Neuroblastoma is a paediatric tumour that develops from embryonic neural crest cells that give rise to the sympathetic nervous system. Whilst most neuroblastomas initially respond well to induction chemotherapy, 50–60% of patients with high-risk disease relapse with aggressive disease. Acquired drug resistance is a major therapeutic challenge in the treatment of such patients.
Using an array of 96 genes related to cancer stem cells in three-paired parental and Vincristine drug resistant neuroblastoma cell lines, this study outlines 11 prominent genes across all cell lines; ABCG2, ERBB2, LIN28A, SOX2, WEE1 MYC, SNAI1, LIN28B, ABCB5, DLL1 and ITAG2 (Fig. 1B). Differing expressions of the same genes across the cell lines, highlights the hallmark heterogeneity of disease.
Using complex bioinformatics analysis, we created a regulatory network composed of 12 genes, 8 of which were highlighted within our study and relating genes discovered from the database (Fig. 2A&B). The network displayed a core role for MYC directly influencing genes associated with oncogenic proteins (ERBB2), self-renewal (SOX2) and cell cycle (WEE1). Indirectly, repression of MYC appears to cause over expression of hallmark drug-resistance genes such as SNAI1 and ABCG2, whilst decreasing expression of NOTCH signalling protein DLL1 and integrin protein ITGA2 (Kang et al. 2009; Chen et al. 2012; Ongaratti et al. 2016). Interestingly MYC (C-MYC) was associated with lower overall survival across four data sets, in contrast to MYCN amplification which is strongly correlated to poor overall and event free survival in high risk neuroblastoma patients (Somasundaram et al. 2019) suggestive of a potential inverse correlation (Westermann et al. 2008).
Whilst most other cancers display a positive correlation between C-MYC overexpression and a number of the genes highlighted in our study, a number of studies have implicated that c- and MYCN can compensate for loss of expression by one another in normal embryonic stem cells(ESCs), induced pluripotent stem cells(IPSCs) and neuroblastoma(Cotterman and Knoepfler 2009; Varlakhanova et al. 2010; Nakagawa et al. 2010). This may explain the findings we have discovered.
Similar to our findings, a study in triple negative breast cancer, showed elevated SOX2 expression alongside overexpression of ABCG2 and the EMT marker Twist-related protein 1 (TWIST1) in drug resistant cells. However, when paclitaxel was administered to SOX2 inhibited mammospheres, ABCG2 and TWIST1 were downregulated, alongside decreased expression of other EMT markers (SNAI1) which induced arrested migration and reduced formation of spheres, indicative of loss of self-renewal (Mukherjee et al. 2017). This study highlights the potential of the EMT-Stemness axis in therapeutic resistance, where SOX2-dependent TWIST1 overexpression maintains stemness in addition to enabling migration. Subsequent inhibition of the axis causes regained sensitivity.
Here we’ve highlighted deregulation of notch signalling pathway genes in drug resistance cell lines, recent research from clinical samples showed 60% (28/46) of clinical samples from neuroblastoma patients with overexpression of NOTCH1 linked to recurrence and lower overall survival (Metovic et al. 2022).
EMT is a complex process, recent data has shown that Notch signalling directly influences overexpression of Snail and decreased expression of E-cadherin (Timmerman et al. 2004; Saad et al. 2010). Snai1 expression and loss of E-Cadherin are associated with several cancers and causing increased migration and invasion (Chen et al. 2010; Zhang et al. 2017), interestingly a recent study in osteosarcoma found that sub-lethal doxorubicin (Dox) treatments showed significant correlation between EMT and Notch signalling in doxorubicin resistant cell lines (Yang et al. 2017) which supports our findings within the context acquired drug resistance. Delta like 1 (DLL1) has also been associated with disease recurrence in other cancer types, including hepatocellular carcinoma (Ma et al. 2016) and with poor prognosis in non-small cell lung cancer (Pancewicz-Wojtkiewicz et al. 2017). A recent study in Neuroblastoma cell lines, showed that DLL1 was highly expressed in MYCN amplified cells, and could be inhibited by miRNAs(Bettinsoli et al. 2017). In aggressive medulloblastoma, NOTCH signalling pathway has shown to regulate self-renewal and metastasis, with TWIST1 and Polycomb complex protein BMI-1 (BMI1) influential in NOTCH1 induced metastases (Kahn et al. 2018).
Two genes, SOX2 and LIN28A are involved with stem cell pluripotency. Interestingly, Yamanaka’s seminal study on the generation of induced pluripotent stem cells (iPSCs) utilised both SOX2 and LIN28A to reprogram adult somatic cells to an undifferentiated state (Takahashi and Yamanaka 2006). Since then there has been an emerging interest in genes linked to stem cells and pluripotency and their role in tumorigenesis (Müller et al. 2016), however whether that role is extended to therapeutic resistance is understudied. The up-regulation of both SOX2 and LIN28A across all VCR-resistant cell lines may suggest that the acquisition of a stem cell phenotype is important for the acquisition of drug resistance, which occurs in 50% of patients with high risk tumours.
This trend is not limited to VCR resistance alone however, in doxorubicin resistant gastric cancer stem cells, Sox2 and ABCG2 overexpression was associated with the dox-resistant phenotype (Tian et al. 2012) in addition to elevated LIN28B overexpression in gastric cancer cells with dox-resistance (Teng et al. 2015). DLL1 exerts a role in development of drug resistance in dox-resistant cell lines and biopsies in osteosarcoma (Pu et al. 2017). Interestingly, conflicting studies regarding Snai1 expression in breast adenocarcinoma cell lines in dox-resistant settings,(Lim et al. 2013; Tsou et al. 2015), however Lim et al. ‘s study highlights the importance of EMT-stemness interaction in drug resistance. Highlighting heterogeneity within the ‘same’ cancer. A recent study conducted into whether inhibition of autophagy paired with very intense chemotherapeutic treatments can prevent development of drug resistance in NB, the initial results seem positive (Chen et al. 2022). This is where the balance between vanquishing the tumour without the consequences of aggressive treatment on the patient.
A recent study has highlighted an axis between BMP4-wnt-notch, whereby BMP has a tumour suppressor role, this same work showed that induction of notch proteins halted growth of NB in IMR32 and SH-SY5Y (Szemes et al. 2020). Low BMP4 and notch expression leads to an aggressive NB phenotype, perhaps there is a complexity within the signalling pathways that mean upregulation in notch signalling still leads to the aggressive phenotype. The research landscape has also associated wnt signalling in the development of therapeutic resistance across different cancer types(Forgham et al. 2015; Vieira et al. 2015; He et al. 2018; Fu et al. 2019; Yang et al. 2019; Clark-Corrigall et al. 2022). This underlines the necessity of a global transcriptome of the cell lines in our study in order to further understand the molecular changes in therapeutic resistance.
There have been some studies investigating the relationship between CSC phenotype and EMT (Prieto-Vila et al. 2017), where activation of stem cell related pathways directly promote the transition to a mesenchymal profile. Previously, Notch signalling was widely understood as a regulator of cell fate decisions, self – renewal, maintenance of tissue stem cells and tissue wound healing but more recently as a regulator of survival and regeneration of CSCs (Capaccione and Pine 2013).
However, a definitive molecular mechanism behind the CSC-EMT relationship still remains elusive (Shibue and Weinberg 2017).
Data from in-silico analysis (Supplementary Fig. 1 &Table 1) showed that the worst probabilities for event free survival in late stage (INSS Stage 4) neuroblastoma were ERBB2 overexpression (10% after 3 years in the Versteeg data set – 88, 15% reach 5 year survival in the Kocak dataset (649)), ITGA2 overexpression (0% were predicted to reach 2 year follow up in both the primary NRC dataset – 283 and Versteeg dataset − 88), SOX2 overexpression (10% at 2 year follow up in the Versteeg dataset, 0% reached 4 year follow up Kocak dataset (649) and DLL1 overexpression (predicts 20% reach 2 year follow up in the Versteeg (88) and NRC (283). Confirmation of the link with important clinical parameters in four patient data sets further supports the potential role of these genes in the development of acquired therapeutic resistance in patients with neuroblastoma and potentially other childhood cancers.
Furthermore, in support of our data a landmark study by van Groningen et al investigating the role of super-enhancers on intra-tumoral heterogeneity identified two distinct lineages; adrenergic (ADRN) and mesenchymal (MES) differentiation with the Notch pathway as a key driver of dedifferentiated mesenchymal identity. In both in vivo and in vitro studies MES cells were shown to have enhanced migration, be more resistant to common neuroblastoma treatments and increased prevalence of relapse disease, which reinforces the idea that treatment exerts selective pressure much like the cancer stem cell hypothesis postulates (Groningen et al. 2017). Since then it has been shown that Notch 3 signalling is a key driver in the reprogramming of ADRN cells to a more mesenchymal state (Groningen et al. 2019) and that MES cells are now known to utilise retinoic acid, an agent used in a maintenance dose to prevent relapse, to proliferate and mobilise (Groningen et al. 2021). A recent study from Newcastle has found that TOP2B expression was needed to maintain the ADRN state, and also lowered notch signalling factors like Notch 1–3 (Khazeem et al. 2022).These studies show that ADRN cells can reprogram themselves and become a more dedifferentiated cell type using NOTCH signalling.
To further support this mouse models, have implicated Notch signaling in the maintenance of neural stem cells maintenance in the fetal brain. Such observations from developmental biology have shown that Notch exerts similar roles which are vital to the tumourigenicity of cancer stem cells particularly in solid tumors including glioblastoma, ovarian cancer, and breast cancer (Aster et al. 2017; Yi et al. 2019; Miao et al. 2020; Yang et al. 2021). Unfortunately, due to the complexity of NB there is no perfect mouse model but even with the advent of iPSC based practices researchers still trying to elucidate this in neuroblastoma(Gonzalez Malagon and Liu 2022).
Our research data warrants further study into this panel of genes in acquired drug resistance in neuroblastoma.