Centromeres: genetic input to calibrate an epigenetic feedback loop

Centromeres are chromatin domains maintained by a self‐templating feedback loop based on nucleosomes bearing the histone H3 variant CENP‐A. The underlying centromeric DNA sequence is largely dispensable, yet paradoxically, it has highly conserved features. Hoffmann et al (2020) now uncover that when the epigenetic chromatin cycle falters, a genetically hardwired mechanism offers robustness to a dynamic epigenetic feedback loop ensuring long‐term centromere inheritance.

New work shows that upon failure of epigenetic centromere maintenance, a DNA sequence‐based mechanism ensures robustness of long‐term centromere inheritance.

Centromeres are specialized chromosomal loci that link sister chromatids and form the attachment site for spindle microtubules in mitosis. Centromeres serve as a paradigm for chromatin‐based epigenetic memory, as their mitotic and meiotic transmission is largely sequence‐independent, but strongly reliant on the presence of nucleosomes containing the H3 variant CENP‐A (Murillo‐Pineda & Jansen, 2020).

Classic modes of heritable gene regulation include Polycomb and Trithorax proteins that lead to gene silencing and activation (Steffen & Ringrose, 2014). Both these systems rely on specific DNA sequences that target these chromatin regulators. Such regulation is considered epigenetic, because the very same sequence elements can propagate different heritable chromatin states. Centromeres are assembled on repetitive DNA sequences, called alpha‐satellites in humans. The AT‐rich and repetitive nature of these sequences is a conserved feature of centromeres in many organisms, yet its role has been enigmatic and difficult to dissect because, unlike the case of the bistable ON/OFF gene expression switch, centromeres are constitutive and have proven largely independent of specific centromeric DNA sequences.

Key evidence for the dispensability of specific sequences comes from genetic trickery in cells. For example, artificial tethering of CENP‐A or its assembly chaperone HJURP to a naïve chromosomal location leads to the formation of a new active centromere. Once established, these nascent centromeres no longer depend on the initial seed and become self‐propagating (Barnhart et al, 2011; Mendiburo et al, 2011). The CENP‐A feedback loop appears to be relatively simple in its basic structure, as highlighted in recent heterologous reconstitution experiments. Recruitment of just three Drosophila melanogaster proteins CENP‐A (CenH3), the CENP‐A binding partner CENP‐C and Cal1 (the HJURP equivalent) to a human chromosomal locus is to some extent sufficient for entering a simplified closed epigenetic loop (Roure et al, 2019).

These studies appear to underscore that centromeric alpha‐satellite DNA is not strictly necessary for centromere function and identity. Then, why are centromeres typically associated with AT‐rich repetitive alpha‐satellite sequences? Particularly, enigmatic is the presence of numerous copies of a motif specifically bound by the centromere protein CENP‐B. Thus far, we knew that satellite DNA is special in that it can initiate a centromere de novo and that this property is dependent on CENP‐B binding, suggesting an ability to nucleate centromere components (Ohzeki et al, 2002). Yet, once centromeres are formed, alpha‐satellite DNA or CENP‐B becomes dispensable.

In this issue of The EMBO Journal, Fachinetti and coworkers (Hoffmann et al, 2020) help to elucidate the role of centromeric DNA at existing functional centromeres. Their approach is to abruptly deplete CENP‐A, thereby turning off the CENP‐A epigenetic feedback loop. In a previous study, they had established a strategy to do this by modifying endogenous copies of CENP‐A to allow their rapid degradation upon addition of the plant hormone auxin, referred to as the CENP‐A^off/on system (Hoffmann et al, 2016). In the current study, they ask, once CENP‐A is depleted and then re‐expressed, can it find its way back to the centromere? The answer is yes, and CENP‐A reloads at endogenous centromeres with normal cell cycle timing.

If not CENP‐A itself, then what is driving new CENP‐A recruitment to centromeres? CENP‐B's ability to directly recognize and bind centromeric sequences makes it an attractive candidate to link genetic specification with the epigenetic loop of CENP‐A. Indeed, previous work showed that CENP‐B, which is otherwise non‐essential, is important to maintain mitotic centromere function when CENP‐A is depleted, indicating that CENP‐B serves in redundant fashion to CENP‐A (Hoffmann et al, 2016). In the current work, when depleting CENP‐A and re‐expressing it in CENP‐B deficient cells, only a fraction of cells assembled new CENP‐A. This indicates that CENP‐B is indeed necessary for efficient reloading of CENP‐A when centromeres are devoid of pre‐existing CENP‐A.

To understand by what means CENP‐B contributes to CENP‐A assembly, the authors explored its relationship with another key centromere protein, CENP‐C. By artificially targeting CENP‐B to a non‐centromeric locus, they found this to be sufficient to recruit CENP‐C even in the absence of CENP‐A, indicating a direct interaction between CENP‐B and CENP‐C. Importantly, when CENP‐A was turned on again it was successfully recruited to this non‐centromeric DNA site through CENP‐C, as cells depleted of CENP‐C were not able to recruit CENP‐A. This indicates that CENP‐B recruits CENP‐C which, in turn, recruits CENP‐A.

To test whether CENP‐B alone can initiate CENP‐A chromatin assembly at native centromeres, the authors engineered off/on cells in which both CENP‐A and CENP‐C could be rapidly degraded and re‐expressed. As expected, depletion of both CENP‐A and CENP‐C impaired centromere formation. However, CENP‐A and CENP‐C re‐expression resulted in reassembly of both proteins but this assembly is fully dependent on CENP‐B. These experiments tell us that CENP‐B has the ability to promote centromere formation that is dependent on CENP‐A loading machinery but independent of pre‐existing CENP‐A and CENP‐C.

The key experiment performed next aimed to discover the mechanism of how CENP‐B promotes de novo reloading of CENP‐A and CENP‐C. In this case, again both CENP‐A and CENP‐C were depleted but rather than re‐expression of both, only one or the other was re‐expressed from a different locus. CENP‐A re‐expression in cells lacking CENP‐C did not result in CENP‐A loading, while CENP‐C re‐expression did result in partial CENP‐C detection at centromeres, even without CENP‐A. This reiterates a model where CENP‐B can maintain centromere position by recruiting CENP‐C, which in turn recruits the CENP‐A loading machinery all independent of any previous CENP‐A present at the centromere.

So far, we have thus learned that CENP‐B is non‐essential but has the ability to recruit CENP‐C and in turn CENP‐A when the latter two are lost. These experiments, although insightful, do, however, not reflect the physiology of a cycling cell that would not naturally face a sudden drop in CENP‐A or CENP–C levels. Does a scenario like this occur in nature? To explore this question, the authors analyzed quiescent circulating T‐lymphocytes and found a subset of these to lack CENP‐A, but to still maintain both CENP‐B and CENP‐C at their centromeres. Conveniently, these cells can re‐enter the cell cycle upon activation of their T‐cell receptor in vitro, which resulted in de novo assembly of CENP‐A. Thus, the authors uncover a natural setting in which the CENP‐A feedback requires a kick‐start, likely via genetically encoded CENP‐B binding.

By analogy, we can consider the International Space Station (ISS) (NASA, 2015) that is currently in orbit around Earth (Fig 1A). The ISS will stay in orbit for long periods without external correction, due to its transit in a near vacuum, analogous to the CENP‐A‐based epigenetic feedback mechanism that can stay “in orbit” on almost any chromosomal position. However, every so often, in response to drag and stochastic fluctuations such as space debris, a propulsion system fires up and pushes the space station back into orbit to keep it from falling back to Earth and to prevent inevitable destruction. The genetic memory at the centromere via CENP‐B binding works in a similar fashion (Fig 1B). When the centromere is reaching its lower “orbit” of CENP‐A occupancy, a genetically encoded CENP‐B‐based memory is able to rescue the centromere by boosting CENP‐A levels and save it from destruction. The self‐templating model for CENP‐A inheritance remains key to continuous centromere inheritance, but occasional genetic input allows such inheritance to be robust and resistant to fluctuations or error (Fig 1). Importantly, rather than viewing CENP‐B as a fail‐safe mechanism, it likely continuously contributes to maintain centromere position and calibrates centromeric chromatin especially at evolutionary timescales.

Figure 1. Keeping the CENP‐A feedback loop in orbit.

(A) The International Space Station stays in orbit indefinitely, if unperturbed. However, due to a less than perfect vacuum and space debris, an occasional boost is needed to stay in orbit. (B) Similarly, the centromere is maintained through cell generations by a largely autonomous histone CENP‐A‐dependent epigenetic feedback loop, but relies on a genetically encoded calibration mechanism, which engages when CENP‐A levels drop.

Finally, the results by Hoffmann et al (2020) indicate that if sufficient CENP‐A is resynthesized following depletion, then CENP‐A can be reloaded to pre‐depletion levels within a single cell cycle. This observation challenges the classic self‐templating model of centromere inheritance, in which “old” chromatin‐bound CENP‐A is critical to recruit new CENP‐A as recently formalized by Pan et al (2019), at least in the strict stoichiometric sense. We knew that CENP‐A levels vary between a certain range (Bodor et al, 2014), but what we did not know is that this range is encoded in the centromere even without CENP‐A being present. If not CENP‐A itself, then what controls CENP‐A levels and restricts it to the centromere domain? The answer may not be a singular one, and centromeric chromatin assembly is likely not a simple all‐or‐nothing process. Possibly, CENP‐B and CENP‐C may emerge as important components of the centromeric rheostat. Also, one cannot exclude that CENP‐A itself may ultimately control aspects of centromeric chromatin homeostasis in a more indirect manner. While uncovering such regulation awaits experimental testing, the work of Hoffmann and colleagues elegantly reveals how genetically encoded centromere specification can act to guide or correct the epigenetic feedback cycle ensuring long/term propagation of centromeres.

The EMBO Journal (2020) 39: e106638