DNA replication: an opportunity to establish asymmetric epigenome

Canonical histones are mainly transcribed and translated endogenously during S phase [62,63]. In principle, both old histones and new histones are present at the replication fork. Even though the mechanisms behind new histone incorporation during DNA replication have been widely studied [64,65], less had been understood regarding how old histones are recycled until recently [57,66–68]

Here we will discuss recent findings with a focus on the patterns and mechanisms underlying old histone recycling in different systems. Recently, a series of genomics studies have shed light on such patterns and mechanisms in symmetrically dividing cells, such as mouse embryonic stem cells (mESCs) and budding yeast. Using the SCAR-seq (sister chromatids after replication by DNA sequencing) technique in mESCs, the old histone deposition pattern is tracked with de novo DNA replication at a genome wide level. Results from this have revealed that old histones H3 and H4 are recycled to both leading and lagging DNA strands, with a slight bias towards the leading strand (Figure 1B) [66]. A recent study using a similar method in human retinal pigment epithelial (hRPE-1) cells indicates this bias could be even stronger [69]. MCM2 is an essential subunit of the CMG (Cdc45–MCM–GINS) DNA helicase and can also act as a chaperone in recycling old histones toward the lagging strand. Mutations at the histone binding sites of the MCM2 have significantly increased the amount of old histone recycling towards the leading strand [66]. Using the eSPAN (Enrichment and Sequencing of Protein-associated Nascent DNA) technique, similar studies have been performed in budding yeast. These results have further substantiated that replication proteins could function as histone chaperones in regulating the Replication-Coupled Nucleosome Assembly (RCNA) process at the replication fork. For example, two subunits of the leading-strand DNA polymerase ε (Pol ε), DPB3 and DPB4, facilitate old histone deposition onto the leading strand (Figure 1C) [68]. Loss-of-function of either subunit leads to the incorporation of old histones biased toward the lagging strand. Furthermore, it has been demonstrated that PCNA (Proliferating Cell Nuclear Antigen), a processivity factor for DNA polymerases [70], coordinates histone deposition at the replication fork [71,72]. In addition, CTF4, Pol α and PCNA were found enriched at the lagging strand [73], leading to the hypothesis that this axis could serve to deposit old histones onto the lagging strand (Figure 1C) [72,74]. Recently, the Mcm2–Ctf4–Pol α axis has been found to facilitate the re-distribution of parental H3–H4 tetramers to lagging-strand DNA at replication forks (Figure 1C) [67]. Consistent with the results in mESCs, mutating the histone-binding domain of the Mcm2 in yeast leads to an enrichment of parental H3–H4 on the leading strand. Similar effects were also observed with the Ctf4 and Pol α primase mutants (Figure 1D) [67].

Open in a separate window

Figure 1.

Models for replication-coupled nucleosome assembly (RCNA) highlighting old histone H3–H4 recycling.

(A) In Drosophila male GSCs, old (H3–H4)₂ are preferentially recycled to the leading strand with yet-to-be identified mechanisms [12]. By default, newly synthesized (H3–H4)₂ are incorporated by the lagging strand, thereby generating a biased histone incorporation in a strand biased manner. (B) In mESCs, the role of Mcm2 is shown to recycle old (H3–H4)₂ towards the lagging strand [66]. (C) In yeast, old histones recycle towards leading strand via interactions with Dpb3/Dpb4-Polε [68], while the MCM2-ctf4-Polα axis is shown to recycle old (H3–H4)₂ towards the lagging strand [67].

Regarding the dynamics of nucleosomes during the RCNA process, it has been shown that most of the old (H3–H4)₂ tetramers remain intact during DNA replication. This suggests that nucleosomes contain either old tetramers or new (H3–H4)₂ tetramers, rather than a tetramer containing both [75,76]. In contrast, most of the old (H2A–H2B)₂ are split as dimers, suggesting that the newly formed nucleosomes should contain a mixture of old and new H2A–H2B, rather than solely one or the other. Therefore, the old intact (H3–H4)₂ tetramer can only be inherited by one strand (either the leading or lagging strand), leading to more asymmetric inheritance patterns at the replication fork. On the other hand, the presence of two split old H2A–H2B dimers gives an equal probability of being incorporated by either the leading or lagging strand, thus resulting in a tendency of more symmetric inheritance pattern. In addition to the splitting mode differences, the (H3–H4)₂ tetramers are relatively more stable in the nucleosome, while the (H2A–H2B) dimers display a more rapid exchange during the post-replication chromatin maturation process [77–79]. This could also explain the differential distribution patterns seen on replicated sister chromatids that enter mitosis.

Additionally, an imaging-based technique named Super-Resolution of Chromatin Fiber (SRCF) has been developed to directly visualize distribution of old versus new histones at the replication regions by combining high spatial resolution imaging and chromatin fiber techniques [12,80]. Using early germline-derived chromatin fibers that express old and new H3 or H4 labeled with two distinct fluorescent tags, the recycling pattern of the old and new H3 or H4 at the replication regions can be resolved. SRCF revealed that the old and new histones H3 and H4 are recycled with a leading strand bias, whereas new H3 and H4 are incorporated towards the lagging strand (Figure 1A). In contrast, both old and new histones H2A showed a largely symmetric incorporation pattern at the replication fork [12]. Such differential histone incorporation patterns indicate the presence of molecular specificities during RCNA [11,12,81–83] Furthermore, DNA fibers that are sequentially labeled by the thymidine analog EdU and BrdU facilitate the investigation of directionality in replication fork movement. Using symmetrically dividing somatic cells as a control, a significantly higher incidence of unidirectional fork movement was detected in early germ cells. The strand biased old versus new histone incorporation, coupled with predominantly unidirectional fork movement, could act together to establish an asymmetric histone incorporation pattern in GSCs. While the SRCF is a powerful technique that is compatible with a few cells such as adult stem cells, improvement on cell type and cell stage specificities is needed to label or isolate fibers precisely from stem cells. Future studies will help further our understanding of the molecular mechanisms underlying such a biased incorporation pattern in the Drosophila male GSC system. Consistent with the asymmetric old and new H3 and H4 incorporation during DNA replication, such asymmetry is reflected by a global difference between sister chromatids during and after Drosophila male GSC asymmetric divisions (Figure 2A). Conversely, old and new H2A that show symmetric incorporation at replication regions display largely symmetric inheritance patterns during and after ACD of male GSCs. In summary, these results demonstrate that the distinct histone incorporation patterns at the replication fork underlie their different inheritance modes, as observed in asymmetrically dividing stem cells and post-mitotic daughter cells. Therefore, DNA replication establishes the asymmetry of the epigenome. In contrast, such an asymmetric histone inheritance pattern is not detected in the symmetrically dividing progenitor germ cells, indicating that the processes of RCNA are likely subject to cell type or stage-specific regulation during tissue homeostasis.

Open in a separate window

Figure 2.

Schematic illustration of the different histone inheritance pattern in different stem cell system.

(A) Drosophila male GSC: sister chromatids are either enriched with old histone (green) or new histone (red) H3 and H4. Histone segregates asymmetrically into the daughter cells; a representative example of global histone asymmetry. Also, CENP-A segregates asymmetrically into daughter cell, represent centromere epigenetic asymmetry and non-random sister chromatids segregation. (B) Drosophila female GSC: sister chromatids are overall carrying a symmetric level of old and new histone. However, a local asymmetry between old and new histone can be seen at stem cell maintenance genes and differentiation gene loci; a representative example of gene-specific histone asymmetry (also found in mouse embryonic stem cells that are induced to undergo ACD). Interestingly, despite of the overall symmetric histone segregation, CENP-A segregates asymmetrically into daughter cell, suggesting sister chromatids still carry epigenetic information differentially and segregate non-randomly.

Mitotic machinery distinguishes epigenetic differences between sister centromeres

Centrosomes are the microtubules organization center (MTOC) in a cell. It is interesting to note that asymmetric inheritance of centrosomes has been reported during ACD of different stem cell types. For example, Drosophila male GSCs and mouse neural glial progenitor cells inherit the mother centrosome, while the differentiating daughter cell inherits the daughter centrosome [22,25]. On the other hand, Drosophila female GSCs and neuroblasts inherit the daughter centrosome, and the differentiating daughter cells inherit the mother centrosome [23,24]. Such a biased inheritance of age-different centrosomes provokes the speculation that nonrandom sister chromatids segregation is tightly coordinated with differential centrosome inheritance. Questions remained, for example, is there a differential crosstalk between the microtubule nucleation from mother versus daughter centrosomes and epigenetically distinct sister centromeres? Because microtubule-centromere interactions are highly dynamic, a high spatial and temporal resolution live cell imaging approach is necessary to study their communications. However, live cell imaging of the mitotic machinery components in cells from intact tissue with high spatiotemporal resolution has been technically challenging [108]. Recently, a Super-Resolution Live Snapshots (SRLS) method has been developed to visualize mitotic spindle and sister centromere dynamics at a high spatial resolution in intact tissues [109]. High spatial resolution snapshots of living cells were collected and arranged according to their sequential orders during cell cycles to investigate highly dynamic cellular events in detail. Using SRLS, it has been shown that the MTOC activity between mother and daughter centrosomes is temporally asymmetric in Drosophila male GSCs. The mother centrosome is highly active during the late G2 phase, prior to the daughter centrosome, which only becomes active during the G2-to-M phase transition (Figure 3A). These temporally asymmetric microtubules lead to a polarized nuclear membrane breakdown (Figure 3B), which happens first at the stem cell side during the G2–M transition and later at the differentiating daughter side in prometaphase GSCs (Figure 3C). These data suggest that asymmetric microtubule activity and polarized nuclear membrane breakdown work in tandem to bias microtubule-kinetochore attachment. First, microtubules emanated from the mother centrosome on the stem cell side preferentially attach to the stronger sister centromere during prophase (Figure 3B). Later, microtubules emanated from the daughter centrosome on the differentiating daughter side subsequently attach to the weaker sister centromere during prometaphase (Figure 3C) [20]. It is worth noting that the stronger sister centromere, together with the stronger sister kinetochore, show higher affinities with microtubules, leading to more microtubule attachment (Figure 3B–D). This differential attachment is likely stabilized, resulting in non-random sister chromatid segregation as observed in both Drosophila male and female GSCs (Figure 2A,B) [20,106]. Therefore, the mitotic machinery forms an asymmetric axis in a spatiotemporally regulated manner during ACD of GSCs as follows: centrosome>microtubules>nuclear membrane>kinetochore>centromere. This ensures that the sister chromatid inheritance will be nonrandom. On the other hand, disruption of the temporally asymmetric microtubules by nocodazole, a microtubule depolymerizing drug, results in the loss of preferential microtubules and sister centromere attachment, and subsequent randomized sister chromatid segregation [20]. These discoveries demonstrate that a cascade of tightly coordinated spatially and temporally asymmetric cis factors (e.g. sister centromeres) and trans events (e.g. microtubules) are essential for asymmetric epigenetic inheritance and cell fate differences. These highly coordinated events in asymmetrically dividing cells have been termed as ‘mitotic drive’ [11,20] (Figure 3 and Supplementary Movie 1).

Open in a separate window

Figure 3.

Mitotic drive phenomenon in Drosophila male GSCs.

Schematic diagram showing one pair of sister chromatids. (A) Late G2: active microtubules from the mother centrosome start poking nuclear membrane toward the stem cell side. At the same time, centromeres are clustered at the stem cell side where microtubules poke in. However, sister centromeres are not resolved at this time. (B) Early prophase: highly active microtubules from the mother centrosome locally break the nuclear membrane toward the stem cell side, allowing microtubules to enter into the nucleus. The stronger sister centromere is stably attached with more microtubules. At the same time, daughter centrosome nucleates active microtubule that start poking nuclear membrane toward the differentiating daughter cell side. (C) Prometaphase: active microtubules emanated from the daughter centrosome locally break the nuclear membrane toward the differentiating daughter cell side. The weaker sister centromere is attached by daughter centrosome-emanating microtubules. (D) Metaphase: a biased bi-oriented spindle with the stronger centromere attached by the mother centrosome-emanating microtubules, whereas the weaker centromere attached by the daughter centrosome-emanating microtubules. (E) Anaphase: the biased microtubule-kinetochore attachment leads to non-random sister chromatid segregation, the stronger centromere at the old histone enriched-chromatid segregates to the self-renewing stem cell, whereas the weaker centromere at the new histone-enriched chromatid is segregated into the differentiating daughter cell side.

A ‘meiotic drive’ phenomenon has been reported in meiosis I during the development of a mouse egg [110,111]. This observation is called ‘centromere drive’ when centromeres serve as a ‘selfish’ element. When mice strains with different kinetochore strengths breed, the oocyte inherits the stronger kinetochore and centromere while the weaker kinetochores and centromeres are inherited by the polar bodies, which subsequently degenerate [112–114] Here, the stronger centromeres have 6 to10 fold more minor satellite repeats, allowing them to recruit more kinetochore proteins. In addition, the meiotic spindle shows asymmetric PTMs, such as high tyrosinated microtubules towards the polar body side compared with the egg side due to the polarized CDC42 signaling [115]. This spindle asymmetry and polarized cortical CDC42 signal were found to be essential to flip and re-orient the stronger centromere towards the egg side. These results suggest that the polarized cortical CDC42 signals serve as a spatial cue for selfish centromeres to distinguish between egg and polar bodies to ensure their preferential retention during meiosis I. Furthermore, these studies show that the flipping and reorientation events of centromeres on homologous chromosomes depends on the amount of destabilizers, such as chromosome passenger complex (CPC) and mitotic centromere-associated kinesin (MCAK) [116]. Stronger centromeres recruit more destabilizing factors (MCAK and CPC), increasing the opportunity of the stronger centromeres to flip, especially when they are attached with tyrosinated microtubules. In addition, ‘meiotic drive’ might involve non-centromeric loci, such as the knob domain in Maize, which acts as a neocentromere by attaching to the meiotic spindle without a kinetochore, leading to the preferential retention of knob domain chromosomes in progenies [110,117,118]. ‘Meiotic drive’ and ‘mitotic drive’ each have distinct features, which are summarized in Table 1 and in recent reviews [11,119].