Genomic variation within alpha satellite DNA influences centromere location on human chromosomes with metastable epialleles

Abstract

Alpha satellite is a tandemly organized type of repetitive DNA that comprises 5% of the genome and is found at all human centromeres. A defined number of 171-bp monomers are organized into chromosome-specific higher-order repeats (HORs) that are reiterated thousands of times. At least half of all human chromosomes have two or more distinct HOR alpha satellite arrays within their centromere regions. We previously showed that the two alpha satellite arrays of Homo sapiens Chromosome 17 (HSA17), D17Z1 and D17Z1-B, behave as centromeric epialleles, that is, the centromere, defined by chromatin containing the centromeric histone variant CENPA and recruitment of other centromere proteins, can form at either D17Z1 or D17Z1-B. Some individuals in the human population are functional heterozygotes in that D17Z1 is the active centromere on one homolog and D17Z1-B is active on the other. In this study, we aimed to understand the molecular basis for how centromere location is determined on HSA17. Specifically, we focused on D17Z1 genomic variation as a driver of epiallele formation. We found that D17Z1 arrays that are predominantly composed of HOR size and sequence variants were functionally less competent. They either recruited decreased amounts of the centromere-specific histone variant CENPA and the HSA17 was mitotically unstable, or alternatively, the centromere was assembled at D17Z1-B and the HSA17 was stable. Our study demonstrates that genomic variation within highly repetitive, noncoding DNA of human centromere regions has a pronounced impact on genome stability and basic chromosomal function.

Figure 1.

Homo sapiens Chromosome 17 (HSA17) centromeric satellite organization. (A) HSA17, illustrated by the UCSC Genome Browser chromosome ideogram, has three distinct alpha satellite arrays with different monomer organizations of higher order repeats (HORs). D17Z1 (blue bar) is comprised of canonical/wild-type 16-monomer (16-mer) HORs (large blue arrowheads) that are operationally defined by EcoRI restriction sites. D17Z1-B (gray bar), located toward the short arm side of the centromere region, is based on a 14-mer HOR (large gray arrowheads) that is also defined by EcoRI sites. D17Z1-C, a third array oriented toward the long arm side of the centromere (light red bar) is also defined by a 14-mer HOR (light red arrowheads) but is less well characterized and not a focus of this manuscript. Individual monomers (green or yellow arrows) in the HORs are <70% identical to each other but are arranged nonrandomly in the same order (i.e., monomer 1 through monomer 16). The HORs are repeated hundreds to thousands of times to create highly homogenous arrays that span multiple megabases. This high degree of homogeneity has confounded standard genomic assemblies of centromere regions. (B) D17Z1 is an extremely polymorphic alpha satellite array; D17Z1-B exists exclusively as a 14-mer HOR array. In D17Z1, single and multiple monomer deletions caused by repeated rounds of unequal crossing over and/or gene conversion produce HOR variants that differ in length by an integral number of monomers. In the general population, HOR variants range from 15-mers to 12-mers, with rare 11-mers (not shown) occurring in isolated populations or individuals (Warburton and Willard 1995). Two major D17Z1 haplotypes (I, II) exist in the population and are primarily distinguished by the presence or absence of a 13-mer, created by a three-monomer deletion (3-mer indel). Additional restriction enzyme polymorphisms (DraI in monomer 5 and a second EcoRI SNP in monomer 13) are in linkage disequilibrium with the 13-mer HOR (Warburton et al. 1991). Individual monomers are denoted by numbered arrows. Shading/lightness indicates monomers that are deleted in specific HOR variants.

Figure 2.

Extensive D17Z1 variation is associated with centromeric epialleles. (A) D17Z1 variation was detected using PCR followed by restriction digestion to reveal HOR size variation as well as identify HORs containing the EcoRI SNP that segregates with the 13-mer/indel HOR. D17Z1 variation within two generations of the three generation CEPH/Utah 1345 family is shown (for data on the third generation, see Supplemental Fig. S2). This family has individuals that are centromeric functional heterozygotes (half-shaded circles or squares): One homolog assembles the centromere at D17Z1 (Z1) and the other homolog assembles the centromere at D17Z1-B (Z1-B). Squares represent males; circles represent females. Each family member is numbered according to the original classification of the pedigrees (Dausset et al. 1990). Agarose gels were imaged as white bands on black background; the images were inverted for presentation purpose only. Quantitation of the amount of wild-type HORs (16-, 15-, 14-mers, Haplotype I) versus variant HORs (13-, 12-, 11-mers, and 13-mer SNP represented by 9 + 4 bands, Haplotype II) was measured. Individuals with HSA17 centromere epialleles (half-shaded squares and circles) had D17Z1 arrays with >80% variation. (B) In CEPH family 1345, the correlation between all individuals with D17Z1 variation and centromeric epialleles (D17Z1-B CEN function on one homolog only) compared to individuals lacking epialleles (D17Z1/D17Z1 CEN function) was statistically significant (asterisks indicate P < 0.001).

Figure 3.

Larger D17Z1 arrays tend to be more homogeneous and are the site of centromere assembly. (A) Although the sizes of many D17Z1 arrays in our data set were already known (Maloney et al. 2012), D17Z1 arrays in new somatic cell hybrid lines were molecularly sized. D17Z1 array sizes were estimated using restriction digestion with enzymes that cut infrequently within the alpha satellite array, followed by resolution of large DNA fragments by pulsed field gel electrophoresis and Southern blotting. Representative Southern blot shows hybridization with a D17Z1-specific DNA probe p17H8. The parental diploid line shows many large DNA fragments from both HSA17 homologs. Individual HSA17 array sizes could only be resolved by moving each HSA17 homolog from the diploid line into the somatic cell hybrid background. Multiple bands were added to estimate the final array sizes. In this example, D17Z1 array size on Homolog 1 is 2.3 Mb and 4.0 Mb on H2. Each sample is shown in duplicate, along with a D17Z1 sizing control (0.7 Mb) for Southern blotting. (B) Because D17Z1-B is a relatively recently identified array, less is known about array size. We measured D17Z1-B array sizes on 12 different HSA17s using stretched DNA fibers and FISH with probes specific to D17Z1 and D17Z1-B (Supplemental Fig. S4). The size of D17Z1 was used as a normalizer to calculate D17Z1-B array size from fluorescent signals on DNA fibers; sizes of both arrays for individual HSA17s were plotted as shown. D17Z1 array sizes ranged from 2.3 to 4.3 Mb, while D17Z1-B sizes ranged from 0.7 to 1.6 Mb. The smaller D17Z1 arrays were associated with HSA17s in which D17Z1-B, not D17Z1, was the functional centromere. HSA17s are named and organized along the x-axis by D17Z1 array size (largest to smallest). Location of the centromere is denoted above the graph. (C) To investigate the correlation between array size and variation, D17Z1 array size and the proportion of wild-type and variant HORs (size + SNP) were plotted, revealing that inactive D17Z1 arrays have higher proportions (>80%) of variant HORs. Centromere location for each HSA17 is denoted above the plot. Z1_4.0 exhibited extensive D17Z1 variation but assembled the centromere at D17Z1. Z1_3.3, Z1_3.1, and Z1_2.6 exhibited moderate variation (∼60%) but still assembled the centromere at D17Z1.

Figure 4.

Centromeres assembled at variant D17Z1 arrays are less stable than homogeneous wild-type arrays. (A) The proportion of variation (wild-type versus variant HORs) in a subset of HSA17s is plotted with stability of the HSA17 (red line). For each line, chromosome stability was determined using FISH with D17Z1 probes and counting the number of HSA17s in 200 cells (stability was defined by maintenance of HSA17 ploidy in each line). When the centromere formed on a large, homogeneous array of D17Z1, such as in Z1_4.3, Z1_3.9, and Z1_3.5, the HSA17 was extremely stable in mitosis. Similarly, when the centromere assembled at D17Z1-B in lines Z1_2.3A and Z1_2.3B (highly variant D17Z1 arrays), HSA17 was very stable. However, when the centromere was assembled on D17Z1 arrays that had moderate or extreme variation, HSA17 was mitotically unstable. Centromere location (D17Z1—white, D17Z1-B—gray) on each HSA17 is denoted above the plot. (B) Line Z1_4.0 had the most variant (98%), yet active, D17Z1 array in our data set, and this HSA17 exhibited chromosome instability. The parental Z1_4.0 (Z1_4.0P) was subcloned to produce multiple, independent versions of the HSA17 (single-cell clones, SCC); subcloning could also account for aberrant behavior in a single cell line that did not reflect inherent behavior of the HSA17. The single-cell clones showed varying levels of chromosome instability, indicating that the unstable phenotype was inherent to this HSA17. The stability of two mouse chromosomes (MMU9, MMU16) was measured to account for genetic background effects that might alter the stability of all chromosomes. (C) CENPA and CENPC (Supplemental Fig. S4E) immunostaining (green) was combined with FISH using D17Z1 probe p17H8 (red) to quantitate the amount of centromere proteins on unstable Z1_4.0. Insets show the HSA17 alone and a single channel image of CENPA staining on the HSA17. Scale bar, 15 µm. (D) The amount of CENPA and CENPC on unstable Z1_4.0 was plotted compared to all other centromeres in the cell. Fluorescence from all centromeres (ALL CENS) was normalized to one, and the fluorescence at Z1_4.0 was calculated according to this normalized value. The amount of CENPA at the Z1_4.0 centromere was half of the amount at all other centromeres in the cell; CENPC was reduced by more than 50%. (E) The unstable Z1_4.0 single cell clone SSC9 also showed reduced amounts of CENPA. By comparison, the amount of CENPA on Z1_2.3B, a stable HSA17 that has a variant D17Z1 array but assembles the centromere at D17Z1-B, was comparable to all the other centromeres in the cell (E′).

Figure 5.

Models for epiallele choice on HSA17 based on CENPB box number or D17Z1 long-range organization. Centromere assembly on HSA17 appears to occur predominantly at large D17Z1 arrays that contain wild-type (invariant, blue blocks) HORs (scenario a). When D17Z1 contains variant HORs (green blocks), centromere assembly either occurs at D17Z1-B (gray blocks, scenario e) and the HSA17 is stable, or at variant D17Z1 and the HSA17 is unstable due to reduction in CENPs (red circles). It is not clear if variant arrays cannot recruit or cannot maintain the appropriate number of CENPs, and the molecular basis for the reduction in CENP molecules is unknown. Long-range organization of D17Z1 might affect CENP recruitment and binding. Large arrays with moderate variation may provide a sufficiently sized domain of homogenous wild-type HORs for centromere assembly (scenario b). However, in the cases of HSA17s that build their centromeres on variant HSA17 arrays and are unstable, CENPA may be distributed across wild-type and variant HOR domains, the latter of which may be less efficient at CENP recruitment/maintenance (scenario c, light red circles). Moreover, irregularity in wild-type and variant HOR organization (i.e., interspersed subarrays of variant and wild-type HORs) may negatively affect centromere function or CENP recruitment (scenario d). It will be important to experimentally discriminate between these organizational scenarios, particularly on HSA17, in order to better understand the spatial relationship between centromere proteins and long-range alpha satellite organization.