Asymmetric cell division (ACD) produces daughter cells with separate distinct cell fates and is critical for the development and regulation of multicellular organisms. Epigenetic mechanisms are key players in cell fate determination. Centromeres, epigenetically specified loci defined by the presence of the histone H3-variant, centromere protein A (CENP-A), are essential for chromosome segregation at cell division. ACDs in stem cells and in oocyte meiosis have been proposed to be reliant on centromere integrity for the regulation of the non-random segregation of chromosomes. It has recently been shown that CENP-A is asymmetrically distributed between the centromeres of sister chromatids in male and female Drosophila germline stem cells (GSCs), with more CENP-A on sister chromatids to be segregated to the GSC. This imbalance in centromere strength correlates with the temporal and asymmetric assembly of the mitotic spindle and potentially orientates the cell to allow for biased sister chromatid retention in stem cells. In this essay, we discuss the recent evidence for asymmetric sister centromeres in stem cells. Thereafter, we discuss mechanistic avenues to establish this sister centromere asymmetry and how it ultimately might influence cell fate.
Asymmetric cell division in stem cells
The asymmetric nature of a stem cell is fundamental to the development of diverse multicellular organisms and stem cell maintenance throughout adult life. In this sense, asymmetric cell division (ACD) allows for organism complexity by generating and maintaining diverse and specialised cell types. Examples of such include haematopoietic stem cells facilitating haematopoiesis of blood cells , and gametogenesis made possible by germline stem cells (GSCs) that divide asymmetrically in the gonads [2,3]. Moreover, ACD is not limited to stem cells, as meiosis is also an inherently asymmetric process [4,5]. ACD generates two distinct cell fates in these cell types. Stem cells divide asymmetrically to produce a daughter cell that self-renews as well as a differentiating daughter cell. Similarly, female meiosis produces the oocyte and polar bodies. These ACDs ultimately present unique biological challenges. Disruption to this balance of self-renewal versus differentiation, and/or gametogenesis, is detrimental to normal tissue homoeostasis and contributes towards the aetiology of diseases such as cancer and potentially infertility [6–8].
Stem cells regulate these distinct cell fate decisions both intrinsically and extrinsically. The stem cell microenvironment, or niche, plays a powerful role in regulating the extrinsic signalling factors that allow the stem cell to make cell fate decisions [9–11]. Signalling ligands, through, e.g. bone morphogenic protein (BMP) and Janus kinase signal transducers and activators of transcription (JAK/STAT) pathways, are well characterised and heavily influence stem cell division from the niche [12–15]. Intrinsic processes have also been extensively studied, mostly pertaining to asymmetric protein/ribonucleic acid (RNA) localisation and cell polarity, differences which ultimately impact gene expression [16–18]. Epigenetics encompasses the heritable changes in gene expression that do not alter the original genetic code. Epigenetic mechanisms inform the transcriptional status of daughter cells, providing additional methods to alter cell fate. To carry this so-called ‘epigenetic memory’ from one cell division to the next, the parental chromatin needs to be inherited along with the duplicated deoxyribonucleic acid (DNA) sequence. The centromere is the primary constriction site of the chromosome and is essential for faithful chromosome segregation at cell division. The centromere itself is epigenetically specified by the histone H3 variant, centromere protein A (CENP-A), providing a structural framework for the efficient inheritance of chromosomes at cell division. In the case of a stem cell, it is postulated that the epigenetic landscape between sister chromatids differ, and in particular at centromeres [19,20]. The incorporation of a histone variant such as CENP-A or the post-translational modification of histones might potentially serve as epigenetic marks that distinguish sister centromeres and chromatids [19–21]. In this mini-review, we discuss the assembly of the centromere as a paradigm of epigenetic inheritance in ACD in stem cells, mechanisms of how sister centromere asymmetry may arise, and possible impacts on cell fate.
Non-random chromosome segregation
Non-random chromosome segregation is a phenomenon in which two supposedly identical sister chromatids can be distinguished and one sister chromatid is selectively segregated towards a specific daughter cell. The non-random segregation of sister chromatids has previously been observed in certain stem cell subpopulations [22–24]. In 1975, Cairns  originally proposed the immortal strand hypothesis as a method of explaining this phenomenon. The hypothesis assumes that adult stem cells retain the parental, older, template DNA as a protective measure to limit the accumulation of aberrant mutations arising from erroneous DNA replication . Although not widely accepted, this hypothesis proved quite resistant to challenge due to limitations in methods available to robustly test it. However, reconciling this hypothesis with our current understanding of DNA repair, stem cell turnover and the overall organisation of the genome it now appears an unlikely prospect . Moreover, studies in the well-established Drosophila GSC system further oppose this hypothesis, concluding that immortal stands are not exhibited by these cells [27–30].
The silent sister chromatid hypothesis, proposed by Lansdorp  in 2007, may help better explain non-random chromosome segregation. This hypothesis states that ACD and cell fate decisions are orchestrated by epigenetic differences between sister chromatids. Indeed, evidence for epigenetically distinct sister chromatids is now emerging, again in Drosophila. In male GSCs, pre-existing (parental) histones H3 and H4 are selectively retained, whereas newly synthesised H3 and H4 are segregated to the daughter cell [31,32]. Furthermore, there is evidence for the selective retention of post-translationally modified histone H3 at threonine 3 (H3T3P) in GSCs .
It has also been reported that pre-existing (parental) CENP-A, the histone H3-variant that defines the centromere , is similarly retained by GSCs and intestinal stem cells (ISCs) [35,36]. Recent studies from the Chen and Dunleavy laboratories have implicated the centromere in the control of non-random sister chromatid segregation in both male and female Drosophila GSCs [35,37]. Both studies show that more CENP-A is present on sister chromatids of the future GSC. Furthermore, the asymmetric distribution of CENP-A correlates with an asymmetric distribution of kinetochore proteins and microtubules. Both studies propose that these asymmetries might ultimately bias sister chromatid segregation (Figure 1). In addition, Dattoli et al.  showed that disruption of sister centromere asymmetry changes the balance of stem versus daughter cell. Here, disruption to the centromeric core possibly regulates cell fate decisions by maintaining stem cells in a self-renewing state .
Centromere and spindle configurations in a symmetrically versus asymmetrically dividing mitotic cell
Sister centromere asymmetry in non-random chromosome segregation
Centromeres constitute the primary constriction site of a chromosome and form the chromatin landscape to which the kinetochore assembles and microtubules subsequently attach . Hence, the centromere is the foundation for faithful chromosome segregation in a dividing cell. Stem cell divisions are asymmetrically orientated by means of both cell and spindle polarity [39–42]. Similarly, oocytes display asymmetric cell and spindle polarities , as well as centromere asymmetries [44,45], proposed to bias chromosome segregation in meiosis [44,46,47]. Hence, it is conceivable that differences in centromere strength between individual sister chromatids might drive stem cell identity by directing non-random sister chromatid segregation.
Asymmetric distribution of kinetochore proteins was first observed in budding yeast post-meiotic lineages . Studies by Ranjan et al.  and Dattoli et al.  have now shown that in both male and female Drosophila GSCs, future stem cell centromeres are ‘stronger’. Specifically, sister chromatids retained in the stem cell contain 1.2–1.4-fold more CENP-A and harbour stronger kinetochores, measured for centromere protein C (CENP-C) and NDC80 [35,37]. Stronger centromeres and kinetochores correlate with the emanation of more microtubules on the stem side [35,37]. In males, the timing of nuclear envelope breakdown facilitates differential microtubule activities, with the GSC-side nuclear envelope breaking down earlier in G2-phase . Conceptually, these stem cell populations display a ‘mitotic drive’ .
Previous studies in Drosophila have shown directionality in the retention of additional cellular components, for example mother and daughter centrosomes. Male GSCs retain the mother centrosome , whereas female GSCs retain the daughter centrosome . Moreover, midbody segregation in male and female GSCs correlates with daughter centrosome inheritance , which is opposite in either sex. Recent observations of centromere and microtubule bias towards the GSC side in males and females do not appear to correlate with mother centrosome inheritance [35,37]. In both cases, GSC centromeres are stronger, independently of mother or daughter centrosome retention. Therefore, we hypothesise that centromere strength might be a driver of asymmetric spindle assembly in these stem cells. Surprisingly, female GSCs are capable of forming a mitotic spindle in the absence of centrioles, in a Drosophila Spindle Assembly Abnormal-4 (DSas-4) mutant background . This raises the possibility that spindle assembly in GSCs could be a more chromatin- or even centromere-driven process, similar to meiotic spindle assembly observed in oocytes of many different systems . It would be interesting to know whether the microtubule strength asymmetry is maintained in acentrosomal DSas-4 mutant GSCs . Nonetheless, now that epigenetically distinct sister centromeres have been identified, questions moving forward should aim to elucidate how the stem cell might mechanistically distinguish such centromeres.
Can CENP-A assembly drive sister centromere asymmetry?
Upon symmetric cell division, total CENP-A is distributed equally between both daughter cells such that each receives 50%. In order to ensure centromere function, newly synthesised CENP-A must be replenished in each cell cycle to 100%, classically measured by an increase/recovery in CENP-A level [54,55]. The majority of symmetrically dividing cells assemble centromeres after mitosis, at early G1-phase . Differing cell cycle timings for centromere assembly have been reported for some organisms and cell types [57–61]. Most interesting are those with an assembly timing before, as opposed to after, chromosome segregation. Of particular interest, gametes display a unique timing for the assembly of CENP-A. Such examples include that of Drosophila spermatocytes and starfish oocytes, which load CENP-A in meiotic prophase I [57,60,61].
In addition to unique centromere assembly timings, exceptions exist also in the amount of CENP-A loaded at centromeres. The centromeric locus comprises interspersed H3 and CENP-A containing chromatin , which allows flexibility in terms of being able to accommodate varying amounts of CENP-A nucleosomes. Strikingly, Drosophila spermatocytes assemble CENP-A to an unexpected level (more than two-fold increase)  – easily enough to compensate for CENP-A dilution by half at premeiotic S-phase. This allows for two consecutive meiotic divisions in the absence of new CENP-A loading. Thus, it is possible that CENP-A assembly is a fluid epigenetic process unbound to the status quo of equal and complete replenishment to 100% CENP-A capacity at each cell division. Rather, centromeres can adapt both CENP-A assembly timing and abundance depending on the requirement of the cell type. This appears again to be the case for asymmetrically dividing stem cells. Ranjan et al.  and Dattoli et al.  have both established the CENP-A assembly timing for male and female GSCs, as well as neural stem cells, with assembly of CENP-A occurring after DNA replication, during G2-phase up to prophase. Significantly, this assembly timing differs from canonical centromere assembly occurring after chromosome segregation in G1-phase.
Unique centromere assembly dynamics in Drosophila stem cells raise two important points. First, centromere assembly in Drosophila GSCs is gradual, occurring from G2-phase up to prophase, the longest cell cycle phase in GSCs . This gradual loading draws some similarity to the long duration of CENP-A assembly observed in spermatocytes at meiotic prophase I, in starfish oocytes and quiescent human cultured cells [57,60,61]. Active CENP-A assembly in quiescent cells epigenetically marks and maintains future proliferative potential. Second, CENP-A levels increase by approximately 30% on an average in female GSCs . Here, CENP-A assembly is potentially asymmetric relative to the complete and equal CENP-A assembly observed in a symmetrically dividing cell . Ultimately, asymmetry in the distribution of CENP-A between sister chromatids might occur in two ways: (i) through the selective retention of ‘old’ CENP-A nucleosomes on one daughter strand at DNA replication and (ii) through the asymmetric assembly of ‘new’ CENP-A nucleosomes after DNA replication up to prophase (Figure 2). However, whether this sister centromere asymmetry pre-exists centromere assembly, or is actively established through the centromere assembly machinery is yet to be clarified.
Current model and outstanding hypotheses regarding the establishment of asymmetric sister centromeres in
Shortened G1-phases have been observed in numerous stem cell populations [63,65–68], perhaps making G2/M a more favourable option for the assembly of the centromere. It has been proposed that a short G1 phase limits the sensitivity of stem cells to differentiation cues [65–67]. Another possibility is that centromere assembly after DNA replication allows the cell to first generate asymmetry before loading the mitotic components. This is also supported by recent work from the Chen lab. Using a sequential nucleoside analogue incorporation assay, Wooten et al.  have elucidated a biased unidirectional replication fork movement in testes-derived DNA and chromatin fibres. These results suggest that replication mechanisms might generate histone asymmetry in asymmetrically dividing cells [32,49]. To this point, the DNA replication marker 5-Ethynyl-2′-deoxyuridine (EdU) is first incorporated at GSC centromeres and pericentromeres . A favourable explanation for this would be the need to establish asymmetric CENP-A (and histone H3/H4) patterning as early as possible in the cell cycle. However, whether this centromere assembly timing is conserved in pluripotent lineages is yet to be elucidated. Indeed, how the epigenetic memory at centromeres reacts and adapts to a pluripotent state remains an important question.
Molecular control of CENP-A asymmetry
Previous studies showed that CENP-A assembly is strongly linked to cell cycle regulation [58,69–72]. In addition, recent findings by Dattoli et al.  proposed the involvement of the cell cycle machinery in establishing sister centromere asymmetry. Specifically, the authors demonstrated that CENP-A assembly is promoted by CYCLIN A during G2, while excessive loading is inhibited by CYCLIN B through the HASPIN kinase, between late prophase and metaphase . Significantly, HASPIN phosphorylates H3 on the threonine 3, a well-known pericentric mark that has been already implicated in regulating ACD in Drosophila male GSCs . In this study, Xie et al.  showed that H3T3P distinguishes pre-existing H3, which is enriched in the stem cell, from newly synthesised H3, which is enriched in the cell that differentiates. Importantly, it provided a key proof of principle that post-translational modifications to histones can epigenetically distinguish sister chromatids. It is possible that the timing of H3T3 phosphorylation, and it asymmetric distribution at pericentromeres, might limit (asymmetric) CENP-A assembly. It would be interesting to examine non-phosphorylatable (H3T3A) and phosphomimetic (H3T3D) mutant lines to determine any effect on CENP-A assembly or asymmetry in this background . However, the point at which sister centromere asymmetry is established in the cell cycle is currently unknown; this might occur in S-phase when the centromere is replicated and parental CENP-A is inherited (Figure 2, Hypothesis 1), or during the time of new CENP-A assembly between DNA replication and prophase (Figure 2, Hypothesis 2).
For a fuller picture of this centromere assembly mechanism and to deduce some key mechanistic targets, consideration from what we already know about canonical (symmetrical) centromere assembly is necessary. Furthermore, more in-depth knowledge about the inherent self-propagation of centromeres is required. Recently, the Heun lab has reconstituted the Drosophila CENP-A assembly cycle in human cells . Here, ectopic targeting of Drosophila proteins, CENP-A, the core centromere component CENP-C and the CENP-A assembly factor, chromosome alignment defect 1 (CAL1)  to human cell LacO arrays reveals an epigenetic loop of assembly and self-propagation. In this model, pre-existing CENP-C provides the recognition site for CAL1 to assemble new CENP-A-H4. Interestingly, CENP-C was identified as a positive hit in a number of stem cell maintenance/differentiation RNA interference (RNAi) screens carried out in Drosophila [75–77]. CENP-C is also asymmetrically distributed in GSCs and ISCs [36,37], and is known to play direct roles in CENP-A assembly and maintenance [69,78,79]. However, there are important differences between GSCs and ISCs in this context. In GSCs, there is a quantitative difference in the amount of CENP-C, with GSCs retaining approximately 1.2-fold more CENP-C compared with the cystoblast (CB) daughter cell (which retains mitotic capacity) . On the other hand, CENP-C can only be detected in ISCs and not in the enteroblast (EB) . The EB can further differentiate into an enterocyte (EC) or enteroendocrine (EE) cell without dividing . The absence of CENP-C here could be explained by the non-dividing and endoreplicating nature of the immature EB. Nonetheless, CENP-C’s asymmetry in favour of the stem cell in both systems suggest a potential role for CENP-C in parental CENP-A maintenance. Clearly, it stands as an interesting candidate to investigate in the context of marking an asymmetric assembly and maintenance of centromeres.
Centromere structure and the dynamics of associated proteins during S-phase may warrant additional investigation in ACD. In Drosophila GSCs, DNA replication begins around the centromere almost immediately after mitosis  – an unusual timing relative to many symmetrically dividing cells. Although the mechanisms of centromere replication remain unclear, the Cleveland lab has recently uncovered a role for an error-correction mechanism to maintain the centromere integrity throughout replication involving CENP-C and the moving replication fork . Chromatin Immunoprecipitation-Sequencing (ChIP-Seq) analysis in human cells revealed that CENP-A is loaded primarily at the centromeric site, but also at a lower level throughout the entire chromosome during G1-phase. Subsequently, the replication fork removes this ectopic CENP-A on chromosome arms, with CENP-C being key to maintaining the integrity of the centromere locus at the replication fork . This may help explain why CENP-C is largely stable at centromeres during mid-late S-phase in human cells  – to protect the integrity of the centromere. However, whether such a mechanism exists in asymmetrically dividing cells is unknown. Interactions between CENP-A and the DNA replication machinery might also generate asymmetry. Recent studies in human cells may indicate a starting point for investigating the handling of CENP-A during replication in ACD. Indeed, the human CENP-A assembly factor, Holliday Junction Recognition Protein (HJURP) (functional homologue of Drosophila CAL1) [83,84] binds pre-existing ‘old’ CENP-A in S-phase, and auxin-induced degradation of HJURP resulted in subsequent loss of CENP-A through S-phase . Furthermore, human CENP-A interacts with the minichromosome maintenance complex component (MCM) helicase component MCM2, and disruption of this interaction decreases CENP-A intensity through S-phase [85,86]. In Drosophila ACD, it is tempting to speculate that modifications to the handling of CENP-A at the replication fork may exist in order to bias the inheritance of ‘old’ CENP-A on one strand. Perhaps CENP-A, CAL1 or CENP-C contain post-translational modifications that allow the replication machinery to preferentially distinguish ‘old’ versus ‘new’ CENP-A in preparation for ACD.
Centromeres epigenetically regulate stem cell fate
ACD is a highly regulated process and errors can lead to cancer and infertility. Given the widespread implications in stem cell biology, understanding the molecular control of self-renewal versus differentiation is a primary motivation of all stem cell biologists. How the epigenetic state of stem cells changes depending on potency, environment and age are all key areas to be fully addressed.
CENP-A was first linked to stem cell self-renewal by Ambartsumyan et al. in 2010 . Surprisingly, CENP-A depletion in induced pluripotent stem cells (iPSCs) still allowed self-renewal. However, when induced to differentiate, CENP-A-depleted cells could then no longer support lineage commitment, undergoing significant p53-dependent apoptosis. Again, we see an example of the fluidity of centromere assembly in response to cellular requirement. In this case, a higher CENP-A threshold needs to be met for differentiation to initiate. However, whether centromere specification can truly be an epigenetic mechanism to direct cell fate, or whether it is simply a marker of proliferation capacity of the stem cell had remained unclear.
First indications of a role for the centromere assembly machinery in stem cell division capacity in a multicellular organism appeared in Drosophila ISCs. Depletion of CENP-A, CAL1 and CENP-C in the midgut epithelium resulted in the loss of ISCs proliferation capacity, as measured by the vast decrease in clonal size . Furthermore, long-term depletion of CAL1 resulted in ISCs loss . In GSCs, CAL1 depletion in males led to reduced numbers of stem cells in the niche, indicating a possible self-renewal failure , while CAL1 knockdown in females blocked GSCs proliferation presumably due to a failure in centromere specification . However, depleting CAL1 in each case does not allow full distinction between a stem-specific role in the germline, versus an inability to divide due to the lack of essential centromere proteins (CENP-A, CENP-C). Hence, overexpression of centromere proteins proved an effective approach to disrupt stem cell self-renewal in this system. Specifically, co-overexpression of CENP-A and CAL1 in female GSCs shifted the distribution of CENP-A between GSCs and CBs from asymmetric (1.2:1) to symmetric (1:1), shifting stem cells towards self-renewal . Similarly, knockdown of HASPIN also disrupted CENP-A asymmetry . In line with these findings, CENP-A asymmetry is lost upon CAL1 depletion in males GSCs . Taken together, accumulating evidence suggests that GSCs require asymmetric sister centromeres to direct non-random sister chromatid segregation and subsequent stem cell fate.
Moving forward, there are many avenues which warrant significant investigation. What relationship does centromere asymmetry have with stem cell polarity cues? Moreover, the extent to which sister centromere asymmetry is observed outside of metazoans should be determined in order to understand how well this phenomenon is conserved. The influence of the centromere as an epigenetic determinant of cell fate also warrants further investigation. How this epigenetically biased segregation ultimately affects cell fate remains elusive. Whether this creates a mechanical asymmetry in multicellular organisms, or indeed whether the gene expression status of stem cells is affected by disrupting centromere asymmetry, are all unanswered questions. Due to the potency of these adult Drosophila stem cell lineages, reservations must be made, as they largely display tissue-specific unipotency. Expansion of these studies into pluripotent lineages in other multicellular organisms is also required.
The silent sister chromatid hypothesis proposes that epigenetic mechanisms regulate stem cell fate.
In a Drosophila model stem cell system, centromeres of sister chromatid pairs show an asymmetric distribution of the centromeric histone CENP-A. Disruption of this sister centromere asymmetry in stem cells perturbs the balance of stem and daughter cells.
The unique timing of centromere assembly in stem cells (between DNA replication and prophase) might serve as a mechanism for asymmetric CENP-A distribution and the epigenetic regulation of stem cell identity.
The authors declare that there are no competing interests associated with the manuscript.
This work was supported by the Science Foundation Ireland-PIYRA [grant number 13/YI/2187 (to E.M.D.)]; the Government of Ireland Postgraduate Fellowship [grant number 2018/1208 (to B.L.C.)]; and the Thomas Crawford Hayes Travel Fund 2019 from the National University of Ireland Galway.
B.L.C. and E.M.D. wrote the manuscript.
We thank Prof Kevin Sullivan, Drs Anna Ada Dattoli and Caitríona Collins for their insights and critical reading of the manuscript.
Asymmetric Cell Division
Chromosome Alignment Defect 1
Drosophila Spindle assembly abnormal-4
Germline Stem Cell
phosphorylated histone H3 at threonine 3
Holliday Junction Recognition Protein
induced Pluripotent Stem Cell
Intestinal stem cell