|Home | About | Journals | Submit | Contact Us | Français|
Loss of Ikaros has been correlated with Notch activation in T cell acute lymphoblastic leukemia (T-ALL), however, the mechanism remains unknown. We identified promoters in Notch1 that drive expression of Notch1 proteins active in the absence of ligand. Ikaros bound to both canonical and alternative Notch1 promoters and its loss increased permissive chromatin, facilitating recruitment of transcription regulators. At early stages of leukemogenesis, increased basal expression from the canonical and 5’-alternative promoters initiated a feed-back loop, progressively augmenting Notch1 signaling. Ikaros also repressed intragenic promoters that are cryptic in wild-type, poised in pre-leukemic, and active in leukemic cells and which also produced ligand-independent Notch1 proteins. Only ligand-independent Notch1 isoforms were required for Ikaros-mediated leukemogenesis. Notch1 alternative-promoter usage was observed at stages of T cell development dependent on Notch signaling and during T-ALL progression. These studies identify a network of epigenetic and transcriptional regulators that control conventional and unconventional Notch signaling during normal development and leukemogenesis.
Regulatory factors that control normal development are frequently implicated in neoplastic transformation. Ikaros, a zinc finger DNA binding protein and a key regulator of lymphopoiesis, is a prime example (Georgopoulos, 2009; Mullighan and Downing, 2008). In the hematopoietic stem cell (HSC) compartment, Ikaros is required for transcriptional priming of a lymphoid gene expression program, enabling lymphoid differentiation to the most primitive of hematopoietic progenitors (Ng et al., 2009; Yoshida et al., 2010). Normal outcome of T cell development also depends on Ikaros activity. Loss of Ikaros accelerates transition through the β-selection checkpoint, causing an aberrant accumulation of T cell receptor (TCR)+ CD4+CD8+ “double positive” (DP) thymocytes that rapidly transit to a leukemic state (Winandy et al., 1999; Winandy et al., 1995). These phenotypes implicate Ikaros as a regulator of signaling pathways and their cellular outcome during T cell differentiation. Consistently, aberrant activation of Notch signaling occurs in Ikaros-deficient T cell leukemias, implicating Ikaros as a negative regulator of this pathway during T cell development and leukemogenesis (Chari and Winandy, 2008; Dumortier et al., 2006; Kleinmann et al., 2008; Mantha et al., 2006).
Notch signaling is another key determinant of lymphopoiesis. The Notch receptors (Notch 1-4) are single-pass membrane molecules that are cleaved by furin-like convertase (S1) within the Golgi secretory pathway and presented as an intramolecular heterodimer on the cell surface (Bray, 2006). Receptor engagement by a ligand presented by a neighboring cell exposes an extracellular ADAM metalloproteinase cleavage site (S2) and a γ-secretase cleavage site (S3) in the transmembrane region, releasing the Notch intracellular domain (ICN), and transmitting Notch signaling into the nucleus (De Strooper et al., 1999). ICN associates with a DNA-binding complex founded on the mammalian homologue of the recombining binding protein suppressor of hairless (RBP-Jκ, also known as CBF1) (Fortini and Artavanis-Tsakonas, 1994) that helps recruit the MAML co-activators to RBP-Jκ binding sites (Nam et al., 2006; Wilson and Kovall, 2006). These events consolidate activation of gene expression programs promoting survival and cell growth, which when deregulated lead to leukemic transformation (Palomero et al., 2006; Weng et al., 2006). Notch signaling in lymphoid progenitors [from early thymic precursor (ETP) through “double negative” 2 (DN2) stage] controls the choice between B, myeloid and T cell fates (Radtke et al., 1999; Sambandam et al., 2005). At the DN3 stage, Notch signaling co-operates with pre-TCR signaling to expand thymocyte numbers undergoing β-selection (Ciofani and Zuniga-Pflucker, 2005; Garbe et al., 2006; Tan et al., 2005). Although Notch1 expression is maintained in DP thymocytes, Notch signaling is blocked at this stage. Signaling resumes in peripheral T cells, where it controls the choice between effector cell fates (Amsen et al., 2009). More than 50% of T-ALL (T-lineage acute lymphoblastic leukemia) have Notch1-activating mutations in the heterodimerization and C-terminal PEST domains (Weng et al., 2004).
Although the connection between Ikaros-loss-of-function and Notch activation is well established in mouse T-ALL, it remains unclear whether this interaction is direct or indirect. Loss of Ikaros may provide aberrant survival properties to differentiating thymocytes, predisposing them to further selection for activating Notch mutations and a malignant phenotype. Here, we described the use of alternative promoters within the Notch1 locus, one of which is upstream of the canonical promoter and a second downstream within an intragenic region encoding trans-membrane domains. These alternative promoters generated transcripts supporting a ligand-independent phase in Notch1 signaling and are active during T cell development and leukemogenesis. We showed that Ikaros directly regulated the epigenetic state and transcriptional output of canonical and alternative Notch1 promoters. Deregulation of this epigenetic process during T cell development caused aberrant activation of Notch1 signaling and rapid development of T-ALL.
To evaluate the role of Ikaros in Notch-mediated transcriptional responses, we combined either of two Ikaros mutant mouse genetic models [Ikaros null (Ikzf1-/-) or Ikaros Dominant Negative (Ikzf1DN/+)] with a conditional inactivation model for Notch1 that deletes a 3.7kb region encompassing the Notch1 promoter and exon1 (Georgopoulos et al., 1994; Radtke et al., 1999; Wang et al., 1996). Thymic Notch1 deletion with Lck Cre (Figure 1, Notch1fl/fl Lck Cre; Notch1-/-), which deletes from DN3 (Lee et al., 2001), or with CD2 Cre (data not shown), which deletes from ETP (de Boer et al., 2003), gave similar results.
T cell development was evaluated in young (3-4 weeks) Notch1 and Ikaros double–deficient (Notch1-/-Ikzf1-/-) and single-deficient mice. Hallmarks of Ikaros-loss-of-function in the thymus are a facilitation of the DN3 to DN4 transition and an increase in CD4 SP thymocytes (Winandy et al., 1999), whereas hallmarks of Notch1-loss-of-function are a partial block at the DN2-4 with accumulation of a DN2-3 intermediate (CD25hi), a decrease in DP and an increase in SP thymocytes (Figure 1A) (Radtke et al., 1999). In the young Notch1-/-Ikzf1-/- mice, a release of the previously described Notch1-loss-of-function mediated DN2-3 block (5.8% vs. 14.4%), a relative increase in DP (72% Notch1-/-Ikzf1-/- vs. 58% Notch1-/- or 59% Ikzf1-/-) and decrease in CD4 SP T cells (17% Notch1-/-Ikzf1-/- vs. 32% Ikzf1-/-) compared to single deficient mice were observed (Figure 1A). Notch1 deletion was detected in most of thymocytes in mutant mice (Figure 1B). As in the Notch1-/- or Ikzf1-/-, a decrease in thymic cellularity was detected in the Notch1-/-Ikzf1-/- mice relative to WT (Figure 1C). Notably, a reduction in CD8 SP was seen in the Notch1-/-Ikzf1-/- thymus with almost all remaining cells displaying an immature ISP phenotype (TCR−/int/HSAhi, Figure 1D) suggesting that Notch1 and Ikaros may cooperate in the development of the CD8 SP lineage (Robey, 1999).
Unexpectedly, from 6 weeks Notch1-/-Ikzf1-/- mice began to develop T cell leukemia at a faster rate than Ikzf1-/- mice (Figure 2A). This effect of Notch1 deletion was observed in all Ikaros mutant genetic backgrounds (Figure 2A and data not shown), and was especially prominent in mice heterozygous for the Ikaros null mutation, which usually undergo normal T cell development and develop disease at low frequency at advanced age (>6 months): the Notch1-/-Ikzf1+/- mutants developed leukemia at high penetrance within 2-5 months. Deletion of one Notch1 allele was sufficient to accelerate leukemogenesis, albeit at an intermediate rate compared with double deletion, indicating a gain of function provided by the mutant locus (data not shown). Unlike Notch1-/-Ikzf1-/- or Ikzf1-/- mice, Notch1-/- mice never developed disease, indicating that the initiating event for transformation is Ikaros-dependent. Accelerated tumor formation in Ikzf1+/- and Ikzf1-/- indicates the presence of an Ikaros-independent component, possibly related to deletion of a second negative regulator operating from the Notch1 promoter.
The leukemic clones arising in the Ikaros-Notch1 compound mutants (Notch1-/-Ikzf1+/-, Notch1-/-Ikzf1-/- and Notch1-/-Ikzf1DN/+) had a prevailing DP-transitional TCRint phenotype similar to that described in Ikzf1-/- mice (Figure 2B and data not shown). Genomic PCR and RT-PCR analysis of these clones confirmed deletion of the Notch1 canonical exon1 (E1c) (Figure 2C-D), and revealed that Hes1 and Il2ra (encoding CD25) are also induced suggesting activation of Notch signaling (Figure 2B and 2D). Notch1-/-Ikzf1-/- thymocytes in the pre-leukemic state (3-4 weeks, polyclonal) and leukemic state (>8 weeks, monoclonal) were further evaluated for Notch1 protein expression. Notch1-/- and pre-leukemic Notch1-/-Ikzf1+/- thymocytes showed a similar reduction in S1 cleaved Notch1 (~120 kDa) below WT amounts (Figure 2E, lanes 2 and 4). In contrast, Notch1-/-Ikzf1-/- or Notch1-/-Ikzf1DN/- leukemic clones expressed high amounts of short protein isoforms (<120 kDa) not detected in WT or in pre-leukemic Notch1-/-Ikzf1+/- thymocytes (Figure 2E, lanes 6, 7, 9). One of these isoforms was recognized by an antibody to ICN (Figure 2E, Val1744, lane 9). Thus, alternative promoters supporting Notch signaling are activated at the mutant Notch1 locus during loss-of-Ikaros-mediated leukemogenesis.
The relevance of alternative Notch1 isoforms to the leukemic phenotype of Notch1-/-Ikzf1+/- clones was investigated. Notch1-/-Ikzf1+/- leukemic clones lacking the canonical Notch1 isoform but expressing high amounts of alternative isoforms (Figure 2 C-E, lanes 6, 7 and 9) were monitored for their ability to grow in the presence of γ-secretase inhibitors (GSI). Reduced proliferation in these clones indicated dependence on the alternative Notch1 isoforms that are also cleaved at S3, like the canonical protein (Figure 2F-H). In contrast, an independently derived Ikzf1DN/+ clone failed to respond to GSI, indicating activation of an alternative transformation pathway (Figure 2F-H).
We next combined Ikzf1-/- or Ikzf1DN/+ mutants with a conditional inactivating mutation in Rbpjfl/fl CD2 Cre (Rbpj-/-), the nuclear effector of Notch signaling (Tanigaki et al., 2004). A marked decrease in thymocyte cellularity and survival was observed in Rbpj-/-Ikzf1DN/+ or Rbpj-/-Ikzf1-/- mice (3-21 weeks) (Figure 2I). Surviving Rbpj-/-Ikzf1-/- thymocytes had a predominant CD4+CD8neg-int TCRint-hi profile, similar to Ikzf1-/- (Figure 2J). However, unlike Ikzf1-/-, Rbpj-/-Ikzf1-/- thymocytes monitored from 3-32 weeks were unable to expand (Figure S1). Rbpj-/-Ikzf1DN/+ mice did not develop detectable lymphoma even after 32 weeks, a time by which most Ikzf1DN/+ mutant mice had succumbed to disease (Figure 2K and data not shown for Ikzf1-/-). In the one case in which an aberrant population was observed, the expanding cells carried a wild-type Rbpj allele (Figure 2K and data not shown). These genetic studies indicate that Notch signaling is required for the transition of Ikaros-deficient thymocytes to a leukemic state, and reveal the existence of alternative promoters in the Notch1 locus that generate signaling-competent Notch proteins that are actively repressed by Ikaros.
Notch1 alternative promoters and transcription start sites (TSS) were evaluated by RNA analysis and 5’RACE of polyA+ RNA from leukemic clones arising in the Notcht1-/-Ikzf1+/- and other genetic backgrounds (Figure 3). Notch1 transcripts (~9.5kb) containing exons 3-6 (Figure 3A and S2A) were detected in WT thymocytes (~85% DP) and in Notcht1-/-Ikzf1+/-, Notcht1+/-Ikzf1+/- and Notcht1-/-Ikzf1DN/+ leukemic clones (Figures 3B and S2B). 5’RACE from exon 3 revealed splicing to exon1a, which is 13kb upstream of exon1 (Figure 3A) (Tsuji et al., 2003). Transcripts (~9.5kB) containing E1a-E3 (E1a-E3 through E34) were detected in leukemic clones but not WT DP thymocytes (Figure 3B, E1a). RT-PCR and sequencing of E1a-E3 transcripts revealed differential use of a second coding exon downstream of exon1a, previously described as exon1b (Tsuji et al., 2003) (Figures 2D, ,3A3A and S2C). Analysis downstream of exon 24 (i.e. E25, E26, E28-31 and 34b), revealed four additional transcripts ranging in size from 3.5-5.5kb (Figures 3B-C and S2A-B). The four major short Notch1 transcripts arise from two intragenic TSS, one in exon 25 and a second in either intron 26 or 27 (Figures 3 and S2A,B), and two alternative polyA+ sites within exon 34 (Tsuji et al., 2003). 5’RACE from exon 34 verified intragenic promoter usage at exon 25 and downstream intronic sites (data not shown).
We also analyzed production of alternative Notch1 transcripts in leukemic clones that arise in Ikaros mutants with an intact Notch1 promoter-exon1 region. Alternative and canonical Notch1 transcripts were detected in most Ikaros mutant leukemic clones (Figures 3C and S2B-C). Activation of alternative Notch1 TSS, either upstream of the canonical promoter or in the exon 25-28 intragenic region, correlates strongly with loss of Ikaros and activation of Notch1 ICN and is independent of deletion of the canonical Notch1 promoter (Figures 3B, C and S2C-D).
We also examined whether the effect of Ikaros on alternative promoter use was specific for Notch1 or also manifested with other Notch family members. Although the canonical Notch3 transcript was elevated (2-5 fold) in Notch1-/-Ikzf1+/- and Ikzf1DN/+ leukemic cells, expression of alternative transcripts from the Notch3 locus was not detected (Figure S2D).
During T cell development, activation of Notch signaling manifests during the DN3 to DP transition, where it contributes to proliferation of immature thymocytes (Figure 4A, I would prefer to leave this one in as it provides a summary of the Ikaros-leukemogenesis process and the markers used to identify its three stages including activation of Notch1 canonical and alternative promoters) (von Boehmer, 2009)). Notch transcriptional targets, such as Hes1, Il2ra and Notch1 itself, are greatly induced during this phase. We examined the activity of canonical and alternative Notch1 promoters at these stages by testing for promoter-specific transcripts. E1c- and E1a-E3-containing Notch1 transcripts were expressed at DN3 but progressively declined at the DN4 and DP stages, similar to Hes1 and other Notch signaling transcriptional targets (Figure 4B-C). Thus, induction of canonical and 5’-alternative Notch1 promoters is part of the Notch transcriptional response at the DN3 stage of T cell differentiation. Quantitative RT-PCR revealed higher expression of E1a-E3 containing transcripts than E1c transcripts at DN3 (Figures 4B and S3A). However, at the DP stage, expression of E1c transcripts was maintained at a higher basal level (Figure 4B). RNA analysis of polyA+ RNA from DN thymocytes also indicated elevated expression of E1a transcripts at this stage (Figure 4D) but did not detect the intragenic transcripts, and further analysis by RT-PCR was not possible due to interference from pre-splicing mRNA (data not shown).
Activation of alternative Notch1 transcripts was next examined during leukemogenesis in Ikaros mutant thymocytes. Three major stages have been defined that demarcate the transition from a pre-transformed to a transformed state and are common to thymocytes with distinct Ikaros mutations (Figures 4A and S3B). Nonetheless, the rate at which different mutant cells undergo this process varies inversely with the levels of residual Ikaros and family members. The first stage is defined by Ikaros mutant thymocytes, which like WT thymocytes consist predominantly of DP with low expression of TCR (SI-DP TCRlo). Thymocytes from Ikzf1-/- mice at 2-3 weeks, Ikzf1DN/+ at <2 months and Ikzf1+/- at < 6 months are frequently classified as SI. The second stage is defined by an aberrant increase of cells with intermediate to high expression of TCR and a variable DP-transitional phenotype (SII- TCRint-hi CD4+CD8+ and CD4lo CD8+ and/or CD4+CD8lo). The third stage is also categorized by a predominant TCRint-hi DP-transitional population (frequently CD4lo CD8+) but with a progressive increase in Notch1 signaling evident from accumulation of ICN and increased levels of its transcriptional targets (SIII early-late TCRint-hi ICNlo-hi, CD25lo-hi, Hes1neg-hi) (Figures 4A, C, E and S3B). Notably, clonality among the expanding TCRint-hi mutant thymocytes was detected both before and upon Notch activation (Figure S3C, SI-III). Thus clonal expansions can be initiated in the absence of Notch signaling; however, the aggressive leukemic clones predominant at late stages are mostly Notch dependent.
Quantitative RT-PCR and RNA analysis of Ikaros mutant thymocytes detected modest elevation of the 5’ canonical and alternative Notch1 transcripts by SII (Figure 4D-F, E1c and E1a, 2-3 fold). By SIII, marked by ICN accumulation, E1a and E1c transcripts were further induced (Figure 4D-F, >10 fold). The increase in basal transcription from Notch1 promoters was also seen in Rbpj-/-Ikzf1DN/+ thymocytes, indicating that this is a direct effect of Ikaros and not caused by Notch signaling manifested in a minority of the population (Figure 4G). Activation of the Notch1 intragenic promoters lagged behind the E1a and E1c promoters, since their corresponding transcripts were first detected at low level in early SIII (Figure 4D, lane 6). These, together with the full-length (i.e. E1c- and E1a-) transcripts, were further induced upon ICN accumulation in late SIII leukemic clones (Figure 4D, lane 7 and 4E, lanes 5-7, 9).
Thus the 5’ alternative and canonical Notch1 promoters are part of a regulatory mechanism operative during both development and leukemogenesis. The increase in basal transcription from the 5’promoters in the pre-leukemic state coincides with reduced Ikaros activity. A further robust induction of both the 5’ and the intragenic Notch1 alternative promoters is detected at later leukemogenesis stages and coincides with ICN accumulation.
The structure and activity of proteins generated by the alternative Notch1 transcripts was assessed in cell culture. The proteins encoded by the canonical Notch1 transcript (E1c: exon1-exon34), the 5’ alternative transcript (E1a-E3: exon1a to exon3-exon34) and the longest intragenic transcript (E25:exon 25-exon 34) were expressed from a construct in which the C-terminus PEST domain was either replaced by a 6-Myc tag (Figure 5A-C, E1a-E3 6MT and E1c 6MT) or left intact (Figure 5D, E25 and E1c and data not shown for E1a-E3) (Kopan et al., 1996; Mizutani et al., 2001). Notch1 ICN with an intact PEST domain served as a positive control (Figure S4, ICN) (Aster et al., 2000). Similar results were obtained with these vectors in HEK 293T (Figure 5) and U2OS cells (data not shown). The low amount of ICN detected upon expression of the canonical Notch1 transcript in HEK 293T indicated low amounts of either ligand-receptor engagement or constitutive Noch1 S3 cleavage in these cells.
The E1a-E3 Notch1 transcript supported higher amounts of ICN production, nuclear localization, and activation of the RBP-Jκ reporter than the canonical transcript, although both expressed recombinant Notch1 proteins at similar levels (Figures: 5B-C and S4). GSI treatment reduced the amounts of proteins generated by the E1a-E3 transcript, indicating dependence of Notch signaling on processing (Figure 5B). In addition to Notch1 pro-protein (yellow star) the E1a-E3 transcript produced a cleaved isoform that migrated faster than the constitutively furin-cleaved protein (S1-green arrow) produced by the E1c transcript (Figures: 5A-B and S4). This protein was specifically recognized by an antibody to the S2 cleaved Notch1 (S2-Val1711-yellow arrow), normally induced upon ligand binding to the receptor (van Tetering et al., 2009). The apparently constitutive processing of the E1a-E3 protein by S2 and S3 (Val1744-purple arrow) was verified by immunoblot and immunofluorescence (Figure 5B-C). GSI increased the S2 product and decreased the S3 product, further supporting constitutive S2 cleavage of this isoform (Figure 5B). Consistent with its S2-S3 processing, the E1a-E3 isoform (but not E1c), was frequently detected in nuclei (Figure 5C).
The E25 transcript generated a range of short proteins that stimulate RBP-Jκ reporter activity through ICN production (Figure 5D). GSI reduced reporter activation by these isoforms, indicating that their activity is also dependent on S3 cleavage. Nonetheless, unlike E1a-E3 there was no accumulation of an S2 product, indicating translation downstream of this site (Figure 5D, and data not shown). The six putative translation initiation methionines in the E25 transcript were next evaluated by mutagenesis (Figures 5A, E and S5). Methionine to alanine substitution at position 1727 (E25: M1727A) abolished the major isoform (Figures 5E, red asterisk and S5). Interestingly, mutations of downstream methionines (M1796A, M1845A and M1848A; blue stars1-2) increased translation from M1727 and reporter activity, indicating that competition between alternative translation sites may modulate Notch signaling driven by the intragenic transcripts (Figure 5E, Figure S5 and data not shown).
The 5’ and intragenic Notch1 promoters thus generate stronger transcriptional activators than the canonical promoter. In both cases this is due to the production of proteins whose cleavage by ADAMs and/or γ-secretase is either accelerated or occurs constitutively when ligand is limiting.
Transcriptional output is partly determined by local enhancer and promoter chromatin regions that control access to the basal transcriptional machinery and regulators. We examined the histone modifications H3K4me3, H3K27me3, H3K36me3 and H3K9/14ac at the Notch1 locus in WT and Ikaros mutant thymocytes thymocytes at stage II (SII-DP TCRint-hi), prior to activation of Notch signaling. Combinations of these modifications provide a measure of transcriptionally active (H3K4me3, K9/K14ac and K36me3), poised (H3K4me3 and K9/K14ac; H3K4me3, K9/K14ac and K27me3; H3K4me3 and K27me3) and repressed (H3K27me3) chromatin (Campos and Reinberg, 2009; Mendenhall and Bernstein, 2008; Schones and Zhao, 2008). We used chromatin immunoprecipitation coupled to high-throughput sequencing (ChIP-Seq) to compare the chromatin status of genes in total thymocytes mostly at the DP stage (~85%) with that of genes expressed in DN, DP or SP populations. Enrichment in H3K36me3, a chromatin marker of long-range transcription elongation, was seen only in genes expressed in DP cells (Figure S6).
Islands of H3K4me3 and H3K9/K14Ac are found primarily in promoter regions and are in part generated by recruitment of the transcription initiation complex (Guenther et al., 2007). In WT DP, a major H3K4me3/H3K9/K14ac island was detected at the canonical Notch1 promoter (Figure 6A). A minor island was detected ~13kb upstream, spanning E1a and further validating this region as an alternative 5’ promoter (Figure 6A). No H3K4me3 or H3K9/K14ac islands were detected in the Notch1 intragenic region in WT DP thymocytes. This indicates that the intragenic promoters are neither active or poised, consistent with Notch1 transcript analysis at the DN and DP stages (Figure 4B-D). H3K36me3, a marker of long-range transcription elongation, was detected downstream of the canonical TSS, in line with its reported accumulation after transition from short-term to long-term elongation (Joshi and Struhl, 2005). H3K27me3, which marks a restrictive chromatin state, was not detected at the Notch1 locus (data not shown).
ChIP-Seq for Ikaros identified two high confidence Ikaros binding sites (IkBS) (p-value 105) ~ 10kb upstream and downstream of the canonical TSS (Figure 6A, IkBSN1 and IkBSN4, purple). The upstream Ikaros site is within 3kb of the E1a promoter. Notably, both sites are in minor H3K4me3 islands. Four lower confidence IkBS (p-value 10-3) were also found: two at the E1c promoter one at the E25 promoter (Figure 6A, IkBSN2, IkBSN3, IKBSN4’, IkBSN5, blue). ChIP-PCR validated the high confidence Ikaros sites as well as two of the lower confidence sites (IkBSN2 and IkBSN5).
We next analyzed the chromatin status of the Notch1 locus in Ikzf1-/- pre-leukemic thymocytes before activation of Notch signaling (Figure 6B, Stage II: ICN-). Cell staging was validated by testing for H3 modifications at Notch1 targets such as Hes1 and Il2rα. Both genes lacked the elongation marker H3K36me3 (Figure 6B). Furthermore, Hes1 retained H3K27me3 and H3K4me3 throughout the locus, consistent with a bivalent and repressed state in DP cells. In contrast, the Notch1 locus displayed an increase in H3K4me3 and H3K9/K14ac islands spanning the 5’ canonical (E1c) and the alternative (E1a) promoters (Figure 6B-C). Increased intensity and range of H3K36me3 was also seen, consistent with the increase in basal transcription from the E1c and E1a promoters in SII mutant thymocytes. In addition to the increase in H3K4me3 and H3K9/K14ac at the 5’end of Notch1, a group of de novo H3K4me3 and H3K9/K14ac islands was detected at the intragenic region, starting at E25 and extending into the 3’ end (Figure 6B). These H3K4me3 and H3K9/K14ac islands encompass the intragenic promoter region that is activated upon ICN accumulation at SIII of leukemogenesis. These data highlight that the genomically-intact Notch1 locus acquires a more permissive chromatin configuration at its canonical and alternative promoters before their robust induction by ICN.
Given reports that Ikaros and ICN work through the same DNA sequences (Beverly and Capobianco, 2003), we examined the distribution of ICN at the IkBS of Notch1 in SIII leukemic cells by ChIP-PCR (Figure 6D). Strong ICN enrichment was detected at IkBSN1 but not at downstream sites (Figure 6D). As previously reported ICN was highly enriched at the Hes1 promoter. Binding of ICN at IkBSN1 was confirmed by ChIP-seq, which also revealed a striking distribution within the intragenic methylation/acetylation islands starting downstream of IkBSN5 in E25 (Figure 6D).
These epigenetic studies provide insight into the transcriptional regulation of Notch1 promoters in WT, pre-leukemic and leukemic thymocytes and establish a direct role for Ikaros in restricting chromatin accessibility at these regulatory sites.
The Notch1 promoter regions and Ikaros binding sites were next evaluated in silico for cis-element composition (Figure 7). A large CpG island surrounded E1c, as is typical of 50-70% of mammalian promoters (Sandelin et al., 2007). Coincident with the CpG island was an array of GC boxes, which can bind the transcriptional activator Sp1. The 5’ alternative promoter E1a did not have these elements, suggesting it is normally weak and may need recruitment of additional factors for activation (Figure 7A). Strong cross-species conservation was detected for the E1a promoter region, supporting its functional importance (data not shown). A set of classic promoter elements is found in the intragenic region, particularly upstream of E28. This intragenic TSS contains a CpG island, TATA, CCAAT and GC boxes at appropriate upstream distances, indicating capacity for strong transcription when this region is accessible (Figure 7A, lower panel). This is consistent with the evidence that binding of Ikaros and possibly other factors is necessary to restrict access to this site.
Sequence analysis identified many potential transcription factor binding sites, including several clusters of E2A motifs within or near IkBS sites (Figure 7A). One site is directly upstream of the canonical promoter close to IkBSN2, two are just downstream of the canonical CpG island and one is within IkBSN4, the IkBS 10kb downstream from the promoter (not shown). Several E2A sites are located directly upstream of the TSS in E1a. In the region of the intragenic promoters, a cluster of E2A sites extends downstream of IkBSN5 in E25 into E26, another cluster is found in intron 26, and another in intron 29. Given the abundance of E2A sites and their proximity to Notch1 regulatory regions, we examined Notch1 promoter use in E2A deficient (Tcf3-/-) primary leukemic clones and cell lines with active Notch signaling (Figures 7B and S7) (Reschly et al., 2006). Alternative Notch1 transcripts predominated in these clones, indicating that Ikaros and E2A participate in a common mechanism of leukemic transformation. E2A binding to multiple regulatory elements in the Notch1 locus, especially sites at the deleted canonical Notch1 promoter region, provide a plausible explanation of the acceleration of leukemogenesis upon Notch1 promoter deletion in Ikaros mutant thymocytes.
Here we showed that the Notch1 locus is regulated by a combination of transcriptional promoters and enhancers supporting a feed-forward loop that likely augments Notch signaling at appropriate stages of T cell differentiation and leukemic transformation. Ikaros is one of the key regulators in this process working at the epigenetic level to limit recruitment of transcription-related factors.
The presence of alternative promoters at the Notch1 locus was first revealed by studies in leukemic cells from mice with combined mutations in Ikzf1 and the Notch1 canonical promoter. Although activity of the alternative Notch1 promoters was augmented in the Notch1-/-Ikzf1+/- leukemic cells, it was also elevated in other leukemia models in which the canonical Notch1 promoter was intact. Nonetheless, combined deletion of the Notch1 canonical promoter and Ikzf1 accelerated this process, indicating participation of additional negative regulatory factors acting through the deleted region. A likely candidate is E2A, because it binds at the canonical Notch1 promoter (Yashiro-Ohtani et al., 2009) and has similar effects as Ikaros on Notch1 transcriptional regulation and leukemogenesis. These studies also show that in T cells, the Notch1 exon1 deletion is not a null allele, particularly under conditions where active Ikaros repression is relaxed. Although it is possible that this allele is an effective null in other contexts, it will be important to assess whether deletion of negative regulatory elements and de-repression of these ligand-independent isoforms occurs in other cell types.
In addition to the canonical promoter, two alternative Notch1 promoter regions were mapped at the 5’ and at an intragenic location. Transcriptional analysis of the Notch1 locus during normal T cell development indicated that at the DN3 stage both the 5’ alternative and canonical promoters were active, whereas at the DP stage only the canonical promoter retained activity. Epigenetic studies in DP thymocytes confirmed the presence of permissive chromatin at the active promoter but also indicated that the inactive 5’ alternative promoter was in a permissive chromatin configuration possibly poised for future transcriptional induction upon activation of Notch signaling. In contrast to the 5’ promoters, the intragenic promoter region was transcriptionally repressed through the DN3 to DP transition, was not marked for transcriptional activation at the chromatin level and displayed extensive Ikaros binding.
During loss of Ikaros-mediated leukemogenesis, all three sets of Notch1 promoters acquired more permissive chromatin prior to activation of Notch signaling. Increase in permissive chromatin was detected at the 5’ alternative and canonical promoters and most notably at the intragenic promoters in pre-leukemic thymocytes on an intact Notch1 locus. The increase in chromatin accessibility at the 5’ Notch1 promoters correlated with an increase in basal transcription detected at the pre-leukemic state that preceded the strong transcriptional induction detected upon ICN accumulation at the leukemic state. Importantly, inhibition of Notch signaling, through γ-secretase inhibition or by Rbpj inactivation, interfered with the robust transcriptional induction at both the canonical and alternative promoters in leukemic cells but did not alter the increase in basal transcription observed in pre-leukemic thymocytes.
Both alternative promoters were responsible for the production of proteins that signal in limiting ligand. The full-length Notch1 protein generated by the E1a promoter lacks the signal peptide present in the canonical unprocessed form and appears to be processed differently. Notably, it is constitutively cleaved by ADAM metalloproteinases (S2) and γ-secretase (S3) in limiting ligand conditions. Thus, alternative 5’ promoter and exon usage may dictate an alternative route of cell trafficking and receptor processing that enhances Notch signaling that is pending further investigation. A range of short Notch1 protein isoforms was produced by the intragenic promoter region. A major isoform was translated from a site downstream of the S2 or ADAM site that lacked the ligand binding and NRR domains, and was a potent activator of the Notch transcriptional response.
Given these data we propose the presence of a feed-back loop in Notch signaling supported by a network of epigenetic and transcriptional regulators and Notch receptors with differential ligand-dependence for activity. Local chromatin at the Notch1 locus controls access to the basal transcription machinery and to enhancer proteins that regulate this process. The presence of Ikaros at binding sites located in proximity to all three Notch1 promoters is responsible for restricting chromatin. Ikaros association with negative chromatin remodeling factors such as Mi-2β and HDACs may restrict access or activity of positive chromatin regulators such as MLL and HATs also recruited to these sites (Kim et al., 1999; Sridharan and Smale, 2007). Loss of Ikaros relieves chromatin restriction increasing access to the basal transcription machinery at both ligand-dependent and -independent Notch1 promoters. This precipitates a forbidden increase in ICN, which even if present at low amounts is more effectively recruited to its target sites that are more accessible due to Ikaros removal. ICN is a potent transcriptional enhancer that can function from at least two regions in the Notch1 locus. One is at the promoter-distal IkBSN1, and the second is at an extended region downstream of the IkBSN5 in E25. Importantly, this ICN binding region shows extensive overlap with the de novo intragenic islands of permissive chromatin arising in Ikaros mutant pre-leukemic cells. Interactions of ICN with HATs (Wallberg et al., 2002) at its target sites may further impact transcription initiation. Transcription of the Notch1 locus is likely effected by a number of regulatory factors. In addition to Ikaros and E2A, our preliminary studies indicate a similar aberrant activation of alternative Notch1 promoters in abnormally expanding DP in an activated Akt2 transgenic model (data not shown) (Malstrom et al., 2001). Thus, both nuclear and signaling factors implicated in leukemogenesis may participate in the regulation of this feed-forward loop in Notch signaling.
Stage-specific activation of the Notch1 promoters may be one key for modulating levels of Notch signaling during development and leukemogenesis. The induction of promoters expressing isoforms with differential ligand requirement may support a feed-forward mechanism that augments Notch signaling required for expansion of immature thymocytes. At the DN3 stage, this may be jump-started by the ligand-dependent canonical Notch1 E1c isoform and propagated by the more active E1a isoform. Subsequently at the DP stage, the E1a promoter is repressed and only the canonical promoter remains modestly active. Together with limited ligand availability, this may restrict Notch signaling to levels supporting cell survival and not proliferation at this stage of differentiation (Laky and Fowlkes, 2008). In the Notch1–deficient cells, deletion of the canonical promoter and lack of expression of the ligand-dependent Notch1 isoform prevented activation of the 5’ alternative promoter under conditions of restrictive chromatin. However, upon loss of Ikaros (and possibly E2A), aberrant increase in chromatin accessibility augments basal transcription at the alternative Notch1 promoters, marking transition to a pre-leukemic state. This allows a progressive increase in ICN accumulation that eventually causes robust induction of both ligand-dependent and -independent Notch1 promoters, thus sealing the transition to the leukemic state. Aberrant activation of the Notch1 promoters collaborates with Notch1 mutations that target the PEST domain in the mouse genetic systems studied here to exacerbate the leukemia phenotype. It will be important to determine whether such a mechanism is also manifested in human leukemia where a predominance of mutations in the heterodimerization and PEST domains has been described. Strong cross-species conservation of both the canonical and alternative promoters gives support. Targeting the transcriptional regulation of Notch1 may open new avenues for leukemia treatment. Small molecules targeting enzymes that control chromatin accessibility in combination with γ-secretase inhibitors may provide better treatment protocols for curing T-ALL.
Ikzf1-/-, Ikzf1DN/+ (Georgopoulos et al., 1994), Notch1fl/fl (Radtke et al., 1999), and Rbpjfl/fl mice (Radtke et al., 1999; Tanigaki et al., 2002; Wang et al., 1996) were bred and housed under pathogen-free conditions. The Notch1 and RBP-Jκ deficient strains were maintained as intercrosses to either lck-Cre or hCD2-Cre (de Boer et al., 2003; Lee et al., 2001).
Notch1-/-Ikzf1+/- and other mutant cell lines established from primary thymic lymphomas were maintained in RPMI media as described in the supplement.
Phenotypic analysis of thymocytes and splenocytes was performed by flow cytometry using FACSCanto (BD) or MoFlo (Cytomation). Data was analyzed with the FloJo software (Tristar). The antibodies used in this study were CD4-PE-Cy5.5 (or CD4-PE-Cy7), CD8α-APC-Cy7, TCRβ-APC, CD5-PE, CD25-PE, HAS-PE, CD69PE, Thy1.2-PE, CD62L-APC and CD44-APC as previously described (Williams et al., 2004).
TCRβ D-J rearrangement as well as deletion at the Notch1 and Rbpj loci were examined by genomic PCR as previously described (Winandy et al., 1995).
The cDNA was synthesized from RNA using SuperScriptII RTase (Invitrogen). For qRT-PCR, transcripts were amplified with HotStart-IT SYBER Green qPCR Master mix (USB) on the ABI sequence detection system (Applied Biosystems). All qRT-PCR reactions were performed in triplicate.
For RACE the first strand cDNA was synthesized with 1 μg polyA+-selected RNA using the SMARTER RACE cDNA Amplification kit (CLONTEC) according to the manufacturer’s instructions. For Northern analysis, standard protocols were used.
The antibodies used for these studies were anti-Notch1, anti-c-Myc (9E10) (Santa Cruz), anti-cleaved Notch1 (S2, Val1711) (van Tetering et al., 2009), anti-Cleaved Notch1 (S3, Val1744) (Cell Signaling Technology, anti-PSF (B92) (Sigma) or anti-Mi-2β (Williams et al., 2004). For immunofluorescence (IF), cells were grown and transfected with the indicated plasmids on chamber slides. Forty eight hours after transfection, cells were fixed with 4% PFA and processed for IF as previously described (Gomez-del Arco et al., 2005). Images were taken with a Nikon A1R confocal microscope.
Ikaros ChIP was performed as described previously (Harker et al., 2002) with some modifications. The following antibodies were used to study histone modifications: anti-H3K4me3 (Abcam ab8580 or Millipore 07-473), anti-H3K9/K14Ac (Millipore 06-599), anti-H3K27me3 (Millipore 07-449), and anti-H3K36me3 (abcam ab9050). For ChIP-sequencing, chromatin from 2×107 to 2×108 cells was used per IP. Precipitated DNA and an equivalent amount of whole genomic DNA were amplified using DNA Sample Prep Kit (Illumina or New England Biolabs). Amplified samples were run on the Genome Analyzer at Systems Biology Lab, Harvard University. Image analysis and base calling were performed using the Illumina Pipeline v1.6. Single end 32bp alignment with mouse mm9 assembly was conducted with ELAND software. Antibody enriched regions and signal peaks were identified by MACS algorithm with input as control (Zhang et al., 2008). For data visualization, properly shifted tags were counted along the chromosome, and tag counts in wiggle format were uploaded into Affymetrix IGB.
The following vectors were used for expression and reporter activity studies: a) pGALuc, a Notch activity reporter plasmid with 6 tandem RBP-J binding sites driving Luciferase expression was kindly provided by Dr. Adolfo Ferrando (Columbia University, NY) with permission of Dr Honjo (Mizutani et al., 2001). b) pCS2-E1c [mNotch1-canonical full Length Notch1] and pCS2-E1c 6MT [canonical full Length Notch1 with 6 tandem myc-tag epitopes inserted after amino acid 2183 at the C-terminal part of the protein] were kindly provided by Dr Kopan (Kopan et al., 1996). c) MIG and MIG ICN retroviruses were kindly provided by Drs. Marisa Toribio and Warren Pear (Aster et al., 2000). d) MIG E1c. The E1c cDNA was cloned from the pCS2 vector into the MIG RI vector. e) MIG E1a, PCS2 E1a and PSC2 E1a 6MT. Notch1 cDNA with the alternative E1a-E3 5’ end was cloned into MIG RI, pCS2 and pCS2 6MT. e) MIG E25 Notch1. A cDNA representative of the intragenic transcript spanning exon 25 through exon 34 was cloned into the MIG RI retrovirus.
The Notch1 expression plasmids were co-transfected in HEK 293 or U2 OS cells, together with the RBP-J Luciferase-reporter pGALuc and 24 h later cells were harvested for luciferase and protein analysis.
All primers used for RT-PCR, 5’RACE, probe generation and candidate ChIP are listed in Supplementary Table 1.
This work was supported by NIH-R01-AI1380342 to K.G and by SAF2009-10708 to CNIC and J.M.R. The CNIC is supported by the Spanish Ministry of Science and Innovation and the Pro-CNIC Foundation. PGA is supported by a Ramón y Cajal Grant and A.F.J. by an NIH training grant in Transplantation Biology. We thank Drs. Honjo, Kopan, Ferrando, Pear, Toribio, Gounari and Dose for providing reagents used in this study and to Bob Czyzewski for mouse husbandry. We also thank Drs. Morgan, Dose, Joshi and Yoshida for review of the manuscript.
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.