Next Article in Journal
Escape Survival and Scale Damage Assessment of Red Mullet (Mullus barbatus Linnaeus, 1758) during Bottom Trawling in the Central Mediterranean Sea
Previous Article in Journal
Therapeutic Potential of a Senolytic Approach in a Murine Model of Chronic GVHD
Previous Article in Special Issue
The Tropomyosin Family as Novel Biomarkers in Relation to Poor Prognosis in Glioma
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Biology
Volume 12
Issue 5
10.3390/biology12050648
biology-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Identification of Prominent Genes between 3D Glioblastoma Models and Clinical Samples via GEO/TCGA/CGGA Data Analysis

by
Brandon Wee Siang Phon
1,
Saatheeyavaane Bhuvanendran
1,
Qasim Ayub
2,3,4,
Ammu Kutty Radhakrishnan
1 and
Muhamad Noor Alfarizal Kamarudin
1,*
1
Jeffrey Cheah School of Medicine and Health Sciences, Monash University Malaysia, Jalan Lagoon Selatan, Bandar Sunway 47500, Malaysia
2
School of Science, Monash University Malaysia, Bandar Sunway 47500, Malaysia
3
Monash University Malaysia Genomics Facility, Monash University, Bandar Sunway 47500, Malaysia
4
Tropical Medicine and Biology Multidisciplinary Platform, Monash University Malaysia, Bandar Sunway 47500, Malaysia
*
Author to whom correspondence should be addressed.
Biology 2023, 12(5), 648; https://doi.org/10.3390/biology12050648
Submission received: 16 March 2023 / Revised: 20 April 2023 / Accepted: 21 April 2023 / Published: 25 April 2023
(This article belongs to the Special Issue Biomarkers and Immunotherapeutic Targets in Glioblastoma)
Browse Figures
Versions Notes

Abstract

:

Simple Summary

Glioblastoma multiforme (GBM) is a deadly brain tumour with little progression in the way of improved quality of life in patients. The possible root cause for this dilemma stems from the inefficiency of traditional two-dimensional (2D) cell-based models used in evaluating potential anticancer agents. In this paper, we investigated the proximity with which three-dimensional (3D) GBM models encapsulate key aspects of clinical GBM samples that have been made available in many genetic data banks. The analysis identified a number of key genes highly expressed in both 3D GBM cell-based models and GBM clinical samples that may play a role in GBM therapy resistance. In conclusion, the findings suggest that 3D GBM cell-based models serve as reliable models for clinical GBM samples.

Abstract

A paradigm shift in preclinical evaluations of new anticancer GBM drugs should occur in favour of 3D cultures. This study leveraged the vast genomic data banks to investigate the suitability of 3D cultures as cell-based models for GBM. We hypothesised that correlating genes that are highly upregulated in 3D GBM models will have an impact in GBM patients, which will support 3D cultures as more reliable preclinical models for GBM. Using clinical samples of brain tissue from healthy individuals and GBM patients from The Cancer Genome Atlas (TCGA), Gene Expression Omnibus (GEO), Chinese Glioma Genome Atlas (CGGA), and Genotype-Tissue Expression (GTEx) databases, several genes related to pathways such as epithelial-to-mesenchymal transition (EMT)-related genes (CD44, TWIST1, SNAI1, CDH2, FN1, VIM), angiogenesis/migration-related genes (MMP1, MMP2, MMP9, VEGFA), hypoxia-related genes (HIF1A, PLAT), stemness-related genes (SOX2, PROM1, NES, FOS), and genes involved in the Wnt signalling pathway (DKK1, FZD7) were found to be upregulated in brain samples from GBM patients, and the expression of these genes were also enhanced in 3D GBM cells. Additionally, EMT-related genes were upregulated in GBM archetypes (wild-type IDH1R132 ) that historically have poorer treatment responses, with said genes being significant predictors of poorer survival in the TCGA cohort. These findings reinforced the hypothesis that 3D GBM cultures can be used as reliable models to study increased epithelial-to-mesenchymal transitions in clinical GBM samples.

1. Introduction

Glioblastoma multiforme (GBM), a grade 4 adult-type diffuse glioma, is the most common malignant and clinically aggressive brain tumour, and it is notorious for its unrelenting aggressiveness. The standard therapy against GBM, known as the Stupp protocol, involves surgical resection ranging from a minimally invasive biopsy to a craniotomy for maximal safe resection followed by radiotherapy and concurrent administration of temozolomide (TMZ) [1]. Reliable as it is, the Stupp protocol has only produced meagre improvements in GBM median survival, i.e., extending the median survival by an additional two months from the initial 12.1 months [1]. Although it has been two decades since the establishment of the Stupp protocol, the median survival of GBM patients has scarcely improved. The latest Central Brain Tumor Registry of the United States (CBTRUS) report documented the lowest overall survival for GBM patients at a mere eight months, a number lower than most brain neoplasms’ overall survival by at least half [2]. The high failure rate of anticancer agents used is one of the stumbling blocks in GBM therapy, as only 10% of drugs in preclinical evaluations make it to the clinic [3].
Across the globe, two-dimensional (2D) cultures remain the in vitro model of choice for preclinical evaluations. Recent advances in the cell culture approach, such as 3D cultures, have placed the widespread use of 2D cultures under intense scrutiny [4,5]. Despite its ease of use, the unnatural geometric and mechanical constraints imposed in a 2D environment come with several drawbacks that result in a lower resemblance of actual GBM phenotypes. These include the absence of tumour cell interactions, low similarity to actual tumour phenotypes, and unrestricted availability of oxygen, nutrients, and metabolites, which do not accurately replicate actual tumour microenvironments. Furthermore, the tumour microenvironment has been postulated to directly influence the genomic landscape of GBM tumours, which underlie the aggressive phenotype that confers higher resistance to therapy [5,6,7]. Additionally, immune cells, stromal cells, endothelial cells, mesenchymal cells, and the extracellular matrix in the cellular milieu also contribute to tumour progression [8]. The lack of these complex biochemical interplays in a 2D environment hence harbours a different microenvironmental pressure, causing cells cultured in 2D to adopt a different transcriptomic landscape [9,10] that eventually differs from that of derived tumours [11,12].
The pitfalls of 2D cultures have led to the rise in popularity of three-dimensional (3D) cultures for several reasons. For one, it is a relatively inexpensive tool that can serve as a bridge between highly restrictive 2D cultures and in vivo xenograft animal models, which have high cost, high maintenance, and also come bundled with increased ethics conundrums. 3D models have been postulated to possess the ability and genomic landscape to better mimic GBM tumour phenotypes, allowing for more robust drug evaluations [13]. In our recent scoping review, we reported on 45 genes (Table 1) that showed different expressions in 3D cultures when compared to 2D cultures, which are also investigated in the current study [14].
With the advent of technological advancements in the field of genomics, genetic data banks have continued to skyrocket in availability and size, with comprehensive data deposited in databases such as The Cancer Genome Atlas (TCGA), Gene Expression Omnibus (GEO), Chinese Glioma Genome Atlas (CGGA), and Genotype-Tissue Expression (GTEx), facilitating the evaluation of novel clinical hypotheses. While there are ample empirical data and investigations on clinical GBMs resulting from these databases, there have not been many analyses performed on the huge amount of transcriptomic data between preclinical GBM models and clinical GBM samples.
In this study, we analysed some of the huge genetic datasets that are available in databases to derive clinically relevant information regarding the transcriptomic landscape observed in 3D GBM cultures. We compared these with the transcriptomic data obtained from brain samples of GBM patients and normal/healthy subjects. We also combined the key findings from our previous analysis of 2D versus 3D cultures (Table 1) with several bioinformatic analyses, thus discovering the synonymous upregulation of genes found in 3D cultures that carries significant clinical significance in GBM patients. Hence, this paper provides evidence to support 3D GBM cultures as suitable preclinical models to more closely emulate GBMs in a clinical setting. Some definitive answers to the debate about the use of 2D and 3D cultures are also discussed.

2. Materials and Methods

2.1. RNA-Seq Expression Data

RNA-Seq expression data from GBM patients and healthy brain samples together with the clinical information from their respective datasets were downloaded from GEO (https://www.ncbi.nlm.nih.gov/geo/, accessed on 4 May 2022). Raw RNA-Seq expression counts for GSE147352 were provided and thus directly obtained from GEO [15]. For GSE145645 [16] and GSE165595 [17], raw RNA-Seq fastq files were downloaded using SRA Explorer in the absence of any provided raw RNA-Seq counts. To obtain raw RNA-Seq counts from the aforementioned expression datasets, STAR [18] was leveraged to align the RNA transcripts to Ensembl’s reference human genome (GRCh38 primary assembly). Quantifications were produced with the help of RSEM [19] using ENCODE3′s STAR-RSEM pipeline parameters. Paired-end alignments were performed for both datasets, with exact codes and all parameters used for the alignments being provided.
For validation, level 3 gene expression profiles of the TCGA GBM patient cohort and healthy brain samples were obtained from the UCSC Xena data portal (https://xenabrowser.net/datapages/, accessed on 16 May 2022) [20]. Specifically, Illumina Hiseq2000 RNA-Seq log2-transformed HTSeq counts were obtained. The clinical data of the GBM patients from the TCGA database, such as gender, age, IDH1R132 mutation status, MGMT methylation status, sample type, survival, and outcome, were obtained from both UCSC Xena [20] and cBioPortal (https://www.cbioportal.org, accessed on 16 May 2022) [21,22]. Utilisation of the Toil pipeline allowed for a unified processing workflow between the TCGA and GTEx datasets, with STAR being used to generate alignments and quantifications obtained using RSEM [23]. The recomputation of the raw RNA-Seq data from the TCGA and GTEx datasets by the UCSC Xena project made the two datasets compatible, allowing for direct expression analyses. Following the acquisition of RNA-Seq data from the TCGA/GTEx cohort, raw RNA-Seq counts of healthy brain samples and GBM patients (Illumina Hiseq; mRNAseq_325 and mRNAseq_693) and the corresponding clinical information were also obtained from CGGA (http://www.cgga.org.cn/download.jsp, accessed on 27 November 2022) [24,25,26,27]. The CGGA database kindly provided RNA-Seq data for two batches (mRNAseq_325 and mRNAseq_693), and data from grade IV GBM patients from both batches were combined and analysed to further eliminate any batch effects. Table 2 summarises the information of the number of GBM patients and normal brain samples in each RNA-Seq cohort.

2.2. Identification of Differentially Expressed Genes

Differential expression analysis was performed in RStudio v4.1.1 using the ‘DESeq2’ [28] and ‘EnhancedVolcano’ packages [29]. |Log2 fold change (LFC)| > 1 and Benjamini–Hochberg adjusted p-values (padj) < 0.05 were set as the cut-offs for screening differentially expressed genes. All R scripts used for the differential expression analysis and subsequent analyses are provided (https://github.com/thegellerbing/2D_3D_Analysis-, accessed on 22 January 2023).

2.3. Construction of Gene Interaction Network

The interaction network was generated using STRING (Search Tool for the Retrieval of INteracting Genes/Proteins) v11 database (https://string-db.org/, accessed on 30 November 2022) [30]. The database predicts interactions between differentially expressed genes based on their physical binding and regulatory interactions. It analyses the network edges of evidence, confidence, and molecular actions between the genes. The minimum required interaction score was set at the highest confidence at 0.900.

2.4. Over-Representation Analysis

Over-representation analysis of the gene sets was performed using the R package ‘clusterprofiler’ [31]. Protein-coding genes that were evaluated in the differential expression analysis were used as the background. Gene annotations of biological processes that had a p-value less than 0.01 and a q-value adjusted using the Benjamini–Hochberg method of less than 0.05 were studied.

2.5. Survival Analysis

Kaplan–Meier survival analysis was performed on the TCGA cohort obtained from UCSC Xena. The analysis was performed in RStudio v4.1.1 using the ‘survival’ [32] and ‘survminer’ [33] packages. Patients from the TCGA cohort were dichotomised into high- and low-expression groups based on the median transcript per million (TPM) value. Patients in the CGGA cohort were stratified based on the fragments per kilobase of exon per million mapped fragments (FPKM) value. The overall survival of patients between the high- and low-gene-expression groups was compared using the log-rank test. A p-value of < 0.05 denoted a statistically significant result.

3. Results

3.1. Commonly Regulated Genes in 3D Cultures Are Replicated in GBM Patients

As a training set, RNA-Seq data of GBM patients compared to normal brain tissues from three different GEO datasets were compared. Out of the 45 genes that we found in our scoping review [14], 32, 35, and 22 genes were differentially expressed, respectively. Table 3 shows the subset of differentially expressed genes in each GEO dataset based on the genes listed in Table 1. Crucially, prominent stem cell markers (PROM1, SOX2, NES), EMT markers (FN1, CD44), angiogenesis and migration markers (VEGFA, MMP2, MMP9), and hypoxia markers (HIF1A) were all upregulated in GBM patients compared to normal human brain samples. Other prominent EMT markers, such as CDH2, TWIST1, and SNAI1, as well as prominent migration marker MMP1 were all upregulated in two of the GEO datasets bar GSE165595. It has been proposed that these markers are the reason for the innate aggressiveness of GBM, an observation that is replicated in 3D cultures when compared to traditional 2D cultures (Table 1). Naturally, even 3D GBM cultures will not be able to outright replicate the transcriptomic landscape of in vivo GBM tumours, with genes such as ABCA2, CYP1A1, and EPCAM being downregulated in GBM patients whereas increased expression of said genes is observed in 3D cultures. Conversely, the expression of CCND1, CDC20, ITGA3, and MYC genes was upregulated in GBM patients compared to normal brain samples, but these genes were downregulated in 3D cultures. The full list of differentially expressed genes for GSE145645, GSE165595 and GSE147352 cohorts can be found in Tables S1–S3, respectively.
To validate the results we obtained from GEO, RNA-Seq data of GBM patients were subsequently obtained from TCGA while RNA-Seq data of normal human brain samples from the cortex and frontal cortex were obtained from the GTEx repository. In total, the gene expression profiles and clinical information of 166 GBM patients were obtained from TCGA. Healthy human brain samples were obtained from both TCGA (n = 5) and GTEx (n = 207). Further validation was obtained from another large-scale sequencing study spearheaded by Zhao et al. [24], which resulted in the compilation of the CGGA database. This allowed differential expression analysis to be performed on brain samples from 388 GBM patients and 20 healthy individuals. The clinical characteristics of the 166 GBM samples from TCGA and the 388 GBM samples from the CGGA are summarised in Table 4.
From a total of 18,347 protein-coding genes, 6704 differentially expressed genes were identified in the TCGA and GTEx cohorts, with 3274 (18%) and 3457 (19%) genes being upregulated and downregulated in GBM patient samples, respectively (Table S4, Figure 1a). Furthermore, 30 of the genes listed in Table 1 met the log2 fold change threshold (padj < 0.05) (Table 5). Meanwhile, the differential expression analysis of the CGGA database saw a total of 3962 upregulated genes (22%) and 1852 downregulated genes (10%) out of 17,748 protein-coding genes (Table S5, Figure 2a), with 28 genes listed in Table 1 meeting the stated thresholds. Table 5 summarises the genes that were differentially expressed for the two different cohorts, while Figure 2b and Figure 3b show the different expression profiles of these differentially expressed genes in brain samples from GBM patients and healthy subjects.
A total of 14 genes were found to be upregulated across all five different cohorts, with most of the prominent stem cell markers, migration and angiogenic markers, EMT markers, and hypoxia markers being upregulated. Similarly, genes that heavily influence the mesenchymal state and migration of GBM (e.g., TWIST1 and MMP1) were upregulated in both the TCGA/GTEx and CGGA cohorts (Table S6). In addition, several genes (ABCA2, CYP1A1, EPCAM, CCND1, CDC20, ITGA3, and MYC) that showed different expression profiles between 3D versus 2D cultures and GBM samples versus healthy brain samples in the GEO datasets were also replicated in the two larger cohorts.
A clear summary of the differentially expressed genes that were subset from genes listed in Table 1 are clearly shown in Table 6. With the exception of a couple of genes, the transcriptomic landscape of GBM patients is largely replicated by 3D cultures in vitro.

3.2. Differentially Expressed Genes in GBM Patients with Different Characteristics

Part of GBM’s aggressiveness stems from its heterogeneity. While GBM appears to be histopathologically similar across tumours, the molecular characteristics of GBM tumours differ from patient to patient. Several distinct molecular characteristics have dictated the response to therapy for most patients, with notable examples being the mutation status of IDH1R132 and the methylation status of MGMT promoters. IDH1R132 mutation status is now used as one of the characteristics to specify between the two grade 4 gliomas, namely GBMs with wild-type IDH1R132 and IDH1R132 mutant astrocytomas. GBMs with wild-type IDH1R132 [34] and unmethylated MGMT promoters [35] are commonly associated with increased resistance against most therapeutic measures. Hence, we investigated whether there were any associations between the genes that were differentially expressed in 3D compared to 2D GBM cultures (Table 1) and the mentioned GBM molecular parameters to determine the effectiveness of 3D GBM cultures as a suitable treatment-resistant cell-based model.
For the GSE147352 cohort, only the IDH1R132 mutation status of the GBM patients was available (n = 69), with 56 patients having wild-type IDH1R132s and 13 patients having mutant IDH1R132s. The differential expression analysis revealed that genes associated with stemness (NES, PROM1), EMT (TWIST1), angiogenesis and migration (VEGFA, MMP9), and genes involved in the Wnt signalling pathway (DKK1, FZD7) were upregulated (Table 7, |log2 fold change| >1, padj < 0.05). The GSE165595 cohort also provided information about the IDH1R132 mutation status (n = 17). A total of 15 GBM patients had wild-type IDH1R132 while 2 GBM patients had mutant IDH1R132. However, differential expression analysis of said cohort yielded no significant results for any of the genes listed in Table 1.
Subsequently, we attempted to stratify GBM patients from the TCGA and CGGA cohorts based on IDH1R132 mutation status (n = 144 and n = 378) and MGMT methylation status (n = 119 and n = 335). Any GBM samples that lacked the IDH1R132 mutation and MGMT methylation status were excluded from the following analysis. We discovered nine genes listed in Table 1 that were upregulated in GBM patients with wild-type IDH1R132 compared to GBM patients with mutant IDH1R132 in the TCGA cohort. These genes are part of the Wnt signalling pathway (FZD7, DKK1) and are mainly involved in EMT (CD44, SNAI1) and migration/invasion (MMP9, VEGFA) (Table 8). The narrative followed through in the CGGA cohort, with notable genes such as FN1, SNAI1, DKK1, FZD7, MMP1, MMP9, and VEGFA having increased expression in wild-type IDH GBMs (Table 9).
Regarding GBM patients with unmethylated MGMT promoters, the expression of TWIST1 and MMP1 genes was found to be upregulated when compared to GBM patients with methylated MGMT promoters in the TCGA and CGGA cohorts, respectively. It is worth noting that GBM patients stratified in terms of their MGMT promoter methylation status yielded a smaller number of differentially expressed genes. More specifically, the TCGA and CGGA cohorts only reported 255 (0.014%) and 618 (0.035%) differentially expressed genes, respectively, between GBM patients with unmethylated and methylated MGMT status.

3.3. Gene Interactions among Differentially Expressed Genes

STRING was utilised to better understand the relationships between the identified differentially expressed genes found in GBM samples when compared to normal brain samples. The subset of 31 genes from Table 1 that were differentially expressed in at least three cohorts was used to construct the interaction network. Figure 3 shows the resulting interaction network between the genes. Notably, there are clear notable interactions between stemness markers (NES, SOX2, FOS), hypoxia marker (HIF1A), EMT markers (TWIST1, CD44, FN1), and migration/angiogenic markers (VEGFA, MMP1, MMP2, MMP9). The mapped interaction chart seems to suggest a relationship between stemness, hypoxia, EMT, and the angiogenic process.

3.4. Functional Enrichment of the Differentially Expressed Genes

To better understand the functions of all 31 genes from Table 1 that were differentially expressed in at least 3 different cohorts, overrepresentation analysis was performed with the 31 genes as a defined gene set. Figure 4a shows the top 30 GO biological processes. Some of the notable biological processes include response to hypoxia and mesenchyme development. The enriched KEGG pathways are displayed in Figure 4b. Table 10 further summarises the statistically significant GO biological process and KEGG pathway annotations involved in GBM development.

3.5. Correlation of Upregulated Genes with GBM Patients’ Overall Survival

To round out the study, Kaplan–Meier survival curves were plotted for genes that were upregulated in GBM samples based on the survival data provided by the TCGA and CGGA databases. Among the 27 genes that were upregulated in GBM patients compared to normal human brain samples within the TCGA cohort, FN1 and TWIST1 were significant predictors of poorer overall survival based on the log-rank test (p < 0.05, Figure 5). On the other hand, 27 genes were upregulated in the CGGA cohort and a run through the list of genes revealed that 19 genes were significant predictors of poorer overall survival. It would seem that several stemness markers, EMT markers, and angiogenic markers corresponded to lower survival probabilities among CGGA GBM patients (Table 11).

4. Discussion

A strong foundation is the key to success. In terms of cancer research, said foundation would refer to the models that are used in preclinical evaluations. In our previous scoping review, we suggested that the paradigms of the more simplistic and reductive nature of 2D GBM culture preclinical models should be shifted in favour of 3D GBM models [11]. The justification for that statement was based on our findings that genes crucial to GBM survival and growth, such as genes related to stemness, the EMT process, angiogenesis, migration, and hypoxia response, were commonly upregulated in 3D models, suggesting that 3D models better define the tumour’s innate aggressiveness. In this paper, we sought to justify our stance and to search for answers regarding the clinical relevance of preclinical 3D models by utilising the vast array of genomic databases available in the information vaults of GEO, TCGA, GTEx, and CGGA. Our differential expression analysis across GEO, TCGA/GTEx, and CGGA asserted our stance vis-à-vis the paradigm shift towards 3D models in preclinical evaluations. EMT-related genes (CD44, TWIST1, CDH2, FN1, VIM, YAP1), angiogenesis/migration-related genes (MMP1, MMP2, MMP9, VEGFA), hypoxia-related genes (HIF1A, PLAT), stemness-related genes (SOX2, PROM1, NES, FOS), and genes involved in the Wnt signalling pathway (DKK1, FZD7) were all upregulated in GBM samples vs. normal brain samples in several cohorts, as seen in 3D cultures when compared to 2D cultures, which were verified in previous biological experiments [36,37,38,39,40].
These genes play a massive role in dictating GBM’s aggressive nature, with NES being a prominent marker of the primitive undifferentiated phenotypes of GBM stem cells. Able to self-renew and differentiate, GBM stem cells have often been cited to be the reason for GBM’s high relapse rate [41,42]. Driven by the Wnt signalling pathway, GBM stem cells often adopt mesenchymal phenotypes through the EMT process, increasing the cells’ migratory and invasive capabilities, which explains the increased expression of genes such as MMP1, MMP2, MMP9, and VEGFA that are involved in facilitating GBM invasion and the vascularisation process [43,44,45,46,47,48,49]. Looking into the GO biological processes, the involvement of FZD7, DKK1, SOX2, CDH2, FN1, SNAI1, TWIST1, HIF1A, MMP1, MMP2, and MMP9 in the Wnt signalling pathway, mesenchyme development, and extracellular matrix disassembly corroborate their roles in the increased invasive capabilities of GBM cells.
Furthermore, Figure 3 also demonstrated the close relationship between the upregulated genes that were found in both GBM samples and 3D cultures. The interaction network showed the distinct relationship between the genes that facilitate the EMT process, such as transcription factor TWIST1 [50] and cell surface glycoprotein CD44 that governs GBM invasion [51], with SOX2. Additionally, the relationship between these genes extended to genes that code for essential collagenases and gelatinases, such as MMP1, MMP2, and MMP9 [52], genes that facilitate vascularisation, such as VEGFA [44], and genes that code for extracellular glycoproteins that bind to membrane proteins, such as FN1 [53]. This was followed by the presence of an interaction between HIF1A, TWIST1, and VEGFA, and reports have indicated that hypoxia is a potent inducer of the EMT process, a phenotype that is easily reproducible with the hypoxic core in 3D cultures [54]. These results cement the notion that 3D cultures are superior in mimicking GBM’s innate aggressiveness compared to 2D cultures, with the formation of a hypoxic core that closely imitates the quiescent nature of the core and aggressive mesenchymal phenotypes of GBMs.
Researchers have been trying to profile the molecular and mutation characteristics of GBMs for decades. This would allow for a more obvious harmonisation of patient cohorts, facilitating more accurate therapy response predictions in the effort to achieve personalised medicine in the future. The most basic of this molecular and phenotypic profiling can be found in the form of IDH1R132 mutation and MGMT promoter methylation. As mentioned earlier, two GBM phenotypes play a role in predicting treatment response and overall survival [35,55]. Consequently, using the clinical data available in the databases, we set out to determine whether 3D GBM cultures can act as reductive models for GBMs with increased treatment resistance. In the process, we observed notable decreases in the expression of genes involved in the EMT process, Wnt signalling, and invasion, such as CD44, TWIST1, MMP1, MMP9, VEGFA, FZD7, DKK1, and SNAI1 in IDH1R132 mutant GBM in the GEO, CGGA, and TCGA/GTEx cohorts, whereas TWIST1 and MMP1 were upregulated in TCGA and CGGA GBM patients with unmethylated MGMT promoters, respectively. A pattern can be seen in the form of increased EMT and invasive behaviour being at the forefront of GBM patients with increased treatment resistance. The IDH1R132 mutation is known to facilitate inhibition of the Wnt signalling pathway, resulting in a less aggressive phenotype, possibly perpetrated by reduced mesenchymal phenotypes within the tumour [56,57]. The transcription factors involved in the EMT process have been found to directly contribute to chemoresistance by modulating the expression of MGMT [58,59]. However, arguments have to be made regarding whether the methylation status of MGMT of GBM patients affects the transcriptomic landscape since the percentage of genes that are differentially expressed in both databases is minuscule to say the least. All the data seem to point to 3D GBM cultures being a suitable substitute for GBMs with wild-type IDH1R132 in preclinical therapy evaluations. Considering most GBM cell lines contain wild-type IDH1R132, 3D models have the capability to further emphasise the resistant nature of said GBM subtypes.
All of these results drive home the need for wider adoption of 3D GBM models for preclinical evaluations. The upregulation of CD44, FN1, MMP1, MMP9, SNAI1, and VEGFA, especially within the wild-type IDH1R132 GBM group, emphasises the suitability of 3D cultures as preclinical substitutes for treatment-resistant models by simulating the EMT process and increased migratory capacity. A move to 3D cultures in preclinical evaluations may well see increased efficiency in weeding out unsuitable anticancer agents, hopefully leading to better clinical evaluation translations for future candidates. Additionally, the PI3K/Akt signalling pathway plays a central role in cell division, migration, adhesion, differentiation, and apoptosis, with its notorious role in activating EGFRs. The close relationship between VEGFA and FN1 (Figure 3) and its involvement in this highly centralising pathway not only signify the importance of inhibiting the said pathway, but it also speaks volume about the pertinence of 3D models for such studies. The enrichment of the Wnt signalling pathway as a result of increased DKK1 and FZD7 expression in GBM patients makes 3D models a prime candidate for preclinical Wnt inhibitory studies. Further evidence for the need for widespread adoption of 3D cultures in preclinical evaluation can be seen in the Kaplan–Meier survival graphs. FN1 and TWIST1, which have been prominent in this discussion, presented a direct correlation between increased expression and decreased overall survival in the TCGA cohort, a fact that has been corroborated by other studies [60,61]. On the other hand, the CGGA cohort indicated a majority of the genes (CD44, CDH2, FN1, FOS, MMP1, MMP2, MMP9, NES, SNAI1, VEGFA, and VIM) that were repeatedly mentioned were similar predictors of poorer survival. The survival analysis of the CGGA cohort also suggested that 3D models can better simulate the increased difficulty of drug delivery to GBM tumours, where patients with increased ABCA1 expression experience lower survival probabilities. Additionally, upregulation of the ABCA1 gene responsible for the increased efflux of chemotherapeutic drugs in 3D cultures was also replicated in every clinical dataset analysed.
However, 3D cultures are not a perfect model by any means. As stated in our previous scoping review, 3D cultures might produce contradictory results for studies targeting the cell cycle [14]. Genes that regulate the cell cycle, such as CDC20, MYC, and CCND1, were indeed upregulated in GBM patients compared to normal human brain samples, but they were found to be downregulated in 3D cultures compared to 2D cultures. This is doubly important for MYC and CCND1 considering their critical contribution to biological processes such as mesenchyme development, response to hypoxia, response to xenobiotic stimulus, and the Wnt signalling pathway, not to mention being involved in the PI3K-Akt signalling pathway. We postulate that an optimum size of around 200~350 μm for the spheroids in 3D cultures simulates the hypoxic core where cells are mostly quiescent, which might affect the cell cycle [62,63]. For said studies that involve the cell cycle, it might behove researchers to vigorously maintain the spheroid size below 200 μm (the diffusion limit for oxygen) in order to better emulate the intermediate layers [64] or procure patient-derived cells to better represent the genetic landscape. Our bioinformatic analysis of both databases clearly indicated increased expression of ITGA3 in GBM samples and in GBM patients with wild-type IDH1R132, whereas our previous analysis found decreased expression in 3D cultures [14]. The contrast in results only serves to drive home the importance of the extracellular matrix dynamics in driving the genetic landscape of GBMs. To date, multiple support systems have been used for 3D GBM cultures, and it is imperative that the properties of the chosen scaffolds be properly considered for the most accurate representation of in vivo GBMs.

5. Conclusions

In this paper, we found evidence of genes commonly upregulated in 3D cultures being replicated in clinical samples. Genes that regulate the EMT process not only have significant prognostic value, but they also seem to be the driving force behind GBM archetypes with poorer treatment responses. In short, we conclude that 3D cultures are superior to 2D cultures, with transcriptomic landscapes that better mimic clinical GBMs, even those with increased treatment resistance (i.e., GBMs with wild-type IDHR132), by mimicking the highly aggressive EMT process that plagues most GBM patients.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/biology12050648/s1, Table S1: Significant differentially expressed genes between GBM samples and healthy brain samples for the GSE145645 cohort; Table S2: Significant differentially expressed genes between GBM samples and healthy brain samples for the GSE165595 cohort; Table S3: Significant differentially expressed genes between GBM samples and healthy brain samples for the GSE147352 cohort; Table S4: Significant differentially expressed genes between GBM samples and healthy brain samples for TCGA/GTEx cohort; Table S5: Significant differentially expressed genes between GBM samples and healthy brain samples for the CGGA cohort. Table S6: Significant differentially expressed genes subset from Table 1 between GBM samples and healthy brain samples for every cohort.

Author Contributions

A.K.R., S.B. and M.N.A.K. conceptualised and administered the project equally. B.W.S.P. performed all the methodology, investigation, and drafted the manuscript with critical contributions from Q.A., M.N.A.K., A.K.R. and S.B. All authors have read and agreed to the published version of the manuscript.

Funding

This project was funded by the Fundamental Research Grant Scheme (FRGS) from the Ministry of Higher Education Malaysia (FRGS/1/2019/STG03/MUSM/01/1) and collaborative research grants from the Jeffrey Cheah School of Medicine and Health Sciences, Monash University Malaysia (Project code: SED-000057 and SED-000032).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

All alignment scripts and R scripts used in this paper can be accessed through the following link: https://github.com/thegellerbing/2D_3D_Analysis- (accessed on 22 January 2023).

Acknowledgments

The authors would like to thank the Jeffrey Cheah School of Medicine and Health Sciences of Monash University Malaysia. The authors would like to acknowledge and express our thanks to Monash University Malaysia’s High-Performance Computing for the provision of computational resources. This review also benefits greatly from the use of TCGA, GEO, and GTEx and gratitude has to be directed to all the personnel involved in the availability of the databases. The authors would also like to extend sincere gratitude to the personnel involved in the CGGA project, especially Tao Jiang for tending to our inquiries.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Rønning, P.A.; Helseth, E.; Meling, T.R.; Johannesen, T.B. A population-based study on the effect of temozolomide in the treatment of glioblastoma multiforme. Neuro Oncol. 2012, 14, 1178–1184. [Google Scholar] [CrossRef]
  2. Ostrom, Q.T.; Cioffi, G.; Waite, K.; Kruchko, C.; Barnholtz-Sloan, J.S. CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2014–2018. Neuro Oncol. 2021, 23 (Suppl. S3), iii1–iii105. [Google Scholar] [CrossRef] [PubMed]
  3. Breslin, S.; O’driscoll, L. Three-dimensional cell culture: The missing link in drug discovery. Drug Discov. Today 2013, 18, 240–249. [Google Scholar] [CrossRef] [PubMed]
  4. Birgersdotter, A.; Sandberg, R.; Ernberg, I. Gene expression perturbation in vitro—A growing case for three-dimensional (3D) culture systems. Semin. Cancer Biol. 2005, 15, 405–412. [Google Scholar] [CrossRef] [PubMed]
  5. Kapałczyńska, M.; Kolenda, T.; Przybyła, W.; Zajączkowska, M.; Teresiak, A.; Filas, V.; Ibbs, M.; Bliźniak, R.; Łuczewski, L.; Lamperska, K. 2D and 3D cell cultures—A comparison of different types of cancer cell cultures. Arch. Med. Sci. 2018, 14, 910–919. [Google Scholar] [CrossRef] [PubMed]
  6. Cooper, L.A.; Gutman, D.A.; Chisolm, C.; Appin, C.; Kong, J.; Rong, Y.; Kurc, T.; Van Meir, E.G.; Saltz, J.H.; Moreno, C.S.; et al. The Tumor Microenvironment Strongly Impacts Master Transcriptional Regulators and Gene Expression Class of Glioblastoma. Am. J. Pathol. 2012, 180, 2108–2119. [Google Scholar] [CrossRef]
  7. Galon, J.; Pagès, F.; Marincola, F.M.; Thurin, M.; Trinchieri, G.; Fox, B.A.; Gajewski, T.F.; Ascierto, P.A. The immune score as a new possible approach for the classification of cancer. J. Transl. Med. 2012, 10, 1. [Google Scholar] [CrossRef]
  8. Hanahan, D. Hallmarks of Cancer: New Dimensions. Cancer Discov. 2022, 12, 31–46. [Google Scholar] [CrossRef]
  9. Roschke, A.V.; Tonon, G.; Gehlhaus, K.S.; Mctyre, N.; Bussey, K.J.; Lababidi, S.; Scudiero, D.A.; Weinstein, J.N.; Kirsch, I.R. Karyotypic complexity of the NCI-60 drug-screening panel. Cancer Res. 2003, 63, 8634–8647. [Google Scholar]
  10. Ross, D.T.; Scherf, U.; Eisen, M.B.; Perou, C.M.; Rees, C.; Spellman, P.; Iyer, V.; Jeffrey, S.S.; Van de Rijn, M.; Waltham, M.; et al. Systematic variation in gene expression patterns in human cancer cell lines. Nat. Genet. 2000, 24, 227–235. [Google Scholar] [CrossRef]
  11. Masters, J.R.W. Human cancer cell lines: Fact and fantasy. Nat. Rev. Mol. Cell Biol. 2000, 1, 233–236. [Google Scholar] [CrossRef] [PubMed]
  12. Jacks, T.; Weinberg, R.A. Taking the Study of Cancer Cell Survival to a New Dimension. Cell 2002, 111, 923–925. [Google Scholar] [CrossRef] [PubMed]
  13. Gómez-Oliva, R.; Domínguez-García, S.; Carrascal, L.; Abalos-Martínez, J.; Pardillo-Díaz, R.; Verástegui, C.; Castro, C.; Nunez-Abades, P.; Geribaldi-Doldán, N. Evolution of Experimental Models in the Study of Glioblastoma: Toward Finding Efficient Treatments. Front. Oncol. 2021, 10, 614295. [Google Scholar] [CrossRef]
  14. Phon, B.W.S.; Kamarudin, M.N.; Bhuvanendran, S.; Radhakrishnan, A.K. Transitioning pre-clinical glioblastoma models to clinical settings with biomarkers identified in 3D cell-based models: A systematic scoping review. Biomed. Pharmacother. 2021, 145, 112396. [Google Scholar] [CrossRef]
  15. Huang, T.; Yang, Y.; Song, X.; Wan, X.; Wu, B.; Sastry, N.; Horbinski, C.M.; Zeng, C.; Tiek, D.; Goenka, A.; et al. PRMT6 methylation of RCC1 regulates mitosis, tumorigenicity, and radiation response of glioblastoma stem cells. Mol. Cell 2021, 81, 1276–1291.e9. [Google Scholar] [CrossRef] [PubMed]
  16. Xu, L.; Chen, Y.; Huang, Y.; Sandanaraj, E.; Yu, J.S.; Lin, R.Y.-T.; Dakle, P.; Ke, X.-Y.; Chong, Y.K.; Koh, L.; et al. Topography of transcriptionally active chromatin in glioblastoma. Sci. Adv. 2021, 7, eabd4676. [Google Scholar] [CrossRef] [PubMed]
  17. Hwang, T.; Kim, S.; Chowdhury, T.; Yu, H.J.; Kim, K.-M.; Kang, H.; Won, J.-K.; Park, S.-H.; Shin, J.H.; Park, C.-K. Genome-wide perturbations of Alu expression and Alu-associated post-transcriptional regulations distinguish oligodendroglioma from other gliomas. Commun. Biol. 2022, 5, 62. [Google Scholar] [CrossRef] [PubMed]
  18. Dobin, A.; Davis, C.A.; Schlesinger, F.; Drenkow, J.; Zaleski, C.; Jha, S.; Batut, P.; Chaisson, M.; Gingeras, T.R. STAR: Ultrafast universal RNA-seq aligner. Bioinformatics 2013, 29, 15–21. [Google Scholar] [CrossRef] [PubMed]
  19. Li, B.; Dewey, C.N. RSEM: Accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinform. 2011, 12, 323. [Google Scholar] [CrossRef]
  20. Goldman, M.J.; Craft, B.; Hastie, M.; Repečka, K.; McDade, F.; Kamath, A.; Banerjee, A.; Luo, Y.; Rogers, D.; Brooks, A.N.; et al. Visualizing and interpreting cancer genomics data via the Xena platform. Nat. Biotechnol. 2020, 38, 675–678. [Google Scholar] [CrossRef]
  21. Cerami, E.; Gao, J.; Dogrusoz, U.; Gross, B.E.; Sumer, S.O.; Aksoy, B.A.; Jacobsen, A.; Byrne, C.J.; Heuer, M.L.; Larsson, E.; et al. The cBio cancer genomics portal: An open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2012, 2, 401–404. [Google Scholar] [CrossRef] [PubMed]
  22. Gao, J.; Aksoy, B.A.; Dogrusoz, U.; Dresdner, G.; Gross, B.E.; Sumer, S.O.; Sun, Y.; Jacobsen, A.; Sinha, R.; Larsson, E.; et al. Integrative Analysis of Complex Cancer Genomics and Clinical Profiles Using the cBioPortal. Sci. Signal. 2013, 6, pl1. [Google Scholar] [CrossRef] [PubMed]
  23. Vivian, J.; Rao, A.A.; Nothaft, F.A.; Ketchum, C.; Armstrong, J.; Novak, A.; Pfeil, J.; Narkizian, J.; Deran, A.D.; Musselman-Brown, A.; et al. Toil enables reproducible, open source, big biomedical data analyses. Nat. Biotechnol. 2017, 35, 314–316. [Google Scholar] [CrossRef] [PubMed]
  24. Zhao, Z.; Zhang, K.-N.; Wang, Q.; Li, G.; Zeng, F.; Zhang, Y.; Wu, F.; Chai, R.; Wang, Z.; Zhang, C.; et al. Chinese Glioma Genome Atlas (CGGA): A Comprehensive Resource with Functional Genomic Data from Chinese Glioma Patients. Genom. Proteom. Bioinform. 2021, 19, 1–12. [Google Scholar] [CrossRef]
  25. Zhang, K.; Liu, X.; Li, G.; Chang, X.; Li, S.; Chen, J.; Zhao, Z.; Wang, J.; Jiang, T.; Chai, R. Clinical management and survival outcomes of patients with different molecular subtypes of diffuse gliomas in China (2011–2017): A multicenter retrospective study from CGGA. Cancer Biol. Med. 2022, 19, 1460–1476. [Google Scholar] [CrossRef]
  26. Wang, Y.; Qian, T.; You, G.; Peng, X.; Chen, C.; You, Y.; Yao, K.; Wu, C.; Ma, J.; Sha, Z.; et al. Localizing seizure-susceptible brain regions associated with low-grade gliomas using voxel-based lesion-symptom mapping. Neuro Oncol. 2014, 17, 282–288. [Google Scholar] [CrossRef] [PubMed]
  27. Liu, X.; Li, Y.; Qian, Z.; Sun, Z.; Xu, K.; Wang, K.; Liu, S.; Fan, X.; Li, S.; Zhang, Z.; et al. A radiomic signature as a non-invasive predictor of progression-free survival in patients with lower-grade gliomas. NeuroImage Clin. 2018, 20, 1070–1077. [Google Scholar] [CrossRef]
  28. Love, M.I.; Huber, W.; Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014, 15, 550. [Google Scholar] [CrossRef] [PubMed]
  29. Blighe, K.; Rana, S.; Lewis, M. Enhanced Volano: Publication-Ready Volcano Plots with Enhanced Colouring and Labeling. 2021. Available online: https://bioconductor.org/packages/devel/bioc/vignettes/EnhancedVolcano/inst/doc/EnhancedVolcano.html (accessed on 2 December 2022).
  30. Szklarczyk, D.; Gable, A.L.; Lyon, D.; Junge, A.; Wyder, S.; Huerta-Cepas, J.; Simonovic, M.; Doncheva, N.T.; Morris, J.H.; Bork, P.; et al. STRING v11: Protein–protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res. 2019, 47, D607–D613. [Google Scholar] [CrossRef]
  31. Wu, T.; Hu, E.; Xu, S.; Chen, M.; Guo, P.; Dai, Z.; Feng, T.; Zhou, L.; Tang, W.; Zhan, L.; et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation 2021, 2, 100141. [Google Scholar] [CrossRef]
  32. Therneau, T.M.; Grambsch, P.M. Modeling Survival Data: Extending the Cox Model; Springer: New York, NY, USA, 2000. [Google Scholar]
  33. Kassambara, A.; Kosinski, M.; Biecek, P. Survminer: Drawing Survival Curves Using ‘ggplot2’. 2021. Available online: https://rpkgs.datanovia.com/survminer/ (accessed on 6 December 2022).
  34. Labussiere, M.; Sanson, M.; Idbaih, A.; Delattre, J.Y. IDH1 Gene Mutations: A New Paradigm in Glioma Prognosis and Therapy? Oncol. 2010, 15, 196–199. [Google Scholar] [CrossRef]
  35. Zhang, K.; Wang, X.-Q.; Zhou, B.; Zhang, L. The prognostic value of MGMT promoter methylation in Glioblastoma multiforme: A meta-analysis. Fam. Cancer 2013, 12, 449–458. [Google Scholar] [CrossRef]
  36. Jia, W.; Jiang, X.; Liu, W.; Wang, L.; Zhu, B.; Zhu, H.; Liu, X.; Zhong, M.; Xie, D.; Huang, W.; et al. Effects of three-dimensional collagen scaffolds on the expression profiles and biological functions of glioma cells. Int. J. Oncol. 2018, 52, 1787–1800. [Google Scholar] [CrossRef] [PubMed]
  37. Gomez-Roman, N.; Stevenson, K.; Gilmour, L.; Hamilton, G.; Chalmers, A.J. A novel 3D human glioblastoma cell culture system for modeling drug and radiation responses. Neuro Oncol. 2016, 19, 229–241. [Google Scholar] [CrossRef]
  38. Wang, K.; Kievit, F.M.; Erickson, A.E.; Silber, J.R.; Ellenbogen, R.G.; Zhang, M. Culture on 3D Chitosan-Hyaluronic Acid Scaffolds Enhances Stem Cell Marker Expression and Drug Resistance in Human Glioblastoma Cancer Stem Cells. Adv. Health Mater. 2016, 5, 3173–3181. [Google Scholar] [CrossRef]
  39. Florczyk, S.J.; Wang, K.; Jana, S.; Wood, D.L.; Sytsma, S.K.; Sham, J.G.; Kievit, F.M.; Zhang, M. Porous chitosan-hyaluronic acid scaffolds as a mimic of glioblastoma microenvironment ECM. Biomaterials 2013, 34, 10143–10150. [Google Scholar] [CrossRef] [PubMed]
  40. Pedron, S.; Becka, E.; Harley, B.A. Regulation of glioma cell phenotype in 3D matrices by hyaluronic acid. Biomaterials 2013, 34, 7408–7417. [Google Scholar] [CrossRef] [PubMed]
  41. Tang, D.G. Understanding cancer stem cell heterogeneity and plasticity. Cell Res. 2012, 22, 457–472. [Google Scholar] [CrossRef]
  42. Zhang, M.; Song, T.; Yang, L.; Chen, R.; Wu, L.; Yang, Z.; Fang, J. Nestin and CD133: Valuable stem cell-specific markers for determining clinical outcome of glioma patients. J. Exp. Clin. Cancer Res. 2008, 27, 85. [Google Scholar] [CrossRef]
  43. Kalluri, R.; Weinberg, R.A. The basics of epithelial-mesenchymal transition. J. Clin. Investig. 2009, 119, 1420–1428. [Google Scholar] [CrossRef]
  44. Jain, R.K.; Di Tomaso, E.; Duda, D.G.; Loeffler, J.S.; Sorensen, A.G.; Batchelor, T.T. Angiogenesis in brain tumours. Nat. Rev. Neurosci. 2007, 8, 610–622. [Google Scholar] [CrossRef] [PubMed]
  45. Tompa, M.; Kalovits, F.; Nagy, A.; Kalman, B. Contribution of the Wnt Pathway to Defining Biology of Glioblastoma. NeuroMolecular Med. 2018, 20, 437–451. [Google Scholar] [CrossRef] [PubMed]
  46. Qiu, X.; Jiao, J.; Li, Y.; Tian, T. Overexpression of FZD7 promotes glioma cell proliferation by upregulating TAZ. Oncotarget 2016, 7, 85987–85999. [Google Scholar] [CrossRef] [PubMed]
  47. Zhou, W.; Yu, X.; Sun, S.; Zhang, X.; Yang, W.; Zhang, J.; Zhang, X.; Jiang, Z. Increased expression of MMP-2 and MMP-9 indicates poor prognosis in glioma recurrence. Biomed. Pharmacother. 2019, 118, 109369. [Google Scholar] [CrossRef] [PubMed]
  48. Mikheeva, S.A.; Mikheev, A.M.; Petit, A.; Beyer, R.; Oxford, R.G.; Khorasani, L.; Maxwell, J.-P.; Glackin, C.A.; Wakimoto, H.; González-Herrero, I.; et al. TWIST1 promotes invasion through mesenchymal change in human glioblastoma. Mol. Cancer 2010, 9, 194. [Google Scholar] [CrossRef] [PubMed]
  49. Myung, J.K.; Choi, S.A.; Kim, S.-K.; Wang, K.-C.; Park, S.-H. Snail plays an oncogenic role in glioblastoma by promoting epithelial mesenchymal transition. Int. J. Clin. Exp. Pathol. 2014, 7, 1977–1987. [Google Scholar] [CrossRef]
  50. Iwadate, Y. Epithelial-mesenchymal transition in glioblastoma progression. Oncol. Lett. 2016, 11, 1615–1620. [Google Scholar] [CrossRef]
  51. Mooney, K.L.; Choy, W.; Sidhu, S.; Pelargos, P.; Bui, T.T.; Voth, B.; Barnette, N.; Yang, I. The role of CD44 in glioblastoma multiforme. J. Clin. Neurosci. 2016, 34, 1–5. [Google Scholar] [CrossRef]
  52. Murphy, G.; Nagase, H. Progress in matrix metalloproteinase research. Mol. Asp. Med. 2008, 29, 290–308. [Google Scholar] [CrossRef]
  53. Pankov, R.; Yamada, K.M. Fibronectin at a glance. J. Cell Sci. 2002, 115, 3861–3863. [Google Scholar] [CrossRef]
  54. Monteiro, A.R.; Hill, R.; Pilkington, G.J.; Madureira, P.A. The Role of Hypoxia in Glioblastoma Invasion. Cells 2017, 6, 45. [Google Scholar] [CrossRef] [PubMed]
  55. SongTao, Q.; Lei, Y.; Si, G.; YanQing, D.; HuiXia, H.; XueLin, Z.; LanXiao, W.; Fei, Y. IDH mutations predict longer survival and response to temozolomide in secondary glioblastoma. Cancer Sci. 2011, 103, 269–273. [Google Scholar] [CrossRef] [PubMed]
  56. Cui, D.; Ren, J.; Shi, J.; Feng, L.; Wang, K.; Zeng, T.; Jin, T.; Gao, L. R132H mutation in IDH1 gene reduces proliferation, cell survival and invasion of human glioma by downregulating Wnt/β-catenin signaling. Int. J. Biochem. Cell Biol. 2016, 73, 72–81. [Google Scholar] [CrossRef] [PubMed]
  57. Yao, Q.; Cai, G.; Yu, Q.; Shen, J.; Gu, Z.; Chen, J.; Shi, W.; Shi, J. IDH1 mutation diminishes aggressive phenotype in glioma stem cells. Int. J. Oncol. 2017, 52, 270–278. [Google Scholar] [CrossRef] [PubMed]
  58. Liang, H.; Chen, G.; Li, J.; Yang, F. Snail expression contributes to temozolomide resistance in glioblastoma. Am. J. Transl. Res. 2019, 11, 4277–4289. [Google Scholar]
  59. Siebzehnrubl, F.A.; Silver, D.J.; Tugertimur, B.; Deleyrolle, L.P.; Siebzehnrubl, D.; Sarkisian, M.R.; Devers, K.G.; Yachnis, A.T.; Kupper, M.D.; Neal, D.; et al. The ZEB1 pathway links glioblastoma initiation, invasion and chemoresistance. EMBO Mol. Med. 2013, 5, 1196–1212. [Google Scholar] [CrossRef]
  60. Wu, S.; Liu, C.; Wei, X.; Nong, W.-X.; Lin, L.-N.; Li, F.; Xie, X.-X.; Liao, X.-S.; Luo, B.; Zhang, Q.-M.; et al. High Expression of Fibronectin 1 Predicts a Poor Prognosis in Glioblastoma. Curr. Med Sci. 2022, 42, 1055–1065. [Google Scholar] [CrossRef]
  61. Mikheev, A.M.; Mikheeva, S.A.; Severs, L.J.; Funk, C.C.; Huang, L.; McFaline-Figueroa, J.L.; Schwensen, J.; Trapnell, C.; Price, N.; Wong, S.; et al. Targeting TWIST 1 through loss of function inhibits tumorigenicity of human glioblastoma. Mol. Oncol. 2018, 12, 1188–1202. [Google Scholar] [CrossRef]
  62. Singh, S.K.; Abbas, S.; Saxena, A.K.; Tiwari, S.; Sharma, L.K.; Tiwari, M. Critical role of three-dimensional tumorsphere size on experimental outcome. Biotechniques 2020, 69, 333–338. [Google Scholar] [CrossRef]
  63. Chaicharoenaudomrung, N.; Kunhorm, P.; Promjantuek, W.; Rujanapun, N.; Heebkaew, N.; Soraksa, N.; Noisa, P. Transcriptomic Profiling of 3D Glioblastoma Tumoroids for the Identification of Mechanisms Involved in Anticancer Drug Resistance. In Vivo 2020, 34, 199–211. [Google Scholar] [CrossRef]
  64. Mehta, G.; Hsiao, A.Y.; Ingram, M.; Luker, G.D.; Takayama, S. Opportunities and challenges for use of tumor spheroids as models to test drug delivery and efficacy. J. Control. Release 2012, 164, 192–204. [Google Scholar] [CrossRef] [PubMed]
Biology 12 00648 g001 550
Figure 1. (a) Volcano plot of differentially expressed genes in the TCGA/GTEx cohort. (LFC = log2 fold change). (b) Heatmap of commonly regulated genes in 3D subset from Table 1 between GBM samples and normal brain tissues in the TCGA cohort.
Figure 1. (a) Volcano plot of differentially expressed genes in the TCGA/GTEx cohort. (LFC = log2 fold change). (b) Heatmap of commonly regulated genes in 3D subset from Table 1 between GBM samples and normal brain tissues in the TCGA cohort.
Biology 12 00648 g001
Biology 12 00648 g002 550
Figure 2. (a) Volcano plot of differentially expressed genes in the CGGA cohort (LFC = log2 fold change). (b) Heatmap of commonly regulated genes in 3D subset from Table 1 between GBM samples and normal brain tissues in the CGGA cohort.
Figure 2. (a) Volcano plot of differentially expressed genes in the CGGA cohort (LFC = log2 fold change). (b) Heatmap of commonly regulated genes in 3D subset from Table 1 between GBM samples and normal brain tissues in the CGGA cohort.
Biology 12 00648 g002
Biology 12 00648 g003 550
Figure 3. Interaction network of the subset of 31 genes from Table 1 that were differentially expressed between GBM samples and healthy brain samples in at least three or more cohorts. These represent genes with high-confidence interactions with each other.
Figure 3. Interaction network of the subset of 31 genes from Table 1 that were differentially expressed between GBM samples and healthy brain samples in at least three or more cohorts. These represent genes with high-confidence interactions with each other.
Biology 12 00648 g003
Biology 12 00648 g004 550
Figure 4. Overrepresentation analysis of enriched (a) GO biological processes and (b) KEGG pathways of the subset of 31 genes from Table 1 that were differentially expressed in at least three different cohorts.
Figure 4. Overrepresentation analysis of enriched (a) GO biological processes and (b) KEGG pathways of the subset of 31 genes from Table 1 that were differentially expressed in at least three different cohorts.
Biology 12 00648 g004
Biology 12 00648 g005 550
Figure 5. Survival curves of 163 GBM patients and count plots between normal brain samples and GBM samples for genes (a,b) FN1 and (c,d) TWIST1. GBM patients were dichotomised into two groups for the log-rank tests based on the median TPM value of each respective gene.
Figure 5. Survival curves of 163 GBM patients and count plots between normal brain samples and GBM samples for genes (a,b) FN1 and (c,d) TWIST1. GBM patients were dichotomised into two groups for the log-rank tests based on the median TPM value of each respective gene.
Biology 12 00648 g005
Table
Table 1. The common functions of regularly upregulated and downregulated genes in 3D GBM models compared to 2D GBM cultures found in our previous scoping review [11].
Table 1. The common functions of regularly upregulated and downregulated genes in 3D GBM models compared to 2D GBM cultures found in our previous scoping review [11].
GenesFunction
Upregulated Genes
PROM1Stemness-Related
NES
SOX2
TAZ
POU5F1
NANOG
FOS
MSI1
CD44EMT-Related
TWIST1
SNAI1
FN1
VIM
CDH2
YAP1
MMP1Angiogenesis/Migration
MMP2
MMP9
VEGFA
EPHA3
ABCG2Drug Efflux
ABCB1
ABCA1
ABCA2
ABCC7
SLC17A3
GFAPECM-Related
ITGA6
EPCAM
HIF1AHypoxia
PLAT
DKK1Wnt Signalling
FZD7
RELBRegulation of Gene Expression
MAML1
IKBKBNFκB Signalling
CDKN1BCell Cycle
EDNRBCell Division
CYP1A1Drug Response
NOTCH2Notch Signalling
Downregulated Genes
CDH1EMT-Related
ITGA3ECM-Related
CCND1Cell Cycle-Related
CDC20
MYC
Table
Table 2. Number of GBM patients and healthy brain samples in each dataset used.
Table 2. Number of GBM patients and healthy brain samples in each dataset used.
DatasetGBM PatientsHealthy Brain Samples
GSE145645323
GSE1473528515
GSE1655951717
CGGA38820
TCGA/GTEX166212
Table
Table 3. Differentially expressed genes in each GEO dataset based on the genes listed in Table 1.
Table 3. Differentially expressed genes in each GEO dataset based on the genes listed in Table 1.
GSE145645GSE147352GSE165595
Upregulated Genes
ABCA1ABCA1ABCA1
CCND1CCND1CCND1
CD44CD44CD44
CDC20CDC20CDC20
CDH2CDH1FN1
EDNRBCDH2FOS
EPHA3DKK1FZD7
FN1EDNRBHIF1A
FOSFN1MMP2
FZD7FOSMMP9
HIF1AFZD7MSI1
ITGA3GFAPMYC
MAML1HIF1ANES
MMP1IKBKBPLAT
MMP2ITGA3PROM1
MMP9ITGA6RELB
MSI1MMP1SOX2
MYCMMP2VEGFA
NESMMP9VIM
NOTCH2MSI1
PLATMYC
PROM1NES
RELBNOTCH2
SNAI1PLAT
SOX2PROM1
TWIST1RELB
VEGFASLC17A3
VIMSNAI1
YAP1SOX2
TWIST1
VEGFA
VIM
YAP1
Downregulated Genes
ABCA2CYP1A1ABCA2
EPCAMEPCAMEPCAM
CYP1A1 CYP1A1
Table
Table 4. Clinical characteristics of the TCGA and CGGA GBM patient cohorts.
Table 4. Clinical characteristics of the TCGA and CGGA GBM patient cohorts.
Total Number (n = 166)Total Number (n = 388)
Sex
Male
Female
Not Reported
106
59
1
235
153
-
Age
>65
<65
Not Reported
101
54
11
38
350
-
IDH Status
Wild-Type
Mutant
Not Reported
136
8
22
288
90
10
MGMT Methylation Status
Methylated
Unmethylated
Not Reported
54
65
47
172
163
53
Sample Type
Primary
Recurrent
153
13
255
133
Event
Living
Deceased
Not Reported
48
106
12
53
322
13
Table
Table 5. Subset of differentially expressed genes from the TCGA/GTEx and CGGA cohorts based on Table 1.
Table 5. Subset of differentially expressed genes from the TCGA/GTEx and CGGA cohorts based on Table 1.
TCGA/GTExCGGA
Upregulated Genes
ABCA1ABCA1
CCND1CD44
CD44CDC20
CDC20DKK1
CDH2EPHA3
DKK1FN1
EDNRBFZD7
EPHA3HIF1A
FN1ITGA3
FZD7MMP1
GFAPMMP2
HIF1AMMP9
ITGA3MSI1
MMP1MYC
MMP2NANOG
MMP9NES
MSI1NOTCH2
MYCPLAT
NESPROM1
PLATRELB
PROM1SLC17A3
RELBSNAI1
SLC17A3SOX2
SOX2TWIST1
TWIST1VEGFA
VEGFAVIM
VIM
Downregulated Genes
ABCA2ABCA2
CYP1A1EPCAM
EPCAM
Table
Table 6. The subset of genes based on Table 1 that are differentially expressed in 3D cultures vs. 2D cultures and across all datasets used in the differential expression analysis.
Table 6. The subset of genes based on Table 1 that are differentially expressed in 3D cultures vs. 2D cultures and across all datasets used in the differential expression analysis.
Genes3D vs. 2DGSE145645GSE165595GSE147352CGGATCGA/GTExGene Function
FOS↑↑↑↑ Stemness-Related
MSI1↑↑↑↑↑↑
NANOG↑ ↑
NES↑↑↑↑↑↑
PROM1↑↑↑↑↑↑
SOX2↑↑↑↑↑↑
CD44↑↑↑↑↑↑EMT-Related
CDH2↑↑ ↑ ↑
FN1↑↑↑↑↑↑
SNAI1↑↑ ↑↑
TWIST1↑↑ ↑↑↑
VIM↑↑↑↑↑↑
YAP1↑↑ ↑
EPHA3↑↑ ↑↑Angiogenesis/Migration
MMP1↑↑ ↑↑↑
MMP2↑↑↑↑↑↑
MMP9↑↑↑↑↑↑
VEGFA↑↑↑↑↑↑
ABCA1↑↑↑↑↑↑Drug Efflux
ABCA2↑↓↓ ↓↓
SLC17A3↑ ↑↑↑
GFAP↑ ↑ ↑ECM-Related
ITGA6↑ ↑
EPCAM↑↓↓↓↓↓
ITGA3↓↑ ↑↑↑
HIF1A↑↑↑↑↑↑Hypoxia
PLAT↑↑↑↑↑↑
DKK1↑ ↑↑↑Wnt Signaling
FZD7↑↑↑↑↑↑
MAML1↑↑ Regulation of Gene Expression
RELB↑↑↑↑↑↑
IKBKB↑ ↑ NF-κB Signalling
CCND1↓↑↑↑ ↑Cell Cycle-Related
CDC20↓↑↑↑↑↑
MYC↓↑↑↑↑↑
EDNRB↑↑ ↑ ↑Cell Division
NOTCH2↑↑ ↑↑ Notch Signaling
CYP1A1↑↓↓↓ ↓Drug Response
↑ indicates upregulation of the gene, ↓ indicates downregulation of the gene.
Table
Table 7. Differentially expressed genes between wild-type IDH1R132 GBM vs. IDH1R132 mutant astrocytoma in the GSE147352 cohort.
Table 7. Differentially expressed genes between wild-type IDH1R132 GBM vs. IDH1R132 mutant astrocytoma in the GSE147352 cohort.
IDH1R132 Wild-Type vs. IDH1R132 Mutant
NESUpregulated
PROM1Upregulated
TWIST1Upregulated
VEGFAUpregulated
MMP1Upregulated
MMP9Upregulated
PLATUpregulated
DKK1Upregulated
FZD7Upregulated
EPHA3Upregulated
ITGA3Upregulated
Table
Table 8. Differentially expressed genes between GBM samples with different characteristics in the TCGA cohort.
Table 8. Differentially expressed genes between GBM samples with different characteristics in the TCGA cohort.
IDH1R132 Wild-Type vs. IDH1R132 Mutant
CD44Upregulated
MMP9Upregulated
VEGFAUpregulated
ITGA3Upregulated
PLATUpregulated
FZD7Upregulated
SNAI1Upregulated
SLC17A3Upregulated
DKK1Upregulated
EDNRBDownregulated
Unmethylated MGMT GBMs vs. Methylated MGMT GBMs
TWIST1Upregulated
Table
Table 9. Differentially expressed genes between GBM samples with different characteristics in the CGGA cohort.
Table 9. Differentially expressed genes between GBM samples with different characteristics in the CGGA cohort.
IDH1R132 Wild-Type vs. IDH1R132 Mutant
DKK1Upregulated
FN1Upregulated
FZD7Upregulated
ITGA3Upregulated
MMP1Upregulated
MMP9Upregulated
SLC17A3Upregulated
SNAI1Upregulated
VEGFAUpregulated
CCND1Downregulated
Unmethylated MGMT GBMs vs. Methylated MGMT GBMs
MMP1Upregulated
Table
Table 10. GBM development-related GO biological process categories of the 31 genes that were differentially expressed in at least three different cohorts.
Table 10. GBM development-related GO biological process categories of the 31 genes that were differentially expressed in at least three different cohorts.
Biological Processesp-ValueFDRGenes **
Response to Hypoxia (GO:0001666)8.51 × 10−64.31 × 10−6FOS, TWIST1, MMP2, VEGFA, HIF1A, PLAT, MYC, CYP1A1
Mesenchyme Development (GO:0060485)9.71 × 10−59.71 × 10−5CDH2, FN1, SNAI1, TWIST1, EPHA3, HIF1A, MYC
Negative Regulation of DNA Damage Response (GP:0043518)2.25 × 10−61.54 × 10−4CD44, SNAI1, TWIST1
Mesenchymal Cell Differentiation (GO:0048762)1.54 × 10−42.92 × 10−4CDH2, FN1, SNAI1, TWIST1, EPHA3, HIF1A
Mesenchymal Cell Migration (GO:0090497)2.92 × 10−42.94 × 10−4CDH2, FN1, TWIST1, HIF1A
Response to Xenobiotic Stimulus2.94 × 10−42.35 × 10−4FOS, MMP2, ABCA2, ITGA3, CCND1, MYC, CYP1A1
Positive Regulation of MAPK Cascade (GO:0043410)4.44 × 10−44 × 10−4SOX2, CD44, CDH2, VEGFA, DKK1, FZD7, NOTCH2
Stem Cell Development (GO:0048864)7.52 × 10−47.52 × 10−4CDH2, FN1, TWIST1, HIF1A
Wnt Signalling Pathway (GO:0016055)2.75 × 10−31.46 × 10−3SOX2, CDH2, ITGA3, DKK1, FZD7, CCND1
Positive Regulation of Epithelial Cell Migration (GO:0010634)2.89 × 10−31.5 × 10−3MMP9, VEGFA, ITGA3, HIF1A
Extracellular Matrix Disassembly (GO:0022617)3.22 × 10−31.7 × 10−3MMP1, MMP2, MMP9
Epithelial to Mesenchymal Transition (GO:0001837)3.34 × 10−31.8 × 10−3SNAI1, TWIST1, EPHA3, HIF1A
KEGG Pathwayp-ValueFDRGenes **
MicroRNAs in Cancer6.58 × 10−54.4 × 10−6MMP9, CD44, VIM, MYC, VEGFA, CCND1, NOTCH2
Wnt Signalling Pathway3.04 × 10−30.023FZD7, MYC, DKK1, CCND1
ECM-Receptor Interaction3.98 × 10−30.027CD44, FN1, ITGA3
Transcriptional Misregulation in Cancer4.94 × 10−30.029MMP9, MYC, PLAT, PROM1
PI3K-Akt Signalling Pathway8.63 × 10−30.028FN1, MYC, VEGFA, ITGA3, CCND1
** Genes were analysed via overrepresentation analysis using the R package ‘clusterprofiler’ and only categories with False Discovery Rate (FDR)-adjusted p-value of < 0.05 related to cancer development are reported.
Table
Table 11. Univariate survival analysis of 374 CGGA GBM patients based on the upregulation of genes in GBM patients compared to normal human brain samples. Patients were dichotomised into two groups based on the median FPKM value.
Table 11. Univariate survival analysis of 374 CGGA GBM patients based on the upregulation of genes in GBM patients compared to normal human brain samples. Patients were dichotomised into two groups based on the median FPKM value.
Gene Hazard Ratio95% CIp-Value
ABCA11.4027851.13~1.750.002547
CD441.4080991.13~1.760.002341
CDC201.264851.01~1.580.036394
CDH21.3230251.06~1.650.012566
FN11.4039161.13~1.750.002561
FOS1.3827861.11~1.720.003926
ITGA31.3046761.05~1.630.01783
MMP11.2497281~1.560.046397
MMP21.3082851.05~1.630.016909
MMP91.2493621~1.560.04784
MSI11.3065591.05~1.630.017031
MYC1.2890351.04~1.610.023265
NES1.3374331.07~1.670.009639
PLAT1.4078021.13~1.760.002363
RELB1.2501621~1.560.046568
SNAI11.4219351.14~1.770.001759
VEGFA1.3539841.09~1.690.006891
VIM1.4222331.14~1.770.00175
YAP11.2487861~1.550.047023
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Phon, B.W.S.; Bhuvanendran, S.; Ayub, Q.; Radhakrishnan, A.K.; Kamarudin, M.N.A. Identification of Prominent Genes between 3D Glioblastoma Models and Clinical Samples via GEO/TCGA/CGGA Data Analysis. Biology 2023, 12, 648. https://doi.org/10.3390/biology12050648

AMA Style

Phon BWS, Bhuvanendran S, Ayub Q, Radhakrishnan AK, Kamarudin MNA. Identification of Prominent Genes between 3D Glioblastoma Models and Clinical Samples via GEO/TCGA/CGGA Data Analysis. Biology. 2023; 12(5):648. https://doi.org/10.3390/biology12050648

Chicago/Turabian Style

Phon, Brandon Wee Siang, Saatheeyavaane Bhuvanendran, Qasim Ayub, Ammu Kutty Radhakrishnan, and Muhamad Noor Alfarizal Kamarudin. 2023. "Identification of Prominent Genes between 3D Glioblastoma Models and Clinical Samples via GEO/TCGA/CGGA Data Analysis" Biology 12, no. 5: 648. https://doi.org/10.3390/biology12050648

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop