Chromosome togetherness at the onset of ESC differentiation

Abstract

Pairing of homologous alleles is a phenomenon generally associated with imprinted and monoallelically expressed loci. In this issue, Hogan et al. (2015) examine the earliest steps between pluripotency and lineage commitment in ESCs and find a critical role for transient pairing of Oct4 alleles in exiting the pluripotent state.

Spatial organization of the genome contributes to the regulation of gene expression. It is well known that in interphase cells the genome is organized into separate chromosome territories. Recent observations using Hi-C also support this view by showing that genomic regions within the same chromosome interact with high frequency and inter-chromosomal interactions are rare (Lieberman-Aiden et al., 2009). Although sequence identity between homologous chromosomal regions complicates the ability of Hi-C approaches to reliably uncover inter-homolog interactions, FISH studies showed that homologous chromosomes occupy separate territories in the nucleus suggesting, that interaction between homologs should be rare ( Cremer and Cremer, 2010).

Although pairing of homologous chromosomes is not common in mammalian cells, it has been described in a number of cases. For example, pairing occurs between X chromosomes at the initial stages of X inactivation, which is critical for silencing one of the alleles (Anguera et al., 2006). Other reports suggest that other mono-allelically expressed regions also pair in the nucleus. For example, homologue pairing is critical for allelic exclusion in immunoglobulin loci (Hewitt et al., 2009) and the imprinting of PWS/AS loci in brain and lymphocytes (Thatcher et al., 2005). Hogan et al (2015) now show that homologue pairing has implications for transcription and is involved in simultaneous repression of Oct4 alleles during ESC differentiation.

Studies of genome wide chromatin interaction frequencies before and after ESC differentiation into discrete lineages have revealed considerable differences, but how these differences are dynamically established is much less clear. To address this issue, Hogan, et al used multi-color DNA fluorescence in situ hybridization (FISH) to localize Oct4, Sox2 and Nanog (OSN) alleles at the earliest stages of mouse ES cell differentiation away from the pluripotent state. OSN are co-regulated and their transcription is necessary to maintain pluripotency and decreases as ES cells transition into lineage specification. During a four day time course after removal of LIF, heterologous allele pairing or homologous pairing of Sox2 or Nanog alleles were not observed. Strikingly, Oct4 alleles were observed to be paired specifically at one time point, on day 3 after LIF removal, and not before or after. It is unclear over how many hours the pairing was present. No pairing at any time point was observed among the more than 20 other genes that were examined.

To ask if the observed pairing is associated with exit from the pluripotent state, the authors used repression of Oct4 transcription as the metric. First, they performed RNA FISH in combination with DNA FISH at the time of pairing and found that more than 60% of pairs had ceased transcribing Oct4 from one or both alleles. This suggested that pairing was contemporaneous with downregulation of Oct4 expression. Next, they used alternative differentiation regimens and again observed transient Oct4 pairing but at a different time point. They observed a remarkable correlation between the day of Oct4 allele pairing, and the day of greatest decrease in Oct4 transcription. Finally, the connection between Oct4 allele pairing and Oct4 repression was examined in vivo in post-implantation mouse embryo epiblast cells. As Oct4 is being downregulated in ectoderm/neuroectoderm and the neural lineage is being specified, 25-30% of the Oct4 alleles were observed to be paired. No pairing was observed at earlier or later stages of development or for Sox2 or Nanog alleles. Taken together, these experiments suggest that in ESCs and in vivo Oct4 allele pairing could be directly associated with Oct4 repression or that the pairing and repression could both be associated with a particular stage of differentiation during which Oct4 is being downregulated.

To delve more deeply into the mechanism underlying pairing, a 10 kb region containing the gene and its upstream distal (DR) and proximal (PR) enhancers was divided into four fragments and ES lines were established with each fragment inserted into the same transgenic position by homologous recombination. The authors found that the transgenic locus containing the DR, PR and the Oct4 promoter paired with the WT Oct4 allele at the same frequency as WT Oct4 alleles paired with each other, while none of the other fragments supported pairing over the background. Next, to address what DNA binding proteins might mediate Oct4 allele pairing, motifs of proteins suggested by other studies to be involved in allele pairing were identified by bioinformatic analysis and mutated. Mutation of Oct4/Sox2 motifs resulted in loss of pairing between the mutant and WT alleles, while mutation of CTCF, YY1 or E2A sites had no effect on pairing. These data implicate OCT4/SOX2 as necessary for pairing. However, they are unlikely to be sufficient because they robustly occupy the regulatory DE site before differentiation when there is no pairing, whereas, pairing occurs as their occupancy diminishes. Moreover, OCT4/SOX2 bind to regulatory sites of non-pairing Sox2 and Nanog alleles (Chew et al., 2005; Levasseur et al., 2008). Further investigation is needed to illuminate this issue.

Thus far, the results support a correlation between Oct4 pairing and silencing as differentiation proceeds, but do not resolve causation. To address this issue, the authors examined epigenetic modification in cells carrying pairing competent and pairing incompetent transgenic Oct4 regulatory regions using allele specific ChIP. The histone modification H3K9me2 is associated with silencing and is known to be a robust indicator of Oct4 silencing. In these cells, WT Oct4 alleles have ample opportunity to pair and are silenced concurrent with increasing H3K9me2 modification from day 3, when pairing occurs, until day 6, as expected. The transgenic locus with YY1 sites mutated also pairs with WT loci, as shown by PE-4Cseq, a 3C variant, and is silenced with the same kinetics as WT loci. However, at pairing incompetent Oct4/Sox2 mutated alleles, H3K4me2 modification does not progress after day 3. The picture that emerges is that transient pairing of Oct4 alleles at day 3 is necessary for the progress of silencing.

As the new kid on the block, it will be informative to compare the pairing phenomenon described by Hogan et al (2015) to the mechanism of X chromosome inactivation during differentiation of ES cells (Anguera et al., 2004). Transient pairing of X chromosomes is a prerequisite to repression of one of the X homologs. Pairing also depends on Oct4 protein and a specific sequence (pairing region) which can cause pairing if inserted into an autosome (Xu et al., 2006; Augui et al., 2007).

The big mysteries remaining include how homologues in different chromosome territories come to pair, what exactly is happening at pairing to cause the downstream effects and what additional proteins beyond Oct4/Sox2 contribute to pairing? Further, how is it that one chromosome is silenced in X-inactivation while both alleles of Oct4 are silenced? Finally, it is intriguing that the regions necessary for pairing in X-inactivation and at Oct4 both contain enhancers. Enhancers loop to target genes and genome wide they also loop to other enhancers. The chromosome pairing data may suggest new roles for enhancers beyond transcription activation.

Figure 1. Relationship between Oct4 allele pairing and repression during differentiation.

During differentiation of ESC away from pluripotency, Oct4 transcription diminishes. Hogan et al (2015) identified transient pairing between Oct4 homologues that occurred at a particular point in ESC differentiation. Oct4/Sox2 sites within regulatory regions of Oct4 were required for the pairing. In the absence of pairing, Oct4 regulatory sequences failed to be modified by H3K9me2, which normally occurs during Oct4 silencing.