The IKAROS Interaction with a Complex Including Chromatin Remodeling and Transcription Elongation Activities Is Required for Hematopoiesis

Summary

Perturbation of the expression level of IKAROS, a transcription factor critical during hematopoiesis, is associated with malignant transformation in mice and humans. The importance of IKAROS expression levels for the control of target-gene regulation was addressed in hematopoietic progenitor cells. The collaboration between IKAROS and the Nucleosome Remodeling and Deacetylase (NuRD) complex can promote gene activation or repression. IKAROS can also interact with the Positive-Transcription Elongation Factor b (P-TEFb) and the Protein Phosphatase 1 (PP1), an important P-TEFb regulator. Immunoaffinity purification of IKAROS interacting proteins and Fast Protein Liquid Chromatography analysis revealed a dynamic interaction between IKAROS, PP1 and the newly defined NuRD-P-TEFb complex. This complex can be targeted to specific genes in cells expressing high levels of IKAROS to promote productive transcription elongation. Based on our results we suggest that, in addition to P-TEFb, the NuRD complex and PP1 are required to facilitate transcription elongation of IKAROS-target genes and normal differentiation of hematopoietic progenitor cells.

Introduction

The tumor suppressor IKAROS is a transcription factor critical for hematopoietic multi-lineage priming, cell fate and lineage determination [1]–[5]. Mice homozygote for the Ikaros null mutation (Ik^NULL) display severe defects in lymphocyte development and function, and develop leukemias and lymphomas with complete penetrance [6]. These phenotypes reflect the requirement of IKAROS to activate the lymphoid program in hematopoietic progenitor cells (HPCs) [4]. IKAROS is also involved in transcriptional regulation of erythroid- and myeloid-specific genes [7]–[13]. The hematopoietic differentiation is affected not only by the presence or absence of IKAROS, but also by its relative expression level [14]. In particular, during B-cell progenitor differentiation, dynamic change of IKAROS expression level has been identified as a key regulator for the expression of multiple target genes [15], [16].

IKAROS controls chromatin organization mainly through association with the Nucleosome Remodeling and Deacetylase (NuRD) complex [5], [17], [18]. NuRD was initially identified as a repressive complex but it was demonstrated afterwards to promote transcription of specific genes as well [5], [19]–[22]. It remains to be defined how this HDAC-containing complex activates transcription. IKAROS contributes to the assembly and stability of the pre-initiation complex (PIC) at promoters [13], [23]–[26] and interacts directly with CDK9, the catalytic subunit of P-TEFb (Positive-Transcription Elongation Factor b) [27]. CDK9 phosphorylates the C-terminal domain (CTD) of the large subunit of RNA Polymerase II (POL II) on Ser2 as well as the SPT5 subunit of DSIF and the E subunit of NELF. These events are required to release promoter-proximal paused POL II and thus, allow productive transcription elongation. Most nuclear P-TEFb is sequestered in the 7SK snRNP repressive complex. This repressive complex is characterized by the snRNP molecule and the proteins HEXIM (HEXIM1 or 2), LARP7 and MePCE [28]. Of interest here, is the dissociation of the P-TEFb from this repressive complex promoted by the sequential activity of the protein phosphatase 2B (PP2B) that favors conformation changes of the 7SK snRNP and protein phosphates 1α (PP1α), involved in CDK9 dephosphorylation at different residues including Thr186 and Ser175 [29], [30]. Dephosphorylated CDK9/P-TEFb is preferentially recruited to promoters by the general factor BRD4 or specific transcription factors such as HIV TAT [31]–[33]. Then, CDK9/P-TEFb becomes catalytically active and promotes the release of promoter-proximal paused POL II when it is “re-phosphorylated” by the TFIIH associated CDK7 [30]. PP1α is one of the three catalytic subunits (α, β or γ) which, together with a regulatory subunit, forms each PP1 enzyme [34]. Interestingly, IKAROS interacts with PP1 and is dephosphorylated by this phosphatase [35]. Whether the IKAROS-PP1 interaction is important for Cdk9/P-TEFb activation and thus, transcription elongation of IKAROS-target genes is not known.

Here, we sought to define the importance of these protein associations for IKAROS and NuRD to function as transcriptional activators. We demonstrate that IKAROS is an adaptor molecule required for the recruitment of the newly identified NuRD-P-TEFb complex to IKAROS-target genes. IKAROS binding to the promoter region of specific genes is also associated with the local recruitment of the CDK9/P-TEFb activator, PP1α. Interestingly, the Mi2/NuRD occupancy at IKAROS-target genes is enhanced when transcription elongation is proficient, and the release of POL II promoter-proximal pausing is decreased in the absence of Mi2. Our data also reveal that the dynamic interaction of IKAROS with the NuRD-P-TEFb complex depends on IKAROS protein levels. Low-levels of IKAROS suffice for Mi2/NuRD recruitment to promoters, but higher levels of IKAROS are required for NuRD-P-TEFb and PP1α to chromatin, CDK9/P-TEFb activation, and productive transcription elongation. Finally, we demonstrate that IKAROS-PP1α protein interaction is required for normal hematopoietic differentiation of primary HPCs.

Results

IKAROS interacts with the newly characterized NuRD-P-TEFb complex

Several studies have demonstrated physical interaction between IKAROS and NuRD [5], [17], [18], [36] or IKAROS and P-TEFb [23], [27]. To define how IKAROS coordinates chromatin remodeling and transcription elongation activities, we first conducted tandem immunoaffinity purification and mass spectrometry (LC-MS/MS) analysis in Jurkat cells expressing Flag- and HA-tagged IKAROS (Flag-HA-Ik) (Figures 1A, S1A). LC-MS/MS of Jurkat/Flag-HA-Ik eluates indicated that IKAROS associates with PP1α NuRD (Mi2, MTA2, RBBP4 and MBD3) and P-TEFb (CDK9, CYCT1) components (Table 1), whereas the P-TEFb inhibitor HEXIM1 and the negative elongation factor NELF were not amongst the purified proteins. In addition, HEXIM1 did not immunoprecipitate with IKAROS (Figure S1B), thus suggesting that IKAROS interacts with the elongation-competent and catalytically active P-TEFb.

To determine whether IKAROS interactions with Mi2/NuRD and CDK9/P-TEFb define a single complex or multiple complexes, Flag-HA-Ik immunoaffinity-purified complexes were analyzed by size exclusion chromatography. Flag-HA-Ik co-fractionated with NuRD and P-TEFb components as a major peak of ∼2 MDa (Figure 1B). Additionally, Mi2, CDK9, CYCT1, PP1α and IKAROS co-eluted in nuclear extracts of wild type Jurkat cells (Figure 1C). Since IKAROS mainly eluted at 0.5 MDa (Figure 1C), it is likely that only a minority of the total nuclear IKAROS stably associates with the NuRD-P-TEFb complex. Accordingly, silver staining of IKAROS-purified complexes (Figure 1A) allowed the detection of two weak, but above-background, protein bands migrating at the range of the two known CDK9 isoforms [37] (Figure S1C). That only a small fraction of the nuclear CDK9/P-TEFb complex is highly active and capable of stimulating transcription elongation rate has also been shown in different systems [38], [39].

Next, we investigated whether the interaction of IKAROS with CDK9 or Mi2 could be differently modulated by fluctuation of IKAROS protein levels. Protein co-immunoprecipitation (co-IP) experiments were carried out in Ikaros knock-down Jurkat cells (Figure S1D). Co-IP results indicated that IKAROS-CDK9 interaction is more sensitive to IKAROS levels than the IKAROS-Mi2 interaction because IKAROS-CDK9 interaction is impaired in Ikaros knock-down Jurkat cells whereas IKAROS-Mi2 interaction is comparable in non-target as well as Ikaros knock-down Jurkat cells (Figure S1E). This suggests that in hematopoietic cells a fraction of nuclear IKAROS proteins can be dynamically assembled in a multifunctional protein complex containing both NuRD and P-TEFb components and that formation and/or stability of IKAROS-P-NuRD-P-TEFb interaction is influenced by the expression level of IKAROS.

Protein interactions between NuRD and P-TEFb components were then assessed by protein co-IP of Jurkat nuclear extracts. The NuRD-associated proteins Mi2, RBBP4, MTA2, and MBD3 interacted with the P-TEFb components CDK9 and CYCT1 (Figure 1D, S1F), whereas none of these factors interacted with the negative control nuclear protein PCNA (Figures 1D, E). Among the NuRD-associated proteins, only RBBP4 weakly interacted with the general P-TEFb activator BRD4 [40] (Figures 1D, S1G). CDK9, CYCT1, Mi2, MTA2, RBBP4, and MBD3 also co-immunoprecipitated with IKAROS (Figure 1D). DNase1 did not affect Mi2-CDK9 protein interaction (Figure S1H), indicating that DNA is dispensable. Thus, tandem immunoaffinity purification, size-exclusion chromatography and co-IP experiments demonstrate that IKAROS can assemble with the newly defined NuRD-P-TEFb multifunctional complex (Figure 1F).

The simultaneous recruitment of P-TEFb and NuRD is IKAROS dose-dependent

We examined whether IKAROS could be a licensing factor for NuRD-P-TEFb complex recruitment to chromatin and whether the variation of IKAROS thresholds, which naturally occurs during hematopoiesis [5], [15], [41], could influence NuRD-P-TEFb association with IKAROS and thereby, regulate transcription elongation. To clarify this issue we studied two IKAROS-target genes, c-Kit and Flt3 (Figure S2A), which are cell receptors critical for HPC survival that are downregulated in Ik^NULL HPCs [42]. We used: (i) bone marrow lineage negative (lin⁻) HPCs obtained from Ikaros wild type (Ik^WT), Ikaros heterozygote null (Ik^HT) or Ikaros homozygote null (Ik^NULL) mice [6]; and (ii) the mouse G1E2 cells, which are proliferating hematopoietic progenitor-like cells [43]. The expression level of IKAROS in G1E2 cells was sufficient to detect Mi2-CDK9 interaction (Figure 2A).

c-Kit and Flt3 expression correlated with the level of Ikaros expression in lin⁻ HPCs (Figure 2B). Similarly, the expression level of c-Kit and Flt3 was significantly decreased in Ikaros knock-down G1E2 cells (Figure 2C). However, the expression level of additional c-Kit regulators, such as Gata2 [44], Cdk9 and Mi2 (see below) was not reduced in Ik^NULL or Ikaros knock-down G1E2 cells (Figures 2B, C and S2B).

To investigate whether reduced c-Kit expression is a direct effect of the Ikaros loss, we carried out chromatin immunoprecipitation (ChIP) assays. ChIP experiments indicated that IKAROS was recruited to critical regulatory regions of the c-Kit locus [44] e.g., the −114 Kb enhancer, the c-Kit Transcriptional Start Site -TSS-, and +1 Kb, +5 Kb as well as +58 Kb regions of the Open Reading Frame (ORF) in Ik^WT lin⁻ HPCs and G1E2 cells (Figure 2D, E). The same regions were occupied by CDK9, and Mi2 in Ik^WT lin⁻ HPCs, whereas the reduced level of IKAROS affected CDK9 and Mi2 recruitment at c-Kit (Figure 2D, E) but not at c-Fos promoter (Figure S2C), which was selected since c-Fos gene is expressed in HPCs but it is not regulated by IKAROS [45]. ChIP performed with TBP or POL II specific antibodies revealed that the PIC is assembled at c-Kit TSS regardless of Ikaros expression level (Figure 2D, E). However, the recruitment of POL II and POL II with phosphorylated Ser2 at the C-terminal domain (PCTD) at c-Kit ORF was reduced in Ik^HT, Ik^NULL lin⁻ HPCs as well as the Ikaros knock-down G1E2 cells (Figure 2D, E), suggesting that IKAROS can positively control c-Kit transcription elongation. This assumption was supported by the significant decrease of CDK9 binding to c-Kit TSS and ORF (Figure 2D, E) and the consequent reduction of c-Kit nascent transcript levels towards the 3′ end of the gene (Figure 2F, G) in Ikaros-deficient cells.

The correlation between IKAROS expression level and the release of promoter-proximal paused POL II was further demonstrated by the analysis of c-Kit traveling ratio, which is determined by the relative ratio of POL II density in gene ORF vs. promoter-proximal regions [46], [47]. As indicated in Table 2, c-Kit traveling ratio decreased in Ik^HT and Ik^NULL lin⁻ HPCs, when compared to Ik^WT.

To address whether IKAROS, CDK9 and Mi2 simultaneously co-occupy the c-Kit TSS, sequential ChIP (re-ChIP) [48] was carried out. Sequential ChIP analysis provided the evidence that in hematopoietic progenitors, these proteins can associate to the c-Kit TSS together (Figure 2H), hence suggesting that transcription elongation is promoted by the IKAROS-mediated recruitment of the NuRD-P-TEFb complex.

Collectively, these results indicate that in HPCs, IKAROS facilitates productive transcription elongation at c-Kit (Figure 2) and Flt3 (Figure S2D, E; Table S1) genes and that reduced IKAROS levels, which do not affect the PIC assembly, can perturb transcription elongation.

The IKAROS-NuRD-P-TEFb complex assists POL II during transcription elongation

Since IKAROS, CDK9 and Mi2 were detected at c-Kit TSS and ORF in Ik^WT HPCs (Figure 2D, E), we tested whether these factors play an active role during transcription elongation or associate with chromatin at c-Kit ORF independently of productive transcription elongation. Thus, transcription elongation was inhibited with Flavopiridol, which is a specific inhibitor of CDK9 activity [49]. G1E2 cells were treated with 100 nM Flavopiridol, for a short period of time (2 hours) to avoid indirect effects due to general inhibition of gene expression [50]. As expected, CDK9, POL II and CYCT1 protein levels were not reduced in G1E2-treated cells (Figure 3A), and IKAROS phosphorylation [35] appeared to be unmodified (Figure S3A). However, c-Kit and Flt3 nascent transcript levels were significantly reduced (Figures 3B, S3B). Flavopiridol treatment of G1E2 cells did not affect IKAROS recruitment to the c-Kit (Figure 3C) or Flt3 (Figure S3C) promoter region. Furthermore, decreased detection of POL II within the gene ORF, POL II accumulation at c-Kit TSS, and lack of PCTD association at c-Kit TSS and ORF, confirmed that Flavopiridol did not affect POL II loading, hence PIC assembly (Figures 3C, S3C) but substantially reduced transcription elongation (Table 2). Similar results (reduced c-Kit expression, unchanged IKAROS recruitment and lack of PCTD association to c-Kit promoter) were obtained in Cdk9 knock-down G1E2 cells (Figure S3D–F). Thus, transcription elongation block (obtained either by Flavopiridol treatment or Cdk9 knock-down) does not affect loading of IKAROS, CDK9/P-TEFb and Mi2/NuRD at c-Kit TSS region, but impedes P-TEFb and, surprisingly, IKAROS and NuRD recruitment within the gene ORF.

Since the elongation block by Flavopiridol led to an accumulation of Mi2 at c-Kit TSS and reduced association with the downstream regions (Figure 3C), we assessed whether NuRD could directly influence transcription elongation by knock-down of one of the NuRD components, Mi2, in G1E2 cells. The Mi2 knock-down cells showed a reduction of c-Kit and Flt3 mature as well as nascent transcripts (Figure 3D, E). The fact that Mi2 knock-down had a greater effect on transcripts generated at distal rather than proximal regions, suggests that the NuRD complex can assist and facilitate POL II progression during transcription elongation.

IKAROS contributes to PP1α association with the NuRD-P-TEFb complex

IKAROS is a substrate of PP1 activity [35] and PP1α can act as a positive regulator of CDK9/P-TEFb [29]. IKAROS interacts with the α catalytic subunit of PP1 in G1E2 cells (Figure 4A), in COS-7 cells transfected with IKAROS (Figure S4A) and in Jurkat cells (Table 1). Among the PP1 catalytic subunits, we identified PP1α as the most relevant and abundant interacting partner of IKAROS (Table S2). Thus, we investigated whether, by binding to PP1α and CDK9, IKAROS could deliver PP1α to the large P-TEFb complex, thereby activating CDK9/P-TEFb and favoring transcription elongation. In accordance with this assumption, PP1α recruitment was demonstrated at c-Kit TSS in Ik^WT and much less efficiently in Ik^HT or Ik^NULL lin⁻ HPCs (Figure 4B). PP1α recruitment to c-Kit TSS also decreased in Ikaros knock-down G1E2 clones (Figure 4C), even though PP1α expression was similar to control cells (Figure 4D).

The role of the IKAROS-PP1α interaction for CDK9/P-TEFb activation was further delineated by exploring the influence of Calyculin A, a PP1/PP2A inhibitor [34], on c-Kit and Flt3 transcription elongation. Short-term treatment with Calyculin A did not cause toxic effects or cell death (Figure S4B), but induced significant reduction of c-Kit nascent transcripts (Figure 4E), whereas transcription elongation was not reduced upon treatment with inhibitors specific for PP2A (Okadaic Acid) or PP2B (Cyclosporin A) (Figure S4C, D), which are two phosphatases that do not dephosphorylate IKAROS proteins [35]. ChIP assay revealed that IKAROS was not significantly recruited to the c-Kit locus in Calyculin A-treated cells (Figure 4F). In general, CDK9, Mi2 and PCTD occupancy at c-Kit TSS, +1 and +5 regions, and PP1α recruitment at c-Kit TSS were also diminished, whereas POL II accumulated at TSS and +1 region and TBP recruitment did not vary (Figure 4F). Similar outcomes were observed at the Flt3 locus (Figure S4E–H). Calyculin A treatment did not affect CDK9 or PP1α protein levels, but rather led to the appearance of two slow-migrating CYCT1- and IKAROS-specific bands (Figure 4G). These bands, which are likely corresponding to phosphorylated forms of CYCT1 and IKAROS proteins, were expressed at detectable levels. Thus, it can be concluded that the IKAROS-PP1α interaction: (i) contributes to IKAROS dephosphorylation and efficient binding to chromatin; (ii) allows PP1α association with c-Kit and Flt3 TSS and (iii) promotes CDK9/P-TEFb activation in order to release promoter-proximal paused POL II.

The IKAROS-PP1 interaction is required for normal hematopoietic cell differentiation

The functional role of IKAROS-PP1 interaction during hematopoiesis was further investigated by ex vivo assays in methylcellulose (Figure 5A). Ik^WT and Ik^NULL lin⁻ HPCs were transduced with the control pMSCV empty vector (Ik^WT/GFP or Ik^NULL/GFP); Ik^NULL lin⁻ HPCs were also transduced with pMSCV/Ik1 (Ikaros1 isoform, identified as Ik^NULL/Ik1 HPCs) or pMSCV/Ik1ΔPP1 (a mutant of Ikaros1 carrying the A465/7 mutation, identified as Ik^NULL/Ik1ΔPP1 HPCs). Ik1ΔPP1 is an IKAROS mutant unable to bind PP1 and therefore, resistant to dephosphorylation [35]. This mutant is reported to possess a shorter half-like than wild type IKAROS and thus, we assessed the stability of Ik1ΔPP1 in Ik^NULL/Ik1ΔPP1 HPCs by single-cell immunofluorescence (IF) analysis. Since these transgenes possess a Flag-HA tag, IF analysis was carried out with anti-HA antibodies. IF studies indicated that Ik1ΔPP1 accumulates in the nucleus of Ik1ΔPP1-infected lin⁻ HPCs (Figure S5A). As well, in 293T transfected cells, whereby Ik1ΔPP1 was expressed at slightly lower levels than Ik1, Ik1ΔPP1 could interact with several known IKAROS protein partners (Mi2, CDK9, CYCT1 and MTA2) as demonstrated by protein co-IP experiments (Figure S5B).

In transduced lin⁻ HPCs, Ikaros1 and Ikaros1ΔPP1 transgene expression levels were about 5- and 8-fold higher than endogenous Ikaros (Figure 5B), which is comparable to other IKAROS rescue models [3]. As in Ik^NULL lin⁻ HPCs, c-Kit and Flt3 expression was impaired in Ik^NULL/GFP lin⁻ HPCs. Importantly, infection of Ik^NULL lin⁻ HPCs with pMSCV/Ik1 increased c-Kit expression to 74% and Flt3 expression to 46.7% (if their expression levels in Ik^WT/GFP lin⁻ HPCs is considered as 100%), whereas expression of c-Kit and Flt3 genes was only increased to 33.3% (for c-Kit) and 11.5% (for Flt3) upon pMSCV/Ik1ΔPP1 infection (Figure 5B). Thus, even though Ik1ΔPP1 could slightly increase c-Kit and Flt3 gene expression in lin⁻ transduced HPCs, Ik1 restored transcription elongation of these target genes more efficiently, indicating the functional importance of the IKAROS-PP1 interaction for gene regulation.

Clonogenic assays in methylcellulose revealed that Ik^NULL lin⁻ HPCs displayed decreased Colony Forming Cell (CFC) activity (1/166 Ik^NULL lin⁻ HPCs vs. 1/33 Ik^WT lin⁻ HPCs) as well as reduced colony size, which suggests reduction of cell proliferative potential. Ik^NULL lin⁻ HPCs produced less mixed CFC (CFU-GEMM) and increased erythroid/megakaryocyte (BFU-E/Mk) colonies (Figure 5C).

Although the expression of the c-KIT receptor at the cell surface of Ik^NULL HPCs is minimal compared to Ik^WT HPCs [42], we investigated whether a close-to-normal c-KIT and FLT3 activation could be obtained by the stimulation of Ik^NULL HPCs with increasing amounts of Stem Cell Factor (SCF or KIT ligand) as well as FLT3 ligand (FL). We observed that Ik^NULL lin⁻ HPCs were refractory to SCF and FL treatments. This functional limitation was demonstrated by the absence of: (i) induced cell proliferation (Figure 5D); (ii) c-Kit or Flt3 induction (Figure 5E, F); and (iii) hematopoietic commitment and differentiation (Figure S5C) upon SCF and/or FL treatment of Ik^NULL lin⁻ HPCs. These results along with previously published work [51], suggest that IKAROS regulation of c-Kit and Flt3 expression is critical for CFC activity and differentiation ability of HPCs.

To define the biological contribution of IKAROS-dependent control over transcription elongation, we studied the hematopoietic potential of Ik^NULL lin⁻ HPCs transduced with Ikaros1 or the elongation-incompetent Ikaros1ΔPP1 mutant. Ik^WT/GFP lin⁻ HPCs maintained the same hematopoietic potential than non-transduced Ik^WT lin⁻ HPCs, both in term of colony number as well as colony types. However, Ik^NULL/GFP lin⁻ HPCs had increased CFU-GEMM and BFU-Mk/E and reduced BFU-E potential compared to non-transduced Ik^NULL lin⁻ HPCs (Figure 5C; Ik^NULL vs. Ik^NULL/GFP). The gain in CFU-GEMM might be caused by an advantageous transduction ability of multipotent progenitor cells, whereas the gain in megakaryocyte vs. erythroid component, as indicated by increased BFU-Mk/E colonies, might be related to the Erythropoietin deprivation during the infection procedure (Figure 5A). Infection of Ik^NULL lin⁻ HPCs with Ikaros1 significantly augmented CFC activity of transduced HPCs (Figure 5C, table), which correlates with the increased number of CFU-GEMM, BFU-E/Mk and, moreover, BFU-E colonies (Figure 5C, table; Ik^NULL lin⁻ vs. Ik^NULL/Ik1 lin⁻ HPCs). Interestingly, CFC activity of Ik^NULL/Ik1ΔPP1 lin^- HPCs was lower than in Ik^NULL/Ik1 lin⁻ HPCs (Figure 5C, table), but the hematopoietic potential of Ik^NULL/GFP lin⁻ HPCs and Ik^NULL/Ik1ΔPP1 lin⁻ HPCs was very similar (Figure 5C, table; Ik^NULL/GFP lin⁻ vs. Ik^NULL/Ik1ΔPP1 lin⁻ HPCs). Thus these results suggest that the Ikaros1ΔPP1 mutant is defective in restoring hematopoietic functions and that IKAROS interaction with PP1, which is required for productive transcription elongation, is important for normal hematopoiesis.

Discussion

P-TEFb can be recruited to gene promoters either by the general P-TEFb activator BRD4 [40] or by transcriptional activators that bind to specific consensus DNA sequences [37]. We previously reported that IKAROS acts as a template for the recruitment of P-TEFb to specific genes in hematopoietic cells [23], [27]. The partnership of IKAROS with the NuRD complex, which is critical for chromatin organization is also very-well established [5], [7], [18], [25]. The results presented here show that NuRD and P-TEFb can be associated together in a multifunctional complex, and that IKAROS is required for NURD-P-TEFb complex recruitment to specific genes in HPCs. Additionally, our results suggest that Mi2/NuRD actively participates in relieving POL II promoter-proximal pausing and contributes to the control of transcription elongation of IKAROS-target genes.

IKAROS association with NuRD-P-TEFb complex: A combination of chromatin organization and transcription elongation activities

Size fractionation analysis of nuclear extracts and IKAROS immunoaffinity-purified protein complexes revealed that a modest but significant percentage of nuclear CDK9 and CYCT1 (P-TEFb) is stably associated with IKAROS and NuRD components. This association resembles to the small and highly active portion of P-TEFb that is included in the Super Elongation Complexes (SECs), known to stimulate transcription elongation rate [38], [39]. Interestingly, based on LC-MS/MS analysis, it appears that several subunits of the SECs interact with IKAROS in hematopoietic cells (Table S3). Furthermore, the absence of the transcription elongation inhibitor, NELF, and the P-TEFb inhibitor, HEXIM1, among the IKAROS-NuRD-P-TEFb interacting proteins supports the notion that the newly defined complex favors transcription elongation in hematopoietic cells.

Nucleosomes form barriers for the elongating POL II. The main remodeling complex known to facilitate POL II passage through nucleosomes is the histone chaperone complex FACT [52]. Although we cannot exclude that FACT is implicated together with IKAROS and the NuRD-P-TEFb complex in the control of c-Kit elongation, FACT components were not identified by LC-MS/MS analysis of IKAROS-associated proteins. Instead, our data indicate that POL II release and the rate of transcription elongation of these genes are affected by Mi2 knockdown. In addition, Mi2/NuRD occupancy at c-Kit and Flt3 ORF decreased when transcription elongation was blocked by the CDK9 inhibitor Flavopiridol. Therefore, our results suggest that in HPCs, the NuRD complex can act as a chromatin complex that both destabilizes and restores nucleosomal structure in order to assist and facilitate the passage of POL II during transcription elongation. Accordingly, the NuRD complex can support nucleosome remodeling through the Mi2-associated helicase activity and histone deacetylation through the HDAC-associated activity [53], [54].

Initially regarded as a co-repressing complex because of its association with histone deacetylases and transcriptional repressors, the NuRD complex has also been associated with permissive chromatin and gene activation [5], [19]. More recently, it has been reported that the NuRD-mediated repression of a subset of pluripotency genes in ES cells occurs in a dynamic equilibrium with activation signals in order to fine-tune expression of these genes in response to differentiation stimuli [20]–[22]. Our findings bring a novel perspective to the role of NuRD in transcription regulation since we demonstrate a functional link between chromatin-associated activities that are required for the control of early transcription regulation i.e., NuRD-dependent chromatin remodeling associated with gene priming and PIC assembly [55], and productive transcription elongation whereby Mi2/NuRD contributes to P-TEFb-dependent relief of paused POL II and transcription elongation.

IKAROS recruits the NuRD-P-TEFb complex at chromatin in dosage-dependent manner

During differentiation of B-cell progenitors, the dynamic change of IKAROS expression level has been identified as a key mechanism for multiple target gene expression [15], [16]. Perturbation of IKAROS expression level has deleterious effects in double-positive thymocytes and in pre-B cells whereby reduced IKAROS function is associated with malignant transformation in mice and humans [41], [56], [57]. Furthermore, the IKAROS haploinsufficiency is reported to promote acute lymphoblastic leukemia with a high risk of relapse [58]. Our results suggest that chromatin association of the NuRD-P-TEFb complex to IKAROS target-genes depends on IKAROS expression level, hence providing a reasonable explanation for IKAROS dosage effects observed in hematological malignancies. In Ik^WT lin⁻ HPCs and G1E2 cells, a fraction of IKAROS associates with P-TEFb and NuRD and this complex is recruited to c-Kit and Flt3 genes. In Ik^HT lin⁻ HPCs and Ikaros knockdown G1E2 cells, IKAROS, Mi2/NuRD and CDK9/P-TEFb are recruited less efficiently to gene regulatory regions and transcription elongation is affected although transcription initiation occurs normally and POL II is in a promoter-proximal paused configuration. Finally, in Ik^NULL HPCs, transcription initiation occurs normally but P-TEFb and NuRD recruitment to c-Kit and Flt3 loci is impaired. Then, the enrichment of elongation-competent Ser2-phosphorylated POL II at gene ORF drops to almost background values. As a result, transcription elongation is profoundly reduced.

A large set of hematopoietic lineage-specific genes are characterized by permissive chromatin organization and PIC assembly at their promoters in HPCs [59]–[65]. IKAROS is required for the establishment of lineage-specific transcriptional programs [4], [5], [51]. Based on our results, we posit that when expressed at higher levels in HPCs and lymphoid progenitors, IKAROS induces rapid and dynamic expression of multiple genes poised for expression through the relief of the promoter-proximal paused POL II and transcription elongation. During alternative lineage-specification, IKAROS levels decrease, the IKAROS-NuRD-P-TEFb complex associates less efficiently to chromatin, and transcription elongation of different activated genes targeted by IKAROS declines.

IKAROS-PP1α functional interaction

PP1α is involved in CDK9 dephosphorylation at Thr-186 and Ser-175 [30]. Dephosphorylation of Thr-186 and Ser-175 facilitates P-TEFb dissociation from the 7SK snRNP repressive complex. While the importance of Ser-175 phosphorylation/dephosphorylation is debated, dephosphorylation of Thr-186 is known to promote binding of the CDK9/P-TEF complex at genes TSS [29]. At the TSS, Thr-186 is phosphorylated by the TFIIH-associated CDK7, an event required for the catalytic activity of CDK9 (P-TEFb activation) and thus, the release of promoter-proximal paused POL II [66]. Hyperphosphorylation of IKAROS negatively affects its stability and DNA binding affinity [67], [68]. IKAROS dephosphorylation by PP1 [35] enhances IKAROS binding to DNA and, based on results presented here, it can also favor the transfer of this phosphatase to CDK9, thereby contributing to P-TEFb release from the 7SK snRNP and CDK9 activation [29]. Indeed, immunoaffinity purification and FPLC analyses suggest that PP1α associates with IKAROS, NuRD and P-TEFb in a 2 MDa complex that does not include subunits of the 7SK snRNP repressive complex. Since we found that PP1α is particularly abundant in close proximity to the TSS of the transcriptionally active c-Kit and Flt3 genes, and PP1α recruitment to these TSS is highly influenced by IKAROS concentration, it can be argued that IKAROS contributes to chromatin recruitment of PP1α at the TSS of target genes hence, facilitating CDK9/P-TEFb association through dephosphorylation of CDK9 Thr-186 and possibly, Ser-175 [29], [69]. Furthermore, although CDK9 phosphorylation by CDK7 activates CDK9 and promotes POL II release, it has been demonstrated that imbalanced CDK9 hyperphosphorylation by CDK7 can have the opposite effect [30]. Thus, the relative higher abundance of PP1α at the TSS (when compared to the other regions bound by the IKAROS-NURD-P-TEFb) might be required to prevent excessive phosphorylation of TSS-bound CDK9 by the TFIIH-associated CDK7.

The importance of the IKAROS-PP1α network was demonstrated by results obtained with the PP1α inhibitor Calyculin A and with the Ik1ΔPP1 mutant. First, Calyculin A treatment, which leads to IKAROS hyperphosphorylation and reduced DNA binding [35], resulted in decreased recruitment of PP1α, Mi2 and CDK9 to c-Kit and Flt3 promoters but did not affect TBP and POL II recruitment. Thus, PP1α activity is critical for IKAROS as well as NuRD and P-TEFb recruitment to these promoters and for transcription elongation, while being dispensable for PIC formation. Second, the rescue of Ik^NULL HPCs attempted with the Ik1ΔPP1 mutant, which interacts with Mi2, MTA2, CYCT1 and CDK9 (Figure S5B) but does not efficiently interact with PP1 [35], demonstrated that without the interaction with PP1, IKAROS cannot exert its normal function during hematopoiesis.

It is worth noting that PP1 can contribute to co-transcriptional pre-mRNA splicing control [70], [71], an event that could also favor transcription elongation of IKAROS target genes. In fact, LC-MS/MS analysis suggested that some proteins implicated in the process of transcription termination can interact with IKAROS in hematopoietic cells (Table S3). Based on our results we suggest that IKAROS, NuRD and P-TEFb are not only active at the TSS but they facilitate transcription elongation and could also influence transcription termination.

Thus, IKAROS control over transcription elongation at genes such as c-Kit and Flt3 is likely to provide a rapid adjustment of their expression levels, and be a critical mechanism affecting HSC/HPCs interaction with the niche, cell fate and stress response [72]–[75].

In conclusion, we have demonstrated that IKAROS is the DNA-binding subunit of the newly characterized NuRD-P-TEFb multifunctional complex, a chromatin-associated complex that contains chromatin remodeling (NuRD) as well as gene transcription elongation (P-TEFb) activities. IKAROS is important for CDK9 and Mi2 interaction and combined recruitment to actively transcribed genes in hematopoietic cells. We demonstrate that low IKAROS expression level does not preclude appropriate promoter organization but impairs productive elongation, whereas higher IKAROS levels are necessary to relieve promoter-proximal paused POL II and efficient transcription elongation of target genes (Figure 6). Our results suggest that NuRD can assist POL II transcription elongation complex throughout the ORF of IKAROS target genes and provides mechanistic cues to explain the central role played by IKAROS as transcriptional activator during hematopoietic lineage decision, differentiation, and interaction with the niche.

Materials and Methods

Protein co-IP, ChIP, re-ChIP, qPCR and qRT-PCR assays

These assays were done essentially as previously reported [7], [23], [27], [48].

Fast Protein Liquid Chromatography (FPLC)

Protein complexes were purified from Jurkat nuclear extract or IKAROS immunoaffinity-purified complexes on an AKTA Purifier FPLC System (GE Healthcare) using a Superose 6 10/300GL column (GE Healthcare).

Sequential immunoaffinity purification of IKAROS-associated proteins

Flag-HA sequential immunoaffinity (or Tap-Tag) purification was carried out as described in Nakatani et al. [76] starting from nuclear extracts prepared from pOZ-N-Flag-HA-IRES-ILR2 (mock sample) or pOZ-N-Flag-HA-IKAROS-IRES-ILR2- (Flag-HA-Ik sample) Jurkat cell clones. LC-MS/MS analysis was performed at the Taplin Mass Spectrometry Facility (Harvard University, Boston).

Mouse transgenic lines and primary cell isolation

Homozygote (Ik^NULL) and heterozygote (Ik^HT) Ikaros null mice were genotyped by PCR as described [6]. Animals were sacrificed by cervical dislocation. Animal experiments were conducted in accordance with the Canadian Council on Animal Care (CCAC) guidelines and approved by the Maisonneuve-Rosemont Hospital animal care committee (approval numbers 2011-16 and 2011-17). Lineage negative (lin⁻) HPCs were purified using the easySep mouse hematopoietic progenitor enrichment kit (StemCell Technology).

Retrovirus infection and clonogenic assays in methylcellulose

pMSCV/Ik1 and pMSCV/Ik1ΔPP1 vectors were generated by cloning the murine Ikaros1 cDNA or the mutant Ikaros1ΔPP1 cDNA, both with a Flag and HA tag at their N-terminal regions, into the MSCV-pgk-EGFP vector (Dr G. Sauvageau, IRIC). Retroviral infection of lin⁻ HPCs was carried out as described [77]. Lin⁻ HPCs were cultured for 3 days in medium containing 50 ng/ml SCF, 10 ng/ml IL3, 10 ng/ml IL6 and 5×10⁻⁵ M β-mercaptoethanol, without Erythropoietin. Transduced lin⁻/GFP⁺ HPCs were isolated on a FACS Aria III sorter (BD Biosciences) based on green fluorescence. HPCs were seeded in triplicate on complete methylcellulose medium (MethoCult Stemcell Technologies). Colonies were scored at day 14.

Statistical analysis

Unpaired Student's t-test was used to determine the level of statistical significance (P-value).

Additional Materials and Methods information can be found in Text S1.

Supporting Information

Attachment 1

Attachment 2

Attachment 3

Attachment 4

Attachment 5

Attachment 6

Attachment 7

Attachment 8

Attachment 9