Background
Glioblastoma multiforme (GBM) is the most common and lethal tumour of the central nervous system in adults [
1]. Despite decades of efforts to tackle this disease, the median survival rate of GBM patients is still not improving [
2]. GBM patients have an average life expectancy of 15 months post-diagnosis and the 5-years survival rate is less than 3% [
3]. The standard-of-care GBM treatment generally consists of maximal safe surgical resection followed by radiotherapy and concomitant chemotherapy. However rapid post-treatment relapse and high intra-tumoral heterogeneity that could either arise naturally during disease progression or treatments-induced have made this disease intractable and more challenging to treat [
4,
5]. Therefore, there is a pressing need for better and efficient diagnostic and therapeutic strategies for this disease.
Temozolomide, an orally administered DNA-alkylating drug, is the current and commonly used chemotherapy agent to treat GBM in the clinic [
6]. This combination treatment of temozolomide and radiotherapy is referred to as the Stupp regimen and it is widely used as the standard-of-care for the treatment of GBM. The landmark study showed that the combination of radiotherapy and concomitant chemotherapy with temozolomide improve the patient’s prognosis compared to radiotherapy alone (median survival of 14.6 months vs 12.1 months, respectively) [
6]. Alternative GBM treatment options such as the VEGF-targeting monoclonal antibody Bevacizumab, other DNA alkylating agents such as lomustine and carmustine implants, alternating electric field therapy and the checkpoint blockade inhibitor have thus far yielded low efficacy in treating GBM [
2,
7,
8]. The Cancer Genome Atlas (TCGA) comprehensive GBM molecular characterizations have identified significant genetic alterations in several important oncogenic signalling pathways such as the RTK/Ras/PI3K (88%), p53 (87%) and pRB signalling pathways (78%) in GBM patients [
9]. Several clinical trials are currently ongoing that aim to target these altered GBM oncogenic signalling pathways components using small molecule inhibitors and/or monoclonal antibodies. However, the results thus far were far from satisfactory [
10]. This seems to suggest that instead of using a single agent targeting a specific component or pathway, novel treatments should consider the administration of several inhibitors targeting multiple different pathways.
The cell surface proteins or surfaceome serve as an information gateway that integrates and transduces extracellular cues into intracellular signalling cascades or vice versa. Surfaceome also play important role in cell adhesion and migration which are among the critical processes during tumorigenesis. Indeed, aberrant surfaceome expression and activity are frequently observed in many cancer types and therefore are good candidates for cancer diagnostic or biomarkers as well as therapeutic targets. Recent evidence has demonstrated that 56% of cell surface proteins are differentially expressed in GBM which are also present in cerebrospinal fluid or plasma, suggesting their potential use as biomarkers [
11]. Of note, surfaceome expression is more dynamic than intracellular proteins and they could be sometimes cell type-specific [
12,
13]. Mass spectrometry analysis showed that the average surfaceome size in brain cancer cell lines is higher than in other cancer types [
12]. Thus, surfaceome genes in GBM may hold the key to understand GBM pathogenesis and drug responsiveness, in which targeting these genes may unravel potential ‘
druggable’ stage in GBM pathways.
A comprehensive overview of the GBM surfaceome landscape has not been fully defined. Therefore, this study aimed to characterize the GBM surfaceome genes expression profile by unifying the two large RNA-Seq datasets from the TCGA (GBM) and GTEx (normal brain). We integrated and performed differential gene expression analysis on these two datasets because of the low number of normal brain tissue samples in the TCGA database. A previously annotated surfaceome gene set was employed to filter and identify the significant differentially expressed surfaceome genes in GBM. To further prioritize the high-confidence GBM cell surface signature, we integrated our transcriptomics analysis with GBM tissues and cell surface proteomics, and PPI hub gene analysis. Collectively, we identified a list of upregulated surfaceome genes in GBM that include CD44, PTPRJ and HLA-DRA in which their biological relevance in supporting GBM pathogenesis could be comprehensively investigated in future studies for the development of novel GBM diagnostic/prognostic or therapeutic strategies.
Discussion
The surfaceome comprise cellular frontiers that permit/inhibit signal transduction as well as playing important roles in modulating cells proliferation, migration and invasion, and cells-cells interaction. The surfaceome can organize itself at a nanoscale resolution [
38]. This spatiotemporal nanoscale organization could define the cell identity and phenotypes, and capacity to communicate with microenvironments such as the extracellular matrix, growth factors, hormones and drugs. Due to their accessibility on the cell membrane, surfaceome proteins are ideal candidates for biomarkers and often targeted for drugs development. Over 50% of drugs curated in the DrugBank target the surfaceome. In addition to their ubiquitous expression on the plasma membrane, the extracellular stalks of these cell surface proteins can be cleaved and released into the bloodstream, making them suitable targets for blood-based diagnostics. Surfaceome can also be draped with glycans during post-translational modifications, which will mediate their interaction with other proteins that reside on either the same or neighbouring cells as well as with the microenvironments [
38]
. Dysregulated surfaceome expressions and functions have been shown to promote tumour formation and progression [
39]. Therefore, scientists have begun profiling and cataloguing surfaceome in various types of cancers [
40‐
43]. These cell surface proteins can be elevated in cancer cells in which they can respond to the increased level of growth factors, rendering cancer cells to sustain their infinite proliferative capabilities [
44] and interact with the microenvironment that could either directly or indirectly modulate the tumour growth and metastatic capabilities [
45].
The GBM transcriptomics dataset has been previously utilized to uncover genes that support GBM pathogenesis as well as genes that have potential prognostic values [
46‐
48]. For example, Nicolasjilwan et al. analyzed the TCGA database to predict the survival of GBM patients based on clinical features, MRI images genomics alterations [
46]. However, most TCGA GBM differential genes expression analyses either relied on a low number of normal brain tissue samples, in which the TCGA GBM cohort contained only 5 normal brain tissues RNA-Seq data, or the data were combined with the GBM TCGA microarray data. This might create an imbalance that would lead to inaccuracy or bias in the downstream analysis. Hence, to increase the robustness of this study in identifying the significantly upregulated GBM surfaceome repertoire, we included the normal brain tissues GTEx RNA-Seq database TCGA in our analysis. On a similar scale, the GTEx studies have performed genes expression profiling in more than 11,000 samples across multiple human tissues from nearly 1000 healthy donors. We compared the TCGA GBM and normal cortex GTEx RNA-seq data and identified 2381 significant differentially expressed genes in GBM, in which 648 were upregulated and 1733 downregulated genes. In agreement with the previous GBM proteomics profiling study [
12], the GO cellular compartment analysis showed that most of the dysregulated genes in GBM encode for the cell surface proteins, suggesting the importance of cell surface proteins in GBM pathogenesis.
Of the 2381 significant DEGs in GBM, 395 genes encode for cell surface proteins, in which 124 and 271 genes were found to be significantly upregulated and downregulated, respectively. Interestingly, receptor subclass was the predominant dysregulated genes in GBM, suggesting the crucial roles of cell surface receptors in supporting GBM pathogenesis. This was indeed in line with several studies reporting the implications of cell surface receptors dysregulation in the pathogenesis of many cancer types [
49]. For this reason, the development of cancer treatment strategies has been revolved around targeting the cell surface receptors such as the receptor tyrosine kinases (RTKs) [
50] and G protein-coupled receptors (GPCRs) [
51]. Therefore, targeting the cell surface proteins particularly the receptor subclass could potentially be further explored as novel GBM therapeutic options.
Robust cancer biomarkers are those that could be reproducibly identified by multi-omics platforms or reported in several different studies. To this end, we integrated the analyzed transcriptomics data with publicly available GBM proteomics data to prioritize high-confidence cell surface proteins. Also, due to post-transcriptional and post-translational modifications, the mRNAs expression level is sometimes not correlated with their respective protein expression levels [
52]. After mapping the prioritized genes from the transcriptomics-proteomics integrative analysis with the PPI network analysis data, we identified 6 genes;
HLA-DRA, CD44, SLC1A5, EGFR, ITGB2, PTPRJ, whereby we considered these genes as the high-confidence GBM predictive surface markers. Previously integrated transcriptomics based on bulk expression profiles suggested that GBM is heterogeneous and can be clustered into at least three subtypes namely pro-neural, classical; and mesenchymal [
28]. Recently, scRNA-seq analysis confirmed the intra-tumoural heterogeneity of GBM in which it can exist in multiple states with distinct cells and transcriptional programs that can be dynamically transitioned into different subtypes [
27]. Most of the identified 6-gene signature belongs to macrophage cell type while only EGFR is specific to GBM (Fig.
4C and Supplementary Fig.
S4). Although most of the hits are not GBM-specific genes, these different cells are part of the GBM microenvironment or tumour niche, which are equivalently important in driving GBM pathogenesis. Hence, regulation of these genes within the tumour microenvironment recapitulates the cellular program, plasticity and genetic drivers of GBM. Overall survival analyses revealed that there was no significant difference in the overall survival between patients who had high and low expression of these 6 genes, either the genes were analyzed individually or when combined. However, when looking at the disease-free survival, patients who had high expression of
CD44, PTPRJ, and
HLA-DRA, either individually or as a group, had significantly poor disease-free survival (Supplementary Fig.
6 and
8B) compared to subjects with low expression of the genes. These findings indicate that these 3 genes,
CD44, PTPRJ, and
HLA-DRA, could potentially be developed as GBM prognostic markers in the clinic.
In addition to identifying the already known GBM drivers like CD44 and EGFR, our integrative analysis approach has also enabled us to identify potential novel genes that have not either been reported or thoroughly discussed in the context of GBM. For instance, within the 6 GBM signature genes, ITGB2 has not been widely associated with the pathogenesis of GBM. ITGB2 encodes for cell surface protein that is important in regulating cell adhesion and cell-surface mediated signalling [
53]. Hence overexpression of this protein is relevant in promoting cancer growth possibly by modulating cancer cells adhesive and migratory properties, and the pro-oncogenic signalling cascades. Though there are
in-silico and in-vitro studies that associated the ITGB2 as one of the important genes in cancer, the exact mechanisms of how this gene promotes GBM remains elusive and worth to be investigated in the future [
11,
54,
55]. Human leukocyte antigen (HLA)-DRA is a classical major histocompatibility complex (MHC) class II molecule that plays important role in immune responses modulation. High expression of the HLA-DR gene family has been associated with more aggressive tumour grade in gliomas and poor prognosis [
56,
57]. Nonetheless, the functions of HLA-DRA in driving GBM growth has not been fully elucidated.
PTPRJ gene is a member of the protein tyrosine phosphatase (PTP) family whose substrates include the RTKs such VEGFR, PDGFR and EGFR [
58]. Since the RTKs pro-oncogenic properties are well-established in which their activation largely depends on phosphorylation, PTPRJ is thus deemed to function as tumour suppressor proteins due to its function as a phosphatase that can negatively regulate the signalling pathway. This was also evidenced by the ectopic expression of PTPRJ in in-vitro models that resulted in cell growth inhibition [
59,
60]. In contrast to these previous reports, we found that
PTPRJ expression was upregulated in GBM and led us to suggest that PTPRJ might have a pro-oncogenic role in GBM pathogenesis. To our knowledge, there have been no previous reports linking
PTPRJ expression and function with GBM pathogenesis. This notion of PTPRJ potential ‘double-edged sword’ and GBM-specific pro-oncogenic function needs to be investigated further. SLC1A5, another hit target from our analysis, is a neutral amino acid transporter in which its high expression has been implicated in many cancer types including GBM [
61]. In GBM, SLC1A5 expression is under the control of pro-oncogenic c-Myc protein but how this transporter supports the tumour cells proliferation and growth remain poorly understood [
62].
As highlighted above, the identification of CD44 and EGFR in this present study is expected because they have been previously described as one of the key targets for GBM [
32,
63]. This validates the robustness of our approach in the sense that not only our analysis identified several novel genes, but also the findings overlap with previous studies. Since EGFR pro-oncogenic roles have been widely implicated in many cancer types and several drugs have been developed and clinically approved to target EGFR [
37,
64,
65], we focused our analysis on CD44. The CD44 encodes for transmembrane glycoprotein that serves as the receptor for hyaluronic acid, a component of the extracellular matrix, and several other ligands including osteopontin, fibronectin and collagen [
32]. The CD44 antigen has been implicated in modulating tumorigenesis in many cancer types in which high expression of this CD44 increases cancer cells proliferation, motility and survival as well as promoting cancer metastasis [
66]. In GBM, high expression of CD44 was identified in the proteogenomic profiling of GBM tissues [
23] and further classified as a GBM cell surface antigen in a systematic analysis [
31]. Interestingly, this transmembrane glycoprotein can be cleaved and secreted into the vasculatures, suggesting its potential to be developed as a diagnostic marker [
67]. It has been reported that the activation of CD44 by its ligand promotes cancer stem cell-like phenotypes in GBM and increased therapeutic resistance [
68]. Consistent with this, drugs targeting CD44 are currently in clinical trials, and so far the results are promising in that CD44 inhibition impede GBM cells growth [
69]. Our co-expression network analysis using graph-based analytics [
15] demonstrated that genes connected to CD44 were also highly co-expressed in GBM compared to normal brain tissues, suggesting that the CD44 signalling axis is important in GBM tumorigenesis.
The currently approved therapies to treat GBM are far from satisfactory and have remained unchanged for more than a decade [
70]. This includes the alkylating agent temozolomide, which is the first line of drug used in treating GBM. Therefore, there is a need for novel or alternative treatment strategies for GBM. Due to the upregulated expression of CD44 in GBM, drugs targeting CD44 are currently undergoing clinical trials and the results are thus far promising in that CD44 inhibition impedes GBM cells growth [
69]. In addition to this, our drug mapping analysis revealed hyaluronic acid as an actionable CD44 binding molecule. It is therefore appealing to investigate the activity and potential use of this existing drug to treat GBM in the future, which has yet to be comprehensively studied. Within the CD44 co-expressed interactome, three additional targets already have drugs that can modulate them namely the C1R, CALR and TNFSR1A (Table
2). Based on our knowledge, the activity and efficacy of these drugs have not been tested in any in-vitro or in-vivo GBM models yet. Also, studying a combination of these available drugs targeting our GBM signature or the CD44 co-expression network could disrupt the aberrant hub gene interactome and potentially enhance GBM treatment efficacy.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.