Discovery, characterization and potential roles of a novel NF-YAx splice variant in human neuroblastoma

Background Identification of novel cancer-associated splice variants is of potential diagnostic, prognostic and therapeutic importance. NF-Y transcription factor is comprised of NF-YA, NF-YB and NF-YC subunits, binds inverted CCAAT-boxes in ≈70% of gene promoters, regulates > 1000 cancer-associated genes and proteins involved in proliferation, staminality, differentiation, apoptosis, metabolism and is subject to component alternative splicing. RT-PCR evaluation of alternative NF-YA splicing in primary human neuroblastomas (NBs), led to discovery of a novel NF-YAx splice variant, also expressed during mouse embryo development and induced by doxorubicin in NB cells. Here, we report the discovery and characterisation of NF-YAx and discus its potential roles in NB. Methods NF-YAx cDNA was RT-PCR-cloned from a stage 3 NB (provided by the Italian Association of Haematology and Paediatric Oncology, Genova, IT), sequenced and expressed as a protein using standard methods and compared to known fully-spliced NF-YAl and exon B-skipped NF-YAs isoforms in: EMSAs for capacity to form NF-Y complexes; by co-transfection, co-immunoprecipitation and Western blotting for capacity to bind Sp1; by IF for localisation; in AO/EtBr cell-death and colony formation assays for relative cytotoxicity, and by siRNA knockdown, use of inhibitors and Western blotting for potential mechanisms of action. Stable SH-SY5Y transfectants of all three NF-YA isoforms were also propagated and compared by RT-PCR and Western blotting for differences in cell-death and stem cell (SC)-associated gene expression, in cell-death assays for sensitivity to doxorubicin and in in vitro proliferation, substrate-independent growth and in vivo tumour xenograft assays for differences in growth and tumourigenic capacity. Results NF-YAx was characterized as a novel variant with NF-YA exons B, D and partial F skipping, detected in 20% of NF-YA positive NBs, was the exclusive isoform in a stage 3 NB, expressed in mouse stage E11.5–14 embryos and induced by doxorubicin in SH-SY5Y NB cells. The NF-YAx protein exhibited nuclear localisation, competed with other isoforms in CCAAT box-binding NF-Y complexes but, in contrast to other isoforms, did not bind Sp1. NF-YAx expression in neural-related progenitor and NB cells repressed Bmi1 expression, induced KIF1Bβ expression and promoted KIF1Bβ-dependent necroptosis but in NB cells also selected tumourigenic, doxorubicin-resistant, CSC-like sub-populations, resistant to NF-YAx cytotoxicity. Conclusions The discovery of NF-YAx in NBs, its expression in mouse embryos and induction by doxorubicin in NB cells, unveils a novel NF-YA splice mechanism and variant, regulated by and involved in development, genotoxic-stress and NB. NF-YAx substitution of other isoforms in NF-Y complexes and loss of capacity to bind Sp1, characterises this novel isoform as a functional modifier of NF-Y and its promotion of KIF1Bβ-dependent neural-lineage progenitor and NB cell necroptosis, association with doxorubicin-induced necroptosis and expression in mouse embryos coinciding with KIF1Bβ-dependent sympathetic neuroblast-culling, confirm a cytotoxic function and potential role in suppressing NB initiation. On the other hand, the in vitro selection of CSC-like NB subpopulations resistant to NF-YAx cytotoxicity not only helps to explain high-level exclusive NF-YAx expression in a stage 3 NB but also supports a role for NF-YAx in disease progression and identifies a potential doxorubicin-inducible mechanism for post-therapeutic relapse.


Background
Alternative gene splicing is a fundamental physiological mechanism for the differential expression of proteins from the same gene coding sequence and is largely responsible for the increased proteomic complexity of higher organisms that cannot be explained by differences in individual gene numbers alone [1]. Aberrant alternative splicing has been reported in cancer and is promoted by stress within the tumour microenvironment, oncogenic viral infection, gene translocation and oncogene de-regulation of splice factor expression. Cancerassociated alternative splicing has been shown to inactivate onco-suppressors and activate oncogenes, making the identification of novel cancer-associated splice isoforms of potential diagnostic, prognostic and therapeutic importance [1][2][3][4][5].
Alternative splicing of the NF-YA gene has been implicated in the regulation of cell staminality, differentiation, apoptosis and transformation. NF-YAs forms part of the stem cell (SC) transcriptional circuitry, predominates in embryonic SCs and is lost upon SC differentiation. In contrast, NF-YAl promotes differentiation and loss of NF-YA expression induces senescence or apoptosis. Alternative NF-YAs splicing is promoted by the oncogenic polyomavirus SV40 and by v-ras oncogene and converts tumorsuppressing, differentiation-promoting NF-Y complexes predominated by NF-YAl into tumor and CSC promoting complexes predominated by NF-YAs [8,[18][19][20][21][22][23].
Neuroblastomas (NB) are aggressive embryonic tumours of neural crest origin, derived from immature sympathetic neuroblasts [24]. These primitive tumours initiate under conditions that impair sympathetic neuroblast culling during development, reported to depend upon either loss of the KIF1B gene associated with chromosome 1p36deletion, germline KIF1B mutations or Nmyc amplification [25][26][27][28][29][30][31][32][33]. NF-Y involvement in NB pathogenesis and progression, however, has received scant attention. In the few existing reports, NF-Y has been shown to be critical for expression of soluble guanyl cyclase in NB cells required for cGMP production and differentiation [34] and is involved in elevated glypican 3 expression in NBs [35]. NF-Y and Sp1 transcription factors combine to promote tetramethylpyrazine-induced neuronal differentiation of NB cells [36] and regulate expression of the α3 Na+, K + -ATPase subunit, essential for maintaining electrochemical gradients across cell membranes [37]. Suboptimal NF-Y function in NB cells has also been implicated in de-regulating the matrix metalloproteinase and tissue inhibitor of metalloproteinase equilibrium, resulting in invasion [38] and increased expression of the NF-YA subunit has been reported to differentiate between aggressive stage 4 NBs and stage 4S NBs that exhibit spontaneous regression [39].
Considering the relative absence of studies of NF-Y expression in NB, combined with reports associating fully spliced NF-YAl with cellular differentiation and reduced malignancy and associating alternative exon B spliced NF-YAs with cellular staminality and increased malignancy [18][19][20][21][22][23], we initiated a study of NF-YAl and NF-YAs expression in human primary NBs. This led to the unexpected discovery of a novel NF-YAx splice variant, with NF-Y functional modifying activity, which forms the subject of this report.

Aim, design and setting
The aim of this study was to report the discovery and characterization of the novel alternative NF-YAx splice variant, discovered as the exclusive NF-YA isoform expressed in an advanced stage 3 NB, expressed during mouse embryo development and induced in NB cells by doxorubicin, and to provide insights into its potential function in NB pathogenesis and progression. Experimental design included cloning and sequence characterization, expression vector construction, protein expression and characterization, transient and stable transfection and biological characterizations in terms of growth, cytotoxicity and tumorigenicity.

Transient transfections
Cells at 1 × 10 5 /ml were grown overnight to ≈80% confluence on 6 or 24 well plates, transfected with NF-YA variant or empty plasmid DNA (1 μg/ml) in Fugene HD, as directed (Promega, Madison, WI), washed at 6 h, grown in complete medium, monitored at 6, 12, 24 and 48 h for evidence of cytotoxicity and photographed. At 48 h, suspension and adherent cell populations were separated and either analysed in AO/EtBr cell-death assay, used for RNA purification and RT-PCR analysis or extracted for Western blotting. Transfection efficiency was estimated using pEGFP-N1 green fluorescent protein reporter plasmid (Clontech, Mountain View, CA).

Direct RT-PCR sequencing
Agarose gel-purified RT-PCR product sequencing was performed using BigDye Direct Cycle Sequencing kit, as directed (Thermo-Fisher Scientific, Waltham, MA). Briefly, cDNAs (final concentration 4 ng/ml) were added to reaction mixtures containing forward or reverse primer (0.8 μM), BigDye direct PCR master mix, deionized water in a final volume of 20 μl and subjected to 35 PCR cycles (94°C and for 10 min and 96°C for 10 s, 55°C for 5 min and 60°C for 4 min). PCR reactions were mixed with 2 ul of NaAc and 50 μl absolute ethanol, incubated for 10 mins, spin-dried at 120000 rpm for 20 mins and sequenced in a mono-capillary sequencer, as directed (ABI PRISM 310, Life Technologies, Monza, IT).

Tumor growth in soft agar
Single-cell suspensions (passed through a gauge × 18 syringe needle) of 5 × 10 4 cells were mixed in a 33% solution of agar (BiTec; Difco) in RPMI containing 5% FCS at 37°C and layered onto a solid 0.6% agarose substrate prepared in the same growth medium. Following agar solidification, complete medium was added, replaced

Tumor growth in NGS mice
Tumorigenesis in vivo was performed as previously described [3]. Stable-transfected SH-SY5Y cells, prepared as a single-cell suspension in PBS without Ca 2+ and Mg 2+ , were injected subcutaneously into the flanks of anesthetized, 6-week-old female NGS mice (Charles River, Calco, Italy), at a concentration of 1 × 10 7 cells in 200 μl per site. Tumor initiation was recorded at a minimum volume (tumor length × [tumor width] 2 × 0.44) of 12 mm 3 and animals sacrificed at 21 days. Groups consisted of 5 animals and the assay performed once (n = 5). All experiments were performed in accordance with Italian national and Rome University guidelines.

Cell death assays
Cell death assays were as previously described [43]. Suspension or adherent cells, detached in ice cold PBS containing 1 mM EDTA, were transferred into sterile tubes, pelleted at 1000 x g at 4°C, washed in ice cold PBS, re-pelleted, re-suspended in 25 μl of PBS containing 2 μl of acridine orange/ethidium bromide solution (100 μg/ml acridine orange and 100 μg/ml ethidium bromide in PBS), mounted onto glass slides, examined immediately under a Zeiss "Axioplan-2" fluorescence microscope, digitally photographed and dead (orange/red nuclei) and live cells (green nuclei) counted. Assays were performed in duplicate and repeated 3 times (n = 6).

Indirect immunofluorescence
Cells grown on Nunc glass chamber slides (Sigma-Aldrich, St Louis, MI) were washed in PBS, fixed and permeabilized in 100% ice cold methanol (− 20°C), incubated for 1 h in blocking solution (1% bovine serum albumin in PBS-0.03% TX100), incubated for 2 h with primary antibody diluted in blocking solution at room temperature, washed, incubated with secondary fluorochrome-conjugated antibody diluted in blocking solution, for 1 h at room temperature, mounted with VectaMount (Vector Laboratories, Berlingame, CA) and observed using a Zeiss Axioplan 2 fluorescence microscope with digital camera and Leica M500 Image Manager software.

Results
This study, originally designed to evaluate fully-spliced NF-YAl and alternatively exon B spliced NF-YAs mRNA expression in primary human NBs, resulted in the unexpected discovery of a novel NF-YAx splice variant.  . 1a) [15].
The 338 bp product, characterized as NF-YAx by direct PCR sequencing (not shown), was the exclusive isoform expressed in a stage 3 NB and a minor isoform, together with NF-YAl and NF-YAs, in two stage 2 NBs but was not detected in either stage 1 or 4 NBs. RNAs from SK-N-MC, SK-N-SH, SH-SY5Y, KCNR, IMR-32 and SK-N-BE NB cell lines, HEK293 embryonic neuronal-lineage kidney cells and human neonatal brain stem cells (SCs) exhibited predominant NF-YAs but not constitutive NF-YAx expression; CHP-126, LAN-1 and LAN-5 NB cell lines exhibited equivalent NF-YAs and NF-YAl but not NF-YAx expression and SHEP NB cells exhibited predominant NF-YAl but not NF-YAx expression. Predominant NF-YAs with minor NF-YAx expression characterized ST14A embryonic striatal neuronal progenitors under progenitor-maintaining (33°C) and differentiation-inducing (39°C) conditions [40] ( Fig. 1c and d). Western blots confirmed predominant NF-YAs protein expression in SH-SY5Y cells, predominant NF-YAl expression in SHEP cell extracts (Fig. 1e). NF-YA isoforms translated in vitro were characterized by SDS-PAGE with approximate molecular masses of 46 kDa (NF-YAl), 42 kDa (NF-YAs) and 35 kDa (NF-YAx) (Fig. 1f).
Samples were not available for NF-YAx protein analysis in primary NBs.

NF-YAx selects tumorigenic, doxorubicin-resistant cancer SCs
Duplicate stable SH-SY5Y NF-YAx-transfectants were established and compared to stable control, NF-YAl and NF-YAs-transfectants. Stable NF-YAl, NF-YAs, NF-YAx1 and x2-transfectants expressed similar levels of NF-YAl, NF-YAs and NF-YAx (Fig. 3c) and did not differ significantly in either proliferation or mitotic rates, assessed by MTS and thymidine-incorporation assays (Fig. 7a). NF-YAl, NF-YAs and NF-YAx stable transfectants all formed neuro-spheres in neural stem cell assays in vitro (Fig. 7b) and similar numbers of similar sized tumor spheroids in soft-agar tumorigenesis assays in vitro (Fig. 7c). In xenograft tumorigenesis assays in NSG mice, pcDNA, NF-YAs, NF-YAx1 and x2-transfectants formed sub-cutaneous tumours of similar dimensions, significantly larger than tumours formed by stable NF-YAl -transfectants (> 2 fold smaller, p < 0.049 compared to all other stabletransfectants, n = 5 per group) (Fig. 7d). In cell death assays, doxorubicin induced similar levels of stable pcDNA, NF-YAl and NF-YAs-transfectant cell death at all concentrations at 6, 24 and 48 h. In contrast, stable NF-YAx1 and x2-transfectants exhibited significantly enhanced survival at 6 but not 24 or 48 h (p < 0.0001, n = 6 for both) in the presence of 10 μM doxorubicin, at 6 and 24 but not 48 h (p < 0.0001, n = 6 for both) in the presence of 5 μM doxorubicin and at all time points in the presence of 0.01, 0.1 and 1 μM doxorubicin (p < 0.0001, n = 6 for both at all time points and doses) (Fig. 7e).
In RT-PCR assays NF-YAx1 and x2 transfectants exhibited higher levels of p75 NTR , Nanog, Nestin and EglN3 expression than NF-YAl and NF-YAs-transfectants, similar levels of Sox-2, CD133 and CD117 expression to NF-YAs-transfectants, elevated above control and NF-YAl-transfectants, high-level Bmi1 expression but did not express KIF1Bβ. In contrast, NF-YAl-transfectants expressed lower levels of p75 NTR and Nanog than the other transfectants. Bcl2, Mcl1, Bcl-xL, PUMA, BAD and Bax expression levels did not differ between stable transfectants (Fig. 8a, b and c). RNAs purified from xenograft tumours exhibited a similar pattern of CSC gene expression to corresponding cell cultures (Fig.  8d). PTC-209 (10 μM for 24 h) abrogated proliferation of all stable transfectants (Fig. 9a) but did not reduce Bmi1 expression nor induce KIF1Bβ mRNA expression in NF-YAx1 or x2-transfectants (Fig. 9b).

Discussion
We report a novel development and genotoxic stressregulated alternative splice mechanism for promoting embryonic neural-lineage progenitor and NB cell-death, characterized by a switch to alternative NF-YA splicing and expression of a novel cytotoxic NF-YAx extra shortform variant. This novel isoform, originally discovered in human primary stage 2 and stage 3 NB RNAs, was the exclusive NF-YA isoform expressed at a high-level in an advanced stage 3 NB and was characterised as a novel NF-YA splice variant, exhibiting in-frame exon B, D and partial F skipping, responsible for truncating NF-YA transactivation domain sequence. NF-YAx readily competed with fully-spliced NF-YAl in CCAAT-box binding NF-Y complex formation but in contrast to NF-YAl and NF-YAs isoforms did not bind Sp1 and, therefore, represents a functional modifier of one of the more important physiological and cancer-associated transcription factors. In mouse embryos, NF-YAx expression coincided with the reported phase of neurotrophin-regulated KIF1Bβ-dependent sympathetic neuroblast-culling, unrestrained NF-YAx expression induced KIF1Bβ-dependent necroptosis in neurallineage progenitors and NB cells and association between doxorubicin-induced NF-YAx expression and necroptosis in NB cells, supports a pro-necroptotic cytotoxic function for NF-YAx and a potential role in KIF1Bβ-dependent suppression of NB initiation, during development. On the other hand, propagation through selection of tumorigenic, doxorubicin-resistant CSC-like stable NF-YAx SH-SY5Y transfectants, resistant to NF-YAx cytotoxicity, not only helps to explain the high-level exclusive NF-YAx expression detected in an advanced stage 3 NB but also supports an additional potential role for NF-YAx, within the tumour context, in disease progression and identifies a potential mechanism for doxorubicin-induced post-therapeutic relapse, through CSC selection. NF-YAx cDNA was cloned from a stage 3 NB and sequence characterised as a novel NF-YA splice variant with in-frame exons B, D and partial F sequence skipping, adding to existing NF-YAl, NF-YAs and NF-YA L2-6 variants [8,12,13,17], with potential implications for NB pathogenesis and progression. NF-YAx expression, detected in 20% (3/15) NF-YA-positive NB RNA samples, was the exclusive high-level NF-YA mRNA isoform expressed in a stage 3 NB but was not detected in stage 1 and 4 NBs, human neonatal neural stem cells, HEK-293 human neurallineage embryonic kidney cells and 14 human NB cell lines. Although samples were not available for NF-YAx protein detection in primary NBs, expression of the endogenous NF-YAx protein was confirmed in SH-SY5Y cells, following treatment with doxorubicin, corroborating the characterisation of NF-YAx cDNA as a complete, inframe, non-mutated NF-YA splice variant that is readily translated into the NF-YAx protein in vitro and in vivo.
Considering the relatively small number of NB samples analysed in this study, however, the possibility that NF-YAx mRNA expression is restricted to localized NB disseminated at most to local lymph nodes (Stages 2 and 3) [47] and may differentiate stage 3 from other disease stages, must await confirmation in a future larger NB cohort study.
NF-YAx was significantly more cytotoxic to neuralrelated progenitor and NB cells than either NF-YAl or NF-YAs, contrasting with a previous report that unrestrained NF-YAl expression is highly cytotoxic and induces p53-dependent apoptosis [9]. This discrepancy can be explained by fact that p53 is compromised by SV40 large T-antigen in ST14A cells [40], by adenovirus-5 in HEK-293 cells [41,42] and by different mechanisms in SH-SY5Y cells [67][68][69], suggesting that this novel NF-YAx-dependent necroptosis mechanism may be restricted to genotoxic-stress under p53 compromised conditions. Furthermore, the sensitivity of ST14A neural progenitors, which exhibit low-level constitutive NF-YAx expression, to NF-YAx-induced necroptosis, confirms that this mechanism depends upon predominant NF-YAx expression. Unrestrained NF-YAdn expression, which contains a DNA bindingdomain mutation that prevents NF-Y binding and transcription [70], also induced HEK-293 and SH-SY5Y necroptosis. This indicates that survival can switch to necroptosis in these cell types when either NF-Y binding is prevented (i.e. with NF-YAdn) or when a change in NF-Y function reaches a critical level (i.e. with NF-YAx ). Similar necroptotic-like neuronal death has also been reported during development [25,48,54,62,71,72]. A pro-necroptotic role for NF-YAx was also supported by the association between doxorubicin-induced NF-YAx expression and SH-SY5Y cell-death, which was also characterised by vacuolation, swelling and cell lysis and significantly inhibited by the necroptosis inhibitor necrostatin-1 [46], consistent with a percentage of doxorubicininduced necroptosis. Confirmation of the role of NF-YAx in this necroptotic proportion of doxorubicininduced death, however, must await development of reagents specific for NF-YAx depletion.
Doxorubicin induction of alternative NF-YAx splicing not only implicates the DNA damage-associated alternative splice mechanism [73] and a role for NF-YAx in the response to genotoxic stress but also identifies NF-YAx as a potential biomarker of response to genotoxic therapy, suggesting that NF-YAx expression in primary NBs could also reflect neo-(See figure on previous page.) Fig. 7 Stable NF-YAx transfectants are tumorigenic and doxorubicin-resistant. a Line graphs demonstrating similar proliferation rates in MTS (left) and 3 H-thymidine incorporation assays by stable pcDNA, NF-YAl, NF-YAs, NF-YAx1 and x2 SH-SY5Y-transfectants, displayed as mean (±SD) absorbance at 429 nm for MTS assays and cpm per cell in 3 H-thymidine incorporation assays, in three independent assays performed in duplicate. b Micrographs demonstrating similar neuro-sphere growth by stable pcDNA, NF-YAl, NF-YAs, NF-YAx1 and x2-transfectants in neural stem cell assays (bar = 1 mm). c) Micrographs demonstrating similar soft agar spheroid-growth in vitro by stable pcDNA, NF-YAl, NF-YAs, NF-YAx1 and x2transfectants (bar = 1 mm). d Subcutaneous xenograft tumours formed by stable pcDNA, NF-YAl, NF-YAs, NF-YAx1 and x2-transfectants in NGS mice (bar = 1 cm), plus histogram demonstrating mean (± SD) tumor volumes (mm 3  adjuvant genotoxic chemotherapy [74]. Furthermore, the propagation of tumorigenic, doxorubicin-resistant, CSC-like stable NF-YAx SH-SY5Y transfectants, resistant to NF-YAx cytotoxicity, unveils an additional potential role for NF-YAx in genotoxic drug-induced, post-therapeutic relapse through CSC selection and maintenance. Moreover, the fact that NF-YAx expression was not induced by agents that promote ERstress, hypoxic-stress, differentiation or malignant behaviour, suggests that this alternative splice mechanism may be relatively restricted to conditions of genotoxic stress in NB cells. NF-YAx substitution of NF-YAl in DNA-binding NF-Y complexes and the capacity of NF-YAl and NF-YAs but not NF-YAx to bind Sp1 (this study, [11,44,75]), characterises NF-YAx as a potential modifier of both NF-Y and NF-Y-dependent Sp1 function. NF-YAx loss of Sp1 binding results from truncation of the aa 55-139 binding site for Sp1 in NF-YAl and NF-YAs to aa 55-103 in NF-YAx (this study, [44,76]). This truncation may also compromise NF-Y interaction with ZHX transcriptional repressors that bind NF-YA aa's 31-140, with potential to also alter NF-Y/ZHX-regulated genes expression, including MDR-1 chemotherapeuticcytotoxicity-regulator and polo kinase-1 mitosis-regulator [77][78][79][80][81]. Transactivation-domain truncation may also weaken NF-Y function and interaction with other factors and the reduced size of NF-YAx, particularly in complexes with smaller NF-YB and NF-YC variants [82], may de-regulate transcription from promoters with precisely-spaced NF-Y-dependent transcriptional domains, e.g. ER-stress response gene promoters [83]. Although, NF-YAx-specific antibodies are not yet available for confirmation, by chromatin immunoprecipitation assay, that NF-YAx is recruited to gene promoters in vivo, the fact that NF-YAx contains an intact DNA binding domain, competes with other isoforms in NF-Y complex formation and binds double stranded inverted CCAAT-box oligonucleotides in NF-Y complexed-form in vitro with similar kinetics to other NF-YA isoforms, strongly supports this probability. Although, we are only beginning to understand how NF-YAx may influence transcription, transient NF-YAx expression reduced Bmi1 and induced KIF1Bβ expression in neural-lineage progenitors and NB cells, and stable NF-YAx expression was associated with enhanced p75 NTR , SOX2, Nestin, Nanog, CD117, CD133 and EglN3 expression.

Conclusions
The discovery of NF-YAx mRNA in primary stage 2 and 3 NBs, its expression as the exclusive isoform in an advanced stage 3 NB, expression in stage E12.5 to E14.5 mouse embryos and induction by doxorubicin in NB cells, unveils a novel NF-YA splice mechanism and variant that is both regulated by and involved in NB, development and conditions of genotoxic-stress. NF-YAx substitution of other isoforms in NF-Y complexes and loss of Sp1 binding capacity characterises this novel isoform as a functional modifier of one of the more important physiological and cancer-associated transcription factors. NF-YAx induction of KIF1Bβdependent embryonic neural-related progenitor and NB necroptosis, association with doxorubicin-induced necroptosis and expression in murine embryos at times corresponding to the phase of neurotrophinregulated KIF1Bβ-dependent sympathetic neuroblast culling, supports a predominant pro-necroptotic cytotoxic function for NF-YAx in neural progenitors and NB cells, with potential to suppress NB initiation during development. On the other hand, propagation by selection of tumorigenic, doxorubicin-resistant CSClike stable NF-YAx expressing SH-SY5Y transfectants, resistant to NF-YAx cytotoxicity, not only helps to explain the exclusive high-level NF-YAx expression in an advanced stage 3 primary NB but also suggests that NF-YAx expression within the tumour context may promote disease progression and provides a possible doxorubicin-inducible mechanism for post-therapeutic relapse, through CSC selection and maintenance, with potential to enhance survival within stressful chemotherapeutic tumour microenvironments [62,103].