal violet. The imaging was observed under a microscope and analyzed using ImageJ software.

Western blot

The cells were lysed using RIPA lysis buffer (Beyotime, P0013B, China) and the total protein concentration was determined using the BCA protein quantification kit (Beyotime, P0009, China). Equal amounts of 30 μg protein samples were separated by 10 % SDS-PAGE gel electrophoresis, and the protein was transferred onto a PVDF membrane (Millipore, IPVH00010, USA). After the transfer, the membrane was placed in a 5 % skimmed milk solution (Beyotime, P0216–300 g, China) at room temperature for 1 hour of blocking, and then incubated with the primary antibody at 4 °C overnight. The next day, the membrane was washed three times with TBST buffer, and then incubated with the HRP-labeled secondary antibody (Donkey Anti-Rabbit IgG (Beyotime, A0239, China, 1:5000)) at room temperature for 1 hour. Finally, the membrane was washed thoroughly with TBST and imaged using the Super Sensitivity ECL Chemiluminescence Substrate Kit (Beyotime, P0018S, China) on the ChemiDoc MP imaging system. The information on the primary antibody is shown in Table 4.

Table 4.

Information Table of WB Primary Antibodies.

Gene	Manufacturer	Country	Item Number	Host	Dilution ratio
pY397-FAK	abcam	UK	ab81298	Rabbit	1:2000
FAK	Proteintech	China	Ag3331	Rabbit	1:20,000
Cleaved-caspase-3	Proteintech	China	230527B2	Rabbit	1:10,000
Cleaved-PARP	abcam	UK	ab32064	Rabbit	1:2000
BAX	Proteintech	China	50,599-2-Ig	Rabbit	1:60,000
BCL2	Proteintech	China	12,789-1-AP	Rabbit	1:9000
GAPDH	Proteintech	China	10,494-1-AP	Rabbit	1:5000

Annexin V/PI double staining method for detecting cell apoptosis

We collected cells that had been treated in different ways and washed them twice with cold PBS. After resuspending, a cell suspension (1 × 10⁶ cells) was prepared. According to the instructions of the kit, we added 5 μL Annexin V-FITC and 10 μL PI staining solution (BD Biosciences, 560,931, USA) and incubated them at room temperature in the dark for 15 min. Immediately after incubation, we used a flow cytometer (Agilent, USA) for detection. Based on the fluorescence signals, we distinguished live cells (Annexin V⁻/PI⁻), early apoptotic cells (Annexin V⁺/PI⁻), and late apoptotic/necrotic cells (Annexin V⁺/PI⁺). The proportion of apoptotic cells was calculated to reflect the level of cell apoptosis.

Immunofluorescence co-localization

After discarding the cell culture medium, we washed the cells 1–2 times with PBS and then fixed them with 4 % paraformaldehyde. The blocking solution was used for immunostaining (Beyotime, P0102, China) to incubate at room temperature for 1 hour. After discarding the blocking solution, we added the primary antibodies (Anti-FAK antibody, Proteintech, 66,258-1-Ig, Mouse, China) (Anti-NRP2 antibody, Proteintech, 86,126-3-RR, Rabbit, China) at 4 °C for overnight incubation. The corresponding secondary antibodies Goat Anti-Mouse IgG H&L (Alexa Fluor® 647) (Abcam, ab150115, UK) and Goat Anti-Rabbit IgG H&L (Alexa Fluor® 488) (Abcam, ab150077, UK) were added at room temperature for 45 min. The slides were mounted with a fluorescence quenching mounting solution containing DAPI. We took pictures using a Zeiss microscope (Germany).

Immunoprecipitation

After collecting cells treated by different methods, they were lysed using cell lysis buffer (Sigma, R0278, USA) and centrifuged to obtain protein supernatants, which were divided into three groups: Input, IgG, and IP. The Input group was not added with antibodies and was stored at −20 °C; the IgG group was added with rabbit anti-IgG antibody (Abcam, ab172730, UK); the IP group was added with the target antibodies Anti-FAK antibody(Proteintech, 12,636-1-AP, Rabbit, China) and Anti-NRP2 antibody (Proteintech, 27,193-1-AP, Rabbit, China), and incubated at 4 °C for overnight rotation to form complexes between the antibodies and the target proteins. The next day, pre-treated Protein A/G magnetic beads (Thermo Scientific, USA) were added to the antibody-protein complexes and slowly shaken at 4 °C for 2 h to couple the antibodies with the magnetic beads. The magnetic beads were collected and washed with IP lysis buffer, and the target proteins were obtained after boiling with loading buffer and used for WB experiments for verification.

Nude xenograft mouse model construction and administration method

This animal experiment was approved by the Animal Care Welfare Committee of the Guizhou Medical University (NO. 2502698). 9 female nude mice (4 weeks) were purchased from Shanghai SLAC Laboratory Animal Co., Ltd. (Shanghai, China) and raised under sterile conditions (12 h black-and-white light cycle, 25 °C, 60 %−70 % humidity). The nude mice were randomly divided into three groups (n = 3/group): sh-NC group, sh-NRP2 group, and sh-NRP2 + Adhesamine group. T98G cells stably expressing sh-NC or sh-NRP2 were subcutaneously injected into the abdomens of the nude mice to establish xenograft tumor models. Starting from the 6th day after tumor formation, only the sh-NRP2 + Adhesamine group of nude mice received FAK agonist Adhesamine (25 mg/kg) orally, twice daily for 24 consecutive days; the sh-NC group and sh-NRP2 group were given the same volume of solvent as a control. Tumor volumes were monitored at 5, 10, 15, 25, and 30 days, and the calculation formula was: tumor volume = (length × width²)/2. After the experiment, all nude mice were euthanized (under excessive isoflurane anesthesia followed by cervical dislocation), and tumor tissues were excised for weighing and photography for the record.

Immunohistochemistry (IHC)

The freshly collected animal tissues were immersed in 4 % paraformaldehyde (Beyotime, China) for fixation for 48 h. The fixed tissues were then embedded in paraffin to form tissue wax blocks. Subsequently, the wax blocks were placed on a microtome and cut into 4 μm sections. The sections were then deparaffinized in xylene, hydrated with gradient ethanol (100 %, 95 %, 85 %, 75 %), and then subjected to antigen retrieval with Tris-EDTA buffer (Beyotime, China) for 20 min at room temperature. The sections were then blocked with 10 % goat serum (Beyotime, China) for 30 min at room temperature and treated with 3 % H₂O₂ for 20 min to block endogenous peroxidase activity. Subsequently, the sections were added with the corresponding primary antibodies and incubated overnight at 4 °C. The next day, a secondary antibody was added and the sections were incubated at room temperature for 1 hour. Finally, the sections were treated with 2,3-diaminobenzidine (DAB, Yeasen, China) for 5 min, stained with hematoxylin (Beyotime, China) for 1 min, dehydrated in gradient ethanol (75 %, 85 %, 95 %, 100 %), transparented with xylene, and sealed with neutral gum. The sections were then observed and imaged randomly in five fields under a microscope (Carl Zeiss, Germany). Antibody information is presented in Table 5.

Table 5.

IHC Antibody Information.

Gene	Manufacturer	Country	Item Number	Host	Dilution ratio
Ki67	Abcam	UK	ab16667	Rabbit	1:200
BAX	Abcam	UK	ab32503	Rabbit	1:500
BCL2	Abcam	UK	ab182858	Rabbit	1:1000

Statistical analysis

Two-sample MR analysis was achieved using R (version 4.3.1) software and the R packages “Two Sample MR” and “MR-PRESSO” (version 1.0). All wet test data were processed by GraphPad Prism 8.2.1 software. All experiments were done at least three times, and the results were expressed as mean ± standard deviation. Differences between groups were compared using t-tests or one-way ANOVA. P < 0.05 indicates a statistically significant difference. * indicates P < 0.05.

Results

Research approach

The technical route of this study is shown in Fig. 1. Firstly, a two-sample MR analysis method was employed to evaluate the causal relationship between brain imaging features and GBM. On this basis, eQTL data related to key DEBIRGs were used for MR analysis with GBM to screen key genes, and further GSEA analysis was conducted to explore their potential pathway mechanisms. Finally, by integrating in vivo and in vitro experiments, the impact of key genes and their enriched pathways on the malignant progression of GBM was deeply verified.

Fig. 1.

Technical roadmap of the research.

Brain imaging features and MR results of GBM

Two-sample MR analysis was undertaken using brain imaging features as exposure and GBM as outcomes. IVW results demonstrated that 267 brain imaging features had significant causal associations with GBM (P < 0.05) (Supplementary Table 3).

The sensitivity analysis indicated that, after excluding the 3 features with significant heterogeneity and the 9 features with multi-level effects, the results of the remaining 255 brain imaging features were all robust and reliable. These results passed the directional test of causal association, confirming the preset direction of the causal relationship (Supplementary Table 4). The leave-one-out test further showed that all the identified causal associations were not driven by a single SNP, ensuring the reliability of the MR analysis results.

Gene mapping and screening of DEBIRGs

The 255 brain imaging feature-related SNPs were mapped to genes using the SNPnexus database, and a total of 2154 genes were identified (Supplementary Table 5). Subsequently, the GSE68848 dataset was standardized, and a differential expression gene analysis was conducted between GBM samples and control samples (Fig. 2A). The same method was used for the GSE4290 dataset for differential expression gene analysis (Fig. 2B). Subsequently, the DEGs from the two datasets were intersected with the brain imaging feature-related genes, resulting in 382 key DEBIRGs (Supplementary Table 6), including 114 up-regulated genes (Fig. 2C) and 268 down-regulated genes (Fig. 2D).

Fig. 2.

Screening of DEGs. (A) Volcano map of difference analysis in GSE68848; (B) Volcano map of difference analysis in GSE4290; (C) Venn diagram of up-regulated genes; (D) Venn diagram of down-regulated genes.

MR results for key DEBIRGs and GBM

To explore the association between DEBIRGs and GBM, we obtained 382 brain tissue eQTL data of DEBIRGs and used them as IVs for MR analysis with GBM. Through the Wald ratio or IVW method, we initially found 14 genes that had significant causal associations with GBM. After further eliminating the genes whose expression trends (upregulation/downregulation) were inconsistent with the causal effect direction in Section 2.2, we finally identified 9 key genes that had robust causal associations with GBM (Fig. 3). Among them, 5 genes (HSPG2, DSE, NRP2, IL15RA, and SCIN) were significantly positively correlated with GBM; 4 genes (BASP1, INPP5A, STX1A, and TULP4) were significantly negatively correlated with GBM. Sensitivity analysis supported the reliability of this result.

Fig. 3.

MR result comparison between brain imaging feature-related genes and GBM.

GSEA enrichment analysis of key genes

To further elucidate the biological functions of the key genes, we conducted single-gene GSEA enrichment analysis. Further, we explored the pathways regulated by the 9 key genes through single-gene GSEA enrichment analysis. Among them, the cell cycle (Cell-cycle), ECM and receptor interaction (ECM-receptor interaction) pathways were significantly enriched in the HSPG2 high-expression group (Fig. 4A) and the BASP1 low-expression group (Fig. 4B). In the DSE high-expression group and the NRP2 high-expression group, the complement-and-coagulation-cascades, cytokine-cytokine receptor interaction pathways were significantly enriched (Fig. 4C and 4D). In addition, the NRP2 high-expression group also significantly enriched the Focal-adhesion pathway (Fig. 4D). Other GSEA enrichment results can be found in Supplementary Figure 1 and Fig. 2.

Fig. 4.

GSEA of key genes. (A) The significantly enriched pathways in the HSPG2 high-expression group. (B) The significantly enriched pathways in the BASP1 low-expression group. (C) The significantly enriched pathways in the DSE high-expression group. (D) The significantly enriched pathways in the NRP2 high-expression group.

NRP2 mediates the Focal-adhesion pathway to regulate the malignant progression of GBM cells

Among the key genes we identified, NRP2 has a relatively high OR value (2.117), suggesting that NRP2 might be a potential risk gene for GBM. Additionally, it is found that when the function of NRP2 is disrupted, it can inhibit the proliferation of glioblastoma cells [25]. Therefore, we conducted in vitro experiments to verify the role of NRP2 in GBM. The analysis of the GSE68848 and GSE4290 datasets revealed that NRP2 was significantly upregulated in tumors (Fig. 5A). In response to this, we transfected sh-NRP2 into GBM cell lines to knockdown it and confirmed the knockdown effect (Fig. 5B). Since NRP2 was enriched and upregulated in the Focal-adhesion pathway in the previous study, we hypothesized that NRP2 might exert its function through this pathway. Therefore, we applied the FAK activator Adhesamine in sh-NRP2 cells and finally divided the cells into the negative control group (sh-NC), the NRP2 knockdown group (sh-NRP2), and the activator intervention group (sh-NRP2 + Adhesamine). Western blotting (WB) analysis revealed that knockdown of NRP2 significantly inhibited the phosphorylation form of FAK, and this effect could be weakened by Adhesamine (Fig. 5C). The results of CCK-8, colony formation, scratch assay, Transwell, flow cytometry, and WB showed that knockdown of NRP2 significantly inhibited the proliferation, migration, invasion, and anti-apoptotic ability of the cells. Compared with the sh-NRP2 group, the proliferation, migration, and invasion abilities of the cells in the sh-NRP2 + Adhesamine group were significantly restored, and the cell apoptosis level was reversed (P < 0.05) (Fig. 5D-I). These results indicate that NRP2 may mediate the regulation of Focal-adhesion pathway to control the proliferation and migration of GBM cells.

Fig. 5.

NRP2 mediates the FAK signaling pathway to regulate the malignant progression of GBM cells. (A): NRP2 expression levels in patients from the GSE4290 and GSE68848 data sets. (B): The knockdown effect of sh-NRP2 in U251 and T98G cells was detected by qRT-PCR. (C): WB was used to detect the expressions of pY297-FAK and FAK. (D): The CCK-8 method was used to detect the cell viability of GBM cells in different treatment groups (sh-NC, sh-NRP2, sh-NRP2 + adhesamine). Adhesamine is a FAK activator. (E): The colony formation experiment was used to detect the number of colony formations of GBM cells in different treatment groups; (F): The scratch healing experiment was used to detect the migration ability of GBM cells in different treatment groups; (G): The Transwell experiment was used to detect the invasion ability of GBM cells in different treatment groups; (H): The Annexin V/PI double staining method was used to detect the apoptosis level of cells in different treatment groups. (I): Western blot was used to detect the protein expression of pro-apoptotic indicators clever-caspase3, clever-PARP, BAX and pro-apoptotic indicator BCL2 in different treatment groups. * indicates P < 0.05, **** indicates P < 0.0001.

NRP2 affects the Focal-adhesion pathway through FAK to regulate the progression of GBM

The phosphorylation of key proteins FAK (and its homolog Pyk2) in adherens junction pathways is significantly upregulated in recurrent glioblastoma, and their activity is closely related to tumor regeneration and proliferation [26]. Therefore, we hypothesized that NRP2 might affect the adherens junction pathway by regulating key protein FAK and thereby influence the malignant phenotype of GBM. To further verify this hypothesis, we examined the effects of NRP2 knockdown on the expression of FAK and its phosphorylated protein. The results showed that the total expression of FAK did not change significantly, while the expression of its phosphorylated protein pY397-FAK significantly decreased after NRP2 knockdown, indicating that NRP2 might mediate the effect of FAK on the progression of GBM (Fig. 6A). We observed a clear co-localization between NRP2 and FAK, and the binding between them decreased after NRP2 knockdown, suggesting a spatial association between the two (Fig. 6B). Additionally, co-immunoprecipitation results showed that NRP2 and FAK had a significant binding in the sh-NC group, and the binding effect weakened after NRP2 knockdown (Fig. 6C). In conclusion, we can draw the conclusion that NRP2 affects the Focal-adhesion pathway through FAK and thereby regulates the progression of GBM.

Fig. 6.

NRP2 affects the Focal-adhesion pathway through FAK to regulate the progression of GBM. (A): Western blot was used to detect the expression levels of FAK and its phosphorylated proteins in GBM cells of different treatment groups (sh-NC, sh-NRP2). (B): Immunofluorescence co-localization was used to verify the binding of NRP2 and FAK. (C): Immunoprecipitation was used to verify the binding of NRP2 and FAK in different groups.

NRP2 promotes the malignant progression of GBM by activating FAK

Based on this, we further conducted in vitro verification by constructing a cell model co-transfected with sh-NRP2 and oe-FAK. After confirming the transfection efficiency through qRT-PCR (Fig. 7A), we analyzed the expression changes of pathway-related proteins using WB. The results showed that knockdown of NRP2 significantly inhibited the expression of pY397-FAK, while overexpression of FAK could partially reverse the decrease in FAK phosphorylation caused by NRP2 knockdown (Fig. 7B). The cell function experiments also indicated that the overexpression of FAK could significantly restore the reduced cell viability, decreased proliferation ability, restricted migration and invasion, and decreased anti-apoptotic ability caused by NRP2 knockdown (Fig. 7C-F). In summary, the above results further suggest that NRP2 may promote the progression of GBM by activating the FAK signaling axis.

Fig. 7.

NRP2 promotes the malignant progression of GBM by activating FAK. U251 and T98G cell groups: sh-NC, sh-NRP2, sh-NRP2 + oe-FAK. (A): qRT-PCR was used to detect the mRNA level of FAK; (B): WB was used to detect the protein expression of pY397-FAK and FAK; (C): CCK-8 was used to detect cell viability; (D): Scratch test was used to detect the migration ability of cells; (E): Transwell test was used to detect the invasion ability of cells; (F): WB was used to detect the protein expression of apoptotic proteins clever-caspase3, clever-PARP, BAX and anti-apoptotic protein BCL2. * indicates P < 0.05.

The nude mouse experiment demonstrates that NRP2 mediates the Focal-adhesion pathway to regulate the growth of GBM tumors

We further constructed a nude mouse model to verify the regulatory effect of NRP2 on the Focal-adhesion pathway. T98G cells stably transfected with sh-NC and sh-NRP2 were subcutaneously injected into the abdomens of nude mice to establish a xenograft model. The nude mice with the sh-NRP2 model were divided into the adhesamine treatment group and the untreated group for FAK activator adhesamine treatment. The experimental results showed that the tumor size, volume and weight of the nude mice in the sh-NRP2 group were significantly lower than those in the sh-NC group (P < 0.05). The tumor size, volume and weight of the sh-NRP2 nude mice treated with FAK activator adhesamine were restored (P < 0.05) (Fig. 8A-C). The IHC detection results further showed that knockdown of NRP2 could inhibit the expression of proliferation marker protein Ki67 in tumor tissues, while promoting the upregulation of apoptosis-related protein BAX and the downregulation of anti-apoptotic protein BCL2 (Fig. 8D). The WB analysis results of FAK in the tissue were consistent with previous findings, and knockdown of NRP2 could significantly inhibit the phosphorylation level of FAK, while overexpression of FAK could significantly reverse this effect (Fig. 8E). These results further verified that NRP2 mediates the regulation of the Focal-adhesion pathway in controlling the growth of GBM tumors.

Fig. 8.

The nude mouse experiment demonstrates that NRP2 mediates the Focal-adhesion pathway to regulate the formation of GBM tumors. (A): Nude mice were subcutaneously injected with sh-NRP2 and sh-NC stably transfected T98G cells. Comparison of tumor size in sh-NRP2 nude mice treated with FAK activator adhesamine. (B): Tumor weights of different treatment groups of nude mice. (C): Tumor volumes of different treatment groups of nude mice. (D): IHC detection of the expression of Ki67, BCL2 and BAX in tumor tissues; (E): WB detection of the expression of pY297-FAK and FAK in tumor tissues.

Discussion

Glioblastoma is the most common and aggressive subtype of glioma. It grows rapidly and has a high degree of malignancy. Due to its significant tumor heterogeneity and resistance to conventional treatments, the treatment of GBM remains a major challenge in the field of neuro-oncology [27]. In recent years, advancements in brain imaging technology have not only significantly improved the surgical treatment outcomes of GBM, but also provided important information for patient management [[28], [29]]. However, the causal relationship between these brain imaging features and GBM still requires more genetic evidence to be clarified.

This study first based on large-scale genetic data of brain imaging features, systematically screened molecular signals with causal associations to GBM from an imaging perspective. Through a two-sample MR analysis of 3935 brain imaging features in the UK Biobank and GBM, we identified 255 brain imaging features with robust causal associations to GBM. Further, we mapped the SNPs related to these imaging features and combined the causal inference of transcriptional differences in GBM and brain tissue eQTL data, ultimately screening out 9 key genes with stable causal associations to GBM (HSPG2, DSE, NRP2, IL15RA, SCIN, BASP1, INPP5A, STX1A, and TULP4).

The study found that BASP1 is a protective factor for GBM, and its expression is negatively correlated with the risk of GBM. Bioinformatics analysis indicated that BASP1 is associated with a favorable prognosis for GBM patients, and knocking down BASP1 can promote the proliferation and migration of GBM cells [30]. The mechanism may be related to the pathways downregulated by BASP1: high expression of BASP1 can inhibit the synthesis of extracellular matrix (ECM) [31], and the activation of the ECM-receptor interaction pathway can promote the progression of GBM [32], which is also consistent with our GSEA enrichment results.

On the contrary, HSPG2, DSE, and NRP2 were identified as risk genes that may promote the progression of GBM. HSPG2 (Heparan sulfate proteoglycan 2) encodes key components of the basement membrane and ECM [33]. Compared with normal brain tissue, the RNA level of HSPG2 in GBM samples was significantly increased [34], and its high expression can promote tumor development by altering the ECM in the brain [35]. DSE (Dermatan Sulfate Epimerase) encodes an enzyme that synthesizes dermatan sulfate (DS) [36], and it is highly expressed in GBM cell lines. Knocking out DSE can inhibit the proliferation of GBM cells [37]. A study has shown that DS proteoglycans can affect the proliferation and migration of tumor cells through specific interactions with cytokines (such as vascular endothelial growth factor) [38]. Moreover, our results and previous studies both show that the cytokine-cytokine-receptor-interaction pathway is involved in the malignant development of GBM [39].

It is worth noting that among the multiple candidate risk genes, although HSPG2, DSE, and NRP2 may all be involved in the development of GBM, multiple pieces of evidence indicate that NRP2 has more prominent research value. At the genetic level, the MR analysis shows that NRP2 has a robust and consistent causal association with various brain imaging features (manifested by a higher OR value and a significant positive correlation with the risk of GBM), suggesting that it may be a key molecule connecting the imaging phenotype and the occurrence of the disease. Secondly, at the expression and function levels, NRP2 is significantly highly expressed in GBM tissues, and participates in regulating the malignant phenotypes of tumor cells, such as cell proliferation [25]. Based on the above genetic evidence and its biological rationality, we determined NRP2 as the core candidate gene for this study and further systematically explored its functional role and molecular mechanism in GBM. NRP2 (neuropilin-2) is an important type I transmembrane glycoprotein receptor that plays an important role in tumor occurrence [40]. It has been reported that the expression level of NRP2 in GBM tissues is 46 % higher than that in normal brain tissues, and this abnormally high expression is closely related to the malignant progression of the tumor [41]. Clinical pathological analysis further confirmed that high expression of NRP2 is an independent predictor of poor prognosis in GBM patients, and its expression level is positively correlated with the invasiveness and recurrence risk of the tumor [42]. At the molecular mechanism level, highly expressed NRP2 in GBM cells exhibits unique adhesion characteristics. Compared with the cell population with low NRP2 expression, the NRP2-high-expressing population forms more Focal-adhesion on laminin, indicating that NRP2 may participate in the migration and invasion process of tumor cells by regulating the interaction between cells and the extracellular matrix [43]. Further research has found that the formation of these Focal-adhesion is closely related to the activation of the FAK signaling pathway, and activating the FAK pathway in GBM can significantly promote the proliferation and migration ability of tumor cells [44]. This suggests that there may be a functional association between NRP2 and the Focal-adhesion pathway. To clarify the relationship between NRP2 and FAK in the malignant progression of GBM, we conducted a series of rigorous cell experiments. The results showed that inhibition of NRP2 could significantly reduce the proliferation activity of GBM cells, which directly confirmed the crucial role of NRP2 in maintaining the malignant phenotype of GBM cells. More importantly, when we added a FAK-specific agonist in the context of NRP2 inhibition, a significant phenotypic reversal was observed. Specifically, FAK agonist treatment not only restored the proliferation ability of GBM cells but also significantly enhanced the migration and invasion activity of the cells. At the same time, the results of cell apoptosis detection indicated that the increase in cell apoptosis caused by NRP2 inhibition could be effectively reversed by FAK agonist, further supporting the view that NRP2 regulates FAK to affect the survival of GBM cells. At the molecular level, we detected changes in the phosphorylation status of FAK and found that the phosphorylation level of FAK protein was significantly downregulated after inhibition of NRP2 expression, which further indicates that NRP2 regulates the FAK protein and thereby affects the Focal-adhesion pathway, leading to the malignant phenotype of GBM. This result was also verified at the animal level, which fully confirmed the key role of NRP2 in driving the malignant progression of GBM by regulating the Focal-adhesion pathway.

However, our research results also have some limitations. Firstly, at the genetic analysis level, the MR analysis is mainly based on GWAS data from the European population, and its applicability in other races needs to be verified; at the same time, due to the aggregated nature of the public data, we cannot conduct stratified analysis on covariates such as age and gender. Moreover, although sensitivity analysis has been conducted, multiple comparisons of a large number of brain imaging features may still introduce a risk of false positives. At the functional validation level, although we have confirmed that NRP2 functions through the Focal-adhesion pathway, the upstream regulatory mechanism of this pathway and the specific effect molecular network of the downstream have not been fully elucidated, which leaves room for further exploration in subsequent research. Moreover, the xenograft tumor experiments are limited by the animal experimental conditions, and the sample size is relatively limited, which may affect the statistical power. Nevertheless, the results of in vivo experiments are highly consistent with those of in vitro experiments at the tumor phenotype and molecular level, which, to a certain extent, support the reliability of the research conclusion. Future research is necessary to conduct more in-depth verification of the related mechanisms on the basis of expanding the sample size, introducing multiple animal models, and using data from different populations.

Conclusion

This study systematically elucidated the causal role and molecular mechanism of brain imaging features in the occurrence and development of glioblastoma (GBM) through integrating genetic analysis and functional experiments. Firstly, based on large-scale genomic data, through a two-sample MR analysis, 255 brain imaging features that were significantly causally associated with GBM risk were identified. Further, 9 key genes, including NRP2, HSPG2, and BASP1, were selected from these features, and a bridge from imaging phenotypes to genetic etiology was constructed. Focusing on the key gene NRP2, a series of in vitro experiments confirmed that inhibiting NRP2 could significantly weaken the proliferation, migration, and invasion abilities of GBM cells and induce cell apoptosis. Further mechanism exploration revealed that NRP2 directly interacted with FAK to regulate its phosphorylation level, thereby activating the Focal-adhesion pathway and driving the malignant progression of tumors. This effect was verified in animal models, indicating that targeting NRP2 effectively inhibited tumor growth in vivo, while FAK agonists reversed this effect. In summary, this study reveals that the gene NRP2, related to brain imaging features, affects the Focal-adhesion pathway through FAK and thereby influences the malignant phenotype of GBM, providing new theoretical basis and potential intervention targets for the early risk warning and targeted treatment of GBM.

Funding

This work is supported by The National Natural Science Foundation of China (81871333, 82260340), Guizhou Province Science & Technology Project ([2020]4Y159 and [2021]430), Discipline leading talent of The Affiliated Hospital of Guizhou Medical University (gyfyxkrc-2023-04), 942 plan for Guizhou Province Key Clinical Specialties.

CRediT authorship contribution statement

Yang Li: Writing – original draft, Software, Resources, Investigation, Formal analysis, Data curation, Conceptualization. Jun Que: Writing – original draft, Validation, Methodology, Formal analysis, Data curation, Conceptualization. Yong Xia: Writing – review & editing, Writing – original draft, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Lei Wang: Writing – review & editing, Visualization, Validation, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Pinshan Zhang: Writing – review & editing, Supervision, Resources, Investigation, Formal analysis, Data curation, Conceptualization. Zhen Cheng: Writing – review & editing, Writing – original draft, Validation, Project administration, Methodology, Funding acquisition, Conceptualization. Bo Gao: Writing – review & editing, Writing – original draft, Supervision, Project administration, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix. Supplementary materials

Supplementary Figure 1 Upregulated pathway for key gene enrichment. (A) BASP1. (B) INPP5A. (C) IL15RA. (D) STX1A. (E) TULP4. (F) SCIN.

Supplementary Figure 2 Downregulated pathway for key gene enrichment. (A) HSPG2. (B) DSE. (C) INPP5A. (D) NRP2. (E) IL15RA. (F) STX1A. (G) TULP4. (H) SCIN.