Chromosome segregation: Brushing up on centromeres

Abstract

Turning centromere DNA into a mechanical spring is central to the fidelity of chromosome segregation. A recent study shows how centromere DNA loops and partitioning cohesin and condensin convert centromeres and pericentromeres into bipartite bottlebrushes.

Centromeres are the region on every chromosome where kinetochores are assembled. Kinetochores, in turn, bind microtubules, providing a dynamic linkage to an ever growing and shortening microtubule plus-end. The mechanism to ensure accuracy in chromosome segregation relies on tension between the sister centromeres. Tension is achieved when each sister kinetochore is attached to a microtubule from opposed spindle pole bodies or centrosomes. Linkage between the sister kinetochores resists microtubule pulling forces, thereby generating the requisite tension. The discovery of cohesin, a member of the SMC (structural maintenance of chromosomes) family of proteins¹, provided a mechanistic solution for tension through linkage of the sister centromere DNA strands. The problem with this solution is one of scale. Cohesin is a ~50 nm ring complex, while sister centromeres are separated by ~1000 nm. The order of magnitude difference in scale needs to be addressed. A second problem in understanding how tension is generated over 1000 nm is the inherent floppiness of DNA. DNA is extremely flexible, highly extensible and not obviously the material of choice to build a spring stiff enough to resist fluctuating microtubule plus-ends while maintaining tension.

Centromeres are enigmatic from the sequence perspective as well. They can be as small as 117 bp (budding yeast), and up to several megabases in humans. They can be built on unique sequences or hierarchical arrays of small satellite repeats (171 bp alpha-satellite). This paradox can be resolved by distinguishing between the platform for kinetochore assembly and microtubule binding from the bolus of DNA used to form the spring. In organisms where the microtubule binding site can be genetically identified, the centromere is only 117 bp. In organisms with regional centromeres, we are unable to identify which repeat forms the microtubule binding site. The situation is clarified if we consider domains of chromatin containing the centromere-specific Histone H3 variant (CENP-A/Cse4). In yeast and humans, the CENP-A region of the chromatin lies on the outer edge of the chromosome, proximal to kinetochore microtubules. The pericentromere is not enriched in CENP-A, but is enriched in the SMC proteins cohesin and condensin. The physical attributes of the centromere and pericentromere are highly conserved. Whether the pericentromere is composed of repeat sequences or not is a secondary feature that differs throughout eukaryotic phylogeny.

Major advances in microscopy that break decades-long limitations in spatial resolution using wavelengths of light compatible with live cell imaging, and computational modeling of DNA as a bead-spring polymer in order to study micron-scale structures, have given us the tools to dissect centromere organization. The discovery of SMC proteins was critical in this endeavor, as they are the major drivers of chromosome and centromere organization.

SMCs form ring complexes^2,3. These rings have the remarkable ability to link sister chromatids (SMC1,3, cohesin), and extrude or stabilize chromatid loops (condensin SMC2,4 and cohesin). Condensin is the major driver of archetypical condensed chromosome structure. Condensin, but not histone proteins, is required to form the underlying X-shaped skeleton of a mitotic chromosome⁴. This function is conserved in condensin from prokaryotes to eukaryotes. These proteins are not sequence specific, relying on secondary features of the chromatid to provide spatial information relative to the DNA.

In a study published recently in Cell, Sacristan, Samejima et al.⁵ have put these tools together, and through single-cell analysis of condensin and cohesin have been able to deduce the physical basis of the centromere spring. The power of single-cell analyses is the ability to deduce structural information that is difficult to deconvolve from population studies. For instance, using chromatin immunoprecipitation to probe the position of cohesin and condensin reveals them to be enriched in the centromere and pericentromere in yeast^6,7. In contrast, single-cell analysis reveals them to be spatially segregated⁸. These data can be reconciled by considering the fluctuations of DNA. The only region of centromere/pericentromere fixed relative to protein binding is the 117 bp centromere. In the pericentromere, a given segment might be bound by cohesin in one cell and bound by condensin in another. Population averaging will reveal the total set of bound DNA, missing the spatial segregation apparent in single cell analysis.

The report from Sacristan, Samejima et al. reveals several novel features of the vertebrate centromere⁵. Firstly, the spatial segregation of condensin and cohesin is conserved from yeast to human. In yeast, condensin lies at the base of loops that are distributed on an axis normal to the spindle axis. In humans, condensin is proximal to the centromere and is split into two domains (bipartite). In contrast, cohesin is radially displaced from the spindle axis in yeast, occupying the tips of pericentromere DNA loops⁸. In humans, cohesin lies between the sister chromatids as well as subjacent to the bipartite centromeric condensin⁹.

The role of condensin as a driver of centromere elasticity has been known for some time. Condensin depletion results in the explosion of DNA loops released from the centromere¹⁰. The centromere (specifically CENP-A chromatin) is disrupted upon depletion of the condensin subunit SMC2. Depletion of cohesin was much less disruptive, leading the authors to conclude that the major driver of the bipartite organization is condensin, not cohesin. These conclusions corroborate work in other organisms showing that depletion of cohesin through removal of the kinetochore receptor for cohesin is much less severe than condensin depletion.

Implementation of bead-spring polymer models allow one to deduce the micron-scale organization of the centromere/pericentromere. A key parameter in the model is the ability to bridge (i.e., cross-link) non-adjacent beads. This implementation is justified based on the biochemical properties of condensin and has a profound impact on the predictive value of the model. Cross-linking across beads is sufficient to segregate the genome into sub-domains, most notably nucleolar and non-nucleolar^11,12. Cross-linking is also sufficient to build gene networks, whose half-life depends on the respective on and off rates of the cross-linker (condensin). In the centromere, cross-linking within two pericentromeres leads to a bipartite organization observed experimentally. The addition of cohesin further refines the model towards reflecting the in vivo situation.

Each of the centromere domains forms what is known as a bottlebrush. A bottlebrush (Figure 1) is the physicist’s solution to transform a floppy polymer into a mechanical spring¹³. Crowding of side chains (DNA loops) from a primary axis reduces fluctuations along the axis and generate tensile strength. As the primary axis bends, the distance between the side chains is reduced in the direction of the bend, increasing crowding between the side chains that in turn resist the bend. In vitro, synthetic bottlebrushes can be built to withstand nano-Newtons of force¹⁴. In vivo, the action of condensin-generated looping in the centromere can easily match the 10–100s of pico-Newton forces acting on the kinetochore^15,16. A second critical feature of bottlebrushes is that the configuration of side chains reduces the extent of entanglements while simultaneously increasing crowding. This functionality leads to strain-stiffening, a property exhibited by biological materials¹³. The bottlebrush configuration also provides an understanding of the conserved distance between separated kinetochores throughout phylogeny¹⁷. Independent of the details of the kinetochore assembly site, the pericentromere spring requires a crowd of DNA loops in a restricted volume to generate the chain crowding required to stiffen the centromere.

Figure 1. Centromere tension via the bottlebrush.

DNA is modeled as a bead-spring polymer (left). A single chain will fluctuate and adopt a random coil configuration over time. In the bottlebrush, the primary axial chain (in orange) is populated with DNA loops (shown as side chains in white, blue, purple and green). Side view of the brush (middle), end-on view (right). As the chains fluctuate, the primary axial chain in orange remains stiff. Bending in any direction leads to crowding of the side chains and forces the axial chain into the entropically favored position.

The domain organization has significant implications regarding error correction and pericentromere/centromere dynamics that ultimately interlace with microtubule dynamics. One of the major sources of chromosome instability is merotelic attachments, in which one centromere attaches to microtubules from oppositely oriented centrosomes¹⁸. The discovery of a means to physically link the centromere subdomains via cohesin highlights an active dynamic response within the centromere itself to rectify these errors.