Turning genes on and off is essential for development and homeostasis, yet little is known about the sequence and causal role of chromatin state changes during the repression of active genes. This is surprising, as defective gene silencing underlies developmental abnormalities and disease. Here we delineate the sequence and functional contribution of transcriptional repression mechanisms at high temporal resolution. Inducible entry of the NuRD-interacting transcriptional regulator Ikaros into mouse pre-B cell nuclei triggered immediate binding to target gene promoters. Rapid RNAP2 eviction, transcriptional shutdown, nucleosome invasion, and reduced transcriptional activator binding required chromatin remodeling by NuRD-associated Mi2beta/CHD4, but were independent of HDAC activity. Histone deacetylation occurred after transcriptional repression. Nevertheless, HDAC activity contributed to stable gene silencing. Hence, high resolution mapping of transcriptional repression reveals complex and interdependent mechanisms that underpin rapid transitions between transcriptional states, and elucidates the temporal order, functional role and mechanistic separation of NuRD-associated enzymatic activities.https://doi.org/10.7554/eLife.22767.001
Transcriptional gene activation and repression are essential for cell commitment, differentiation and homeostasis in metazoans. The dynamics of inducible gene expression have been characterised in exquisite detail (Ballaré et al., 2013; Cosma, 2002; Grøntved et al., 2013; Vicent et al., 2011; Voss and Hager, 2014; Hargreaves et al., 2009; Ramirez-Carrozzi et al., 2009; Smale and Natoli, 2014). Compared to gene activation, the mechanisms that underpin gene repression have received less attention (Brockdorff and Turner, 2015; Cosma, 2002; Feng et al., 2016; Fuda et al., 2009; Grossniklaus and Paro, 2014; Su et al., 2004; Trinh et al., 2001). Understanding transcriptional repression is important because defects in gene silencing result in developmental abnormalities and disease (Denslow and Wade, 2007; Watson et al., 2012). The nucleosome remodeling and deacetylase (NuRD) corepressor complex is critical for developmental gene regulation (Costa et al., 2012; Laugesen and Helin, 2014; Reynolds et al., 2012; Yamada et al., 2014). NuRD malfunction leads to birth defects (Fahrner and Bjornsson, 2014) and underlies disease states ranging from autism spectrum disorders (Yamada et al., 2014; Yang et al., 2016) to cancer (Laugesen and Helin, 2014). NuRD is unique among corepressors in that it brings together histone deacetylation and ATP-dependent chromatin remodeling activities in the same complex, and therefore may have the necessary enzymatic attributes to convert active promoters into densely packed, hypoacetylated chromatin refractory to the transcriptional machinery. It has been argued that understanding the kinetics of changes in nucleosome positioning and modifications that accompany transcriptional repression is important to link the biochemical properties of NuRD with our understanding of chromatin-based gene regulation (Denslow and Wade, 2007). Meeting this challenge is difficult in experimental systems where the kinetics are slow (Costa et al., 2012), or gene silencing is asynchronous (Reynolds et al., 2012).
To investigate the sequence and causality of transcriptional repression mechanisms we focus on target genes of the NuRD-interacting transcriptional regulator Ikaros in B cell progenitors (Ferreirós-Vidal et al., 2013). Ikaros is essential for B cell development (Heizmann et al., 2013; Joshi et al., 2014; Schwickert et al., 2014) and frequently mutated in B cell malignancies (Mullighan et al., 2008, 2009). Myc and the prototypic pre-B cell gene Igll1 are direct Ikaros target genes (Ferreirós-Vidal et al., 2013; Ma et al., 2010; Thompson et al., 2007). Igll1 encodes an essential component of the pre-B cell receptor (pre-BCR), which initially drives pre-B cell proliferation (Melchers et al., 1993), but subsequently triggers the expression of the Ikaros paralog Aiolos (Thompson et al., 2007). Together, Ikaros and Aiolos orchestrate the silencing of Igll1, the termination of pre-BCR expression, cell cycle exit, and differentiation of pre-B cells (Sabbattini et al., 2001; Thompson et al., 2007). Igll1 therefore is part of a regulatory circuit that controls the transition from self-renewal to differentiation of B cell progenitors. Myc is highly expressed in proliferating B cell progenitors and can interfere with B cell progenitor differentiation by overriding cell cycle exit (Ma et al., 2010).
The interaction of Ikaros with the CHD4/Mi-2beta chromatin remodeling subunit of the NuRD co-repressor complex (Kim et al., 1999) is direct (Sridharan and Smale, 2007) and evolutionarily conserved (Kehle et al., 1998). Although histone deacetylation has been linked to Ikaros-mediated repression (Koipally et al., 1999), the mechanisms of Ikaros-mediated gene regulation remain incompletely understood. At the Igll1 promoter the transcriptional activator EBF1 and Ikaros are thought to compete with each other for binding to a composite Ikaros/EBF1 site (Sabbattini et al., 2001; Thompson et al., 2007). Overlap between binding sites for Ikaros and transcriptional activators is also found the Ikaros target gene Dntt (Ernst et al., 1996; Hahm et al., 1998; Trinh et al., 2001), and the Ikaros-related Hunchback protein competes for DNA binding with transcriptional activators at Hox genes in Drosophila (Lawrence, 1992). The concept of Ikaros-activator competition is supported by in vitro gel mobility shift experiments (Thompson et al., 2007; Trinh et al., 2001), but it remains unclear how such competition may play out in living cells.
Ikaros binds sequence motifs of gamma satellites, the core repeat unit of mouse pericentromeric heterochromatin (Brown et al., 1999, 1997; Cobb et al., 2000; Hahm et al., 1998; Klug et al., 1998). Silent genes are often positioned within transcriptionally repressive compartments of the nucleus (Bickmore, 2013), including pericentromeric heterochromatin (Brown et al., 1999, 1997), and the repositioning of Rag and Dntt loci accompanies stable, but not transient gene silencing (Brown et al., 1999; Su et al., 2004). These data suggested the attractive model that Ikaros may position silent genes close to pericentromeric heterochromatin (Brown et al., 1999; Cobb et al., 2000). Surprisingly, however, a direct role for Ikaros or NuRD in the repositioning of target genes has not been demonstrated (Ferreirós-Vidal et al., 2013; Thompson et al., 2007).
To study transcriptional repression at high temporal resolution we utilize the regulated translocation of Ikaros from the cytoplasm to the nucleus (Ferreirós-Vidal et al., 2013). Nuclear translocation results in immediate binding to target gene promoters, providing a clear path to target gene regulation. Using this approach we demonstrate that transcriptional repression is surprisingly rapid: promoters are depleted of RNAP2 and invaded by nucleosomes within minutes. Mechanistically, RNAP2 eviction is controlled by chromatin accessibility, not vice versa. Kinetic analysis reveals an unexpected temporal dissociation between NuRD enzymatic activities: CHD4-dependent remodeling - but not HDAC activity - initiates repression by excluding transcriptional activators and RNAP2. However, stable silencing and pericentromeric repositioning of Ikaros target genes require both nucleosome remodeling and HDAC activity. Hence, our high resolution view reveals that transcriptional repression involves complex and interdependent mechanisms, and that the enzymatic activities of CHD4 and HDACs are selectively required at different times and for discrete aspects of gene silencing.
Ikaros-ERt2 fusion proteins are retained in the cytoplasm of pre-B cells (Ferreirós-Vidal et al., 2013) until their release by the ERt2 ligand 4-hydroxytamoxifen (4-OHT), which initiates the translocation of Ikaros-ERt2 within minutes (Figure 1A, Figure 1—figure supplement 1A). Association with DAPI-dense pericentromeric gamma-satellite regions containing Ikaros binding sites (Brown et al., 1999, 1997; Hahm et al., 1998) suggests that Ikaros binds DNA immediately after nuclear translocation. We exploited the temporal resolution afforded by inducible nuclear translocation to map the in vivo binding dynamics of Ikaros and EBF1 at the Igll1 promoter, and to experimentally test whether Ikaros and EBF1 compete with each other (Sabbattini et al., 2001; Thompson et al., 2007).
ChIP and quantitative PCR showed a steep rise in Ikaros binding to Igll1 promoter within 5 min (Figure 1B). The Igll1 promoter has two Ikaros sites, Ik1 at −83nt and Ik2 at −107nt from the Igll1 TSS. An EBF motif is embedded in the Ik1 site, and gel mobility shift experiments demonstrate mutually exclusive binding of EBF1 and Ikaros (Thompson et al., 2007). Despite the presence of two additional EBF1 motifs at −198 and −228nt, the embedded EBF1 site is critical for Igll1 promoter activity (Mårtensson and Mårtensson, 1997) (Figure 1C). It was therefore surprising that the rise in Ikaros binding 5 min after 4-OHT was not accompanied by a corresponding fall in EBF1 binding (Figure 1B). Increased Ikaros binding without simultaneous eviction of EBF1 5 min after 4-OHT suggests that Ikaros does not directly displace EBF1.
We used kinetic modeling to explore the impact of increased nuclear Ikaros in the presence of a constant concentration of EBF1 and a constant number of Igll1 promoter sites. In a 2-state model where DNA binding sites are either occupied by EBF1 or by Ikaros (EBF-DNA <-> Ikaros-DNA), any increase in Ikaros-DNA binding comes at the expense of EBF-DNA binding (Figure 1D). The observed temporal dissociation between Ikaros binding and EBF1 eviction by ChIP (Figure 1B) is inconsistent with direct competition between Ikaros (red) and EBF1 (green).
Interactions of transcription factors with target DNA are often transient, and not all target sites for DNA binding factors are occupied at any given time (Elf et al., 2007; McNally et al., 2000; Poorey et al., 2013; Voss and Hager, 2014). We therefore considered a 3-state model (Figure 1E) where in addition to sites occupied by EBF1 or Ikaros a fraction of DNA binding sites is free at any given time (EBF-DNA, green <-> free DNA, turquoise <-> Ikaros-DNA, red). In this scenario, the initial increase in Ikaros binding does not displace EBF1 directly, but rather reduces the fraction of unbound sites. Eventually, the reduced availability of unbound sites impacts on EBF1 by reducing the chances of successful re-binding during dynamic cycles of unbinding and re-binding. Consistent with this prediction, EBF1 ChIP measured a gradual reduction in EBF1 binding between 15 and 60 min after 4-OHT (Figure 1B). This revised model explains how increased Ikaros occupancy of the Igll1 promoter does not affect EBF1 binding immediately, but with a significant delay relative to the observed increase in Ikaros binding.
The abundance of nuclear Ikaros rises gradually over time (Figure 1A, Figure 1—figure supplement 1A), yet ChIP shows that Ikaros binding flattens out after 5 min and shows little or no further increase after 15 min, even though declining EBF1 binding should release potential binding sites (Figure 1B). While the models in Figure 1C and D assume a one-step increase in nuclear Ikaros, modeling an incremental rise in nuclear Ikaros predicted a more pronounced increase in Ikaros binding (red) with little or no decrease in EBF1 binding (green) at equilibrium (Figure 1—figure supplement 1B). This does not reflect the experimental observations of a blunted rise in Ikaros binding and reduced EBF binding. The data were captured closely, however, when we allowed for the possibility that in response to the nuclear translocation of Ikaros the pool of DNA target sites available for binding decreases over time (Figure 1—figure supplement 1C, blue). We therefore investigated the impact of Ikaros on the chromatin state of target promoters.
ChIP sequencing (ChIP-seq) showed that the nuclear translocation of Ikaros abolished RNAP2 binding to promoters and gene bodies of the target genes Igll1 and Vpreb1 (Figure 2A, Figure 2—figure supplement 1) that are repressed by Ikaros (Ferreirós-Vidal et al., 2013; Thompson et al., 2007). RNAP2 occupancy remained unchanged at the nearby Top3b promoter, which is not regulated by Ikaros (Figure 2A, Figure 2—figure supplement 1).
We mapped the nucleosome landscape before and after nuclear translocation of Ikaros using micrococcal nuclease digestion and sequencing (MNase-seq). Prior to nuclear translocation of Ikaros the Igll1 promoter showed a nucleosome-free region that was occupied by RNAP2. Upon nuclear translocation of Ikaros this region became occluded by nucleosomes, and RNAP2 occupancy was reduced (Figure 2A, inset and Figure 2—figure supplement 1). Reduced accessibility of the Igll1 promoter is consistent with models where Ikaros and EBF1 compete for a decreasing number of accessible binding sites (Figure 1—figure supplement 1C).
We next explored the dynamics of Ikaros-induced changes in RNAP2 occupancy, chromatin organisation and transcription at the Ikaros-repressed genes Igll1 and Myc. ChIP-PCR demonstrated a rapid decrease in RNAP2 occupancy at the promoters, gene bodies and transcription termination sites (TTS) of Ikaros-repressed genes (Figure 2B). RNAP2 occupancy decreased at promoters, gene bodies, and the 3' end of transcription units (Figure 2B), suggesting that Ikaros represses Igll1 and Myc transcription by interfering with RNAP2 recruitment rather than elongation. Genome-wide, RNAP2 occupancy was significantly reduced at 372 downregulated genes 6 hr after Ikaros translocation (Table 1). The great majority of repressed genes with reduced RNAP2 occupancy showed a global reduction in RNAP2 binding to both the promoter and the gene body (324 of 372 or 87%). Only a small minority of repressed genes showed a block in RNAP2 elongation, as judged by preferential reduction of RNAP2 occupancy over the gene body (15 of 372 or 4% of genes with reduced RNAP2 occupancy). Hence, transcriptional repression of Ikaros target genes occurred by reduced RNAP2 recruitment, and a block in RNAP2 elongation was the exception.
We assessed nucleosome occupancy of promoters and transcription start sites by PCR amplification of fragments protected from MNase digestion. Nucleosome occupancy increased with similar kinetics as RNAP2 eviction (Figure 2C).
We used quantitative RT-PCR with primers spanning intron-exon boundaries to quantify the abundance of unspliced (primary) transcripts, which are short-lived and serve as an indicator of transcriptional activity (Figure 2D). The abundance of Igll1 and Myc primary transcripts was reduced within minutes of Ikaros translocation with similar kinetics as RNAP2 eviction and increased nucleosome occupancy (Figure 2D). These changes were selective for Ikaros-repressed genes, as illustrated by comparison with the Ikaros-induced gene Zfp36 (Figure 2C–D).
We next asked whether nucleosome invasion of Ikaros-repressed target promoters was secondary to reduced RNAP2 occupancy (Gilchrist et al., 2012) (Figure 3A, right). Triptolide triggers the global degradation of RNAP2 (Manzo et al., 2012; Wang et al., 2011) and resulted in the near-complete loss of RNAP2 from whole cell lysates (Figure 3A, left) and from the Igll1 and Myc promoters (Figure 3B, left; the inactive Rex1 promoter is shown as a negative control). Loss of RNAP2 reduced promoter occupancy by the basal transcription factor TFIIB (Figure 3B, center) but, interestingly, the Igll1 and Myc promoters remained sensitive to MNase in the absence of RNAP2 (Figure 3B, right). Hence, accessibility of Ikaros target gene promoters is not simply a consequence of RNAP2 occupancy.
We next compared the impact of Ikaros on target gene promoters in control and RNAP2-depleted cells. Nuclear translocation of Ikaros-ERt2 efficiently reduced the accessibility of the Igll1 and Myc promoters even after RNAP2 depletion (Figure 3C).
To address the formal possibility that the ERt2 component of the Ikaros-ERt2 fusion protein contributed to the Ikaros-induced reduction in promoter accessibility we used a complementary approach where the nuclear translocation of Ikaros is triggered by the inducible cleavage of Ikaros-TEV-ERt2 fusion proteins (Wehr et al., 2006) (Figure 3—figure supplement 1A). This confirmed that nuclear translocation of Ikaros in the absence of ERt2 was able to regulate promoter accessibility (Figure 3—figure supplement 1B).
The demonstration that the Igll1 and Myc promoter regions remain accessible in the absence of RNAP2 mechanistically separates RNAP2 occupancy from promoter accessibility. We therefore focused on the question whether Ikaros controls promoter accessibility through active chromatin remodelling.
We used RNA interference to deplete the NuRD-associated ATPase subunit Mi-2β/CHD4 encoded by Chd4 (Figure 4A). CHD4 depletion substantially reduced the impact of Ikaros on the accessibility of the Igll1 and Myc promoters (Figure 4B), indicating a key role for CHD4 in mediating Ikaros-induced nucleosome remodeling at the Igll1 and Myc promoters. Ikaros-induced RNAP2 eviction and transcriptional repression were significantly delayed by Chd4-RNAi (Figure 4C), consistent with a role for CHD4-dependent chromatin remodeling in the silencing of Ikaros target genes.
We next assessed the impact of Ikaros on the association of NuRD components with the Igll1 promoter. ChIP-PCR showed a rapid increase in the binding of the NuRD subunits CHD4 and MBD3 (Figure 4D). The onset CHD4 and MBD3 recruitment mirrored increased Ikaros binding after 4-OHT (Figure 1B). In contrast to Ikaros, increased binding of CHD4 and MBD3 was transient, returning close to their starting level by 6 hr (Figure 4D).
CHD4 and the SWI-SNF-associated chromatin remodeler BRG1 can have antagonistic effects on transcriptional regulation (Curtis and Griffin, 2012; Gao et al., 2009; Ramirez-Carrozzi et al., 2006). To assess the relationship between Ikaros and BRG1 we compared genes regulated by Ikaros (See Table 1—source data 1) with genes regulated by BRG1 (Bossen et al., 2015). Genes activated by BRG1 were repressed by Ikaros (Odds ratio = 2.97, p=1.7×10−30) and vice versa (Odds ratio = 3.32, p=9.5×10−31). Conversely, Ikaros-repressed genes were depleted among BRG1-repressed genes (Odds ratio = 0.62) and vice versa (Odds ratio = 0.31; Table 2). This analysis indicates antagonistic regulation of target genes by Ikaros and BRG1.
In addition to the chromatin remodeler CHD4, the NuRD complex contains the histone deacetylases HDAC1/2 (Dege and Hagman, 2014). Given the critical role for NuRD-associated chromatin remodeling activity in Ikaros-mediated gene silencing we asked whether HDAC activity was equally involved. Unexpectedly, broad inhibition of HDAC activity by trichostatin A (TSA) did not detectably interfere with Ikaros-induced loss of RNAP2, reduced promoter accessibility and early transcriptional repression of Igll1 and Myc (Figure 5A, Figure 5—figure supplement 1A). Application of the HDAC1/2 inhibitor MS-275 confirmed that Ikaros initiated the transcriptional downregulation of Igll1 and Myc regardless of HDAC1/2 activity (Figure 5B, Figure 5—figure supplement 1B). Similarly, HDAC inhibition did not interfere with the transcriptional downregulation of Ikaros target genes in primary pre-B cells (Figure 5C). To address the possibility that the ERt2 component of the Ikaros-ERt2 fusion protein might account for the apparent lack of HDAC activity in Ikaros-initiated gene silencing we performed experiments where the nuclear translocation of Ikaros was triggered by the inducible cleavage of Ikaros-TEV-ERt2 fusion proteins. TEV-induced nuclear translocation of Ikaros in the absence or ERt2 confirmed that HDAC activity was not required for the loss of accessibility (Figure 5D, left) or the transcriptional silencing of Igll1 and Myc (Figure 5D, right).
These data demonstrate that - unexpectedly - HDAC activity is dispensable for Ikaros-induced repression. Consistent with this conclusion, ChIP experiments showed that although Ikaros reduced the acetylation of histone H3 and H4 at the target genes Igll1 and Myc, the kinetics of histone deacetylation were markedly slower than those observed for reduced promoter accessibility, RNAP2 eviction, and the decline in primary transcripts (Figure 5—figure supplement 1C). We saw no detectable increase in the trimethylation of H3K27 by PRC2 (Oravecz et al., 2015) at the Igll1 and Myc promoters 24 hr after nuclear translocation of Ikaros (Data not shown). Histone modifications are therefore not likely drivers for early events in Ikaros-mediated repression.
While the immediate transcriptional downregulation of Ikaros target genes was unaffected by HDAC inhibition (Figure 5), we found that Ikaros target gene transcription was not stably silenced 24 hr after Ikaros nuclear translocation in the presence of TSA (Figure 6A).
We examined the protection of nucleosome-associated DNA at the Igll1 promoter and TSS using PCR amplicons of defined length (Figure 6B, top). TSA treatment for 24 hr did not affect Ikaros-induced nucleosome protection of short amplicons comprising 90–110 base pairs centred on peaks of MNase (Figure 6B, bottom left). In contrast, TSA reduced Ikaros-induced protection of longer amplicons of 130–140 bp, which represent the near complete length of 147 DNA base pairs associated with full nucleosomes (Figure 6B, bottom right). These data suggest that in the absence of HDAC activity, nucleosomes at the Igll1 promoter and TSS may remain unstable (Henikoff et al., 2011) or less evenly spaced. To address if this is due to missing H2A-H2B dimers we performed ChIP for histone H2B and H3. At the Igll1 promoter and TSS, Ikaros triggered an increase in the occupancy not only of total histone H3, but also in the ratio of histone H2B over H3 (Figure 6C). Prolonged HDAC inhibition significantly blunted this increase in the H2B/H3 ratio (Figure 6C), indicating that HDAC activity facilitated the assembly of H2B-containing nucleosomes.
To test the role of Ikaros in the repositioning of target genes to the transcriptionally repressive environment of pericentromeric heterochromatin we carried out 3D DNA-FISH experiments. In the absence of 4-OHT, only a small minority (8%) of Igll1 alleles (green) were associated with γ-satellite DNA (red), the repeat unit of mouse pericentromeric heterochromatin (Figure 6D). In contrast, most (78%) Igll1 alleles were associated with γ-satellite DNA after 4-OHT (Figure 6D), indicating that increased nuclear Ikaros cells was sufficient for the repositioning of Igll1 to pericentromeric heterochromatin. Ikaros-induced of Igll1 repositioning was reduced to 55% of alleles by low doses (1 ng/ml) of TSA (Figure 6D). We note that under these conditions TSA effectively inhibited histone deacetylation (Figure 5—figure supplement 1A) but did not globally disrupt nuclear organisation, since centromeric heterochromatin retained its compacted state and association with the nuclear periphery (Figure 6D).
Taken together, incomplete transcriptional silencing of Igll1, partial MNase protection of Igll1 promoter DNA and incomplete repositioning of Igll1 alleles to pericentromeric heterochromatin in the presence of HDAC inhibitors indicate that HDAC activity is required for full Igll1 silencing, even though it is dispensable for the early events in transcriptional repression (Figure 5).
The experiments discussed above suggest that Ikaros may utilize silencing mechanisms not just sequentially, but also in a mutually interdependent fashion. We therefore asked whether CHD4-mediated chromatin remodeling contributes to the repositioning of Igll1 loci. We found that the Ikaros-induced repositioning of Igll1 alleles was severely impaired (from 78% to 13%) by knockdown of Chd4 (Figure 7A). Pericentromeric repositioning of Ikaros target loci therefore requires NuRD-associated chromatin remodeling as well as HDAC activity.
To explore the relationship between silencing mechanisms further, we revisited the models for competition between Ikaros and EBF1 for binding to the Igll1 promoter (Figure 1C,D; Figure 1—figure supplement 1C). We compared the kinetics of Ikaros and EBF1 binding to Igll1 in control cells (Figure 7B left) and in Chd4-depleted cells (Figure 7B right, shChd4). Chd4-depleted cells showed a steeper rise in Ikaros binding between 5 and 15 min than control cells. Strikingly, the decrease in EBF1 binding seen in control cells was significantly blunted in the absence of CHD4. In this regard, the binding kinetics of Ikaros and EBF1 in the absence of CHD4 resembled a 3-state model (Figure 1—figure supplement 1B) where an incremental increase in nuclear Ikaros concentration triggers a steep rise in Ikaros binding during the second time interval and results in little or no eviction of EBF1. In control cells, the observed dynamics more closely resembled a 4-state model (Figure 1—figure supplement 1C) where Ikaros and EBF1 compete for binding sites that remain accessible. Consistent with a scenario where Ikaros not only displaces EBF1, but ultimately limits its own binding, integration of MNase-seq with Ikaros ChIP-seq showed that nuclear translocation of Ikaros increases nucleosome occupancy at Ikaros ChIP-seq peaks genome-wide (Figure 7C).
Here we take a high-resolution view at transcriptional repression based on the inducible nuclear translocation of Ikaros. Nuclear translocation of Ikaros results in immediate binding to target gene promoters, providing a clear path to target gene regulation. Kinetic ChIP shows half-maximal binding of Ikaros to the Igll1 target gene promoter within 5 min (Figure 7D). Unexpectedly, increased Ikaros binding did not directly displace the transcriptional activator EBF1 from the Igll1 promoter, ruling out simple 2-state models where activator and repressor directly compete with each other. Ikaros effected half-maximal removal of RNAP2, transcriptional silencing and promoter invasion by nucleosomes within 12 min of nuclear translocation (Figure 7D).
Ikaros recruited the NuRD-associated chromatin remodeller CHD4 to target promoters, and CHD4 was required for the invasion of the Igll1 and Myc promoters by nucleosomes, the exclusion of RNAP2, and the eviction of EBF1. Nuclear translocation of Ikaros reduced the accessibility of Ikaros binding sites genome-wide. Ultimately, Ikaros-induced chromatin remodeling limited the binding of the repressor itself, as evidenced by increased Ikaros binding after knockdown of Chd4. The binding kinetics of Ikaros and EBF1 at the Igll1 promoter are best explained if a fraction of binding sites become inaccessible as a result of Ikaros nuclear translocation. Mechanistically, Ikaros can reduce EBF1 binding to the Igll1 promoter only in the presence of CHD4, and EBF1 is binding is sensitive to local chromatin structure (Treiber et al., 2010).
The synchronous nature of our system enabled the detection of a transient increase in the association of the NuRD components CHD4 and MBD3 to Ikaros target promoters. Despite genetic evidence that NuRD is required for the differentiation of mouse embryonic stem cells, increased NuRD recruitment was not seen during this less synchronous process (Reynolds et al., 2012). Coincident with increased NuRD recruitment, the SWI/SNF-associated chromatin remodeler BRG1 was evicted from the Igll1 promoter. Ikaros and BRG1 were broadly antagonistic in B cell progenitors, since BRG1 was required for the activity of many Ikaros-repressed genes. BRG1 did not return to the remodeled Igll1 promoter even after the binding of CHD4 and MBD3 declined, which may in part explain why promoter accessibility was not re-instated, suggesting that Ikaros-mediated NuRD recruitment and eviction of BRG1 may cooperate in the repression of Ikaros target genes.
Interestingly, our high-resolution view of Ikaros-mediated gene silencing indicates a temporal dissociation of enzymatic activities associated with NuRD where histone deacetylation occurred after chromatin remodelling and transcriptional repression (Figure 7D). CHD4, but not HDAC activity was required for reduced promoter accessibility, the loss of RNAP2, and transcriptional repression. Although dispensable for the initiation of transcriptional repression, HDAC activity did facilitate stable transcriptional silencing, nucleosome occupancy of target promoters, and locus repositioning to pericentromeric heterochromatin.
The silencing mechanisms delineated here are interdependent, as illustrated by the role of HDACs and the involvement of CHD4 in competition between Ikaros and EBF1 for access to the Igll1 promoter, RNAP2 occupancy and locus repositioning to repressive nuclear environments, a hallmark of heritable gene silencing (Brown et al., 1999; Su et al., 2004). Our data link locus repositioning to chromatin state by demonstrating that both histone deacetylation and CHD4-mediated chromatin remodeling contribute to efficient repositioning of the Igll1 locus. Histone deacetylation and nucleosome positioning likely contribute to the compaction of the locus, facilitating 'like-with-like' associations of chromatin states (Jost et al., 2014).
Genes that are induced rapidly in response to TLR signaling in macrophages are often regulated at the level of transcriptional elongation (Hargreaves et al., 2009) and typically do not require ATP-dependent chromatin remodelling (Ramirez-Carrozzi et al., 2009, 2006). At the Ikaros target genes Igll1 and Myc as well as genome-wide, RNAP2 recruitment, rather than RNAP2 elongation, was the primary mechanism by which Ikaros repressed RNAP2-dependent transcription. The role of RNAP2 was not to keep nucleosomes away from the Igll1 and Myc promoters because experimental depletion of RNAP2 did not result in a loss of promoter accessibility. Experimental depletion of RNAP2 and the resulting loss of RNAP2-mediated transcription did not interfere with Ikaros-induced promoter remodelling. This is consistent with a direct role for Ikaros in target gene repression, and reminiscent of primary response gene induction in the absence of protein synthesis. With respect to the role of active chromatin remodeling, however, Ikaros-mediated repression was unlike the induction of primary response genes, and reminiscent of nuclear hormone receptor signaling where several enzymes cooperate to remodel chromatin within minutes of hormone addition (Ballaré et al., 2013; Grøntved et al., 2013; Nacht et al., 2016; Vicent et al., 2011).
The molecular determinants of whether IKAROS binding results in gene activation or repression remain to be explored. For genes that are repressed by Ikaros with fast kinetics, we favour a model where transcriptional repression by IKAROS involves active, CHD4-mediated chromatin remodeling of the Igll1 and Myc target gene promoters. This restricts promoter access, excludes RNAP2 and the transcriptional activator EBF1, and ultimately limits binding of Ikaros itself. The emerging picture is that the rapid repression of developmentally regulated genes utilizes complex and interdependent mechanisms, and that complexity does not come at the expense of speed in transcriptional state transitions.
The B3 cell line was generated and characterised in our lab (Brown et al., 1997). B3 and primary pre-B cells were maintained in Iscove's Modified Dulbecco's Medium (IMDM, Invitrogen), 10% FCS, penicillin and 100 μg/ml streptomycin) (GIBCO, Invitrogen). For primary pre-B cells, media were supplemented with β-mercaptoethanol, L-glutamine 100 U/ml, 5 ng/ml recombinant mouse IL-7 (R and D Systems) and ST-2 feeder cells. HA-Ikaros-ERt2-IRES-GFP or -Cherry in the MSCV retroviral vector was modified by insertion of a TEV protease recognition sequence to yield HA-Ikaros-TEV-ERt2-IRES-GFP or -Cherry and where indicated B3 cells were co-transfected with split-TEV constructs where the N- and C-termini of TEV protease were fused to FRB (FLAG-FRB-N_TEV- IRES-dsRED) and FKBP (FLAG-FKBP-C_TEV-IRES-CFP), respectively. Transduced cells were sorted by flow cytometry for homogeneous expression of constructs. MSCV HA-Ikaros-IRES-GFP and IRES-GFP control vectors, virus production and spin infection have been described.
Nuclear translocation of Ikaros-ERt2 was induced by Ikaros was induced by 0.5 µM 4-hydroxytamoxifen (4-OHT, Sigma-Aldrich) and TEV activity was induced by rapamycin (25 nM, Selleckchem). Nuclear translocation was monitored by anti-HA (Covance MMS-101R) immunofluorescence staining and confocal microscopy as described and images were processed using Leica and Image J software.
Where indicated, cells were treated with 1 μM triptolide (Sigma), 1 ng/ml trichostatin A (TSA, Sigma) or 10 μM MS-275 (Selleckchem).
shRNA for Chd4 (5’-GACTACGACCTGTTCAAGCAG-3’) and pQsupR control plasmid were used as advised by the manufacturer (Invitrogen) and media were supplemented with 20 μM pan caspase inhibitor (R and D Systems, FMK001).
mRNA extraction, reverse transcription PCR and immunoblotting was as described. We used antibodies to histone H3 (Abcam AB1791), acetylated histone H3 (Abcam AB47915), acetylated histone H4 (Millipore 06–866), RNAPII (Santa Cruz sc-899), TFIIB (Santa Cruz sc-225), CHD4 (Abcam AB72418), and tubulin (Sigma T9026).
MNase digestion was done as described (Carey and Smale, 2007; Henikoff et al., 2011). Cells were cross-linked for 10 min at room temperature in 1% formaldehyde in 10% fixation buffer (0.5 mM EGTA pH 8.0, 100 mM NaCl, 1 mM EDTA, 50 mM HEPES pH 8.0, 10% formaldehyde). Fixation was stopped by glycine (140 mM). Cells were washed twice with cold PBS (4°C, 5 min, 900rcf) and resuspended in lysis buffer (5 mM PIPES pH 8.0, 85 mM KCl, 0.5% NP-40) supplemented with protease inhibitor cocktail (Roche), 0.15 mM spermine (Sigma), and 0.5 mM spermidine (Sigma) on ice for 20 min. Sample were washed and resuspended in MNase CaCl2+ buffer (10 mM Tris-HCl, pH 7.4, 15 mM NaCl, 60 mM KCl, 0.15 mM spermine, 0.5 mM spermidine) and digested with MNase (200 U/ml, Worthington) at in room temperature for 10 min. Digestion was stopped by 20 µl 100 mM EDTA, 10 mM EGTA per 2 × 106 cells. As a control, an equal number of cells were treated in parallel without adding MNase. Samples were reverse crosslinked, DNA purified by phenol/chloroform and quantified (PicoGreen dsDNA quantitation kit, Life Technologies).
For MNase-qPCR, 2 ng DNA were used per qPCR reaction. Results were normalised to undigested controls and compared to the inactive Rex1 or αActin promoter.
For MNase-seq, 50 ng DNA after MNase digestion (carrier ethanol treatment for 6 hr or 0.5 μM 4-OHT treatment for 6 hr) were used to prepare MNase-seq samples (Next Ultra, NEB) without size selection steps. Libraries from three biological replicates were sequenced (Illumina Hi-seq 2500, 50 bp paired end) to ~500–600 M reads per condition. Reads were aligned to mouse genome mm9 and reconstructed into complete fragments in silico. To calculate normalised coverage per base for MNase data, fragments with length between 110 bp and 170 bp were normalised to total reads outside of mm9 blacklisted regions.
Ikaros ChIP was as described using a C-terminal Ikaros antibody provided by Dr Stephen Smale that has been extensively characterised in ChIP-seq experiments (Ferreirós-Vidal et al., 2013; Bossen et al., 2015). We used an EBF1 antibody (Santa Cruz sc-137065) previously characterised by ENCODE for EBF1 ChIP-seq. For RNAP2, chromatin remodeler and histone modification ChIP (Brookes et al., 2012) cells were fixed as described for MNase, lysed in 25 mM HEPES, 1.5 mM MgCl2, 10 mM KCl, 0.1% NP-40, pH 7.9 at 50 × 106 cells/ml on ice for 15 min, centrifuged at 900 rcf for 10 min at 4°C, resuspended in 140 mM NaCl, 50 mM HEPES, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate, pH 7.9 and sonicated for 25 min at 4°C with 30 s on, and 30 s off (Bioruptor, Diagenode). Sonicated chromatin was recovered as supernatant after centrifugation at 25,000 rcf for 15 min at 4°C and incubated with antibodies to histone H3 (Abcam AB1791), acetylated histone H3 (Abcam AB47915), acetylated histone H4 (Millipore 06–866), RNAP2 (Santa Cruz sc-899), TFIIB (Santa Cruz sc-225), CHD4 (Abcam AB70469), MBD3 (Bethyl A302-528A) or Brg1 (Millipore, 07–478) on a rotating platform at 4°C overnight, washed sequentially in 140 mM NaCl, 50 mM HEPES, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate, pH 7.9 followed by 500 mM NaCl, 50 mM HEPES, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate, pH 7.9 followed by 250 mM LiCl, 50 mM HEPES, 1 mM EDTA, 0.5% NP-40, 0.5% sodium deoxycholate, pH 8.0 and finally 10 mM Tris-HCl, 1 mM EDTA, pH 7.5. Reverse crosslinked DNA was purified by phenol-chloroform extraction. For ChIP-seq, 10 ng ChIP and input DNA were used to prepare libraries (Next Ultra, NEB) with size selection for fragments between 100 and 300 bp. After 50 bp single-end sequencing (Hi-seq 2500, Illumina, ~155 M reads per condition), reads were aligned to mm9 using BWA version 0.7.5. SAM files were generated, aligned, sorted and indexed with samtools version 0.1.18. Duplicate reads were marked using Picard MarkDuplicates. The quality of the datasets and predicted fragment lengths were analysed using the ChIPQC Bioconductor package. Normalised BigWig files were generated using rtracklayer (version 1.3.2) with signal normalised to total mapped fragments per sample. ChIP-seq sample quality was assessed using ChIPQC (version 1.8.7) (Carroll et al., 2014). For the analysis of differential polymerase signal in genes and TSS regions, gene models for mm9 were retrieved from Ensembl Biomart. Total reads were counted within TSSs and gene bodies and analysis of differential binding performed using DEseq2 with efficiency and library complexity derived from ChIPQC supplied to the DEseq2 model as normalisation factors (Love et al., 2014). MNase signal over Ikaros ChIP-seq peaks (Ferreirós-Vidal et al., 2013) was calculated using the Bioconductor soGGi package (Dharmalingam and Carroll, 2015) and differential MNase signal identified using the DChIPrep package (Chabbert et al., 2016). Differential gene expression was based on a p-value of <0.05 after Benjamini-Hochberg correction for multiple testing (adj. p<0.05).
|Igll1 Prom. short||CTGTGAGTGAAAACAGTTAGGCTTGC||ACCAGCAGGCACACCCCAGTG|
|Igll1 Prom. long||CAAACCCCAGGCTGTCTCTA||GGCAGCTGTGAGTGAAAACA|
|Igll1 TSS short||GGTGGAAACTAGAGACAGCCTGG||CGGCAAAAGGATTGTTCCTCC|
|Igll1 TSS long||TGGCCAAAAGCTATTCCAGT||GGGTAGTCCCTTTGGGAGAG|
|Igll1 +200 nt||TGCTGTTGGGTCTAGTGGATGG||GCCTGTTGCTTCCCACTGAAG|
|Igll1 +1.3 kb||CAGCCAGTATCCCGACAAGT||AGAATCTGCTGGGCCTGATA|
|Myc +350 nt||CCTAAGAAGGCAGCTCTGGA||GCTGATGTTGGGTCAGTCG|
|Myc +1 kb||AACCAGAGGGAATCCTCACA||GAACCGCTCAGATCACGACT|
|Zfp36 +200 nt||TACGCGAGTGACAGCAGTGT||GCTGGCTGGAAATGAGAGAG|
|Zfp36 +0.6 kb||CACTTCACGAGCTTGCCAGT||AGTCAGGTTCTCCCTGGAGTT|
To assess the impact of nuclear Ikaros on Ikaros-EBF1 competition, we implemented kinetic models of increasing complexity. First, we considered a 2-state model where DNAE and DNAI correspond to DNA bond by EBF1 and Ikaros with the binding rate constants are kE and kI (Equation 1).
The corresponding rate equations are:
To solve these rate equations by numerical integration we used KinTek Explorer (KinTek Corporation, Austin, TX; Johnson, 2009; Figure 1D). The 2-state model in Equation 2 cannot recapitulate the features observed in kinetic ChIP experiments (Figure 1B) using the estimated parametes Ikaros kon 0.3 koff 0.052; EBF1 kon 0.036 koff 0.033, or any other parameters tested. We therefore implemented a 3-state model that includes a population of free (unbound) DNA (Equation 3).
In this model, Ikaros and EBF1 can compete for the same binding site only if it is free. The EBF1 and Ikaros dissociation rate constants are k-E and k-I, respectively. The corresponding rate equations are in Equation 4.
Parameters were estimated as Ikaros kon 0.3 koff 0.052; EBF1 kon 0.036 koff 0.033, start conditions: I-DNA 0.1, E-DNA 0.1, free DNA 0.8, similar results were obtained with a range of parameters. The 3-state model recapitulates the primary features observed in the ChIP experiments (Figure 1B,E), but does not account for the gradual increase in nuclear Ikaros observed experimentally (Figure 1A and Figure 1—figure supplement 1B) or the possibility of binding site occupancy by nucleosomes (non-accessible DNA or DNAN). We introduced DNAN in a 4-state model (Equation 5) where kN and k-N are the nucleosome binding and dissociation rate constants, respectively:
We used the parameters Ikaros kon 0.3 koff 0.052; EBF1 kon 0.036 koff 0.033; nucleosome kon 0.13 koff 0.003, start conditions: I-DNA 0.1, E-DNA 0.1, free DNA 0.8, starting concentration of Ikaros: 0.1, Ikaros increment per time interval: 2.0 and obtained the kinetic curves shown in Figure 1—figure supplement 1C. We experimentally tested the model predictions in kinetic ChIP experiments. Data from control cells (Figure 7B, left) correspond well to the 4-state model in Equation 5, suggesting a role for nucleosomes in restricting binding of Ikaros and EBF1. Ikaros and EBF1 binding were increased in the absence of the nucleosome remodeler CHD4 (Figure 7B, right), in line with predictions by the 3-state model in Equation 4 without DNAN.
The spatial organization of the human genomeAnnual Review of Genomics and Human Genetics 14:67–84.https://doi.org/10.1146/annurev-genom-091212-153515
Micrococcal Nuclease-Southern blot assay: I. MNase and restriction digestionsCold Spring Harbor Protocols 2007:pdb prot4890.https://doi.org/10.1101/pdb.prot4890
Targeting of ikaros to pericentromeric heterochromatin by direct DNA bindingGenes & Development 14:2146–2160.https://doi.org/10.1101/gad.816400
The chromatin-remodeling enzymes BRG1 and CHD4 antagonistically regulate vascular wnt signalingMolecular and Cellular Biology 32:1312–1320.https://doi.org/10.1128/MCB.06222-11
Mi-2/NuRD chromatin remodeling complexes regulate B and T-lymphocyte development and functionImmunological Reviews 261:126–140.https://doi.org/10.1111/imr.12209
soGGi: Visualise ChIP-Seq, MNase-Seq and Motif Occurrence as Aggregate Plots Summarised Over Grouped Genomic Intervals, version 1.4.4R package version.
A potential role for Elf-1 in terminal transferase gene regulationMolecular and Cellular Biology 16:6121–6131.https://doi.org/10.1128/MCB.16.11.6121
Mendelian disorders of the epigenetic machinery: tipping the balance of chromatin statesAnnual Review of Genomics and Human Genetics 15:269–293.https://doi.org/10.1146/annurev-genom-090613-094245
Transcriptional silencing by polycomb-group proteinsCold Spring Harbor Perspectives in Biology 6:a019331.https://doi.org/10.1101/cshperspect.a019331
Ikaros is absolutely required for pre-B cell differentiation by attenuating IL-7 signalsThe Journal of Experimental Medicine 210:2823–2832.https://doi.org/10.1084/jem.20131735
Fitting enzyme kinetic data with KinTek global kinetic explorerMethods in Enzymology 467:601–626.https://doi.org/10.1016/S0076-6879(09)67023-3
Modeling epigenome folding: formation and dynamics of topologically associated chromatin domainsNucleic Acids Research 42:9553–9561.https://doi.org/10.1093/nar/gku698
The Making of a Fly: The Genetics of Animal DesignOxford England; Cambridge, Mass., USA: Blackwell Science.
Ikaros and aiolos inhibit pre-B-cell proliferation by directly suppressing c-Myc expressionMolecular and Cellular Biology 30:4149–4158.https://doi.org/10.1128/MCB.00224-10
Deletion of IKZF1 and prognosis in acute lymphoblastic leukemiaNew England Journal of Medicine 360:470–480.https://doi.org/10.1056/NEJMoa0808253
Early B cell factor binds to a site critical for lambda5 core enhancer activityEuropean Journal of Immunology 27:315–320.https://doi.org/10.1002/eji.1830270145
Predominant interaction of both ikaros and helios with the NuRD complex in immature thymocytesJournal of Biological Chemistry 282:30227–30238.https://doi.org/10.1074/jbc.M702541200
Dynamic assembly of silent chromatin during thymocyte maturationNature Genetics 36:502–506.https://doi.org/10.1038/ng1351
Dynamic regulation of transcriptional states by chromatin and transcription factorsNature Reviews Genetics 15:69–81.https://doi.org/10.1038/nrg3623
Nuclear hormone receptor co-repressors: structure and functionMolecular and Cellular Endocrinology 348:440–449.https://doi.org/10.1016/j.mce.2011.08.033
David N ArnostiReviewing Editor; Michigan State University, United States
In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.
Thank you for submitting your article "A high-resolution map of transcriptional repression" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom, David N Arnosti (Reviewer #1), served as Guest Editor, and the evaluation has been overseen by Jessica Tyler as the Senior Editor.
The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.
Three reviewers have read your manuscript, and we agree that this work is potentially suitable for publication in eLife, after you have addressed points raised in the review process.
1) Data reproducibility. A major point raised in the review process was whether the ChIP data shown represents biologically independent measurements or technical replicates from single preparations of chromatin. A number of experiments (Figure 2, 3, 4, 5, 6) show detailed measurements about protein occupancy or histone modification, but in some cases, the authors do not provide enough information about the data presented to allow the reader to judge if the changes measured are convincing demonstrations of biological processes. Specifically, for most of the ChIP measurements, the figure legend notes n=3; is that three measurements of single chromatin preparation, or three independent preparations, independently measured? Certain experiments (e.g. 4D) are described as "independent experiments", thus, the former cases may represent a single biological experiment (chromatin prep), with independent chromatin IP measurements. One line of evidence supporting this view is that individual figures have fairly tight error bars, yet vary quite a bit between experiments, e.g. RNAP drop in Figure 3B,C is 3-5 fold, while in Figure 4C the change is ~2 fold. The paper should clearly distinguish technical from biological replicates, and indicate if the results are supported by more than one experiment. Clearly, the strength of the conclusions is related to the reproducibility of the effects from separate observations.
2) Relationship of findings to prior knowledge about Ikaros. A second point from the review was need to set the authors' findings in a more general context, to show how information about HDAC, NuRD, and chromatin structure in regulation of Myc and Ipll1 supports or contradicts previous studies on Ikaros action. For instance, earlier work from Georgopoulos showed that on the CD4 intronic enhancer, Ikaros and NuRD appear to antagonize each other's activity, such that loss of CHD4 results in enhanced, not reduced Ikaros, activity. Similarly, previous studies have implicated basal factor pTefb-Ikaros interactions, as well as polycomb. Other studies have shown concomitant binding of Ikaros and Ebf1 at the Igll1 promoter in large pre-B cells where the gene is expressed, raising the question of whether Ikaros activates in this context. The paper would be much more valuable to the community if the authors integrate their findings better with previous work, and in the Discussion propose how similarities or differences are to be reconciled. As it stands, the main summary concluding the paper is rather superficial.
3) Presentation and use of genome-wide data. The reviewers found that the integration of genome-wide information was potentially useful but incomplete; the quality of genome-wide information should be indicated with standard statistical measures (e.g. how much agreement is found among the triplicate MNase experiments? reproducibility of peaks etc.), even if the analysis of the entire dataset is being prepared in a separate study. MNase digestion patterns at two loci discussed, but overall significance is difficult to assess; is variation indicative of a change in nucleosome position, occupancy or other aspect of the individual experiments? It is not clear if the RNA pol II genome-wide study was similarly performed in triplicate. A related point to consideration of genome-wide data is related to point 2; it is reported that out of the 924 genes repressed upon increase in Ikaros expression, 372 show a reduction in RNAP2. How are the other 550 genes regulated? Are these not directly repressed by the Ikaros-NuRD complex? It would help this paper to provide enough information to understand Table 1 and Supplementary Data Table 1, which enumerate and list genes affected by Ikaros. For instance, the two genes studied in detail here are direct targets of Ikaros, and show a loss of RNA Pol II. A large number, but not the majority of other repressed genes also show a loss of Pol II; which of these are expected to be direct targets? An additional point regarding the table is the lack of clarity of how genes are assigned to it – is for example a log2 value of 0.05 even a significant change?
4) Modeling. The reviewers had differing opinions about the modeling; some asked for further validation through additional measurements of actual nucleosome positioning and Ikaros occupancy at individual sites, while another view was that the models are useful enough to point toward future studies of mechanism. A justification of the application of the modeling would strengthen this aspect of the study. In addition, a clarification of the actual mechanism suggested is requested: Ikaros has been variously described as a recruiter of NuRD, or an antagonist, or binding in an overlapping manner at promoters but not enhancers. The kinetics of binding shown in 4D indicate that a model with simple, stable recruitment of NuRD by Ikaros is too simple. But it is not clear what mechanism the authors are proposing, based on their findings.
5) Endogenous Ikaros vs. induced form. An increase in Ikaros nuclear localization is assumed from the immunofluorescence data but this is not a very quantitative method. A western blot analysis for Ikaros in the nuclear vs, cytoplasmic fractions at the different time points should be performed. The way the ChIP data is presented, it is not clear how much endogenous Ikaros is already present at the target promoters before the induction of the tamoxifen-regulated protein, and how this compares with the induced levels. Previous studies have already noted that Ikaros is present at these genes, so how much of the modifications are due to overexpression of the protein? Why would nucleosome changes results only from overexpression; is this a function of rapid interconversion between active and inactive states on the genes with only endogenous Ikaros? In addition, the paper is silent about the Ikaros-like Aiolos protein; is it present at these genes, and does it play a role in regulation under these conditions?
6) MNase interpretation. The purported identification of "fragile" sites by MNase tests was not viewed as very strong, for a number of reasons. To identify such changes as an alternatively-bound form of the nucleosome, one would wish for a titration of MNase, not just a single digestion. In addition, it would be important to know if the change is due to a nucleosome shifting in position, rather than being weakly bound. Finally, the differential amplification in 6B with "short" and "long" amplicons may or may not be significant – it is not clear if the n=3 is a technical replicate, in which case, the support is weak.
A number of technical points relating to reagents, methods, and data presentation should be addressed:
7) The antibodies for Ikaros and EBF1 are not described; in the Materials and methods, a reference is made to Ferreiros-Vidal 2013, but EBF1 is not used there. What is the measure of specificity for these reagents? Is there other data to support the EBF1 binding shown here?
8) In Figure 1B, "enrichment" is the measure of chromatin binding for Ikaros and EBF1. The Materials and methods don't describe how these calculations were performed. One way might be to measure the signal at some non-bound site, and divide the ChIP signal% input by this background control. If so, what is their defined background? What were the% input values for these promoters and other locations?
9) For Pol II and TFIIB measurements, the manuscript shows directly the% input recovered from ChIP, which is a preferable approach for chromatin aficionados (their values ~1% are quite credible). The Y-axis is labeled "enrichment", however, which is confusing, because presumably the% input was not normalized to signal at another locus.
10) In Figure 3—figure supplement 1, part B, using the TEV-activated Ikaros system, depletion of Pol II does further affect MNase sensitivity, which is different from the system with 4-OHT induced Ikaros. The authors don't comment on this, though they previously make the point that Pol II eviction is not causing the changes in MNase sensitivity.
11) Figure 5—figure supplement A: TSA treatment apparently induces a decrease in histone acetylation levels? Please comment.
12) The differences for pericentric chromatin localization +/- TSA (6D) and loading of Ikaros with or without CHD4 (7B) may or may not be significant; it is not clear how reproducible these differences are in independent experiments. (See Point 1). In 6D, apparently two experiments were conducted; is this data from one?
13) It is not clear why the KinTec modeling used the specific parameters indicated in Materials and methods. Have these been measured in a particular system?
14) In subsection “Loss of RNAP2, reduced promoter accessibility, and transcriptional repression are early and near-simultaneous events” the authors state that they use primer pairs that span introns as a means to measure unspliced transcripts. They presumably mean primers that span individual exon/intron junctions? The primers used should be included in Materials and methods.
[Editors' note: further revisions were requested prior to acceptance, as described below.]
Thank you for resubmitting your work entitled "A high-resolution map of transcriptional repression" for further consideration at eLife. Your revised article has been favorably evaluated by Jessica Tyler as Senior editor and a Reviewing editor.
The manuscript has been improved but there are four remaining issues that need to be addressed before acceptance, as outlined below:
In this revision of Liang et al., the authors have addressed the points raised in the first round of review. Regarding data reproducibility, the authors have indicated in figure legends which datasets are averages of multiple biological experiments, and which are representative. Additional experiments were performed for 3D FISH, and a Figure 2—figure supplement 1 shows the reproducibility of ChIP and MNase experiments. Regarding previous observations about Ikaros action in different cell types and promoters, they clarify that their findings for c-myc and Igll1 promoters may represent just one side of this protein's regulatory activities, which includes activation at Zfp36.
The reviewers had asked for more specific analysis of genome-wide gene expression, since only a minority of genes showed the Pol II decrease found at the two promoters of interest. In the response to the reviewers, the authors reanalyze the data, and conclude that for a majority of promoters, it appears that Pol II decrease at the promoter is the trend, and that differences in sensitivity of measuring mRNA vs. Pol II occupancy may explain some of the discrepancy.
1) In addition to having this information in the letter of response, the manuscript should indicate that they believe the majority of genes are showing reductions in Pol II, but that differential sensitivity may be an issue. Otherwise it seems the authors are content to ignore what might be a mechanism that impacts the majority of repressed genes.
The use and interpretation of the models were questioned; in response, the authors provide more information on modeling, and emphasize that their many biological measurements point away from the simple two-state model, where Ikaros directly dislodges EBF1. This conclusion seems to be strongly supported by the data. They emphasize the failure to find reasonable parameters for the simple model, rather than over-interpreting specific parameters for more complex models, which seems a reasonable conclusion.
The nature and quality of experiments from genome-wide experiments is better explained and illustrated by the additional supplemental figure noted above, by correlation analyses in the response letter, and data deposited in GEO.
In response to the question about how endogenous Ikaros figures into this system, the authors note that they find similar responses in CRISPR'd cells lacking endogenous Ikaros, and show preliminary data in the letter. As asked, they also carry out a Western blot showing the levels of native and induced Ikaros over the time course, showing that the levels of the inducible form is approximately the same as that of the endogenous protein.
2) The Western blot data should be included in the manuscript as part of, or attached to, Figure 1, where the system is introduced.
The interpretation of a "fragile" nucleosome state was questioned, based on the limited probing with one concentration of MNase. The authors carry out a titration, showing a similar trend, but also soften their conclusion, noting that it appears that it is either nucleosome depletion or movement that is affecting the repressed promoter.
A number of minor points were addressed including provenance of antibodies, figure labeling, and explanation in the figure legend that nascent RNA was measured with intron-exon boundary spanning primers.
3) The text still refers to "intron-spanning primers" – that needs to be fixed.
4) Typo Figure 7 Average nuceosome profilehttps://doi.org/10.7554/eLife.22767.035
- Ziwei Liang
- Brian Hendrich
- David Rueda
- Amanda G Fisher
- Matthias Merkenschlager
- Matthias Merkenschlager
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We thank Drs Jorja and Steven Henikoff for help with MNase-seq, Drs John Schwabe and Ross Chapman for helpful discussion, the STATegra consortium and in particular Drs David Gomez-Cabrero, Peri Noori, Sunjay Fernandes, Mathilda Eriksson and Jesper Tegner for sharing RNA-seq data prior to publication, Drs Pierangela Sabbatini and Niall Dillon for Igll1 cosmids, Dr James Elliott and Thomas Adejumo for cell sorting, and Dr Laurence Game and her team for sequencing. This work was funded by the Medical Research Council UK, Wellcome, and UK-China Scholarships for Excellence (a scheme run jointly by the Department for Business, Innovation and Skills and the China Scholarship Council).
- David N Arnosti, Reviewing Editor, Michigan State University, United States
© 2017, Liang et al
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.