Current Research – continued

Strikingly, parental nucleosomes are re-deposited to the same chromatin domains after DNA replication in the case of repressed, but not active genes (Figure 5). In the latter case, nucleosomes disperse and their fate is undetectable. Importantly, parental nucleosomes are no longer inherited when transcription from a repressed gene is activated by retinoic acid (RA) addition (Figure 5), underscoring that only silenced chromatin is epigenetic. This finding fits well with the ability of PRC2 to “write-and-read” H3K27me3⏤PRC2 recognition of parental nucleosomes comprising H3K27me3 DNA promotes its catalysis of H3K27me3 on naïve nucleosomes that are incorporated on daughter DNA during DNA replication. We are currently investigating the intriguing possibility that specific and novel histone chaperones control the segregation of nucleosomes associated with repressive chromatin.

Figure 5
Figure 5. Local re-deposition of pre-existing nucleosomes occurs in repressed, but not active, chromatin domains. ChIP-qPCR of “marked” nucleosomes as a function of time of release into S-phase of the cell-cycle. The positional inheritance of parental nucleosomes is evident at repressed genes (-RA), but is lost when these genes are activated by retinoic acid addition (+RA).

Halting the spread of repressive chromatin The HOXA cluster, pivotal to development, is completely repressed and engaged by PRC2 in ESC. Notably, the presence of intergenic DNA-binding sites for the transcription factor and insulator protein, CTCF, within the HOXA cluster are key to maintaining boundaries between repressed and active genes during development, and CTCF is known to interact with cohesin in establishing topologically-associated domains (TADs). We find that disruption of one such CTCF site in the HOXA cluster gives rise to an aberrant gene activation during ESC differentiation, leading to a homeotic transformation. This phenomenon is also evidenced by an invasion of H3K4me3 (associated with transcriptionally active chromatin) into the normally repressed domain containing H3K27me3 (Figure 6).

Figure 6
Figure 6. Loss of CTCF binding causes transcription to spread within the HOXA cluster. Top, Schematic showing the linear, spatiotemporal expression of genes within the HOXA cluster as a function of development. Bottom, ChIP-seq tracks for CTCF and the indicated active (H3K4me3) and repressed (H3K27me3) chromatin modifications along the HOXA cluster. DNA binding sites for CTCF are displayed on top. Deletion of the CTCF site between Hoxa5 and Hoxa6 (Δ5|6) results in H3K4me3 spreading into the neighboring gene flanked by a CTCF site in cervical motor neurons (MNs).

Importantly, the CTCF boundaries in the HOXA cluster function in a cell-type specific manner, yet CTCF remains bound to its sites during differentiation, thus, we predicted that factors in addition to cohesin would specify CTCF-boundary formation. By devising a loss-of-function genetic screen, we identified genes whose deletion leads to an aberrant gene activation similar to that observed upon deleting the CTCF binding site, yet CTCF binding was unaffected. We are currently pursuing a selected set of such genes that point to boundary formation being a non-trivial process that, in addition to CTCF, entails other factors that appear to be site-specific.

Additional headway comes from our discovery that CTCF interacts with RNA, fostering CTCF oligomerization. Two CTCF Zinc Fingers are important for RNA interaction and remarkably, this interaction is important for establishing appropriate boundaries within TADs, long-range interactions via chromatin loops and appropriate gene expression. Remarkably, CTCF mutant in one such ZF affects cohesin association with chromatin at some CTCF sites, yet the boundary is maintained (Figure 7), reinforcing that not every CTCF site functions identically.

We hypothesize that the many CTCF-mediated interactions/insulation with its cognate DNA sites carry intrinsic specificity mediated by specific factors, including RNA, that participate in establishing long-range interactions and TAD formation, and are actively pursuing this prospect.

Figure 7
Figure 7. RNA binding mutants of CTCF. Top, left, Schematic representation of known CTCF protein domains with its 11 zinc fingers (ZFs) shown numbered. Top, right, Table summarizing differences between the ZF mutants (ZF1∆ and ZF10∆) that are deficient in RNA-binding. Bottom, ChIP-seq for CTCF or Cohesin (Rad21 subunit) showing binding to chromatin at insulation sites in cells expressing the following versions of CTCF: WT (gray), ZF1∆ (black), or ZF10∆ (red). Below the ChIP-seq data are graphical representations of an insulated domain formed by CTCF: WT (left), ZF1∆ (middle) or ZF10∆ (right). 2 CTCF proteins (green) and cohesin rings (blue) are stabilized by an RNA in the case of WT CTCF (left). The following outcomes are exhibited by the RNA binding-deficient mutants at a subset of domains: 1) both ZF1∆ proteins remain bound and the domain remains insulated in the absence of cohesin, possibly with the help of an unknown factor (middle) and 2) both cohesin and ZF10∆ have lost one insulation site flanking the domain, disrupting the insulation and resulting in aberrant contacts (long-range interactions) with other regulatory regions, such as enhancers (right).

PRC1 The other major PRC complex is PRC1. Mammalian PRC1 is quite heterogenous. Our proteomic studies identified 6 major, distinctive PRC1 complexes and the PcG proteins comprising them. We also found some unexpected constituents. We discovered a PRC1 complex (PRC1-AUTS2) that harbors non-PcG proteins: the kinase CK2 and strikingly, AUTS2. The function of AUTS2 is unknown, but it is putatively involved in Autism Spectrum Disorders (ASD) and/or neurological functions. Remarkably, in contrast to the accepted role of PRC1 in repression, we demonstrated that PRC1-AUTS2 activates transcription in vivo and elucidated the mechanism underlying this conversion of PRC1 through its interaction with AUTS2. We are fully engaged in this promising, new investigation and have initiated studies to understand the functional role of PRC1-AUTS2 and other novel PRC1 complexes in neurodevelopment, as well as phenotypes associated with ASD. We are testing an exciting hypothesis that entails a switch in PRC1 composition from one (PRC1-AUTS2) that activates neurodevelopmental genes in the embryo expressing AUTS2 to another that represses these genes in adulthood when AUTS2 is absent. We have used CRISPR to tag or delete components of PRC1-AUTS2 in ESC and will investigate how the differential composition and genomic localization of these complexes impact transcriptional programs as ESC differentiate into cortical neurons. In tandem, we are analyzing how PRC1-AUTS2 regulates neurological functions and behavior with our conditional AUTS2 knockout mice. We have already found profound behavioral alterations indicative of ASD and are beginning to link these behavioral phenotypes to brain pathology and aberrations in neuronal function.

Epigenetics in a model organism

To fully grasp the epigenetic basis of gene programming in the context of a whole organism, we are exploring the blueprint of epigenetic features in experimentally approachable, eusocial creatures: ants.  Having sequenced the genomes of two ant species, we are now tackling the alterations in epigenetic features that accompany changes in ant morphology and behavior with an eye towards unraveling the epigenetic programming of human behavior and the aging process.

Figure 8
Figure 8. Behavioral memory and reversible longevity in the ant Harpegnathos saltator (Hsal). (A) In an Hsal colony with queen and worker castes, the worker fecundity is suppressed by queen pheromones. The Hsal caste is extremely plastic and easily manipulated as a function of the presence of a queen. Without the queen, a worker increases in fecundity and switches to a gamergate with full-grown ovaries. The gamergate can be fully reverted to a worker with worker-like ovaries, when returned to a queened-colony (revertant). Longevity and reproduction are positively correlated in ants: the disparate lifespans of queen, worker, gamergate and revertant are indicated. Note that the lifespan after switching a worker to a gamergate is significantly extended and then, significantly diminished when a gamergate is reverted. The revertant can restore its fecundity and become a gamergate for a second time (2nd Gamergate) when placed in a queen-less colony. (B) Strikingly, the revertant retains “gamergate memory” upon its conversion to 2nd gamergate⏤it lays eggs earlier than does a gamergate.

We identified a set of genes whose expression is altered during the first transition from worker to gamergate and importantly, the expression of some of these genes remains altered upon reverting the gamergate to a full worker. This altered transcription program will be fully explored to study longevity (acquired in the first transition) and epigenetic memory (acquired in the second transition). We are also studying the process of aging. We find that an interplay between high insulin levels to promote oogenesis and anti-insulin factors to prevent aging in the fat body/adipocytes is responsible for the unusual correlation of high reproduction with extended longevity in ants.

Harpegnathos saltator (Hsal) colonies contain a queen and a single caste of workers, but possess a natural adult plasticity in caste switching: without the queen, a non-reproductive worker can convert into a reproductive “surrogate” queen aka gamergate (Figure 8A). This process accompanies a life-span extension: a queen lives over 5 years, workers normally 7 months and a gamergate can live more than 3 years! This process can be reverted when the gamergate is placed in a queened-colony, including: partial loss in longevity, renewed foraging activity and most importantly, the loss of production of queen pheromones and of the ability to produce oocytes and lay eggs (Figure 8A). Most strikingly, a second transition to a gamergate proceeds at a much faster rate than the initial transition, as evidenced by earlier times of egg laying (Figure 8B). We hypothesize that an epigenetic memory is established on genes regulating caste-specific phenotypes during the first transition from worker to gamergate that enhances the responsiveness of the 2nd transition, compared to converting a naïve worker.