|Home | About | Journals | Submit | Contact Us | Français|
Self-renewal and proliferation of neural stem cells and the decision to initiate neurogenesis are the crucial events directing brain development. Here we show that the ubiquitin ligase Huwe1 operates upstream of the N-Myc-DLL3-Notch pathway to control neural stem cell activity and promote neurogenesis. Conditional inactivation of the Huwe1 gene in the mouse brain caused neonatal lethality associated with disorganization of the laminar patterning of the cortex. These defects stemmed from severe impairment of neurogenesis associated with uncontrolled expansion of the neural stem cell compartment. Loss and gain of function experiments for Huwe1 in the mouse cortex demonstrated that Huwe1 restrains proliferation and enables neuronal differentiation by suppressing the N-Myc-DLL3 cascade. Notably, human high-grade gliomas carry focal hemizygous deletions of the X-linked Huwe1 gene in association with amplification of the N-myc locus. Our results indicate that Huwe1 is a master regulator of the balance between proliferation and neurogenesis in the developing brain and this pathway is subverted in malignant brain tumors.
During development of mammalian neocortex, generation of neurons (neurogenesis) is preceded by proliferation of neural stem cells and committed progenitors in the germinal compartments lining the cerebral ventricles – the ventricular zone (VZ) and the subventricular zone (SVZ) (Doe, 2008; Gotz and Huttner, 2005). Neocortical neurogenesis in the VZ of the normal mouse extends from E11 through E18 (Takahashi et al., 1993, 1994, 1995a, b). During this time the cell cycle length increases, the proliferating fraction of neuroepithelial progenitors in the VZ gradually decreases, whereas the fraction of cells exiting the cell cycle and undergoing differentiation progressively increases (Dehay and Kennedy, 2007; Gotz and Huttner, 2005; Takahashi et al., 1995a, b). The increasing fraction of progenitors withdrawing from cell cycle is paralleled by activation of differentiation programs for the neuronal lineage, thus ensuring that the correct number of neuronal subtypes are consistently generated and form the functional circuits that orchestrate the laminar organization of the cortex (Kintner, 2002). Protein ubiquitination has recently emerged as a rapid and efficient post-translational modification of substrates that regulates many biological conditions. Specificity for distinct substrates is conferred by the E3 ubiquitin ligase subunit. When ubiquitin tagging to intracellular substrates occurs through Lys48-linked polyubiquitin chains, proteins are labeled for 26S proteasome-mediated recognition and proteolytic destruction. Little is known about the nature and dynamics of E3 ubiquitin ligases that, in response to antimitogenic and differentiative signals, modify key transcriptional regulators to instigate the transition from dividing neural progenitors to post-mitotic neurons.
Huwe1 encodes a HECT domain ubiquitin ligase. Through covalent interactions with the catalytic cysteine residue, HECT domain ligases accept ubiquitin from E2 ubiquitin-conjugating enzymes and transfer the ubiquitin to specific substrates (Huang et al., 1999). Huwe1 is a large protein (~ 500 kD) that has attracted considerable interest because several and quite disparate substrates have been assigned to this E3 (histones, Mcl1, p53, c-Myc, cdc6, N-Myc) (Adhikary et al., 2005; Chen et al., 2005; Hall et al., 2007; Liu et al., 2005; Zhao et al., 2008; Zhong et al., 2005). Not surprisingly, the biological functions of Huwe1 remain controversial (Bernassola et al., 2008). We have generated mice in which the region coding for the HECT domain of Huwe1 has been floxed. Targeted inactivation of floxed Huwe1 in the mouse brain disclosed that Huwe1-mediated ubiquitination is essential to restrain proliferation, elicit neurogenesis and establish the laminar structure of the cortex. This activity is abolished in malignant brain tumors containing genetic or epigenetic inactivation of Huwe1.
The pattern of expression of the Huwe1 protein in the nervous system is characterized by an increasing gradient, with the lowest amounts in the regions that contain neural stem cells/undifferentiated progenitors (VZ/SVZ) and the highest levels in differentiated neurons (CP) (Zhao et al., 2008). To determine the requirement for Huwe1 during development of the nervous system, we used a conditional Huwe1 mutant allele in which exon 80 and exon 82, which encode the HECT domain of the Huwe1 protein, were flanked by LoxP sites (Zhao et al., 2008). To inactivate Huwe1, we used a Cre deleter strain that was driven by the Nestin promoter, resulting in expression throughout the nervous systems from embryonic day 10.5 (Lothian et al., 1999). Huwe1 gene is X-linked, thus we performed our analysis on male offspring because inactivation of the single Huwe1 allele in this gender results in the Huwe1 null genotype (Fig. S1A). Crossings between Huwe1Flox/X or Huwe1Flox/Flox females and Nestin-Cre transgenic males generated Huwe1Flox/YNestin-Cre animals (hereafter referred to as Huwe1F/YNes). We observed efficient deletion of Huwe1 in the nervous system (Fig. S1A-B). Western blot and immunofluorescence analysis of Huwe1 protein using two antibodies recognizing the N-terminus (Huwe1 N-ter) or the HECT domain at the C-terminus (Huwe1 HECT) confirmed that the HECT-domain was absent in the brain of Huwe1F/YNes pups. Moreover, the mutant Huwe1 protein was markedly reduced, suggesting that C-terminal deletion destabilized the residual polypeptide (Fig. S1C, D). Huwe1F/YNes mice were born with expected ratio. However, all Huwe1F/YNes neonates died within 24 h of postnatal life (Table S1).
The brains from Huwe1F/YNes newborns were slightly smaller than those from Huwe1F/Y or Huwe1+/YNes and wild type littermates. Histologic analysis of the mutant brains at E18.5 and P0 revealed distinct abnormalities in various regions of the brain, including cerebral cortex, hippocampus and cerebellum (Fig. S2A, S2E, S2F). Mutant mice had a poorly developed dentate gyrus and an immature, very small cerebellum (Fig. S2E, S2F). In Huwe1F/YNes mice cortical thickness was significantly reduced, the laminar organization was altered and superficial neuron layers were not clearly identifiable (Fig. S2C). Conversely, the germinal layers (VZ/SVZ) that contain proliferating neural stem/progenitor cells were expanded and extended into regions normally occupied by differentiating cells. Despite the overall reduction in cortical size, cellular density throughout the cortex was increased and intervening neuropil decreased (Fig. S2A, S2C). Furthermore, mutant brains contained ectopic cellular clusters in differentiated striatal regions, or more rarely in other post-mitotic areas of the brain (Fig. S2B, S2D). At E15.5 histology abnormalities were barely discernible in mutant brains.
The increased cellular density in the neonatal cortex suggested that neuronal production was abnormal. To study cell proliferation, we performed quantitative analysis of the proliferation marker Ki67 at different embryonic stages (E13.5, E15.5 and E18.5). At each stage of development Huwe1 knockout cortices contained significantly higher numbers of proliferating cells (Fig. 1A, S3A, S3B). Germinal areas (VZ and SVZ) were primarily involved in the proliferative expansion. However, the ectopic clusters dispersed throughout post-mitotic and differentiated regions of the Huwe1 knockout brain were also invariably positive to Ki67 (Fig. 1B). Therefore, loss of Huwe1 deregulates the growth control in the nervous system and causes expansion of the germinal layers and ectopic clustering of proliferating cells. Also the higher number of Ki67-positive cells in Huwe1F/YNes E13.5 and E15.5 embryos indicated that hyper-proliferation precedes cortical abnormalities in Huwe1-null brains.
Cortical neurogenesis is associated with progressive lengthening of the cell cycle and results in increased fraction of progenitors that exit the cell cycle and differentiate (Dehay and Kennedy, 2007; Gotz and Huttner, 2005). We asked whether the expansion of the neural progenitor pool in Huwe1F/YNes animals is consequent to defective withdrawal and/or shortening of the cell cycle in comparison with normal brains. We analyzed the probability to withdraw from cycling and the cell cycle timing in Huwe1F/YNes and wild type embryos. First, we determined the fraction of cells that exit cell cycle after 24 h labeling with BrdU. The cell cycle exit was estimated as the fraction of BrdU-positive cells that lost positivity for Ki67 (BrdU+/Ki67-) and the total number of BrdU-positive cells. Quantification of these experiments from multiple pairs of Huwe1-null and control brains at E13.5, E15.5 and E18.5 demonstrated that the fraction of progenitors that exited cell cycle was significantly lower in Huwe1-null brains at each developmental age (Fig. 1C, S3C, S3D). To unravel potential changes in cell cycle timing, we determined the proportion of cortical progenitors labeled by 1 hour-pulse BrdU, whereby neural progenitor cells are identified by Ki67 staining. Because the length of S phase remains constant throughout development, a larger fraction of BrdU-positive neural progenitors (Ki67-positive) would indicate shortening of the cell cycle whereas a lower proportion would imply a prolonged cell cycle timing. Quantification of results from three pairs of control and Huwe1-null brains indicated that the length of the cell cycle of the mutants was indistinguishable from the wild type brains at E13.5 and E15.5 (Fig. S4A, B). However, when the same analysis was conducted at E18.5, Huwe1-null brains contained a significantly higher fraction of BrdU+ progenitors, thus indicating that loss of Huwe1 shortened cell cycle timing at the latest stages of neural development (Fig. 1D). Taken together, cell cycle withdrawal and timing analyses suggest that the primary defect caused by loss of Huwe1 is the reduced probability to exit from active cycling. However, as development proceeds, Huwe1 inactivation impairs also the lengthening of cell cycle timing, which is normally associated with the increased rate of progenitors undergoing differentiation during late neurogenesis.
We determined whether the ability to differentiate was also perturbed in Huwe1-null brains. The RNA binding protein HuC/D is an early marker of neuronal differentiation expressed in the developing cortical plate since the initial stages of neurogenesis. It functions to promote neuronal differentiation by modulating the expression of genes required for axonal outgrowth such as the gene encoding for the GAP43 protein (Ekstrom and Johansson, 2003). HuC/D and GAP43 were readily detected in wild type cortices at E15.5 but the expression of the two markers was notably reduced in Huwe1F/YNes brains (Fig. 2A). GAP43 was also diminished in the mutant cortex at E18.5 (Fig. S5A). The developmental progression from a neuroepithelial progenitor/radial glia phenotype, characteristic of the VZ, to an intermediate progenitor phenotype, predominant in the SVZ, is accompanied by down-regulation of the homeodomain transcription factor Pax6 and concomitant induction of the T-box transcription factor Tbr2 (Englund et al., 2005). We compared the size and spatial distribution of these populations between the Huwe1F/YNes and control brains at E18.5. In wild type cortex Pax6 was confined to the thin VZ layer and Tbr2 was detected as a more superficial domain coinciding with the SVZ. Conversely, Huwe1-null cortex exhibited expansion and increased cellular density of the Pax6 expressing domain with consequent superficial displacement of the Tbr2-positive region. The thickness of the Tbr2 domain was similar among wild type and Huwe1-null cortices (Fig. 2B, see also Fig. S5B for P0). These findings indicate that loss of Huwe1 in the cortex induces uncontrolled growth of the Pax6-expressing stem/progenitor population, which remains negative for Tbr2. Next, we used molecular markers to determine the consequences of genetic inactivation of Huwe1 on laminar structure of the cortex. In the normal brain the transcription factor Tbr1 and the Ctip2 protein are expressed in the deep cortical layers and mark early born neurons. As late-born neurons populate the upper layers of the cortex, more superficial layers positive for MAP2, Cux1 and p27Kip2 (layers II-III) juxtapose the Tbr1/Ctip2-expressing domains. Staining for Tbr1 and Ctip2 demonstrated decreased thickness but increased neuron packing density of the Tbr1/Ctip2-positive deep layers in Huwe1F/YNes embryos and neonates (Fig. 2C, Fig. S5C). Moreover, the most superficial layers of the cortical plate of Huwe1-null embryos (layers II-III), marked by expression of MAP2, Cux1 and p27Kip2, reached only a rudimentary development, being mostly occupied by a disordered population of Tbr1/Ctip2-positive neurons (Fig. 2C, D and Fig. S5C). To examine the efficiency of neuronal production and explore the mechanisms that lead to perturbed lamination in the Huwe1F/YNes brain, we determined the position of neurons that became postmitotic at different times using the BrdU “birthdating” method. To examine early neurogenesis (deep layer neurons), BrdU was given to pregnant females when their embryos were E12.5. To examine late neurogenesis (superficial layers), BrdU was given at E15.5; in both cases brains were dissected at P0. BrdU labeling at E12.5 showed no major defect in the position and absolute number of deep layer Tbr1-positive neurons (Fig. S6A, B). However, a differentiation defect was already evident at this stage because mutant brains contained a significantly lower fraction of BrdU-positive neurons that stained positive for Tbr1 (Fig. S6B, right panel). We interpret these findings to indicate that, although the percentage of progenitors undergoing differentiation is reduced in the Huwe1-null cortex, the larger pool of progenitors available in the knockout brain accounts for a comparable number of Tbr1-positive neurons in layer VI. BrdU labeling at E15.5 indicated that the ability to generate neurons that migrate to the most superficial layers was markedly reduced in the absence of Huwe1 (Fig. 2E, F). The observation that the Huwe1-null brain displays normal staining pattern for reelin, a key regulator of radial migration (Fig. S6C), suggests that the increased neuron packing density in the mutant cortex is not secondary to cell intrinsic migration defects. In sum, Huwe1 directs differentiation in the developing mouse cortex and is essential for the neuronal differentiation events that accompany the generation of late-born neurons in layers II-III.
To identify the mechanistic basis of the Huwe1-knockout phenotype, we asked whether loss of Huwe1 resulted in accumulation of N-Myc and uncontrolled activation of molecular events downstream of N-Myc. Immunostaining of control and Huwe1-knockout brain revealed that N-Myc was elevated at E13.5, E15.5 and E18.5 (Fig. 3A, upper panels, S7A, B, upper panels). D-type cyclins are established N-Myc target genes and are expressed in the developing CNS (Glickstein et al., 2007; Oliver et al., 2003). As for N-Myc, Cyclin D1 was also weakly expressed throughout the cortical layers of wild type embryos but it was notably elevated in the absence of Huwe1 (Fig. 3A, lower panels, S7B, lower panels). Although elevation of cyclin D1 (and downregulation of p27Kip2, Fig. S7C) may explain some of the events elicited by deregulated N-Myc in the Huwe1-knockout cortex, Myc transcription factors display remarkable cell-type specificity for target activation (Eilers and Eisenman, 2008). To unravel the complement of N-Myc transcriptional targets (regulon) in the nervous system, we developed a screen for N-Myc target genes based on a reverse engineering approach. This analysis was performed on 176 malignant glioma samples that had been analyzed by global gene expression profiling and the ARACNe algorithm (algorithm for the reconstruction of accurate cellular networks) was used to infer a global transcriptional network (Basso et al., 2005). N-Myc but not c-Myc emerged among the 10% largest transcriptional hubs, displaying an inferred sub-network of 230 highly significant inferred targets (p<0.005, Bonferroni corrected), among which 88 are positively regulated and 142 repressed by N-Myc (Fig. 3B, Table S2). Having identified a neural-specific set of N-Myc-regulated target genes (the N-Myc regulon), we interrogated the expression profile tumor dataset - 236 glioblastoma (GBM) samples - from the Cancer Genome Atlas (TCGA) project to determine whether the N-Myc regulon is differentially regulated in tumor sample subsets where Huwe1 is differentially expressed (i.e. low expression in one subset and high expression in the other). The two subsets were generated by sorting all glioma samples by Huwe1 expression and then by selecting the first and last 83 (35%) samples. Differential regulation of the N-Myc regulon was assessed using an extension of the Gene Set Enrichment Analysis (GSEA) method (Subramanian et al., 2005), called GSEA2, developed to simultaneously analyze the enrichment of both activated and repressed targets of a transcription factor (see methods and Ref. (Lim et al., 2009)). When a GO-derived list of over 1,000 transcription factors was tested for differential activity on their regulon, conditional on Huwe1 expression, N-Myc was ranked among the top 30 (Fig. 3C). The analysis of the GSEA2 plot revealed marked loss of N-Myc transcriptional activity in tumors with high Huwe1 expression, i.e., concurrent down-regulation of N-Myc activated targets and upregulation of N-Myc repressed targets (p<1×10-5, Fig. 3D). We also asked whether Huwe1 controls the activity of c-Myc. The GBM-derived regulon of c-Myc was noticeably smaller than that of N-Myc (97 versus 230 inferred targets, Fig. S8A, Table S3). The GSEA2 plot showed that the c-Myc regulon was similarly inactivated in tumors with high Huwe1 expression (Fig. S8B). Taken together, the computational results indicate that Huwe1 operates as repressor of N-Myc and c-Myc transcription in neural tissue.
One of the highest confidence N-Myc positive targets inferred by ARACNe is the Notch ligand DLL3 (Fig. 3B). Active Notch signaling is required to maintain neural cells in a multipotent, proliferative state and inhibit differentiation (Hitoshi et al., 2002; Mizutani et al., 2007; Nagao et al., 2007; Yoon and Gaiano, 2005). Therefore, we asked whether (i) DLL3 could be validated as a direct N-Myc target gene; (ii) N-Myc activates Notch signaling; (iii) DLL3 is overexpressed in Huwe1 knockout brain and (iv) Notch signaling is activated in Huwe1 knockout brain. The regulatory DNA region adjacent to the DLL3 gene contained N-Myc-binding sites (E-boxes) and chromatin immunoprecipitation from N-myc amplified neuroblastoma cells demonstrated that endogenous N-Myc is physically associated with the DLL3 promoter but not irrelevant genomic DNA sequences (Fig. 4A). Furthermore, ectopic expression of N-Myc induced the expression of DLL3 (Fig. 4B) and Notch1 activation, measured by Western blot analysis for the cleaved and activated intracellular form of Notch1 (Notch1-IC, Fig. 4C). When similar amounts of N-Myc and c-Myc were expressed in the same cells, N-Myc but not c-Myc induced the expression of DLL3 mRNA and DLL3 protein (Fig. S9A, C). However, c-Myc induced efficiently the expression of two classical Myc-target genes (ODC and cdc25A), indicating that the inability of c-Myc to induce DLL3 was not consequent to a generally lower transcriptional activity by c-Myc (Fig. S9A). Accordingly, N-Myc but not c-Myc induced cleaved Notch1 (Fig. S9B). The ability of N-Myc but not c-Myc to induce DLL3 is consistent with the presence of DLL3 in the regulon of N-Myc but not c-Myc (Fig. 3B, Fig. S8A, Tables S2, S3). To provide further validation for the N-Myc-DLL3 axis, we asked whether expression of N-myc and DLL3 were concurrently regulated in neural cells retaining stem cell features as opposed to the same cells growing under conditions characterized by rapid loss of stemness. GBM cultures grown in serum-free Neurobasal medium containing EGF and basic FGF (“NBE”) are similar to normal neural stem cells whereas switching these cells to serum-containing culture conditions leads to rapid loss of their stemness (Lee et al., 2006). By using an entirely independent brain tumor dataset, a recent report compared the gene expression profile of NBE-growing brain tumor stem cells with that from the same cells growing in serum (Lee et al., 2006). Gene expression profile data confirmed that N-myc and DLL3 are strongly co-regulated in this alternative dataset. Additionally, the two genes were shown to be abundantly expressed in NBE-growing brain tumor stem cells but significantly and concurrently depleted in the serum-growing counterparts (Fig. 4D).
Next, we asked whether the elevated levels of N-Myc detected in the Huwe1-null brain increased DLL3 expression and activated Notch signaling. Quantitative RT-PCR expression analysis for the Notch ligands DLL1, DLL3, DLL4, Jagged-1 and Jagged-2 and the Notch1 gene revealed that DLL3 mRNA was the only component of the Notch pathway elevated in Huwe1-null brain at every developmental stage (Fig. 4E, Fig. S9D). Furthermore, immunostaining of Huwe1-knockout and wild type brains for Notch1-IC revealed a marked increase in the absence of Huwe1 at E15.5 (Fig. 4F, upper panels) and E18.5 (Fig. 4F, lower panels). Notably, the Notch1-IC-positive domain overlapped with the region showing the highest expression of the N-Myc protein. Taken together, these findings suggest that an N-Myc-DLL3 cascade may be a key functional target of Huwe1 during neural development.
To explore the functional significance of the activation of the N-Myc-DLL3 axis for the abnormalities of the Huwe1-knockout cortex, we analyzed the consequences of silencing N-myc or DLL3 in Huwe1-null brains. Electroporation of a Cre-GFP expressing plasmid in the cortical germinal layers of Huwe1flox embryos produced efficient recombination of the floxed allele (Fig. S10). We analyzed BrdU incorporation and performed immunostaining for the neuronal differentiation marker TuJ1 and the DLL3 protein in sections obtained from organotypic slice cultures derived from electroporated E14.5 Huwe1flox brains (Hand et al., 2005; Nguyen et al., 2006). Compared to GFP-electroporated cortices, electroporation of Cre-GFP triggered the same abnormalities detected in Huwe1F/YNes brains (increased proliferation, impaired neuronal differentiation and overexpression of DLL3, Fig. 5A-C). Expression of Cre-GFP in the presence of specific siRNA oligonucleotides targeting N-myc or DLL3 but not control siRNAs rescued the hyperproliferation and the differentiation defects caused by Huwe1 inactivation (Fig. 5A, B). Interestingly, the expression of DLL3 was reversed in Huwe1-null cells electroporated with N-myc siRNA oligonucleotides, indicating that loss of Huwe1 triggers elevation of DLL3 through N-Myc (Fig. 5C). Thus, the defects generated by inactivation of Huwe1 require an active N-Myc-DLL3 pathway.
To test whether ectopic expression of Huwe1 in the developing normal brain produces phenotypic and molecular effects that complement those elicited by loss of Huwe1, we electroporated embryo cortices with V5-tagged Huwe1 wild-type plasmids expressing the 500 kD full-length Huwe1 protein (V5-Huwe1) or the catalytically inactive C4341A point mutant (V5-Huwe1-CA). Expression of V5-Huwe1 wild-type but not the V5-Huwe1-CA mutant caused dramatic inhibition of cell cycle progression and significantly reduced expression of DLL3 (Fig. 6A, B). Taken together, these data suggest that Huwe1 controls the timing of cell cycle withdrawal and initiation of differentiation in the developing brain by eliminating the N-Myc-DLL3 cascade from cortical progenitors.
It has been proposed that human brain tumors are initiated and maintained by alterations in pathways that regulate self-renewal and differentiation of neural stem cells (Rich and Eyler, 2008; Stiles and Rowitch, 2008; Sutter et al., 2007). Given the importance of the Huwe1-N-Myc pathway in the biology of neural stem cells and neurogenesis in vivo, we asked whether the genes encoding for N-myc and Huwe1 are targeted by alterations in human brain tumors. Glioblastoma (GBM) is a CNS tumor for which the most likely cells of origin are neural stem cells. The Cancer Genome Atlas (TCGA) project provides data on the DNA copy number of primary GBM samples (Network, 2008). We interrogated the high-density Affymetrix 6.0 platform of SNPs arrays in the Atlas tumor collection for the possible occurrence of genetic alterations in the N-myc and Huwe1 loci by using the Partek Genomics Suite 6.08. Eight of 129 (6.2%) tumors carry amplification of the N-myc gene. Remarkably, we discovered that two of the eight N-myc-amplified tumors (25%) but none of the tumors lacking N-myc gene amplification contain focal deletions of the Huwe1 gene (FET p-value=0.0034, Fig. 7A). Detailed mapping of the breakpoints revealed that both deletions remove a minimal region including coding exons 2-20 of Huwe1 but do not affect the neighboring genetic loci (80 marker probes are deleted in tumor #178 and 31 marker probes are deleted in tumor #289, Fig. 7B). Interestingly, Huwe1 is located on the X chromosome (Xp11.22) and each of the two patients whose tumors contain deletion of Huwe1 is a male (XY). Therefore, the two tumors are converted to a Huwe1-null state in a manner functionally analogous to that of other tumor suppressor genes that undergo complete loss-of-function in glioma (e.g. Ink4a, PTEN) (Network, 2008). Next, we asked whether loss of Huwe1 expression is also associated with brain tumorigenesis. To this end, we used the ONCOMINE database (www.oncomine.com) that contains gene expression data compiled from the microarray analysis of 23 non-tumor human brain samples compared to 77 GBM samples. Interestingly, the GBM samples showed a highly significant (p=9.3E-10) down-regulation of Huwe1 mRNA in comparison to the corresponding brain tissues (Fig. 7C). Thus, malignant brain tumors inactivate the Huwe1 gene through both genetic and epigenetic alterations.
Restriction of self-renewal and proliferation of neural stem cells is coupled to neurogenesis to coordinate cortical architecture during development. In this study, we have conditionally deleted the Huwe1 gene in the mouse brain and demonstrated that this enzyme is essential for the transition from self-renewing and proliferating neural stem/progenitor cells to differentiated neurons. Consequently, loss of Huwe1 severely perturbs neurogenesis and leads to disorganization of the laminar structure of the cortex. We have identified a new “N-Myc-DLL3” pathway as mediator of the abnormalities detected in Huwe1-null brain. The identification of this pathway is the missing link between two crucial transcriptional regulators that provide essential signals for neural stem cell activity and inhibition of neuronal differentiation. Interestingly, a recent study reported loss of Notch activation in the mouse cortex by neural-specific mutation of another E3 ubiquitin ligase, Mind Bomb 1 (Yoon et al., 2008). Together, we suggest that neural development requires a fine balance of ubiquitin ligases with positive (Mind Bomb 1) and negative (Huwe1) signaling ability converging on the Notch pathway. Our work also identifies focal intragenic deletions of the Huwe1 gene in GBM, a human brain tumor sustained by deregulated neural-stem like activity and marked anaplasia. Thus, mutations of an ubiquitin ligase that restrains proliferation, initiates neurogenesis and organizes the laminar patterning of the cortex are selected during oncogenic transformation in the human brain.
The expansion of the neural stem cell compartment elicited by loss of Huwe1 becomes progressively more evident as neural development proceeds and Huwe1-/- neural stem/progenitor cells fail to exit cell cycle and commence neuronal differentiation. The deregulated proliferative activity conferred by loss of Huwe1 together with abnormal cell morphology and loss of “crowd control” ultimately lead to severe perturbation of neuronal differentiation and disorganization of brain architecture (Dehay and Kennedy, 2007). During relatively early stages of neurogenesis (before E14.5), cell cycle timing is not affected by mutation of Huwe1. However, loss of Huwe1 severely impairs lengthening of cell cycle that accompanies the progressive shift from proliferation to differentiation during late neurogenesis.
The phenotype of the Huwe1-mutant brain in the mouse is complementary to that caused by inactivation of transcription factors that expand the neural stem cell compartment and inhibit neurogenesis (N-Myc and Notch) (Hitoshi et al., 2002; Knoepfler et al., 2002; Mizutani et al., 2007). We found that the N-Myc protein markedly accumulated in the Huwe1-null brain and this effect preceded the phenotypic defects. To unravel the identity of the downstream signaling events triggered by aberrant N-Myc in Huwe1-null brain, we designed a novel computational approach to dissect and interrogate the activity of transcription factors following modulation of candidate regulators in a specific cellular context. From this approach, the Notch ligand DLL3 emerged as one of the strongest inferred N-Myc target in the brain and we have confirmed that, during neural development, Huwe1 negatively regulates expression of DLL3 in an N-Myc-dependent fashion. Based on this information, we have experimentally validated DLL3 as a direct transcriptional target of N-Myc and, most importantly, we discovered that the hyperproliferation and neuronal differentiation defects resulting from knocking out Huwe1 in the cortex are fully reversed by silencing the expression of DLL3 in vivo. Thus, the N-Myc-DLL3 cascade is restrained by Huwe1 to set the timing of cell cycle withdrawal and neuronal differentiation in the developing brain. Although our results are consistent with DLL3 activating Notch1 in the neural stem cell compartment at mid-gestation, in other systems DLL3 might also behave as inhibitor of Notch1, possibly through competition with other Notch ligands (Geffers et al., 2007; Ladi et al., 2005).
Activation of stem cell activity and deregulation of N-Myc (and Notch) have been linked to tumorigenesis in the brain (Stiles and Rowitch, 2008; Sutter et al., 2007). High-density gene copy number analysis discovered that at least two tumors (both of which originated in male patients) had highly focal deletions that were entirely contained within the coding region of the Huwe1 gene. The reduced expression of Huwe1 in GBM suggests that epigenetic alterations may be more frequently involved. Thus, Huwe1 is the second example (after WTX) of X-linked tumor suppressor genes in which inactivation of the single allele in tumor-bearing male patients is sufficient to create a complete loss-of-function status in tumor cells (Rivera et al., 2007). Identification of this mechanism is a significant departure form the classical two-hit hypothesis considered to mediate autosomal inactivation of tumor suppressor genes in human cancer. The Huwe1 mutations display an additional unique feature in that they co-exist with N-myc gene amplification in GBM. Although the finalistic concurrence of N-myc and Huwe1 alterations in the same tumors may seem counterintuitive, cooperation between two independent genetic events is consistent with the mechanisms leading to activation of c-Myc in Burkitt's and other lymphomas in which transcriptional deregulation of the c-myc gene requires translocation to the immunoglobulin loci and is associated with escape from ubiquitin-mediated degradation by the c-Myc protein through independent mutation of regulatory phosphorylation sites (Bahram et al., 2000; Bhatia et al., 1993). Thus, the short half-life of the N-Myc protein in neural cells requires that the increasingly synthesized protein in brain tumors be stabilized by independent genetic loss of the gene encoding for the ubiquitin ligase Huwe1. We predict that neural stem cells carrying concurrent deletion of Huwe1 and transcriptional activation of the N-myc oncogene gain the most powerful myc-associated activities necessary for oncogenic transformation.
To generate Huwe1Flox/YNestin-Cre animals (hereafter referred to as Huwe1F/YNes) heterozygous or homozygous floxed Huwe1 females (Huwe1Flox/X or Huwe1Flox/Flox) were bred onto Nestin-Cre heterozygous males. Huwe1 mutants were genotyped by PCR of genomic DNA prepared from tail biopsies using primers described previously (Zhao et al., 2008). All animal experiments were approved by and performed in accordance with the guidelines of the International Agency for Research on Cancer's Animal Care and Use Committee.
Short pulse bromodeoxyuridine (BrdU, Sigma-Aldrich) labeling was carried by injecting pregnant mice with BrdU intraperitoneally (50 μg/g body weight) 1 h prior to sacrifice. For birth dating experiments, timed-pregnant mice (at E12.5 and E15.5) were injected with BrdU intraperitoneally at 40 μg/g body weight. PFA-perfused embryos (E13.5 and E15.5) or dissected brains (E18.5 and P0) were fixed for 24 h in 10% formalin. Five micron paraffin sections were deparaffinized, rehydrated, treated for antigen retrieval, and incubated in 10% serum blocking solution prior to the incubation with primary antibodies at room temperature for 1 h or at 4°C for 18 hr. For BrdU staining, sections were treated with 3 N HCl for 50 min at room temperature followed by two rinses of 0.1 M boric acid (pH 8.0) for 5 min at RT prior to blocking. Fluorescent detection was performed with Cy3 or FITC-labeled secondary antibodies (Jackson Immunoresearch Laboratories), whereas biotin-conjugated secondary antibodies (Vector Laboratories) were employed for Vectastain Elite ABC development. Tyramide amplification system (Perkin Elmer) was used for cleaved Notch1 (Notch-IC), N-Myc and cyclin D1 immunodetection according to the manufacturer instructions. The primary antibodies used are: anti-cleaved Notch1 (Val1744, Cell Signaling Technology), anti-BrdU (Roche), anti- Ki67 (Novocastra), anti-Ctip2 (Abcam), anti-MAP2 (Sigma), anti-GAP43, anti-cyclin D1 and anti-HuC/D (Invitrogen), anti-N-Myc (Calbiochem), anti-HUWE1 HECT-domain (Lifespan Biosciences), anti-HUWE1-Lasu1 (Bethyl Laboratories), anti-p27 (Thermo Scientific), anti-Tbr1, anti-Tbr2, anti-Reelin, and anti-Pax6 (Millipore), and anti-Cux1 (Santa Cruz).
Ex vivo electroporation was performed as described previously with minor modifications (Zhao et al., 2008). Huwe1 floxed embryos were recovered at E14.5. Endotoxin-free GFP plasmid (1μg/μl, MSCV-GFP or MSCV-Cre-GFP) mixed with siRNA (10 μM, N-myc, DLL3 and Control Smart Pool, Dharmacon) were injected into lateral ventricles using a Femtojet microinjector (Eppendorf). In Huwe1 overexpression experiments, pcDNA3.1/V5-His/LacZ, pcDNA3.1/V5-Huwe1WT or pcDNA3.1/V5-Huwe1CA plasmids (2μg/μl) were used for electroporation. Electroporated cerebral hemispheres were dissected and sectioned coronally into 300-μm-thick slices using a vibratome (VT1000S, Leica). The slices were cultured for 2 days in neurobasal medium containing 1% N2, 1% B27, 2 mM L-glutamine, penicillin–streptomycin and fungizone. Medium was changed every day. Slices were labeled with 10 μM BrdU for 2 h followed by fixation in ice cold 4% PFA at 4 °C for 20 h. Fixed slices were incubated for 24 h in PBS containing 15% sucrose and 30% sucrose sequentially and embedded in OCT. For each staining, 10 μm sections of at least 3 brains per group from two independent experiments were used. The antibodies used for double staining on cryosections are: Brdu (Roche), anti-Brdu (Abcam), anti-v5 (Invitrogen), anti-β-III tubulin (Promega), anti-Dll3 (Santa Cruz), and anti-GFP (Invitrogen). Confocal images acquired with a Zeiss Axioscop2 FS MOT microscope were used to score double positive.
Each experiment was performed with samples from at least three animals from two independent litters. In histogram values represents the mean values; error bars are standard deviations. Statistical significance was determined by t test (with Welch's Correction) using GraphPad Prism 4.0 software (GraphPad Inc., San Diego, CA).
Microarray samples are first partitioned into two 35% subsets where Huwe1 is relatively most and least expressed. GSEA2 is used to assess whether transcriptional targets of a transcription factor is statistically enriched in genes differentially expressed between the two subsets. The algorithm considers differential expression of both transcription factor's positive and negative targets as the reflection of changes in the transcription factor's activity. Suppose that activity of a transcription factor is negatively associated with the abundance of Huwe1 (e.g. Huwe1 ubiquitinates and degrades a transcription factor), one would expect the transcription factor-activated targets to be enriched in genes that are downregulated, while the transcription factor-repressed targets to be enriched in genes that are upregulated when Huwe1 is highly expressed. Statistical significance of the enrichment is then determined by comparing the observed enrichment score to the null distribution computed through random permutation of the transcriptional target sets. Details of the algorithm have been previously described in (Lim et al., 2009).
Copy number analysis on the TCGA GBM samples was done using Partek Genomics Suite 6.08 (http://www.partek.com/). Both paired and unpaired analysis was done to get the raw copy numbers since some tumor samples did not have corresponding normal samples. To identify regions of amplification and deletion, the genomic segmentation algorithm of Partek was implemented with the following parameters: minimum of ten genomic markers at a p-value of 0.001 and a signal to noise ratio of 0.3. A p-value of 0.01 was used to filter for regions of interest from the segmentation results.
This work was supported by grants from the NIH-NCI to A.L. (R01CA131126), A.I. (R01CA085628) and A.C. (R01CA109755) and by the National Centers for Biomedical Computing NIH Roadmap initiative (U54CA121852). D.D'A. and M.S.C. are supported by fellowships from Provincia di Benevento/Ministero del Lavoro, Italy.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.