Exome sequencing of glioblastoma-derived cancer stem cells reveals rare clinically relevant frameshift deletion in MLLT1 gene

Since it is one of the most aggressive and recurrent brain tumors, glioblastoma multiforme (GBM) poses a great health problem. The recurrent nature of GBM is principally attributed to the presence of a group of tumor-initiating cells, cancer stem cells (CSC); these are thought to play decisive roles in GBM’s resistance to radio- and chemotherapy, resulting in a typically fatal tumor recurrence ~ 7 mo after diagnosis [1].

In our previous studies, we identified two types of CSC within the tumor core (c-CSC), and in the peritumor tissue of GBM (p-CSC) [2‐4]. The genetic makeup and driver mutations of the primary and recurrent GBM primary tumor cells and their associated CSC is not clarified yet. The presence of a heterogeneous population of cancer cell clones within GBM necessitates a thorough understanding of potential genetic variants and their interrelation to each other. Elucidation of the clonal structure and underlying genetic variants in GBM is thus crucial to develop targeted therapies for lethal cancer, such as GBM.

As the name implies, the term ''multiforme'' indicates the high degree of heterogeneity not only in the histopathological features of glioblastoma but also in its genetic mutational load. Previous studies have classified GBM mutations into two main types: clonal and subclonal ones. The clonal mutations are identified in all tumor cells before the process of transformation, while the subclonal mutations are present only in a subset of tumor cells; and they occur later in tumor growth [5]. Whether or not the GBM CSC are clonal or subclonal mutations is not clear yet, and the role of GBM CSC in either the initiation and/or maintenance of GBM growth is still in need of further investigation.

Three core signaling pathways, namely p53, Rb, and receptor tyrosine kinase (RTK)/Ras/phosphoinositide 3-kinase (PI3K) have been previously implicated as playing a main role in the initiation of GBM growth [6]. Alteration in the core molecular pathways is thought to be coordinated and clearly imprinted in different molecular subtypes of GBM. This is clearly reflected not only inpatient sensitivity to different therapeutic modalities but also it appears to have pronounced effects on the clinical outcome [7].

Previous genomic studies of GBM reveals the presence of 21,540 somatic mutations in 71 GBM-relevant genes of which 20,448 were single-nucleotide variants (SNVs), 39 were dinucleotide mutations, and 1,153 were small insertion and deletion (indel) mutations. The SNVs mutations included 5379 silent, 3901 missense, 831 nonsense, 360 splice-site, and 760 mutations resulting in a frameshift [8]. Several genes were identified as significantly mutated genes in GBM namely PTEN, TP53, EGFR, PIK3CA, PIK3R1, NF1, RB1, IDH1, PDGFRA, LZTR1, SPTA1, ATRX, GABRA6, KEL, BRAF V600E, H3.3 histones, EGFR, MET, CDK6), CDK4, MDM2, PDGFRA, SOX2, MYCN, CCND1, CCNE2, PARK2, QKI, TGFbR2, LRP1B, NPAS3, LSAMP, SMYD3, EGFR, CPM, PRIM2, FAM65B, PPM1H, RBM25, HOMER2, EGFRvIII, PDGFRA, p53 pathway (MDM2, MDM4, and TP53), the Rb pathway (CDK4, CDK6, CCND2, CDKN2A/B, and RB1), PI3K pathway (PIK3CA, PIK3R1, PTEN, EGFR, PDGFRA, and NF1), IDH1, ATRX, TERT [8].

In our previous studies, Notch inhibition significantly impaired cell growth of c-CSC compared to p-CSC pools, with no effects observed in cell cycle distribution, apoptosis, and cell invasion assays. Moreover, there was simultaneous targeting of EGFR and PDGFR, which would be beneficial in the treatment of GBM [3]. In another study, the newly discovered PDGFRα/Stat3/Rb1 regulatory axis may represent a potential therapeutic target for GBM treatment [4]. Moreover, Hes1 seems to be a favorite target but not sufficient itself to target GBM efficaciously; therefore we suggested that any potential pharmacological intervention should provide for the use of anti-Stat3/5 drugs either alone or in combination regimen [2].

The complex interrelated molecular pathway that governs the tumorigenic transformation of GBM c- and p-CSC has prompted us to perform a systemic genomic study using whole-exome sequence (WES), to explore potential differences in the global genetic variants between GBM-associated c-CSC and p-CSC. In this regard, four pairs of c-CSC) or p-CSC tissue-derived cancer stem cells were isolated from 4 different patients with the aims of (1) obtaining the complete sequence of the exome of the 4 pairs of GBM cancer stem cell lines, (2) obtaining, for each CSC, the list of the variants respect to the reference genome, (3) comparing the exome sequences between each pair of c-CSC and p-CSC, identifying the different and common variants, and (4) identifying the variants that affect the functionality of genes known to be involved in cancer origin and development. Our experiments are focused on glioblastoma–IDH-wild type, and no disease-defining alterations were present in histone, BRAF or other genes.

By comparative analysis, we identified common gene single nucleotide variants (SNV) in all GBM c-CSC, and p-CSC in the following genes: TP73, PDE4DIP, FN1, KMT2C, MUC6, CREB3L1, GSE1, APC2, and MUC16. A potentially deleterious variant was a frameshift deletion at Gln461fs in MLLT1 gene, which was encountered only in p-CSC. Our study supports the hypothesis that the varied genetic composition between the GBM-associated c-CSC and p-CSC may be involved in the different therapeutic responses or the recurrent nature of GBM. Moreover, the design of specific targeted therapeutic strategies against the most critical/penetrant variants affecting the functionality of GBM should be directed to the genetic alteration associated with both the primary GBM tumor cells and GBM-associated CSC.

In this study, we used exome sequencing to provide a list of somatic/rare alterations associated with GBM. Specifically, we are interested in highlighting potential variants associated with oncogenic genes both in c- and p GBM CSC. Linking genomic data to clinical information might provide new opportunities to decipher genomics-based biomarkers, and to generate novel hypotheses which might help to highlight novel disease-related mechanisms. Our experiments are focused on glioblastoma–IDH-wild type, and no disease-defining alterations were present in histone, BRAF or other genes.

In the present study, SNV was identified in the oncogene TP73 both in C- and p-GBM CSC. TP73-AS1 constitutes a clinically relevant lncRNA in GBM. Significant overexpression of TP73-AS1 was previously identified in primary GBM samples and comprises a prognostic biomarker in glioma and GBM with high levels of expression, identifying patients with particularly poor prognoses. TP73-AS1 promotes TMZ resistance in GBM CSC and is linked to regulation of the expression of metabolism-related genes and ALDH1A1, a protein known to be expressed in cancer stem cell markers and which protects GBM CSC from TMZ treatment [16].

Several SN heterozygous variants in PDE4DIP have been detected both in c- and p-GBM CSC. In all cases, the genomic alterations were associated with amino acid changes such as: p.Arg1867Cys (chr1:144866643), p.Lys1359Glu (chr1:144879375), p.Ala1066Thr (chr1:144882823), p.His482Arg (chr1:144917841), p.Glu410Val (chr1:144918957), and p.Ser275Leu (chr1:144922583) (Table 3). The role of PDE4DIP in GBM was previously reported: it has been demonstrated that this gene is down-regulated in glioma cell lines treated with dB-cAMP. a hat reduces the invasiveness, proliferation, and migratory properties of glioma cells and increases the survival of glioma cell lines compared to untreated cell lines [17, 18].

A homozygous SNV was detected in FN1 gene in the examined in c- and p GBM CSC. This variant was detected at chr2:216272900 where a homozygous T-G transition that was associated with p.Thr817Pro was identified. Activations of MYC, NFE2L2, FN1, and TGFβ1 and inhibition of TP53 in GBM were previously demonstrated by Halla et al. [19]. FN1 is upregulated by TWIST1, which is known to promote epithelial-mesenchymal transition and/or GBM invasion [20, 21]. Furthermore, FN1 is associated with glioblastoma recurrence and can be regarded as a target for antiangiogenic therapy [22]. COL1A1 and FN1 are associated with migration, invasion, angiogenesis, recurrence, and OS in GBM patients. Thus, these genes may serve important roles in the tumorigeneses of GBM [20].

The present study revealed the presence of heterozygous indel in the KMT2C gene. This indel was detected at chr7:151945071 position, and it was associated with p.Tyr816Ter. KMT2C in a GBM is rare; this mutation occurs in only about 4% of GBMs. However, all sorts of other cancers show this mutation. Alterations of EZH2, KMT2C, and CHD4 at the genetic or protein level could perturb an epigenetic program, leading to malignant transformation in glioma [23].

Four heterozygous SNV were detected in MUC6 gene at chr11:1017183, chr11:1017220, chr11:1017325, and chr11:1017337. These variants were associated with p.Pro1873Gln, p.Thr1861Ala, p.Tyr1826Asp, and.Gln1821_Thr1822delinsHisAla, respectively. The MUC6 gene encodes gastric mucin, a secreted glycoprotein that plays an essential role in epithelial cytoprotection from acid, proteases, pathogenic microorganisms, and mechanical trauma in the gastrointestinal tract [24]. The susceptibility to gastric cancer may be related to variation in MUC6 gene expression [25].

Two homozygous indels were detected in CREB3L1 gene at chr11:46342081, and chr11:46342259. Normal and tumor tissues with similar CREB3L1 expression include ESCA esophageal cancer, GBM glioblastoma multiforme, HNSC head, and neck squamous cell carcinoma, LUAD lung adenocarcinoma, SARC sarcoma, THCA papillary thyroid carcinoma, THYM thymoma, UCEC uterine corpus endometrial carcinoma [26]. CREB3L1 is a member of the CREB/ATF family of transcription factors and functions as a transducer of the unfolded protein response (UPR) [27]. A large fraction of proteins synthesized in the cell undergoes folding and post-translational modification in the endoplasmic reticulum before being released to perform their desired function. This process can be disrupted by endoplasmic reticulum stress resulting from hypoxia, glucose or nutrient depletion, change in calcium homeostasis, or expression of mutant or misfolded proteins, potentially leading to accumulation of unfolded proteins that, if released from the endoplasmic reticulum, can have detrimental effects. The accumulation of unfolded proteins in the lumen of the endoplasmic reticulum initiates the UPR. The UPR works to regain endoplasmic reticulum homeostasis by reducing protein translocation into the endoplasmic reticulum, increasing the protein-folding capacity of this organelle, decreasing translation initiation, and increasing protein degradation [28]. Prolonged activation of the UPR leads to apoptosis [29].

In the present study, a heterozygous SNV was detected at chr16:85667696 in GSE1 gene which results in p.Ala62Thr amino acid change. In previous work, it was demonstrated that engineered candidate cooperating mutations in Gorlin neuroepithelial stem (NES) cells, with mutation of DDX3X or loss of GSE1 both accelerated tumorigenesis. These findings demonstrate that human NES cells provide a potent experimental resource for dissecting genetic causation in medulloblastoma [30].

A heterozygous SNV in APC2 gene was encountered at chr19:1457111 which was associated with amino acid change p.Pro359Gln. Continuous activation of the Wnt/β-Catenin signaling has been reported to play an important role in multiple processes of tumor progression, leading to uncontrolled cancer cell proliferation, growth, and survival. miR-1249 targets and suppresses APC2 expression, an important Wnt/β-Catenin pathway-regulated factor. These data suggest that miR-1249 could be a novel therapeutic target for microRNA-mediated cell proliferation in glioma [31]. MUC16 is overexpressed in multiple cancers and plays an important role in tumorigenicity and acquired resistance to therapy. Apart from its protective role in normal physiology, MUC16 contributes to disease progression and metastasis in several malignancies. Identification of neo-antigenic epitopes in MUC16 that correlate with improved survival has raised hopes for developing MUC16-targeted immunotherapy [32].

A heterozygous indel in MLLT1 gene was recorded in the present study at chr19:6213974; it was associated with amino acid change p.Gln461fs. Interestingly, the MLLT1 indel was identified in GBM derived p-CSC only. KMT2A (MLL) rearrangements are observed in various types of pediatric and adult leukemia, but only one adult case report has so far shown KMT2A (MLL)-MLLT1 gene rearrangements in blastic plasmacytoid dendritic cell neoplasm (BPDCN) [33].

From these findings, we hypothesize that MLLT1 may be of importance in stem cell differentiation/glioma pathogenesis. However, the implications and potential downstream effects of this genomic variant are not explored. The lack of germline sequencing data makes inferences about somatic SNVs highly concerning. This may necessitate more future studies on higher number of GBM samples, and further experimentation would be needed to highlight potential downstream effects of identified indel in MLLT1 gene. In order to fill these caveats, we did an extensive search for the potential role of MLLT1 gene in relation to different malignancies. There were very few data available in previous publication about the implication of this variants in GBM, and potential downstream effects. However, it was demonstrated that mixed lineage leukemia (MLL) fusion proteins are derived from translocations at 11q23 that occur in aggressive subtypes of leukemia, and MLL is joined to different unrelated proteins to form oncogenic transcription factors. Zeisig et al. [34] demonstrated a direct interaction between several nuclear MLL fusion partners and present evidence for a role of these proteins in histone binding. In two-hybrid studies, ENL (the protein product of MLLT1 gene) interacted with AF4 and AF5q31 as well as with a fragment of AF10. Overlay and pulldown-assays finally showed a specific and YEATS domain-dependent association of ENL with histones H3 and H1. These studies support a common role for nuclear MLL fusion partners in chromatin biology (Fig. 3).

MLL rearrangements are also present in about 10% of other pediatric and adult acute myeloid leukemia (AML) and acute lymphoid leukemia (ALL). These translocations and others occurring in early life are associated with a dismal prognosis compared with adult leukemias carrying the same translocations. This observation suggests that infant and adult leukemias are biologically distinct but the underlying molecular mechanisms for these differences are not understood. Sinha et al. have developed a novel MLL-ENL embryonic leukemia model in mice that can be used to study some aspects of infant leukemia ontogeny [35].

So, now the question is: is ENL fused to MLL in our GBM CSC? Is ENL fused/rearranged to other protein to make a new chimeric protein with aberrant oncogenic functions? We are currently performing more experimentation to prove this new hypothesis for GBM pathogenesis. Debernardi et al. [36] demonstrated that AF10 is involved in 2 distinct chromosomal translocations associated with hematologic malignancy. The chimeric fusion proteins MLL/AF10 and CALM/AF10, resulting from the t(10;11)(p12;q23) and the t(10;11)(p12;q14), respectively, consistently retain the leucine zipper motif of AF10. The leucine zipper interacted with GAS41, a protein previously identified as the product of an amplified gene in a glioblastoma. GAS41 shows significant homology to the Saccharomyces cerevisiae protein ANC1 and to the human MLL fusion partners AF9 and ENL. The interaction was confirmed in vivo [36].

In addition to genetic mutations represented mainly as SNV and indels that have been encountered in the present investigation in key oncogene genes (such as TP73, PDE4DIP, FN1, KMT2C, MUC6, CREB3L1, GSE1, APC2, MUC16, and MLLT1) previous studies have elucidated that GBM was associated with alteration in other key signature oncogenes such as EGFR and PI3K, and that in over 40% of GBM carries one or more nonsynonymous mutation among the chromatin-modifier genes [37]. Alterations in chromatic rearrangement have been described for other types of cancer such as ovarian [38] and renal [39] carcinoma.

Based on the genomic profile of GBM c- and p-CSC that has been confirmed in the present investigation, most of the SNV that were detected in both c- and p-CSC are represented by common variants that have previously been recorded in other types of cancers. To our knowledge, none of the identified homozygous and heterozygous SNV were previously linked to chemotherapeutic drugs used to treat GBM. Moreover, none of the identified variants were previously reported to play a decisive role in the recurrent nature of GBM, an observation that might indicate that the GBM CSC did not play a significant role both in GBM resistance to chemotherapy and its recurrent nature. Nonetheless, the mutation load of the GBM seems to be an integral part of GBM mutanome.

Recently, several studies have been devoted to elucidating potential targeted therapies for GBM, and several such strategies were designed against key GBM oncogenic genes/pathways such as BRAF [40] and FGFR1/FGFR2/FGFR3 [40]. In our previous studies, Notch inhibition significantly impaired cell growth of c-CSC compared to p-CSC, Besides, p-CSC are more refractory to anti-EGFR targeting either alone or in combination with the anti-Notch1 drug compared to c-CSC [3], suggesting that p-CSC possess a different genetic background which confers them resistance to the anti-tumor agents [3]. Simultaneous targeting of EGFR and PDGFR negatively impacted both c-CSC and p-CSC. In another study, the newly discovered PDGFRα/Stat3/Rb1 regulatory axis might represent a potential target for the treatment of refractory p-CSC [4]. We also reported that the interference of Notch1 target Hes1 overcomes the resistance of CSC to GSI-X [2].

The majority of GBM tumors had a complex genome and transcriptome, and usually, they were associated with a high frequency of structural variants on the q arm of chromosome 12, involving the MDM2 and CDK4 genes. This may be a functional alteration relevant to GBM [41]. This view didn’t match the SNV identified in the analysis of the genetic mutations of GBM-associated p- and c-CSC, which might indicate that these key GBM oncogenic genes are mainly relevant to the primary GBM tumor cells rather than to the GBM-associated CSC genes.

Nearly half of GBM tumors display a complex alteration in the EGFR genes as represented by fusion and deletion that compose essential features of the somatic mutations associated with GBM [42]. Despite the main role of EGFR deletion/fusion in the survival and growth maintenance of GBM, other different EGFR alterations might also be encountered. Such alterations might induce variable responses to other targeted therapeutic modalities.

Whether or not the identified GBM genomic alterations are in concordance with the proteomic variations as reflected in the downstream molecular pathways, there is still a need for further investigation, and targeting the altered genomic pathways should be directed not only to key oncogenic genes encountered within the GBM primary tumor cells but also to downstream signaling components along a pathway of GBM-associated CSC.

The poor prognostic outcome of GBM is attributed to the presence of cancer stem cells (CSC) which confer resistance against standard chemo- and radiotherapeutics modalities. In the present study, potential differences in genetic variants between c-CSC versus p-CSC derived from four GBM patients have been investigated. By comparative analyses, we identified common gene mutations in all GBM c-CSC, and p-CSC: TP73, PDE4DIP, FN1, KMT2C, MUC6, CREB3L1, GSE1, APC2, MUC16. A potential deleterious variant was a frameshift deletion at Gln461fs in MLLT1 gene that was encountered in p-CSC only. Further investigation is required to confirm the presence of the identified mutations in patient tissue samples, as well as the significance of the frameshift mutation in the MLLT1 gene on GBM biology and response to therapy based on genomic functional experiments.