Pediatric cancers often resemble trapped developmental intermediate states that fail to engage the normal differentiation program, typified by high-risk neuroblastoma arising from the developing sympathetic nervous system. Neuroblastoma cells resemble arrested neuroblasts trapped by a stable but aberrant epigenetic program controlled by sustained expression of a core transcriptional circuit of developmental regulators in conjunction with elevated MYCN or MYC (MYC). The transcription factor ASCL1 is a key master regulator in neuroblastoma and has oncogenic and tumor-suppressive activities in several other tumor types. Using functional mutational approaches, we find that preventing CDK-dependent phosphorylation of ASCL1 in neuroblastoma cells drives coordinated suppression of the MYC-driven core circuit supporting neuroblast identity and proliferation, while simultaneously activating an enduring gene program driving mitotic exit and neuronal differentiation.
These findings indicate that targeting phosphorylation of ASCL1 may offer a new approach to development of differentiation therapies in neuroblastoma.
Neuroblastoma tumor cells resemble malignant neuroblastic developmental intermediates trapped epigenetically in a precursor-like state (1). In common with other childhood tumors that arise from aberrant proliferation of developmental intermediates, reactivation of a latent developmental program of mitotic exit and differentiation may provide a promising therapeutic approach (2); for instance, reactivation of differentiation in the pediatric malignancy acute promyelocytic leukemia has transformed the outcome of this disease from a frequently fatal disease to one with a 90% cure rate (3). Directed reactivation of tumor cell differentiation requires an understanding of the molecular mechanisms that cause stalling of normal development and an ability to reverse that stalling.
In neuroblastoma, recent studies have supported a model of aberrant epigenetic control that locks cells in a progenitor-like state via a network of super-enhancers; genes associated with an early developmental and proliferative program are highly expressed because of associated and mutually regulated oncogenic super-enhancers (4–8). High level of expression of these genes is maintained by elevated super-enhancer–associated MYCN or c-MYC particularly in disease with poor prognosis. Although collapse of this super-enhancer network by targeting transcriptional CDKs can lead to neuroblastoma cell death (9), reactivating the normal developmental processes in neuroblastoma cells is likely to require a more directed approach and modulation of the activity of transcriptional regulators that usually control these processes during development. One such regulator is ASCL1, a bHLH proneural transcription factor that plays a central role in noradrenergic neuron differentiation.
ASCL1 regulates both neural stem-cell maintenance and differentiation during embryonic development (10), whereas its expression can be associated with cancer “stem-ness” in glioblastoma (11, 12), small-cell lung carcinoma (13), medulloblastoma (14), and neuroblastoma (15–17). Conversely in glioblastoma, enhancing ASCL1 activity by Notch inhibition is sufficient to promote cell-cycle exit and attenuate tumorigenicity (12). Therefore, evidence from the developing and adult central nervous system as well as glioblastoma indicates that ASCL1 levels, and therefore activity, may be critical for determining whether ASCL1 supports stem/progenitor maintenance or drives differentiation (10, 12).
Building on our understanding of the regulation of ASCL1 protein in normal development (15, 18), here, we show that preventing ASCL1 phosphorylation in neuroblastoma cells results in cell-cycle exit by direct suppression of the neuroblastic core-regulatory circuit (CRC) and the target genes that drive cell-cycle progression. Alongside a suppression of this pro-proliferative program, we find that un(der)phosphorylated ASCL1 simultaneously activates expression of the CDK inhibitor CDKN1C as well as genes associated with neuronal differentiation and function, driving mitotic exit, and morphological maturation. Thus, we show that post-translational activation of a key developmental regulator in cancer cells can unmask latent tumor-suppressive activity, inhibiting proliferation and promoting differentiation.
Materials and Methods
The neuroblastoma cell line SH-SY5Y was kindly gifted by Prof. John Hardy, UCL. Cells were verified by submitting genomic DNA for short tandem repeat sequencing and compared with the Children's Oncology Group Cell Line Identification database (http://www.cogcell.org/clid.php) to ensure it was genetically matched to standardized cell lines. All cell lines were confirmed to be free of Mycoplasma before use and were tested every 3 months. All cell lines were cultured in DMEM/F-12 with GlutaMAX supplement, 10% tetracycline-free FBS (Clontech), and 100 U/mL penicillin and 100 μg/mL streptomycin. The Lenti-X 293T cell line was cultured in DMEM/F-12 with GlutaMAX supplement, 10% tetracycline-free FBS (Clontech), and 100 U/mL penicillin and 100 μg/mL streptomycin.
Generation of lentivirally transduced cell lines
Lentivirally transduced lines were generated using the pLVX-CMV-Tet3G system, detailed method is available as an associated file at the GEO database https://www.ncbi.nlm.nih.gov/geo/ and assigned the identifier GSE153823.
Proliferation assays were performed in the Incucyte analysis system (IncuCyte Zoom Videomicroscopy, Essen Bioscience Ltd.). Relative confluency was calculated by comparing growth rates of either WT or S-A ASCL1 versus uninduced cells. Cell count experiments were carried out in parallel to evaluate total cell numbers at different time points. Briefly, cell numbers of two independent clones of each of WT and S-A SH-SY5Y were compared with uninduced clonal cell lines (SH-SY5Y cells were induced with 4 ng/mL doxycycline over 96 hours).
Chromatin isolation, western blotting, and phos-tag western blot
Subcellular fractionation was performed using the subcellular protein fractionation kit (Thermo Scientific) according to the manufacturer's instructions. Western blot analysis were performed as previously described in ref. (19). ASCL1 phospho-status was determined in 8% acrylamide gels polymerized with 20 μmol/L Phos-tag reagent (WAKO) and 40 μmol/L MnCl2. After running and before transfer, phos-tag gels were washed three times, 10 minutes each with transfer buffer (25 mmol/L Tris-HCl, 190 mmol/L glycine, 20% methanol) plus 10 mmol/L EDTA, followed by a final wash with transfer buffer. Antibodies used were: anti-ASCL1 (1:250; a kind gift from David Anderson and Francois Guillemot), anti–C-MYC (1:10,000; Abcam), anti–N-MYC (1:500; Abcam), anti-H3 (1:5,000; Abcam), anti-PARP (1:500; BD Biosciences), and anti-GAPDH (1:1,000; Sigma) or anti α-tubulin (1:5,000; Sigma).
IHC was performed as previously described in ref. (19). Antibodies used were anti–β-tubulin (TUJ1; 1:1,000; Covance) or anti-ASCL1 (1:200; Abcam).
Quantitative real-time PCR
Quantitative real-time PCR was performed as previously described in ref. (19). The primer sequences are provided in Supplementary Table S1. Relative quantification was determined according to the ΔΔCt method. Data are presented as means ± S.E.M. of normalized values.
Chromatin immunoprecipitation, ChIP-Seq
ChIP and ChIPseq were performed as described previously (20, 21). Antibodies used were anti-ASCL1 (Abcam) and anti-IgG (Abcam). ChiPSeq data and full-associated methods are both available as an associated file at the GEO database https://www.ncbi.nlm.nih.gov/geo/ and assigned the identifier GSE153823. ChIPseq experiments were performed in two separate clones of each of WT and S-A SH-SY5Y tetracycline-inducible stable cell lines induced with 1 μg/mL doxycycline for 24 hours in four biological replicates. Differential-binding analysis (Diffbind) was performed as described previously (20). ChIP-qPCR primer sequences are provided in Supplementary Table S1.
RNA-sequencing experiments were performed in SH-SY5Y, WT and S-A SH-SY5Y tetracycline-inducible stable cell lines, using two clones of each in at least five biological replicates for each cell line. Full RNAseq data and Materials and Methods for RNA-seq are available at the GEO database https://www.ncbi.nlm.nih.gov/geo/ and assigned the identifier GSE153823. Single-end 50-bp reads generated on the Illumina HiSeq sequencer were aligned to the human genome version GRCh37 and read counts were generated using STAR 2.5.1a (22).
Gene set enrichment analysis
FDR values derived from DESeq2 analyses of the RNA-Seq data for all selected genes were −log10 transformed. These values were subsequently used for ranking and weighting of genes. GSEA Pre-ranked analysis tool from Gene Set Enrichment Analysis (GSEA) software, version 2.2.3 (http://software.broadinstitute.org/gsea/index.jsp) has been used for establishing potential functional relation in gene expression.
Quantification and statistical analysis
Statistical analysis were performed using a two-tailed unpaired student t test (*, P ≤ 0.05; **, P ≤ 0.01; and ***, P ≤ 0.001); standard error of the mean calculated from at least three independent experiments.
The next-generation sequencing data from this publication have been deposited to the GEO database https://www.ncbi.nlm.nih.gov/geo/ and assigned the identifier GSE153823.
Elevated ASCL1 activity promotes mitotic exit of neuroblastoma cells
Neuroblastoma cells frequently express ASCL1 (15), yet remain highly proliferative. As overexpression of ASCL1 can drive ectopic noradrenergic neurogenesis in vivo (15), we questioned whether an increase in ASCL1 activity over the endogenous level might be sufficient to render neuroblastoma cells post-mitotic and activate morphological differentiation. To explore this, we exploited the SH-SY5Y neuroblastoma cell line, which expresses a moderate level of ASCL1 protein (15) and is a representative model of the higher-risk ADRN disease subtype (refs. 4, 5; hereafter neuroblastoma cells). Using isogenically matched ASCL1-inducible neuroblastoma lines to study the effects of different expression levels (Supplementary Fig. S1A) revealed that increasing WT ASCL1 progressively reduced neuroblastoma cell confluency (Supplementary Fig. S1B). Cell apoptosis is accompanied by cleavage of PARP, which can be readily observed after SDS PAGE separation in control neuroblastoma cells treated with the apoptotic inducer Staurosporine (Supplementary Fig. S1C). A PARP cleavage product is absent from Dox-treated ASCL1-expressing neuroblastoma cells, so the reduction in cell number cannot be attributed to an increase in apoptosis, but does indicate that elevating ASCL1 activity suppresses neuroblastoma cell proliferation.
A phosphomutant form of ASCL1 (S-A ASCL1) can be generated by mutating all serine–proline phosphorylation sites to alanine–proline. This phosphomutant form of the protein has previously been used to show that preventing multi-site phosphorylation of ASCL1 promotes transcription factor–mediated reprogramming of embryonic ectoderm and fibroblasts into neurons (18). When expressed in neuroblastoma cells and detected by western blotting after SDS PAGE separation, S-A ASCL1 runs faster than WT ASCL1 but with similar mobility to WT ASCL1 that has been exposed to a broad-spectrum phosphatase (Supplementary Fig. S1D). This indicates that extensive ASCL1 phosphorylation in neuroblastoma cells usually occurs on serine–proline sites and is prevented by introducing phosphorylation site mutations. We next tested whether preventing phosphorylation of ASCL1 on serine–proline sites in neuroblastoma cells affects its ability to potentiate cell-cycle exit and differentiation.
Un(der)phosphorylated ASCL1 drives cell-cycle exit and differentiation
To directly compare the effect of increasing ectopic expression of either wild-type or phospho-mutant ASCL1 in neuroblastoma cells, we induced ASCL1 protein in doxycycline-inducible clonal cell lines expressing either WT or S-A ASCL1 (Fig. 1A and Supplementary Fig. S1B), resulting in nuclear expression of ASCL1 (Supplementary Fig. S1E). When induced at similar levels in neuroblastoma cells, phosphomutant S-A ASCL1 shows greater potency in suppressing cell division than WT ASCL1 (Fig. 1A left, Supplementary Fig. S1B). Proliferative arrest induced by S-A ASCL1 is accompanied by cell aggregation, and also neurite extension normally associated with neuronal differentiation (Fig. 1A, right).
To understand how un(der)phosphorylated ASCL1 inhibits neuroblastoma proliferation and enhances changes associated with differentiation, we interrogated RNAseq data to identify changes in expression of selected key regulators 24 hours after WT and SA-ASCL1 expression, alongside ChIPseq data quantifying ASCL1 binding adjacent to regulated genes. S-A ASCL1 binds more effectively than WT ASCL1 to regulatory regions associated with several prominent pro-proliferative genes, including SKP2, E2F1 and the MYC-target gene CDCA7, and also downregulates their expression (Fig. 1B), as further validated by ChIP-qPCR and real-time PCR analysis (Fig. 1C and D). Other key cell-cycle regulators, including CDK1 and CDK2, are also highly downregulated by un(der)phosphorylated ASCL1 (Fig. 1D).
Un(der)phosphorylated ASCL1 inhibits the MYC-driven CRC that drives neuroblastoma
ADRN-type neuroblastoma is locked into a highly proliferative state by a core oncogenic network of transcription factors, representing the “core-regulatory circuit” (4, 5). This includes the developmental regulators PHOX2B, HAND2, and GATA3 that combinatorially activate high level expression of each other and of downstream target genes via clusters of regulatory elements called “super-enhancers” (4, 5). Importantly, ASCL1 binds in proximity to the genomic locations that encode all three of these genes, suggesting feed-forward regulatory mechanisms (17). We investigated the effect of elevated WT or S-A ASCL1 on expression of these CRC genes. Upregulation of WT ASCL1 results in a modest suppression of the CRC genes GATA3 and HAND2, while having little effect on PHOX2B, whereas S-A ASCL1 induction results in almost total suppression of expression of all 3 of these CRC genes (Fig. 2A).
c-MYC/NMYC binds at super-enhancers associated with the PHOX2B/HAND2/GATA3 CRC and are associated with particularly high levels of gene activation (4). High-risk neuroblastoma can have either MYCN amplification (23), or alternatively elevated MYC (MYC). Elevated MYC functions in place of NMYC in our SH-SY5Y neuroblastoma cell model to maintain their neuroblastic phenotype (24). Consistent with a role for NMYC amplification being replaced by MYC activity in SH-SY5Y cells, we see that NMYC is cytoplasmic (Supplementary Fig. S2A). Moreover, MYCN is known to be directly regulated by ASCL1 in MYCN-amplified cell lines (17), whereas MYCN gene expression is not altered by WT or S-A ASCL1 in neuroblastoma cells (Supplementary Fig. S2B), which is also consistent with MYC and not MYCN regulating the neuroblastic program in SH-SY5Y neuroblastoma cells. We therefore investigated the effect of ASCL1 on MYC expression and its chromatin association in neuroblastoma cells.
Both WT and S-A ASCL1 bind the MYC gene locus and S-A ASCL1 in particular suppresses MYC expression (Fig. 2A). S-A ASCL1 induction also results in reduced C-MYC binding to neuroblastoma cell chromatin (Fig. 2B and Supplementary Fig. S2C), and consistent with this, S-A ASCL1 induction significantly perturbs the expression of hallmark MYC target genes (Fig. 2C), as well as downregulating genes associated with cell-cycle phase transitions (Fig. 2C).
In addition to suppressing pro-proliferative genes, and the CRC that maintains the proliferating neuroblastic phenotype, ASCL1 also binds and activates expression of the CDK inhibitor CDKN1C (Fig. 3A). Thus, S-A ASCL1 can directly drive cell-cycle exit by coordinated downregulation of the MYC-driven CRC that maintains neuroblast identity as well as regulating genes that control the cell-cycle directly. This is in contrast with WT ASCL1 that shows less-repressive activity on MYC and other key cell-cycle target genes.
Decreased proliferation of neuroblastoma cells could be a consequence of a switch to a pro-differentiation state. In support of this hypothesis, we saw that increased ASCL1 activity, and particularly that of phospho-mutant S-A ASCL1, results in enhanced binding and activation of a wide range of direct targets that have previously been associated with neuronal differentiation (Fig. 3A and Supplementary Fig. S3; refs. 25–27), with further expression validation undertaken by ChIP-qPCR and real-time PCR analysis (Fig. 3B and C).
We next explored whether the program of ASCL1-induced mitotic arrest is reversible in SH-SY5Y neuroblastoma cells. When S-A ASCL1 was induced at low levels, marked suppression of cell division was observed (Fig. 1A, Supplementary Fig. S1B and Fig. 3D). When we removed doxycycline after 4 days to turn off expression of S-A ASCL1, we saw a rapid re-entry into cell cycle (Fig. 3D). However, when S-A ASCL1 was expressed continuously for 7 days and then turned off, the re-entry into the cell cycle was substantially slower (Fig. 3D), demonstrating that post-mitotic effects of prolonged expression of un(der)phosphorylated ASCL1 are relatively durable even after ASCL1 removal.
ADRN-type neuroblastoma, the most lethal form of the disease, is driven by high level expression of the CRC of transcription factors, including PHOX2B, HAND2, and GATA3 (4, 5). These factors are usually transiently expressed in sympathetic neuron development, but elevated MYCN and/or MYC results in stabilization of mutually regulated super-enhancers associated with these and other genes that lock cells in a pro-proliferative neuroblastic phenotype (6, 24). Indeed, deregulated MYCN has been directly shown to invade enhancers of developmentally regulated genes, locking cells in pre-established progenitor behavior epigenetically (7). Here, we show that activation of ASCL1 by dephosphorylation drives a direct program of mitotic arrest and gene expression changes consistent with differentiation in neuroblastoma cells that carry high levels of MYC, as well as promoting cell aggregation and neurite outgrowth (Fig. 1A). Neuroblastoma is a highly diverse disease. It will be important to further explore ASCL1's potential as a therapeutic target in this devastating disease by determining which subtypes of neuroblastoma respond to dephosphorylated ASCL1 by undergoing mitotic exit and differentiation.
ASCL1 is expressed in ADRN-type neuroblastoma (5, 17), where selective pressures may favor a level lower of ASCL1 than the threshold required to drive cell-cycle exit and differentiation (Supplementary Fig. S1B). We show that enhancing ASCL1 activity by dephosphorylation can unlock a latent ability of neuroblastoma cells to exit the cell cycle and undergo differentiation manifested in transcriptional and morphological changes, in part by coordinately downregulating the MYC-driven PHOX2B/GATA3/HAND2 CRC that drives the pro-proliferative oncogenic ADRN phenotype of neuroblastoma cells. In addition, un(der)phosphorylated ASCL1 simultaneously directly suppresses pro-proliferative targets such as SKP2 and E2F1 and activates the CDK inhibitor CDKN1C, resulting in exit from cell cycle, as well as activating a program of genes associated with neuronal development and differentiation.
Elevating ASCL1 expression level in the subset of glioblastoma stem cells that usually express ASCL1 only at a low level is sufficient to drive stable cell-cycle exit (12). High levels of ASCL1 also impair reversion to a self-renewing state even when ASCL1 is subsequently turned off (12), mimicking its normal role in development and pointing to a wider ability of ASCL1 to drive a stable state of differentiation in malignant cells. Our data in neuroblastoma cells indicate that activation of ASCL1 may also be achieved by preventing its phosphorylation. ASCL1 has been shown to be phosphorylated by cyclin-dependent kinases in developing Xenopus embryos (18) and by MAP kinases in glioma cells (28). Rapidly proliferating neuroblastoma cells are likely to have high level of both classes of kinases, so CDK inhibitors and MAP kinase inhibitors are good candidates, perhaps in combination, to drive differentiation of neuroblastomas by dephosphorylation of endogenous ASCL1, and it will be important to now test this in vivo. We also note that investigation of trials of CDK4/6-targeted CDK inhibitors as therapeutic agents in neuroblastoma are already underway (29) and possible effects on differentiation as well as mitotic exit/cell death should be explored. There is also interest in the therapeutic use of CDK inhibitors that primarily target the transcriptional CDKs, CDK7/9 (9). These appear to bring about cell death by targeting the high level expression of the CRC genes and MYCN mediated by high level super-enhancer activity. We note that inhibitors that generally repress transcription are very likely to impede re-imposition of a differentiation program, so inhibitors with specificity for mitotic CDKs are better candidates for use in therapies designed to promote differentiation.
Proneural transcription factors are increasingly identified as playing a crucial role in lineage-specific oncogenesis. For example, ASCL1 is expressed in a number of tumors, including neuroblastoma, glioblastoma, and small-cell lung cancer (11, 13, 15, 16), whereas other proneural proteins act in an oncogenic context in medulloblastoma, and many endocrine neoplasias (2, 30). The relative importance of these proneural factors in defining the cell of origin and context of oncogenic drivers versus an active role in driving tumorigenicity is generally unclear (2), although an active pro-tumorigenic function is likely in some contexts; for instance, ASCL1 transcriptionally upregulates oncogenic and anti-apoptotic drivers such as MYCL1, SOX2, NF‐IB, and BCL2 in small-cell lung cancer (13, 31).
Although a picture is emerging in multiple cancers of the function of these master regulator proneural factors being subverted to promote oncogenesis, their activity in tumors still retains echoes of their normal developmental role in controlling the balance between proliferation and differentiation (2, 30). Phosphorylation of proneural proteins is a widely conserved regulatory mechanism controlling the differentiation activity of many members of this transcription factor family in development (e.g. in refs. 18, 19, 32, 33). In neuroblastoma cells, we see that a more normal developmental function of driving differentiation can be re-imposed on ASCL1 protein by manipulating its post-translational modification. We propose a wider model that the phosphorylation of proneural factors by elevated CDKs and other proline-directed kinases in rapidly dividing cancer cells may act in multiple tumor types to profoundly affect the spectrum of proneural target gene expression, favoring those that potentiate stem/progenitor maintenance, rather than those that would drive differentiation. Given ASCL1's role supporting oncogenesis in other cancers (11, 13, 15, 16, 34), and the fact that different types of aggressive tumors converge on a neuroendocrine identity that is associated with enhanced accessibility of ASCL1-responsive elements (35), it is tempting to speculate that inhibiting ASCL1 phosphorylation could be exploited to drive differentiation for therapeutic benefit in a range of cancers.
F.R. Ali reports grants from The Terry Fox Foundation, MBRU College of Medicine Internal grant award (MBRU-CM-RG2019-14), and MBRU-ALMAHMEED Collaborative Research Award (ALM1909) during the conduct of the study. D. Marcos reports grants from Neuroblastoma UK, as well as non-financial support from Wellcome Trust and MRC Cambridge Stem Cell Institute during the conduct of the study. L.M. Woods reports grants from Cancer Research UK during the conduct of the study. L.M. Parkinson reports grants from Cancer Research UK during the conduct of the study. L.A. Wylie reports an NIH-Oxford Cambridge Scholarship from National Institutes of Health during the conduct of the study. A. Philpott reports grants from Cancer Research UK, Medical Research Council, Neuroblastoma UK, and Wellcome Trust during the conduct of the study. No disclosures were reported by the other authors.
F.R. Ali: Conceptualization, formal analysis, validation, investigation, visualization, methodology, writing-review and editing. D. Marcos: Conceptualization, formal analysis, validation, investigation, methodology, writing-review and editing. I. Chernukhin: Formal analysis, investigation, methodology. L.M. Woods: Formal analysis, investigation, methodology. L.M. Parkinson: Formal analysis, investigation, methodology. L.A. Wylie: Conceptualization, formal analysis, investigation, methodology. T.D. Papkovskaia: Formal analysis, investigation, methodology. J.D. Davies: Formal analysis, investigation, methodology. J.S. Carroll: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, visualization, methodology, writing-original draft, project administration. A. Philpott: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing.
We would like to sincerely thank Sabine Dietmann, Mohammadmersad Ghorbani for help with bio-informatic analysis, Geetha Sankaranarayanan for technical support. Carol Thiele, Louis Chesler, John Gurdon, Guy Blanchard, Clive DeSantos, Eva Papachristou, and all Philpott laboratory members for helpful discussions. We would like to thank Prof. John Hardy for generously supplying the neuroblastoma cell line. The work was supported by Cancer Research UK Program grant A25636 and Wellcome Trust Investigator award 212253/Z/18/Z (to A. Philpott), MRC Research grant MR/L021129/1 (to F.R. Ali and A. Philpott), Neuroblastoma UK (to D. Marcos, T.D. Papkovskaia, and A. Philpott), Cancer Research UK Cambridge Center Pediatric Program (to L.M. Parkinson), The Terry Fox Foundation (to F.R. Ali), MBRU College of Medicine Internal grant award MBRU-CM-RG2019–14 (to F.R. Ali), MBRU-ALMAHMEED Collaborative Research Award ALM1909 (to F.R. Ali) and core support from the Wellcome Trust and the MRC Cambridge Stem Cell Institute (to F.R. Ali, D. Marcos, J.D. Davies, and A. Philpott), and Cancer Research UK Cambridge Institute (to I. Chernukhin and J.S. Carroll).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.