Y-Box Binding Protein-1: A Neglected Target in Pediatric Brain Tumors?

Brain and central nervous system tumors represent the most common childhood solid tumors. Comprising 21% of all pediatric cancers, they remain the leading cause of cancer-related mortality and morbidity in childhood. Due to advances in neurosurgical technique, radiotherapy and the use of combination therapy, survival rates have generally increased. However, by cause of the lesion itself, its surgical removal and subsequent treatment, survivors are at high risk of long-term neurocognitive sequelae and secondary cancer. Clearly, improvements in diagnosis and treatment are needed. Accordingly, current treatment is evolving away from conventional, uniform therapy and towards risk-stratified regimens and molecularly-targeted therapies, with the aim of diminishing adverse side effects while minimizing the risk of disease recurrence. The multifunctional oncoprotein Y-box binding protein 1 (YB-1) may serve as one such molecular target. Increased YB-1 levels have been reported in a number of pediatric brain tumors, where YB-1 appears to facilitate the advancement of malignant phenotypes. These include proliferation, invasion, and resistance to therapy, as well as the maintenance of brain tumor-initiating cells. Here we evaluate the current literature and show how YB-1 modulates signaling pathways driving each of these phenotypes. We also review the regulation of YB-1 at a transcriptional, translational, posttranslational and subcellular level and argue that there is strong and sufficient evidence to support the development of YB-1 as a biomarker and future therapeutic target in childhood brain tumors.


Introduction
Cancer represents the leading disease-related cause of death in children residing in high-income countries. Brain and central nervous system (CNS) tumors are the most common solid childhood malignancies and the second most prevalent cancer in children after leukemia, accounting for a quarter of all childhood cancers diagnosed each year in the United Kingdom. Considerable advances in imaging, neurosurgical technique, radiotherapy, and the advent of combination chemotherapy means that survival rates for children with brain and CNS tumors has greatly improved over recent years. However, even with such advances, survival for children with brain and CNS tumors remains highly variable, with 5-year survival ranging from less than 10% to over 90% depending on tumor type (1). Furthermore, longterm pediatric CNS tumor survival still remains lower than that of leukemia and other solid pediatric malignancies (2). In addition, research has shown that 80% of survivors go on to suffer significant, long-term consequences of exposing the developing brain to the aforementioned medical interventions including physical, cognitive, neurological and endocrine complications, and an increased risk of secondary cancer (3). The long-term sequelae of treatment and the high rates of relapse observed in a number of pediatric brain tumor subtypes demands the development of alternative treatment strategies.
In recent years, advances in transcriptomics and DNA methylation profiling have transformed our understanding of the molecular underpinnings of a number of childhood brain tumors. Where tumors such as medulloblastoma (MB) were historically categorized into risk groups based upon histology, metastatic status, and resection status, it is now recognized that these tumors encompass at least four distinct molecular subgroups which differ genetically and phenotypically and are associated with different outcomes (4). Accordingly, at present, clinical trials are moving away from treating brain tumors as a uniform disease and towards differential therapy for groups ascribed on the basis of molecular and clinical risk features.

Medulloblastoma
MB, a malignant embryonal cerebellar neoplasm, is the most frequent brain tumor of childhood, comprising approximately 20% of all pediatric CNS tumor cases. Contemporary treatment protocols include maximal safe surgical resection with a combination of craniospinal radiotherapy and adjuvant chemotherapy (5). MB can occur at all ages, although the peak age of diagnosis is approximately 6 to 8 years and incidence declines with increasing age. Advances in molecular and genetic profiling uncovered substantial heterogeneity within MB and led to the identification of four core distinct molecular subgroups-Wingless (WNT), Sonic Hedgehog (SHH), Group 3, and Group 4-which differ in their patient demographics, prognoses, and metastatic status (4,6,7). More recently, on the basis of significant intra-subgroup heterogeneity, variances in age of onset and disparate prognoses, the existence of additional subtypes beyond the four consensus subgroups has been reported (8)(9)(10)(11). Such variations, particularly in clinical response between both subgroups and subtypes, has promoted re-evaluation of patient risk stratification. Accordingly, it is now proposed that patients are stratified into low, standard, high, and very high clinical risk groups and treatment strategies altered to reflect this (12). Only one current clinical trial (NCT01878617) takes into account both subgroups and risk stratification. In this trial, patients with low-risk WNT, receive de-escalated treatment with the aim to reduce treatment-related morbidities, whereas skeletally mature standard-and high-risk patients with SHH receive standard-of-care therapy with the addition of smoothened receptor inhibitors. Standard and high-risk patients with non-SHH/-WNT are then prioritized for intensified treatment.
Despite phenotypic and genetic variances, all MB tumors are classified by the WHO as Grade IV (high-grade) neoplasms, owing to their aggressive and malignant behavior (13). Indeed, a poor prognostic factor in nearly all patients with MB (except those with WNT MB, where metastasis is rare) is the high frequency of metastasis, with up to 40% of pediatric patients exhibiting metastases at diagnosis and almost all patients metastatic at relapse (14). Indeed, survival rates of SHH, Group 3, and Group 4 patients exhibiting metastases at diagnosis range from under 50% to 75%, lower than that of nonmetastatic patients which can range from 75% to over 90% (12). Patients with recurrent MB have poorer outcomes still with a median survival of less than 1 year (15). Clearly, an improved understanding of the molecular events driving treatment resistance and recurrence represents a critical research area within the MB field.

Pediatric Glioma
Glioblastoma multiforme (GBM) is the most frequent and aggressive malignant primary brain tumor in adults. Accounting for 16% of all primary brain tumors, it is a deadly disease with a survival rate of just 14 to 15 months post-diagnosis, despite aggressive treatment. Likewise, pediatric GBM (pGBM) remains one of the few incurable pediatric cancers, with a long-term survival rate of less than 10% (16). Accordingly, it is classified as a WHO Grade IV tumor. Due to a lower incidence in children, studies have traditionally combined pGBM with anaplastic astrocytoma and diffuse intrinsic pontine glioma (DIPG), as high-grade gliomas. Together, these high-grade gliomas constitute 15% of pediatric CNS tumors.
Although pGBM are indistinguishable histologically from adult tumors, recent studies have considerably improved our understanding of the molecular biology both separating and underlying adult and pediatric disease. In 2012, Sturm and colleagues identified six distinct GBM subgroups based upon global DNA methylation patterns: IDH, K27, G34, RTK I, RTK II (classic), and mesenchymal. Although only the RTK II group is devoid of pediatric tumors, the majority of pGBM are found in the K27, G34, and RTKI subgroups, where K27 and G34 are defined by mutations in H3F3A and RTKI by PDGFRA amplifications (17). Indeed, PDGFRA amplification, frequent chromosome 1q gain, and less frequent chromosome 7 gain and 10q loss appear to represent markers of pediatric disease (18). An additional gene expression study performed specifically in pGBM revealed two molecularly diverse disease subsets (one with RAS and AKT pathway activation, one without) with differential survival within pediatric cohorts; importantly, these were also distinguishable at the molecular level from adult samples (19). Despite such significant variances in comparison to adult disease, current treatments for pGBM are predominantly based upon studies conducted in adults. As such, conventional therapy includes postoperative nonselective radiotherapy and chemotherapy (20). A better understanding of the molecular pathogenesis of pGBM is desperately needed to identify new targets for individualized or molecularly stratified therapies.
As previously described, pediatric high-grade gliomas encompass both anaplastic astrocytoma and DIPG in addition to pGBM. Representing approximately 10% of all pediatric CNS tumor cases, DIPG is a WHO grade IV tumor (13). Unlike other pediatric tumors, in which survival rates have improved substantially over the last four decades, outcome for patients with DIPG remains dismal with a median survival of less than 1 year (21). This is largely due to the lack of therapeutic options available for treatment of this aggressive tumor. Recently, increased tumor tissue availability and next-generation sequencing methods has enabled the identification of two histone gene aberrations-point mutations in H3F3A and HIST1H3B, encoding the histone variants H3.3 and H3.1. These mutations are present in $80% of patients with DIPG, differ in age and prognoses and importantly, may represent promising therapeutic targets (16). Anaplastic astrocytomas encompass 2% of all pediatric CNS tumor diagnoses and are categorized as WHO grade III tumors. Five-year survival for this malignant tumor is just 31% (22). Like pGBM, anaplastic astrocytomas are histologically indistinguishable from their adult counterparts, however exhibit significant differences at a molecular level. For example, where adult anaplastic astrocytomas commonly present with EGFR amplification and IDH1 mutations, H3F3A mutations are common to pediatric disease (ref. 22 and cited therein).

Ependymomas
Ependymomas are the third most common CNS tumor in children, accounting for 6% to 12% of cases. They can exist as WHO Grade I, II, or III, although a number of studies have noted a lack of correlation between tumor grade and outcome (13,16). Despite recent advances in therapeutic approach, prognosis for this tumor type has remained relatively poor, particularly for infants in which 5-year survival is just 42% to 55% (23). In addition, one-third of ependymoma cases relapse, which like in MB, correlates with dismal prognosis (16). Historically, ependymomas were considered a single entity based on their morphological appearance. However, it is now clear that ependymomas arising within posterior fossa, supratentorial, and spinal anatomical locations represent distinct biological entities. As such, molecular profiling studies have further classified ependymomas into nine molecular subgroups-three in each anatomical compartment-to fully reflect the heterogeneity across and within compartments (23).

YB-1 in Brain Tumors: A Neglected Therapeutic Target?
The identification and development of molecularly-targeted therapy represents a critical area of pediatric neuro-oncology research, both to minimize adverse long-term sequelae associated with current regimens and to improve treatment efficacy, potentially reducing the rate of disease recurrence with its well-described dire prognosis. Y-box binding protein-1 (YB-1), encoded for by the YBX1 gene on chromosome 1p34.2, may represent one such target. Although originally identified by its ability to bind specifically to Y-boxes (5 0 -CTGATTGGT/CT/CAA-3 0 ) in the promoters of target genes, it has since been demonstrated that YB-1 can interact with numerous RNA and DNA sequences and as such, YB-1 is now considered a multifunctional protein with extensive roles beyond its transcription factor functionality (24). In fact, YB-1 has been implicated in almost all DNA-and mRNA-dependent processes within the cell, including mRNA translation and packaging, DNA repair, proliferation, pre-mRNA splicing, and DNA replication (25).
Due to the multifunctional nature of YB-1 and its position upstream of numerous cellular signaling pathways, including E2F, PI3K/AKT/ mTOR, RAS/RAF/MEK/ERK, and P53 pathways (discussed extensively in ref. 26), it is perhaps not surprising that YB-1 has been heavily linked to cancer. Overexpression has been recorded in numerous tumor types including breast cancer, osteosarcoma, cervical cancer, and lung cancer where expression, and in particular nuclear expression, correlates with poor prognosis, proliferative, and metastatic potential and drug resistance (ref. 27 and cited therein).
Despite considerable research in solid tumors, there is a significant lack of studies exploring the role of YB-1 in pediatric brain tumors or indeed in brain and CNS tumors in general. Elevated YBX1 expression has been reported in both pGBM and MB (19,28,29). Further, analysis of large-scale, publically available datasets shows YBX1 expression is substantially increased not only in MB and pGBM, but also in DIPG, anaplastic astrocytoma and ependymoma (Fig. 1A). In both MB and pediatric high-grade glioma, elevated YBX1 expression correlates with poor survival outcome ( Fig. 1B and C). The high expression and prognostic significance of YB-1 in pediatric brain tumors therefore suggests that it may represent a clinically relevant target worthy of further investigation. In this review, we discuss YB-1 structure and regulation and explore how its diverse cellular functions may facilitate disease progression in pediatric brain tumors.

Structural Organization of YB-1
YB-1 is a 324 amino acid protein with a molecular mass of 35.9 kDa (Fig. 2). It comprises of three domains: a short alanine/proline-rich N-terminal domain (A/P domain), a highly conserved cold shock domain (CSD), and an elongated C-terminal domain (CTD) with alternating clusters of positively and negatively charged amino acid residues (30). The three-dimensional structure of the CSD domain has been resolved by NMR and is known to adopt a classical oligonucleotide/oligosaccharide (OB)-fold consisting of a closed five-stranded antiparallel b-barrel (31). The CSD contains ribonucleoprotein (RNP)-1 and RNP-2 RNA binding motifs and is responsible for both specific and nonspecific RNA binding and specific ssDNA binding. Recently, it was demonstrated that the YB-1 CSD forms a homodimer, both in a crystal structure as well as in solution. This has been proposed to facilitate the RNA binding ability of YB-1 and its multimerization through the CTD-a process involved in mRNA packaging and translational control (32).
Contrastingly, the 3D structures of both the CTD and A/P domains are unknown, likely due to their predicted intrinsically disordered nature (33). The A/P domain contains binding sites for various proteins, including serine/arginine rich splicing factor 9 (SRP30C; also interacts with the CSD), cyclin D1, actin, and P53 (34)(35)(36)(37). Likewise, the CTD has been reported to bind a number of proteins, including P53, DNA repair protein NEIL2 and YB-1 itself (38-41). The CTD also contains several proposed nuclear localization signals (NLS), all of which are denoted in Fig. 2. Originally, the YB-1 NLS was identified at residues 183-202 (42). More recently, a further set of NLS have been proposed: NLS-1 at residues 149-156, NLS-2 at residues 185-194, and NLS-3 at residues 276-292 (43). Finally, a proline-tyrosine (PY)-NLS was identified at residues 174-202, closely matching the position of the originally identified NLS (44). Two cytoplasmic retention sites (CRS) have also been suggested, residing at residues 247-267 and 264-290 (42,45). The CTD also has nucleic acid binding capabilities. Although the CSD appears to have a more significant role in RNA binding, it is thought that DNA binds with highest affinity to the CTD (46).

Regulation of YB-1 expression
In spite of interest in YB-1 and its functions, our understanding of the regulation of YBX1 expression remains vague. Several transcription factors have been shown to regulate YBX1 through interaction with a number of different motifs throughout both the YBX1 promoter and 5 0 untranslated (UTR) region ( Table 1). Correspond-ingly, control of YBX1 expression is likely to be highly contextdependent and may be triggered in response to various stimuli. Six E-boxes are located throughout the YBX1 promoter. E-box binding transcription factor C-MYC has been shown to interact with the YBX1 promoter and drive transcription in a P73-dependent manner in response to DNA-damaging stimuli (47). In addition, basic-helixloop-helix transcription factor TWIST1 has been implicated in E-boxdependent YBX1 transcriptional control (48), a finding further validated by the involvement of TWIST1 binding proteins p300/CBPassociated factor (PCAF) in YBX1 transcriptional activation and programmed cell death protein 4 (PCDP4) in YBX1 downregulation (49,50). A 5 0 UTR GGATAA element has also been shown to activate the YB-1 gene promoter through binding of both GATA-1 and GATA-2 to this region, although interestingly not to GATA elements present within the promoter (51).
Posttranscriptional regulatory mechanisms of YB-1 are better described previously (Table 1). YB-1 translation is controlled by an auto-regulatory feedback loop in which YB-1 binds to a $80-nucleotide regulatory element in the 3 0 UTR of its own mRNA, which encompasses overlapping binding sites for both YB-1 and poly(A)binding protein (PABP). PABP binds to a 50-nucleotide A-rich sequence within the regulatory element, stimulating YBX1 mRNA translation in a poly(A) tail-independent manner. Conversely, YB-1 selectively inhibits its own synthesis through interaction with two sequences within the regulatory element, both containing the same 8-nucleotide motif -UCCAG/ACAA. The binding sites of YB-1 and PABP overlap, meaning that the two proteins compete for YBX1 mRNA binding (40,52). YB-1 is a major protein in both polysomal and free messenger RNPs (mRNP) and plays an extensive role in the regulation of overall protein synthesis within the cell, depending on the amount of YB-1 associated with mRNA. In the complete absence of YB-1, or conversely at high YB-1 concentrations where mRNA is saturated with YB-1, translation is suppressed, whereas at relatively low YB-1 levels, translation is stimulated (25). Thus, YB-1 appears to auto-regulate its own synthesis at a concentration that is optimal for the translation of other cellular mRNAs. Moreover, YB-1 translation also appears to be sensitive to signaling through the mTOR pathway, dictated by sequences in the 5 0 UTR of YBX1 mRNA. Regulation in this way appears to be highly dependent on cell division rate. When cells are slow-dividing/serum starved, mTOR signaling is attenuated and hence YB-1 translation is inhibited (53). Various families of noncoding RNA (ncRNA) have also been implicated in the regulation of YBX1 expression, including miR-137 and miR-216a, the lncRNA GAS5 and MIR22HG, reviewed in ref. 54.
YB-1 is known to undergo various posttranslational modifications ( Table 1; Fig. 2), with phosphorylation being the focus of the majority of research to date. Indeed, there is evidence to suggest that phosphorylation "activates" YB-1, promoting both nuclear transport and facilitating DNA binding. The most extensively studied phosphorylation site in YB-1 is serine-102 (S102) in the CSD, which kinases AKT and RSK have been shown to target, implicating both the MAP kinase pathway and PI3K cascade in the control of YB-1 activation (55,56). Other confirmed phosphorylation sites include: S165 (CSD), which appears to be crucial for transcriptional activation of nuclear factor kappa (NF-kB); tyrosine-162 (Y162) (CTD), phosphorylation of which is mediated by FGFR1 and Y188 and Y281 (CTD), where Y281 phosphorylation appears to correlate with YB-1 nuclear shuttling (57). Acetylation sites have also been reported, including lysine-81 (K81) in the CSD as well as K301 and K304 in the CTD. Although the functional significance of K81 acetylation is unknown, acetylation at positions K301 and K304 has been suggested to promote microvesicle-mediated secretion of YB-1 from the cell (57,58). Ubiquitination sites have also been identified in YB-1 and retinoblastoma binding protein 6 (RBBP6), an E3 ubiquitin ligase, has been shown to interact with YB-1, leading to ubiquitination and proteasomal degradation (57). In addition, the peptide bond linkage between glutamate-219 (Glu219) and glycine-220 (Gly220) has been reported to be targeted by the 20S proteasome for endoproteolytic cleavage in an ubiquitin-independent manner and may also be involved in YB-1 nuclear shuttling (43,59).

Regulation of YB-1 cellular localization
YB-1 is a cytoplasmic-nuclear shuttling protein. In nonmalignant cells, over 90% of YB-1 is located in the cytoplasm where it is associated mRNPs. The NLS and CRS located in the CTD of YB-1 appear to regulate YB-1 distribution between the nucleus and the cytoplasm. The CRS is thought to be dominant, which combined with the affinity of YB-1 to mRNA and its interaction with cytoplasmic partner proteins, is responsible for YB-1 cytoplasmic retention (39). The exact mechanism of YB-1 translocation to the nucleus is not fully understood (Fig. 3) and appears to be a dynamic process that occurs on account of YB-1 being able to undergo structural rearrangements in response to various stimuli, resulting in exposure of NLS to transport proteins. Indeed, transportin-1 has been proposed to mediate YB-1 nuclear import through interaction with the PY-NLS (44).
It has been suggested that changes in YB-1 nuclear localization occur in a cell-cycle-dependent manner, with nuclear accumulation observed during the G 1 -S transition. Two regions in the CTD, encompassing NLS-2 and NLS-3, were shown to mediate translocation in this context (60). YB-1 subcellular localization may also be mediated, in part, by interaction with other proteins. YB-1 is known to play a role in pre-mRNA splicing, however, YB-1 appears to interact directly with splicing factor SRP30C via the A/P domain in an RNA-independent manner. Colocalization of YB-1 and SRP30C has been shown to promote significant nuclear shuttling of YB-1, an effect which is reversed upon sequestering of SRP30C into stress-induced nuclear bodies following heat shock (35). Another study, undertaken in breast cancer cells, highlighted the importance of WAVE3, a member of the WASP/WAVE actin-cytoskeleton remodeling protein family. The WAVE3-YB-1 interaction appears to be facilitate the nuclear translocation of YB-1, which in this context promoted the transcriptional regulation of cancer stem cell (CSC)-specific genes implicated in self-renewal and expansion (61).
Clear evidence exists for YB-1 nuclear translocation in response to cellular stress, for example upon exposure to UV radiation, oxidative stress, hyperthermia, and chemotherapeutics (62,63) and literature cited therein. A 2005 study revealed that wild-type, transcriptionally active P53 is necessary for efficient nuclear translocation of YB-1, likely via the trans-activation of a P53 target "effector" gene (64). Comparatively, a 2006 study suggested that YB-1 accumulates in the nucleus by complexing directly with P53 and WRN, a process thought to facilitate WRN-mediated DNA repair (65). As previously described (see section "Regulation of YB-1 Expression"), proteolytic cleavage between Glu219 and Gly220 by the 20S-proteasome under cellular stress may also trigger YB-1 nuclear localization through the production of a fragmented YB-1 protein lacking a CRS (43,59). However, the proteasomal theory of YB-1 nuclear accumulation is still disputed, largely due to the lack of a suitable YB-1 antibody (see section "YB-1 as a Biomarker"; refs. 45, 66). Finally, RSK1 and RSK2 have been shown to directly phosphorylate YB-1 at S102, with knockdown/drug-mediated inhibition diminishing YB-1 promoter binding (55). Experiments using breast and ovarian cancer cell lines have shown phosphorylation by AKT at S102 promotes nuclear translocation; disruption of which, either by mutation or inhibition, prevents nuclear localization and negatively affects cell growth and the expression of drug resistance genes (56,67). Contrastingly, some groups have found no evidence of AKT-mediated phosphorylation at S102 influencing cellular localization at all, with mutation of the S102 site resulting in unchanged nuclear/cytoplasmic distribution of YB-1 (ref. 68 and studies cited therein). Alternatively, phosphorylation at Y281 within NLS-3 may act dominantly on the subcellular localization of YB-1, promoting nearly exclusive nuclear localization of full-length YB-1 (43). In addition, a recent report proposed that rather than cytoplasmic RSK promoting the phosphorylation and resultant nuclear shuttling of YB-1, nuclearphosphorylated YB-1 (pYB-1) is instead a product of the nuclear translocation of RSK, which then phosphorylates and activates preexisting nuclear YB-1 in response to cellular stress (69).

YB-1 and proliferation
Sustaining proliferative signaling via dysregulation of the cell cycle is arguably the most important hallmark of a tumor cell. Reduction of YB-1 results in diminished proliferation and apoptosis in cancer cell lines and in vivo models, with a number of studies revealing a reduction in cyclin and cyclin-dependent kinase (CDK) level upon YB-1 knockdown (60,67,70,71). In MB, YB-1 appears to be critical for sustaining proliferation both in cells and tissues derived from SHH MB mouse models and in cerebellar granule neuron precursors (CGNP; the proposed cells-of-origin for SHH MB) in vitro and ex vivo (28). Likewise, a study examining the oncogenic functions of YB-1 in pGBM revealed silencing of YB-1 significantly reduced SF188 cell growth in monolayer and soft agar and delayed tumor growth in mice (72).
Several pathways that promote cancer cell proliferation are activated by YB-1, including E2F, PI3K/AKT/MTOR, and RAS/ RAF/MEK/ERK (26). A prime example is the activational role played by YB-1 in the regulation of IGF2 expression in CGNPs and MB cells ( Fig. 4A; ref. 28). The insulin-like growth factor (IGF) signaling pathway has been reported in MB, is required for SHH MB formation and MB proliferation control and has been associated with metastatic progression (14,73). For example, activation of the IGF receptor by IGF1/2 results in downstream activation of PI3K signaling, leading to inhibition of GSK3b, a kinase responsible for blocking cell-cycle progression in CGNPs (28,73). Interestingly, in the aforementioned study, YB-1 was shown to be induced by Sonic Hedgehog (Shh) in CGNPs, demonstrating co-operation between SHH and IGF-mediated PI3K signaling and identifying the SHH: YB-1:IGF2 axis as a powerful target for therapeutic intervention in MB. Furthermore, a 2007 study assessing molecular pathways in pGBM identified at least two disease subsets, one poor prognosis associated with a proliferative phenotype and RAS and AKT pathway activation and one good prognosis without. Of particular interest, the subset with RAS and AKT pathway activation exhibited nuclear YB-1 expression which was associated with elevated EGFR expression, whereas the good prognosis subset exhibited predominantly cytoplasmic YB-1 expression (Fig. 4A). Thus, in pGBM YB-1 may undergo AKT-mediated phosphorylation, resulting in nuclear translocation (Fig. 3) and concordant transcription factor Proposed mechanisms of YB-1 nuclear translocation. Simplified overview of potential YB-1 nuclear transport mechanisms. YB-1 has been suggested to translocate to the nucleus upon phosphorylation by RSK/AKT or conversely is activated within the nucleus by nuclear-translocated RSK. Interaction with different proteins including splicing factor SRP30C or actinremodeling protein WAVE3 may also promote nuclear localization. Involvement of the 20S proteasome has also been proposed, producing either a Cor N-terminal fragment (CTF; NTF) which translocates to the nucleus. Functional P53 protein has also been suggested to promote nuclear translocation, both by direct interaction with YB-1 and by the transcriptional activation of an "effector protein," which then facilitates YB-1 nuclear transport. Dashed lines represent nuclear translocation. functionality while relieving translational repression of numerous pro-mRNAs, hence contributing to increased EGFR levels, RAS activity, and gliomagenesis (19). A similar finding was reported in a lung cancer study, where nuclear YB-1 localization was associated with EGFR expression and poor prognosis (74).

YB-1 and invasion/metastasis
Metastasis is the leading cause of cancer mortality. At the point of diagnosis, up to 40% of patients with MB and 17% of patients with pGBM display clinically detectable metastatic disease (14,75). Metastasis is a complex process requiring a number of events which can vary between cancer types, however common to all are changes in cell-cell and cell-matrix adhesion, cell polarity, and cytoskeletal organization. A number of studies have revealed a clear role for YB-1 in invasion and metastasis. Bioinformatic analysis of a large breast cancer microarray displayed a strong association between YBX1 mRNA expression and distant metastasis formation (76). In vitro studies have demonstrated that YB-1 depletion impedes the invasive capabilities of a number of cancer cell lines (77,78). Likewise, knockdown of YB-1 both stably and transiently in pGBM cell line SF188 significantly reduces cell invasion in transwell invasion assays ( Fig. 4B; ref. 72). The mechanisms underlying this process have not been studied in pGBM, perhaps owing to the fact that the molecular pathways involved in metastasis in pediatric brain tumors remain largely unclear. However, in other cancer types YB-1 has been shown to regulate multiple stages of the metastatic cascade at the level of transcription, pre-mRNA splicing, and translation.
One of the first steps in metastatic dissemination is loss of polarity and cellular adhesion within the primary tumor mass. A key step in the loss of adhesion in epithelial cells is the inactivation of adherens junction protein E-cadherin by epithelial-to-mesenchymal (EMT)inducing transcription factors, such as SNAI1 and TWIST. Although only recently has the EMT process in nonepithelial tumors been considered important for tumor progression, contemporary glioma research has named the EMT, a key player in glioma invasion, with "mesenchymal" a known GBM subtype associated with poor outcome (79). Interestingly, malignant CNS tumors frequently exhibit minimal E-cadherin expression, with one study demonstrating a marked reduction in E-cadherin in adult and pediatric high-grade glioma and MB compared with low-grade gliomas, perhaps indicative of the highly aggressive, malignant nature of these tumors (80). Gainand loss-of-function studies in cervical and breast cancer lines have revealed that YB-1 plays a key role in the translational regulation of proteins implicated in the acquisition of a migratory mesenchymal

YB-1 in brain tumors
NucleoƟde Excision Repair (NER) Proposed oncogenic functions of YB-1 in brain tumors. YB-1 has been shown to facilitate the advancement of several key malignant phenotypes in brain tumors, including proliferation (A), invasion (B), resistance to therapy (C), and the maintenance of BTICs (D). Oncogenic functions presented in black type have been proven experimentally in brain tumor patient samples and cell lines, whereas those in grey type have been investigated in other tumor types but are yet to be explored in brain tumors.

YB-1 in Pediatric Brain Tumors
AACRJournals.org phenotype including SNAI1/2, ZEB2, and TWIST1 (81,82). Correspondingly, overexpression of YB-1 in breast cancer cells is associated with loss of E-Cadherin and tight junction protein ZO-1, increased expression of N-cadherin and vimentin, and the emergence of a mesenchymal phenotype (81). As both SNAI1 and ZEB2 have been associated with invasion, migration, and EMT in pediatric and adult glioma and glioblastoma (Fig. 4B), studying the role of YB-1 in this context in brain tumors may yield interesting results (83,84).
The basement membrane (BM) and extracellular matrix (ECM) represent a substantial physical barrier to the migration of cancer cells. BMs in the CNS include the pial BM, the vascular BM, and the BMs that are associated with Schwann cells. Both MB and pGBM can disseminate through passive shedding into the cerebral spinal fluid (CSF) system. Although the mechanism by which this occurs is not fully understood, in order for invading cells to extend into the CSF, they would have to cross one or more CNS BMs. To do this, multiple proteolysis pathways are likely to become activated. Studies have shown YB-1 to regulate the expression of numerous matrix metalloproteinases. YB-1 knockdown in pancreatic and melanoma cancer cell lines results in significant downregulation of MMP-11, MMP-14, and MMP-2 (77,85). Moreover, in ER-positive breast cancer cells, YB-1 appears to enhance the presentation of membrane type-1 (MT1)-MMP at the sites of cell invasion, namely the cell membrane, where it can degrade the ECM (86). MMPs have been shown to be associated with glioblastoma and MB invasiveness (Fig. 4B). In particular, (MT1)-MMP expression is increased in >75% of MB tumor samples and MMP2 expression is associated with pediatric high-grade gliomas (87,88).
Further, YB-1 induces transcriptional activation of CD44, a nonkinase transmembrane glycoprotein overexpressed in CSCs and implicated in tumor cell invasion and drug resistance (89). YB-1 can also stimulate alternative splicing of CD44 pre-mRNA, resulting in exon v4 inclusion and the production of CD44v4. The CD44v4 variant has been shown to promote both cancer cell invasion and chemoresistance in various cancer cell lines (90,91). Although yet to be researched in specifically pGBM, increased CD44 and CD44 variant expression has been described in GBM (particularly in the mesenchymal subgroup), where expression is associated with poor survival (Fig. 4B). In vitro, the extracellular domain of CD44 appears to facilitate GBM cell invasion through direct interaction with hyaluronic acid and other matrix factors, promoting cell motility through ECM adherence (92).

YB-1 and multidrug resistance
The development and acquisition of multiple drug resistance (MDR) in cancer cells remains a major obstacle in the treatment of metastatic disease. YB-1 knockdown has been shown to increase cellular sensitivity to numerous cytotoxic drugs including cisplatin and etoposide (85,93). Correspondingly, YB-1 knockdown in pGBM SF188 cells, as well as GBM U251 cells renders cells sensitive to both temozolomide and taxol in an O 6 -methylguanine-DNA methyltransferase (MGMT)-independent manner (72). As previously mentioned (see section "Regulation of YB-1 Subcellular Localization"), exposure of various epithelial cancer lines to UV radiation, oxidative stress, hyperthermia, and chemotherapeutics promotes YB-1 nuclear localization (refs. 62 and 63 and literature cited therein). YB-1 nuclear translocation has also been recorded in Group 3 and SHH MB cell lines in response to treatment with cisplatin and vincristine (unpublished observations; Louisa Taylor, Ian Kerr and Beth Coyle). Moreover, in brain tumor-initiating cell (BTIC) lines derived from patients with primary GBM, YB-1 is highly phosphorylated and localized to the nucleus in temozolomide-resistant lines, whereas in lines with low resistance, YB-1 is predominately located in the cytoplasm (94). Such observations indicate that cancer cells may increase nuclear YB-1 expression/promote nuclear YB-1 translocation as a protective measure in response to extracellular stress.
Perhaps the best known mechanism for the development of MDR is the enhanced synthesis of certain members of the ATP-binding cassette family, most notably ABCB1 (P-glycoprotein). ABCB1 has been implicated in the cellular export of a wide range of chemotherapeutics including vinka alkaloids, anthracyclines, taxanes, and protein kinase inhibitors. Of particular note, a significant correlation between ABCB1 expression and high-risk MB has been demonstrated, indicating that overcoming/inhibiting ABCB1 may represent a mechanism to enhance the efficacy of current chemotherapy regimens in MB (95). Indeed, correlation between ABCB1 expression and nuclear YB-1 translocation has been demonstrated in patients with breast cancer treated with paclitaxel and in patients with prostate cancer following neo-adjuvant hormone therapy (62,96). Early studies suggested YB-1 may transcriptionally regulate ABCB1 expression by interacting with a Y-box element present in the ABCB1 promoter region in response to cellular stress ( Fig. 4C; refs. 63, 97). Controversially, other studies have disputed this. A 2016 study using triplenegative breast cancer lines demonstrated that although high YB-1 expression correlates with increased resistance, this occurs in an ABCB1-independent manner (98). Likewise, others have argued that there exists more evidence to support nuclear factor Y (NF-Y) as a Y-box element-binding transcription factor in cancer cells, suggesting a post-transcriptional role for YB-1 in the regulation of ABCB1/other Y-box element containing genes (99).
Another mechanism of resistance to cytotoxic therapy is the activation of DNA repair mechanisms and the evasion of druginduced apoptosis. YB-1 has long been suggested to be part of the DNA repair machinery (Fig. 4C). YB-1 possesses high affinity to cisplatin-modified DNA and DNA containing abasic sites/mismatches and possesses intrinsic exo-and endo-nuclease activity. It also interacts with several key components of base-and nucleotide-excision repair pathways (reviewed in ref. 25). Such observations, combined with observed YB-1 nuclear translocation in response to cytotoxic therapy, supports the theory that YB-1 is involved in the repair of DNA lesions imparted by genotoxic therapy, conferring drug resistance to tumor cells. At the same time, YB-1 has been proposed to regulate key pro-apoptotic and immune response genes including FAS, MHC Class I and II, and notably key tumor suppressor gene TP53. Indeed, YBX1 knockdown in SHH MB cell lines promotes apoptosis through the deregulation of heterochromatin-regulated genes associated with inflammatory response, apoptosis, and death receptor signaling. This occurs through a concurrent reduction in CDX5 expression, a heterochromatin-associated protein regulated posttranscriptionally by YB-1, which represses transcription of apoptosis-related genes through Histone 3 K9 trimethylation (H3K9me3) interaction (29).
Perhaps counterintuitively given the role of P53 in YB-1 nuclear translocation (Regulation of YB-1 Cellular Localization), YB-1 can both repress transcription of TP53 and inhibit the ability of P53 to transactivate cell death genes BAX (Bcl2-associated X protein) and NOXA (NADPH oxidase activator); although interestingly not cellcycle-associated gene CDKN1A (cyclin-dependent kinase inhibitor 1A; refs. 64,70). The mechanism by which this occurs likely requires direct interaction between P53 and YB-1, reducing its affinity for promoter binding and causing promoters with a low binding affinity for P53 (i.e., pro-apoptotic genes) to be disabled (64). More recently, overexpression of YB-1 in GBM cell lines U87 and DK-MG was shown to promote temozolomide resistance through direct interaction with the MDM2/P53 signaling pathway. YB-1 overexpression results in MDM2 activation and the resultant ubiquitination and proteasomal degradation of P53, inhibiting P53-mediated apoptosis ( Fig. 4C;  ref. 100). In support of this, overexpression of MDM2 was recorded in 67% of patients in a pediatric high-grade astrocytoma cohort, with P53 tumor suppressor pathway inactivation, either by single or combinatorial events taking place in >95% cases (101). Furthermore, in vivo studies have highlighted a role played by MDM2 in connecting SHH and P53 pathways in CGNPs, suggesting that MDM2 may be required for SHH MB tumorigenesis (102). Notably, SHH patients harboring germline TP53 mutations represent a very high-risk group in MB, with survival rates of less than 50%; further highlighting the importance of P53 pathway disruption in MB. Taken together, targeting genes implicated in the P53 tumor suppressor pathway, such as YBX1, is likely to be of high therapeutic value for the treatment of pediatric brain tumors.

YB-1 and BTICs
As described previously, frequent relapse represents a serious obstacle to pediatric brain tumor survival. A number of factors are associated with poor responses to therapy and concurrent relapse, including the presence of BTICs, referred to in some studies as brain CSCs. BTICs are multipotent, have the ability to self-renew, form neurospheres and initiate tumor development (103). YB-1 was first associated with TICs in 2010, where it was reported to induce breast TICs to express CD44 and CD49f, leading to enhanced cell growth and drug resistance (89). Both PI3K/AKT and RAS/MAPK pathways are activated in BTICs, in which YB-1 is a downstream phosphorylation substrate. Correspondingly, in BTIC lines derived from patients with primary GBM, both total and pYB-1 was highly elevated compared with normal CNS tissue (94). This is supported by a comprehensive GBM study by Fotovati and colleagues (103). YB-1 was found to colocalize with neural stem cell markers Nestin, SOX2, and BMI-1 in SF188 pGBM cells cultured as neurospheres, as well as patient with GBM derived BTIC isolates (Fig. 4D). YB-1 knockdown was associated with loss of neural stem cell markers, reduced proliferation and differentiation in SF188 neurospheres, whereas forced differentiation of primary BTICs resulted in loss of YB-1 expression. Importantly, high and co-ordinated expression of YB-1, SOX2, and BMI-1 was detected in 67% of GBM cases, which subsequently went on to relapse (103). Consistent with these observations, a recent study which employed CRISPR/Cas9 to knockout YB-1 in melanoma and breast TICs revealed that that YB-1 maintains the stemness of TICs by promoting the expression of stemness-related genes FZP-1, GLP-1, GINS1, and NOTCH2 (104). Collectively, these studies imply that targeting YB-1 will stimulate BTICs to undergo differentiation and suppress their proliferative capacity and hence further supports the development of YB-1 targeted therapies as a novel approach in the management of aggressive pGBM.

YB-1 as a biomarker
There is evidence to support the use of YB-1 as a prognostic biomarker in pediatric brain tumors. Analysis of patient data demonstrates that YBX1 mRNA is upregulated in MB and high-grade pediatric glioma where it correlates with poor survival outcome ( Fig. 1A-C; refs. 19,28,29). Indeed, a number of studies have proposed YB-1 as a novel biomarker of glioma progression. In patient with GBM samples, YB-1 protein expression has been shown to increase with tumor stage (103). Likewise, in a larger study, YB-1 protein/mRNA levels were shown to differ significantly between tumor grades, with Grade I/II tumors presenting with mainly cytoplasmic staining and Grade III/IV presenting with abundant nuclear and cytoplasmic staining, which was found correlate with poor overall survival (105). Of particular note, YB-1 was significantly elevated in the CSF of Grade III/IV patients compared with that of Grade I/II patients, indicating that YB-1 may also represent a promising CSF marker for distinguishing malignant gliomas (105). The aforementioned studies demonstrated that overall YB-1 overexpression is associated with poor prognosis, however it may also be important to consider YB-1 subcellular localization when determining the prognostic value of YB-1 expression. For example, although YB-1 was found to be upregulated in 86% patient samples in a pGBM cohort, high nuclear expression was associated with an AKT-active poor prognosis subgroup, whereas samples exhibiting predominantly cytoplasmic expression were associated with a better prognosis, AKTinactive subgroup (19).
A potential barrier to the use of YB-1 as a prognostic marker in pediatric brain tumors, or indeed in other cancer types in the clinic, has been the variability in available antibodies. As previously reported, there exists controversy surrounding the function of proteolysis in the regulation of YB-1 nuclear translocation (See Regulation of Subcellular Localization), arising from cross-reaction between N-terminal YB-1 antibodies and another protein called heterogeneous nuclear ribonucleoprotein A1 (hnRNPA1; ref. 66). The hnRNPA1 protein is present in both nuclear and cytoplasmic compartments and migrates at $37 kDa (YB-1 migrates at $49 kDa), meaning immunostaining products produced by such cross-reaction are very difficult to interpret (45,66). Despite this issue being raised by research and review articles alike, one commercially available N-terminal cross-reactive antibody (abcam, ab12148) continues to be used, with nine references in 2019 alone (26,45,54,66). Further to this, disparities have also been demonstrated in the ability of YB-1 antibodies to detect nuclear YB-1 protein in addition to cytoplasmic YB-1, a finding that may prove important if the quantification of nuclear expression specifically is required to assess prognosis (45). Standardization of one properly validated antibody is of fundamental importance to develop effective prognostic screens using YB-1 as a biomarker.

YB-1 as a therapeutic target
Given the multifunctional nature of YB-1 and its position upstream of numerous tumor-promoting molecular pathways, YB-1 represents an attractive therapeutic target. One potential therapeutic approach to directly target YB-1 is through oligonucleotide-based methods including siRNA and miRNA. YB-1 knockdown by siRNA has proven effective in vitro, resulting in reduced proliferative and migratory capability and increased drug sensitivity in pediatric brain tumor cells (72,103). However, the delivery of siRNA to tumor cells in the human body, especially those metastasized, represents a major challenge to using such molecules in the clinic. Unmodified siRNA is rapidly degraded in the bloodstream, does not readily enter cells and can be immunogenic. The blood-brain barrier (BBB), comprised of a tight arrangement of endothelial cells, represents a further challenge for efficient siRNA delivery. There are a number of systems which may represent viable options to deliver oligonucleotide-based therapy to brain tumors, including lipid-based, inorganic and polymeric nanoparticles. Poly(b-amino ester) (PBAE)-based nanoparticles have successfully been used to deliver and efficiently release siRNA in a patientderived GBM mouse model, successfully knocking down a reporter gene within GBM cells without systemic toxicity (106). Dual-modified cationic liposome nanoparticles incorporating CD133-targeting ligands and encompassing paclitaxel and survivin siRNA have also proven successful in in vitro and in vivo glioma studies, displaying high specificity for CD133 þ glioma stem cells and low toxicity to brain endothelial cells (107). As an alternative to larger nanoparticles, which frequently accumulate in the reticuloendothelial system of the liver and spleen, a recent study proposed an alternative oligonucleotide stabilization approach using Y-shaped block catiomers (YBC). Notably, the number of positive charges in the YBC can be adjusted to match that of negative charges in each oligonucleotide strand, facilitating selective paring in the bloodstream. The resultant complex is stable in the blood stream and incredibly small ($18 nm), allowing efficient delivery in an in vivo patient-derived GBM model (108). Although further preclinical and clinical trials are required, such studies show oligonucleotidebased therapy, both by nanoparticles and alternative approaches, to be a promising brain tumor treatment strategy.
A number of groups have proposed that the transcription factor functionality of YB-1 is activated by phosphorylation. For this reason, disruption of YB-1 phosphorylation may represent another approach to targeting YB-1 activity. The most widely characterized phosphorylation site in YB-1 is S102 (Fig. 2). Blocking the S102 site by way of a decoy cell permeable peptide led to inhibition of EGFR expression and reduced growth in prostate and breast cancer cell lines, without affecting non-malignant cells (109). Upstream inhibitors targeting RSK and AKT may also be an option. Fisetin is a dietary flavonoid which is able to bind directly with RSK, both suppressing RSK kinase ability and promoting interaction between fisetin and the YB-1 CSD. Together, this led to a decrease in YB-1 phosphorylation and down regulation of total YB-1 protein, with a concurrent decrease in MMP-2, MMP-9, and ABCB1 expression and reduced cell viability in a melanoma model (110). However, it must be noted that some groups have found no evidence of YB-1 phosphorylation influencing nuclear translocation (see section "Regulation of Subcellular Localization"). Further study of the validity of the phosphorylation theory of YB-1 nuclear localization and as well as in which cancer types/cell types this theory holds true will be required prior to the development of this therapeutic approach.
Over recent years, virotherapy has emerged as an alternative treatment for cancer. The recent FDA approval of oncolytic herpes virus T-VEC (Talimogene laherparepvec) for the treatment of metastatic melanoma has confirmed the possibility of using viruses in the clinic. As such, a promising novel treatment approach for brain tumor therapy is YB-1 targeted virotherapy. YB-1 is known to play an important role in the adenovirus life cycle, where post-adenovirus infection, YB-1 can translocate to the nucleus where it regulates the expression of viral polymerase (111). This finding raised the possibility of targeting YB-1 nuclear accumulation with YB-1 dependent oncolytic adenoviruses. Ad-Delo3-RGD is a recombinant adenovirus in which the transactivation domain CR3 of the E1A protein is ablated to enable viral replication solely in YB-1 positive cancer cells. The Ad-Delo3-RGD virus induced significant cell lysis in various GBM cell lines (112). Of particular note, due to the YB-1 nuclear translocation induced by certain cytotoxic drugs, co-treatment with cisplatin and temozolomide significantly enhanced tumor cell killing in vitro and reduced tumor growth rate in a xenograft glioma mouse model (112). Likewise Ad-Delo3-RGD mediated substantial cytolysis in GBM patient-derived BTIC lines, with significantly reduced viral replication in nonmalignant astrocyte cells (94). Although Ad-Delo3-RGD is yet to be tested in pediatric brain tumor lines/models, other oncolytic viruses have yielded promising results in pediatric high-grade glioma, supporting the development of YB-1-targeting oncolytic viruses for patients with these tumors (113).

Conclusion
To conclude, this review has described recent advances in our understanding of YB-1 as a master regulator of cancer biology, with a specific focus on pediatric brain tumors. On account of its multifunctional nature, YB-1 appears to facilitate the advancement of numerous malignant phenotypes in brain tumors, including proliferation, invasion, and resistance to therapy and is a key player in the maintenance of brain tumor-initiating cells. It is also highly expressed in pediatric brain tumors and correlates with poor survival outcome. Taken together, YB-1 clearly represents a protein worthy of further research within the pediatric oncology field. Further, with the advent of YB-1 targeted therapeutic approaches such as YB-1-targeted virotherapy and oligonucleotide therapy, YB-1 has extensive potential for development as both a therapeutic target and novel biomarker for the management of pediatric brain tumors.