α satellite DNA variation and function of the human centromere

ABSTRACT

Genomic variation is a source of functional diversity that is typically studied in genic and non-coding regulatory regions. However, the extent of variation within noncoding portions of the human genome, particularly highly repetitive regions, and the functional consequences are not well understood. Satellite DNA, including α satellite DNA found at human centromeres, comprises up to 10% of the genome, but is difficult to study because its repetitive nature hinders contiguous sequence assemblies. We recently described variation within α satellite DNA that affects centromere function. On human chromosome 17 (HSA17), we showed that size and sequence polymorphisms within primary array D17Z1 are associated with chromosome aneuploidy and defective centromere architecture. However, HSA17 can counteract this instability by assembling the centromere at a second, “backup” array lacking variation. Here, we discuss our findings in a broader context of human centromere assembly, and highlight areas of future study to uncover links between genomic and epigenetic features of human centromeres.

The centromere is the chromosomal locus that is important in chromosome pairing and essential for chromosome segregation during cell division. It is the site of kinetochore formation, the multi-protein structure that attaches chromosomes to spindle microtubules for segregation during cell division. Despite this essential role in chromosome inheritance, the features of centromeres vary among organisms. Centromeres range in size from small point centromeres (∼125bp) in budding yeasts to large regional centromeres (100kb–5Mb) in humans and plants.¹ Despite these genomic disparities, the proteins of eukaryotic centromeres are related, emphasizing the functional importance of the locus. Centromeres are defined by specialized nucleosomes containing the histone H3 variant CENP-A. CENP-A nucleosomes are interspersed with canonical H3 nucleosomes to create a unique type of chromatin that differentiates the centromere from the rest of the chromosome.^2,3 Centromeric (CEN) chromatin also serves as the foundation of the kinetochore, interacting with CENP-C and other members of the constitutive centromere-associated network (CCAN) to assemble the protein network between the DNA and the microtubules.⁴

Assembly of CEN chromatin occurs on DNA sequences that differ among organisms and even within the same individual, suggesting general sequence independence for recruitment of CENP-A and other centromere proteins. The lack of sequence similarities at eukaryotic centromeres has encouraged the current view of centromere identity as an epigenetically defined process, with little contribution from the underlying genomic sequences. However, de novo engineered centromeres (i.e. human artificial chromosomes, HACs) have only been generated from specific sequences, and centromeres in most organisms are consistently maintained at the same genomic location, raising the possibility of sequence-dependent aspects of centromere specification.^5-7 Few groups have studied large regional centromeres from a genomic perspective, primarily because centromeres are located within extensive and complex genomic regions enriched for repetitive DNA and retrotransposons.^8,9 The centromeres of plants (maize, rice, potato, Arabidopsis thaliana) are among the most well-defined regional centromeres, and in fact, several recent studies have produced nearly complete assemblies of a few maize centromeres.^10-13 These impressive advances notwithstanding, we have focused this commentary on human centromeres and our recent studies exploring links between α satellite genomic structure and centromere function.

Human centromeres are located at regions of A-T rich α satellite, a DNA repeat based on a 171 bp monomer. Monomers are 50–70% identical, and a defined number of monomers are arranged tandemly to create a higher order repeat (HOR) unit. The number of monomers and their order in the HOR confer chromosome specificity; that is, the centromeres of different human chromosomes are defined by HORs that are structurally distinct. For instance, the Homo sapiens chromosome X (HSAX) is defined by a HOR of 12 monomers (DXZ1, 12-mer), whereas the HSA8 is defined by a 7-mer HOR (D8Z2). The order and sequence of the monomers differ between chromosome-specific HORs, and thus, α satellite arrays on each chromosome can be distinguished. Despite these structural and organizational features that discriminate between α satellite arrays, the human genome assembly lacks contiguous α satellite sequences at centromeres. Monomers are readily identified by sequencing, but the HORs that define each chromosome-specific α satellite region are reiterated hundreds to thousands of times so that the highly homogeneous arrays (97–100% identical) extend over many megabases. This makes it difficult to accurately assemble long α satellite arrays, especially from short sequence reads.⁸ Without contiguous centromeric genome assemblies, it has been challenging to link specific features of α satellite organization to centromere function. Recent computational efforts have resulted in graphical models of human centromere sequences,^14-16 a first step toward linear centromere maps. This approach has allowed assessment of genetic content within α satellite DNA, revealing some of the diversity in satellites within and among centromeres and reinforcing that many satellites are distinct among different chromosomes. A limitation of these “maps” is that they do not delineate the order of sequences within any given centromere, so the long-range organization of α satellite arrays in a single individual, much less the population, remains largely undetermined.

From these studies, we know that most human chromosomes have multiple HORs, that is more than one multi-megabase higher order α satellite array within the centromere region. Chromosomes like HSA1, HSA5, HSA7, HSA15, and HSA18 have 2 (or more) independent chromosome-specific α satellite arrays. Moreover, HSA1, HSA5, and HSA19, share at least one array (D1Z5), further emphasizing the structural complexity of centromere regions.^17,18 The multi-array organization of human centromere regions presents a new view of normal chromosome biology. Structurally, endogenous human chromosomes closely resemble dicentric or tricentric chromosomes that have been thought to arise primarily through genome rearrangements that fuse 2 or more different chromosomes. Multi-array endogenous chromosomes are generally stable. Conversely, dicentric chromosomes were originally described by Barbara McClintock as unstable chromosomes that were poorly tolerated by the genome.^19-21 Dicentrics that arise by genome rearrangement occur frequently in humans (1 in 1000 individuals) and are stable through mitosis and meiosis.²² This has been thought to be due to the poorly understood phenomenon of centromere inactivation, a process by which one centromere loses its identity and function. Human chromosomes, even multi-array chromosomes, have only one site of centromere and kinetochore formation, so, in a sense, they are normal models for studying active and inactive α satellite arrays. The fact that most endogenous human chromosomes possess more than one α satellite array suggests that the humans may be inherently more tolerant of dicentrics caused by genome rearrangement. Interestingly, when the same acquired dicentric chromosome occurs in multiple individuals, a specific centromere is often inactivated.²³ Results like these have led to models describing differences in the functional potential (i.e., “centromere strength”) of distinct α satellite arrays. It is conceivable that the independent arrays on endogenous chromosomes may also exhibit variable strength or functional capabilities.

HSA17 has 3 α satellite arrays D17Z1, D17Z1-B and D17Z1-C, and is essentially structurally tricentric^24-26 (Fig. 1A). These α satellite arrays (and other DNA segments in the human genome) were named according to established gene nomenclature [D17Z1: DNA segment (D), chromosomal assignment (17), complexity of DNA (Z for repetitive), and sequential number (1, 2, 3…) to confer uniqueness of DNA segment].^27,28 We previously showed that within the population, 70% of individuals carry 2 HSA17s in which D17Z1 is the site of centromere and kinetochore assembly, based on the presence of inner and outer kinetochore proteins²⁹ (Fig. 1B). About 30% of the population has a heterozygous centromere configuration in which D17Z1 is the active centromere and the location of kinetochore assembly on one homolog, while D17Z1-B is the active centromere on the other (Fig. 1B). Such flexibility in centromere location, termed centromeric epialleles, is mitotically and meiotically stable.²⁹ The molecular basis of centromeric epialleles, that is, the ability to switch the position of centromere and kinetochore assembly between α satellite arrays, is not understood. Furthermore, they are not exclusive to HSA17. Our recent studies have identified centromeric epialleles on HSA1 and HSA7 (S McNulty, J Ross, and B Sullivan, unpublished observations), indicating that flexibility in centromere location is an intrinsic property of human chromosomes. In the case of HSA17, D17Z1 appears to be the major site of centromere and kinetochore assembly, suggesting that it is a dominant centromere. Collectively, these observations raise several interesting questions. What would make one centromere stronger than the other? Why might one α satellite array be a preferred centromere? How is the site of centromere assembly chosen when 2 or more competent arrays are available?

Figure 1.

α satellite variation and the molecular basis of centromeric epialleles in humans. (A) The centromere region of human chromosome 17 (HSA17) contains 3 α satellite arrays that are each defined by a different higher order repeat (HOR) unit. α satellite DNA is composed of 171bp monomers (white arrows) that are 50–70% identical. A defined number of monomers are tandemly arranged to create a HOR that is chromosome-specific. D17Z1 (blue), the predominant array is defined by a canonical 16-monomer HOR; EcoRI restriction sites demarcate the first monomer of each HOR. D17Z1-B (green) and D17Z1-C (shaded orange) are each defined by different 14-mer HORs. The monomers are numbered by their order in the HOR, and do not necessarily indicate sequence identity at the same monomer position between HORs of different arrays. (B) In the population, 70% of individuals carry 2 HSA17s that assemble the centromere and kinetochore (red dot) at D17Z1 (Z1/Z1). In 30% of individuals (Z1/Z1-B), D17Z1-B is the active centromere on one HSA17 homolog and D17Z1 is the centromere on the other homolog. No individuals have been identified yet that assemble both HSA17 centromeres at D17Z1-B (Z1-B/Z1-B). (C) D17Z1 is a polymorphic array. Single and multiple monomeric deletions produce HOR variants, including 15-mers, 14-mers, 13-mers, as well as 12-mers and 11-mers (not shown). (D) Some monomers also carry a common SNP in monomer 13 (black arrowhead) that creates an EcoRI site. This SNP is in linkage disequilibrium (LD) with the 13-mer HOR. Arrays containing specific HORs and the SNPs exist as distinct haplotypes in humans. Wild-type haplotype (I) occurs in most individuals and is defined by the canonical 16-mer HOR, as well as rarer 15- and 14-mers (C). Wild-type D17Z1 arrays are usually the site of centromere and kinetochore assembly (red circles) on mitotically stable HSA17. Haplotype II is defined by HOR variants that include a high proportion of 13-mers, many of which contain the SNP. D17Z1 arrays that have a high proportion of variant HORs are less likely to be the site of centromere assembly. Instead, the centromere is formed at “backup” array D17Z1-B and the HSA17 is extremely stable. (E) In a subset of Haplotype II individuals, the proportion of wild-type to variant HORs within the multi-megabase D17Z1 array is nearly equivalent. In these instances, if the centromere forms at D17Z1, the HSA17 is extremely unstable in mitosis due to a deficiency in centromere and kinetochore proteins (small red circles) and abnormal kinetochore architecture.

In answering these questions, one can consider models of gene expression and genetic variation. Gene function can be affected by genomic variation within the gene body or regulatory regions that alter promoter activity, splicing, enhancer activity, or transcription factor binding. Early studies of α satellite arrays uncovered substantial genomic variation at centromeres.^30-32 Within the same chromosome-specific array, this variation exists in multiples forms: as HOR size variation, single nucleotide polymorphisms (SNPs) within HORs, and differences in total array size between homologues and among individuals. HSA17 exhibits extensive α satellite variation.^31,33-35 The major array D17Z1 is classically defined by a HOR of 16 monomers (16-mer) (Fig. 1A, C), however, monomeric deletions have produced variant D17Z1 HORs that contain 15-, 14-, 13-, 12-, and 11-mers.^31,34 These deletions have not swept through entire arrays, but are typically present as a fraction of an array, so that 35% of the human population has hybrid D17Z1 arrays containing both wild-type (16-mer) HORs and variant HORs. D17Z1-B and D17Z1-C appear to be homogenous arrays; both are defined by different, but related, 14-mer HORs.^16,24

In our recent study, we explored the role of α satellite variation in controlling centromeric epialleles.³⁶ We specifically tested the role of total α satellite array size in determining the location of the centromere on HSA17. D17Z1 and D17Z1-B arrays were molecularly sized from multiple individuals whose HSA17 centromeres had been functionally characterized.²⁹ D17Z1 is overall a much larger array, ranging from 2–4 Mb between HSA17s homologues and among different individuals.^36-38 Total array size for D17Z1-B and D17Z1-C is smaller, ranging 0.3–1.5 Mb^24,29,36 (K Chew and B Sullivan, unpublished observations). We found that large D17Z1 arrays (≥ 3 Mb) tended to be the site of centromere and kinetochore assembly. However, when D17Z1 and D17Z1-B were closer in size, D17Z1-B tended to be active. Although array length cannot absolutely predict the site of centromere assembly, our findings imply that a large D17Z1 array may recruit or retain an increased critical mass of centromere proteins, giving it an advantage over a smaller D17Z1-B array.

About 70% of the population has 2 HSA17s in which D17Z1 is the centromere on both homologs, and ∼30% of individuals carry one homolog in which D17Z1-B is the active centromere.^29,36 We were particularly interested in the fraction of HSA17s in which D17Z1 is not the active centromere, and delved deeper into the genomic structure of D17Z1 arrays on these chromosomes. The D17Z1 canonical 16-mer HOR is operationally defined by EcoRI sites that designate monomer one of the HOR and define the boundary between individual HORs.²⁶ D17Z1 has 2 major polymorphisms. First, there is a size polymorphism caused by deletion of 3 monomers, yielding a 13-mer HOR^31,39 (Fig. 1C–E). As already mentioned, additional monomeric deletions have produced HOR variants ranging from 15-mer to 11-mer.^31,34 In addition, a subset of HORs contain a SNP that has introduced an EcoRI site within monomer 13 (Fig. 1D, E).³³ HOR size variants and the presence or absence of the common SNP define distinct haplotypes.³⁴ Haplotype I (wild-type) contains 16-, 15-, and 14-mer HORs and is present in 65% of HSA17s in the population. Haplotype II (variant) is present in 35% of HSA17s in the population. It is defined by the 13-mer HOR that is in linkage disequilibrium with the SNP. Interestingly, the frequencies of wild-type to variant D17Z1 haplotypes (65%:35%) resembles that of active to inactive D17Z1 arrays (70:30).

We also measured D17Z1 variation in the context of centromere location among various individuals and within a multigenerational family. Individuals carrying HSA17 centromeric epialleles exhibited substantial D17Z1 variation. Variant D17Z1 arrays defined by Haplotype II (i.e., containing the 13-mer HOR and the EcoRI SNP) were negatively associated with centromere function. That is, centromere and kinetochore assembly was more likely to be occur at D17Z1-B if D17Z1 was variant. Overall, our findings indicated that a large D17Z1 array composed of > 50% wild-type HORs will typically be the site of centromere assembly (Fig. 1C). However, smaller D17Z1 arrays (< 3Mb) that contain more (> 80%) variant HORs are more likely to be inactive and instead, the centromere will be assembled at D17Z1-B (Fig. 1D). Our studies suggest that D17Z1-B serves as a “backup” array, when the amount of variation within D17Z1 exceeds 80%.

Intriguingly, several D17Z1 arrays exhibiting 50–70% variation were chosen as the site of centromere assembly (Fig. 1E). We found that these HSA17s were highly unstable and showed increased aneuploidy over time. Notably, HSA17s with active, invariant D17Z1 or active D17Z1-B arrays did not exhibit appreciable instability. The HSA17 mutants allowed us to test the causal relationship between D17Z1 variation and HSA17 instability. We measured key centromere proteins on active variant D17Z1 arrays and compared them to active wild-type D17Z1 or D17Z1-B arrays. CENP-A is a variant of histone H3 that creates a unique type of chromatin exclusive to the centromere.^2,3,40 CENP-C is a member of the CCAN that links the inner and outer kinetochore and is important for CENP-A recruitment and kinetochore maturation.^41-43 We observed reduced amounts of CENP-A and CENP-C on variant D17Z1 arrays, but not wild-type D17Z1 or D17Z1-B arrays. These results suggested that the molecular basis for instability of HSA17s with active, variant arrays is an architectural kinetochore defect. Our studies also suggest that there is a critical molecular threshold for α satellite variation and centromere formation. In our data set, centromere formation occurred on D17Z1 arrays with moderate (50–70%) variation at the cost of decreased chromosome stability. It is not clear why a chromosome would continue to assemble the kinetochore at a mutated array. Determining if centromere function on these “threshold” arrays eventually switches to the backup D17Z1-B array to correct HSA17 instability and identifying molecular triggers that stimulate the shift are important next steps in the study of centromeric epialleles.

Overall, our findings indicate that different α satellite sequences do not have equal functional potential and that variation within α satellite organization negatively affects centromere assembly and function. The relationship between long-range organization of α satellite, i.e., where wild-type versus variant HORs are situated across a 4Mb array, is not known. A prior study of 3 HSA17s intimated that HOR variants within D17Z1 are clustered into domains.⁴⁴ We do not know how HOR size variants are organized in active vs. inactive D17Z1 arrays in our functionally characterized data set. A few models could explain how α satellite organization affects centromere assembly and function. Inactive D17Z1 arrays in our study exhibited > 80% variation and had 30–60 times more HOR size/SNP variation compared with HSA17s with active D17Z1 arrays. These inactive arrays are homogenously variant, so the large number of variant HORs dispersed across the entire array may prevent or disfavor CENP recruitment/maintenance, skewing centromere assembly toward D17Z1-B. However, several of the D17Z1 arrays in our data set exhibited an intermediate range of variation and centromere function. The arrays were 50–70% variant and were often chosen as the site of centromere assembly, but the HSA17s were unstable. On variant arrays containing an equal amount of wild-type and variant HORs, the long-range organization of the entire α satellite region and where centromeric chromatin is located may be crucial. α satellite organization in which variant HORs are clustered at one end of the array may be less detrimental because centromere assembly can occur on wild-type HORs concentrated at the opposite end of the array. However, if variant and wild-type HORs are interspersed across the entire array, the irregularity might disrupt structural requirements for kinetochore architecture, such as CENP-C-mediated bridging between nucleosomes.⁴

CENP-B is another constitutive centromere protein with α satellite DNA binding properties. It recognizes the CENP-B box, a 17-bp sequence motif found in a subset of α satellite monomers on all human chromosomes except the Y.^45-47 CENP-B is thought to position CENP-A nucleosomes and to stabilize CENP-A and CENP-C, based on the position of the CENP-B box within the DNA that is wrapped around the nucleosome.^48-53 The number and location of CENP-B boxes might correlate with the ability of an array to achieve the proper higher order structure required for centromere function. Monomeric deletions that gave rise to variant HORs may have altered the number of available binding sites for CENP-B which could destabilize the interactions between CENP-A nucleosomes, CENP-C, and other CCAN proteins. Our rough calculations of the number of CENP-B boxes in highly variant arrays indicate that they have 25–50% fewer CENP-B boxes than wild-type arrays (NG Peterson and BA Sullivan, unpublished observations). Nevertheless, variant D17Z1 arrays have 5–6 times more CENP-B boxes than D17Z1-B arrays, suggesting that a factor other than the overall number of CENP-B boxes affects functional potential of variant arrays. CENP-B boxes within variant arrays could be mutated so that they are not recognized by CENP-B, but without extensive sequence information we cannot test this hypothesis. Alternatively, HORs containing 13-mers are shorter than wild-type (16-mer) HORs and the decreased HOR length might alter positioning of nucleosomes and centromere protein complexes across a variant D17Z1 array.^49,53-55

Centromeric transcription is an integral part of kinetochore assembly and mitosis.^56-63 In humans, each α satellite array produces a unique set of array-specific, long non-coding transcripts (SM McNulty and BA Sullivan, unpublished observations). We speculate that transcription of wild-type vs. variant α satellite HORs are correlated with distinct differences in centromere assembly. Genomic variation within D17Z1 may alter the abundance, stability, and/or structure of long, non-coding α satellite RNAs so that variant transcripts are less stable or cannot interact properly with centromere proteins. Additional studies are needed to distinguish transcription at variant and wild-type D17Z1 arrays and to capture the interaction of these transcripts with centromere protein complexes.

The extent of genomic variation within D17Z1 beyond what we have studied is not known, and much less so are the types and frequency of variation in other α satellite arrays located on different human chromosome. Concerted efforts to expand genomic studies of highly repetitive sequences will allow us to fully uncover links between α satellite DNA organization and centromere function.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Funding

The research described here was supported by March of Dimes grant 1FY13–517 and National Institutes of Health grant 5R01-GM098500.