A genomic and epigenomic view of human centromeres

Abstract

Human centromeres are large, complex chromosomal loci that serve as the foundation for kinetochore assembly, contribute to chromosome architecture and sister chromatid cohesion, and participate in chromosome separation during cell division. Encoded by thousands to millions of base pairs of repetitive DNA, these regions were previously represented as gaps in the human genome assembly due to limitations in sequencing technologies and computational tools that could accurately distinguish and anchor the highly similar repeats within a linear genome assembly. Substantial advances in long-read sequencing over the past 5 years have permitted these large human centromere regions to be spanned, revealing new genomic and epigenomic information and the structural organization of these essential regions. Here, we review these discoveries and discuss knowledge gaps that have been filled and emerging functional questions.

ToC blurb

The complex and repetitive nature of centromeres has historically posed challenges to their genomic assembly and functional understanding, which are now being overcome with long-read sequencing. This Review discusses recent genomic and epigenomic insights into human centromere biology that have positioned the field for future studies of centromere structure in chromosome biology and human disease.

Introduction

Centromeres are specialized chromosomal regions essential for accurate chromosome segregation during cell division. They serve as the attachment point for the mitotic spindle, ensuring that sister chromatids are properly distributed to daughter cells during mitosis and meiosis. Functionally, centromeres are defined by their role in the assembly of the kinetochore, a multi-protein complex involving over 100 proteins that mediates microtubule attachment (Fig. 1a). Centromeres are critical for maintaining genomic stability, as defects in centromere architecture or function can lead to aneuploidy, a hallmark of cancer and congenital disorders^{1–3, 4,5}.

Figure 1. Centromere organization in humans and other organisms.

a. The centromere morphologically appears as a primary constriction on chromosomes (chr). This site is where centromeric chromatin containing the histone H3 variant CENP-A is assembled, leading to the recruitment of other protein factors, including CCAN and KMN, that connect the inner centromere to the spindle microtubules. b. The underlying DNA at human centromeres is α-satellite DNA, composed of a 171-bp α-satellite monomer tandemly arranged into higher-order repeat (HOR) units that are extensively reiterated into large homogeneous arrays that span up to 5 Mb. α-Satellite arrays are flanked by unstructured or less structured monomers and divergent HORs (dHOR) and other repetitive DNAs such as HSat and long interspersed element 1 (LINE-1) elements. Centromeres in other organisms, except for the budding yeast Saccharomyces cerevisiae, vary in size and are defined by different types of repetitive elements, including satellites and transposable elements (TEs), as well as non-repetitive and unique sequences. Only the small 120-bp centromeres of S. cerevisiae are sequence-specific, composed of three DNA elements. CDE, centromere DNA element; Cen, centromeric repeats; cnt, central core region; imr, innermost repeats; dh/dg, outer repeat; rDNA, ribosomal DNA.

Centromeres vary across eukaryotes, reflecting a variety of organizational diversification strategies (Fig. 1b). The unicellular organism Saccharomyces cerevisiae has ‘point centromeres’, small (~125 bp) DNA sequences sufficient to confer centromere function⁶. By contrast, most multicellular eukaryotes, including plants and animals, have larger centromeres that are defined by different DNA sequences and organization. For example, Drosophila melanogaster centromeres are embedded in heterochromatin and are rich in satellite repeats and transposable elements⁷, whereas Caenorhabditis elegans has holocentric centromeres distributed along the entire chromosome length, enabling diffuse kinetochore attachment^8,9. Human centromeres are regional centromeres, with the core centromere assembled on a portion of highly repetitive, tandem α-satellite DNA arrays that span 0.3–5 Mb¹⁰. Overall, eukaryotic centromeres lack shared sequence identity among organisms and exhibit structural diversity.

Despite disparate DNA sequences at regional centromeres, a common feature in most eukaryotes is the presence of a unique histone variant, centromere protein A (also known as cenH3, but referred to as CENP-A hereafter), that replaces canonical histone H3 in a subset of nucleosomes at centromeric chromatin^11,12. Beyond CENP-A, a constellation of approximately 16 centromere proteins, collectively termed the constitutive centromere-associated network (CCAN), forms an essential structural and functional interface that links CENP-A-containing chromatin to the outer kinetochore (Fig. 1a)¹³. The functional core of centromeres, where CENP-A nucleosomes are enriched, is typically flanked by pericentromeric heterochromatin^2,14,15. This heterochromatin contributes to centromere cohesion and kinetochore stability.

The assembly of CENP-A into centromeric chromatin is tightly regulated and occurs primarily during early G1 phase of the cell cycle, following chromosome segregation¹⁶. This process is orchestrated by a dedicated chaperone, Holliday Junction Recognition Protein (HJURP), which escorts CENP-A to the centromere and facilitates its incorporation into nucleosomes^17,18. Unlike replication-coupled core histone deposition, CENP-A loading is replication-independent and relies on pre-existing CENP-A molecules to template new deposition, ensuring centromere propagation across cell divisions. This self-templating mechanism underscores the epigenetic maintenance of centromere identity^1,19,20.

Previous studies have provided an understanding of the biochemical and molecular aspects of centromeres²¹. However, links between α-satellite DNA and the functional aspects of the human centromere remained less clear, largely due to the inability to traverse the large gaps that centromeres have represented in the human genome assembly. The development of long-read sequencing (See Glossary, hereafter denoted [G]) by Pacific Biosciences (PacBio) and Oxford Nanopore Technologies (ONT) opened the possibility of spanning these large genomic gaps that had previously defied resolution by short-read sequencing technologies. By generating highly contiguous and accurate reads, long-read sequencing provided the first comprehensive view of centromere and pericentromere organization, sequence variation and its relationship to kinetochore function and heterochromatin formation.

In this Review, we discuss recent advances in human centromere biology, specifically new genomic and epigenomic information that has transformed our view of this important chromosomal locus and opened new avenues of discovery. We address how the current state of knowledge has positioned the field for future studies of centromere structure in chromosome biology and human disease. We also discuss genome engineering studies that have provided models to test functional elements and behaviour of de novo human centromeres. We refer throughout to relevant centromere studies in other organisms, although we point the reader to recent reviews that address advances in non-human centromere biology more broadly^21–24.

Human centromere genomics

Human centromeres have historically been among the most technically challenging regions of the genome to assemble, largely due to their highly repetitive composition and large size (several hundred kb to multiple Mb). Even after ‘completion’ of the human genome in 2003, centromere regions remained as gaps in the genome assembly. A major milestone was the completion of the first fully assembled human genome, CHM13 [G], by the Telomere-to-Telomere [G] (T2T) Consortium in 2022, which used PacBio high-fidelity (HiFi) and ONT ‘ultra-long’ sequencing to produce the first gapless assemblies of all repetitive sequences, including pericentromeric and centromeric regions. CHM13 provided an accurate genomic assembly for centromeric regions that had previously been absent or only computationally modelled in earlier genomic references. Approximately 76% of the novel assembled sequence in CHM13 included centromeric and pericentromeric satellite arrays. The T2T assembly produced the first comprehensive maps of α-satellite arrays, annotating their internal organization, sequence composition and structural variation of higher-order repeat [G] (HOR) units within arrays across all chromosomes in the human genome. It also clarified length differences and array boundaries that were previously speculative.

Core centromere region

Human centromeres are formed on α-satellite DNA, repetitive sequences defined by a 171-bp monomeric sequence unit (Fig. 1b). Individual monomers are 50–85% identical and arranged tandemly. A defined number of monomers create a HOR unit that is present in hundreds to thousands of copies to produce a large homogeneous array in which the HOR units are 97–100% identical. The number and order of monomers within a HOR unit can confer chromosome specificity and allows individual α-satellite arrays to be molecularly distinguished.

Following the release of CHM13, Altemose et al.²⁵ leveraged the complete centromeric assemblies to characterize the internal organization, patterns of change and adaptation, and potential functional roles of α-satellite arrays. With CHM13 as a reference, the authors developed computational approaches to resolve the structure of centromeric α-satellite arrays at both large and fine scales. This study provided evidence for a ‘layered expansion’ model of centromere changes, in which more homogenized HOR repeats expand outward from the core centromere, while older sequences diverge and erode at the periphery of centromeres (Fig. 2a)²⁵.

Figure 2. Centromere variation in humans.

a. The genetic architecture of a human centromere. From the bottom, recently expanded α-satellite sequences are found in the region of the centromere that is the site of CENP-A nucleosomes where kinetochore assembly occurs. The further from the site of layered expansion, the frequency of variation (shown in blue as increasing divergence from the youngest HORs) increases coincident with an increase in LINE-1 insertions (light green to dark green dots). With this increase in LINE-1 density is an increase in the age of the LINE, with the youngest elements found within the youngest HORs. b. X chromosome α-satellite monomers, classified into HOR variants by sequence and structure (HOR variants 1–7, key at bottom), were compared across the centromeres of seven individuals from different regions (map on right). An African-specific α-satellite variant (brick red) was found in three individuals with recent African admixture. The length of the centromere annotation is shown at the top. Adapted from Altemose et al Science 2022²⁵.

The complete T2T centromere assemblies of CHM13 and more recently CHM1 [G]²⁸ revealed large-scale organization of satellite sequences and layering of HOR α-satellite, divergent α-satellite (less-ordered HOR α-satellite) and other satellite sequences²⁵. In general, centromeric chromatin forming the inner kinetochore is assembled on a HOR α-satellite (the active HOR) that is largely homogeneous and uninterrupted. About 78% of chromosomes contain a second smaller HOR array, which is usually not the site of kinetochore assembly and thus is considered the inactive array. Most structural changes, such as inversions or deletions, are concentrated outside of homogeneous HOR arrays. However, on some chromosomes, such as chr1, chr3 and chr4, the active HOR arrays are split by inversions or insertions of other satellites or repetitive elements²⁵.

Overall, these findings suggests that human centromeres are dynamic and involved in ongoing molecular processes in which structural and functional elements shift and reorganize over time. Notably, the most recently expanded repeats within each α-satellite array were significantly more likely to associate with essential kinetochore protein CENP-A^43,25,29, suggesting that satellite sequence or regions of high homogenization or gene conversion play a role in α-satellite array size and in defining kinetochore positioning and are thus central to centromere specification and function.

Pericentromere regions

Long-read assemblies have also illuminated the contribution of transposable elements to centromere architecture. The regions surrounding the HOR α-satellite DNA arrays include divergent α-satellites, human satellites 1, 2 and 3 (HSat1, HSat2, HSat3), β-satellite DNA, γ-satellite DNA, transposable elements such as long and short interspersed elements (LINEs and SINEs, respectively), endogenous retroviruses and some protein-coding genes. Although some of these sequences are present within HOR α-satellite arrays, they are enriched adjacent to HOR arrays in pericentromeric regions (Fig. 1, Fig. 2a)^30–32. The distribution of these elements varies among chromosomes, although some elements are present on multiple chromosomes, such as β-satellite arrays on chr1 and the acrocentric chromosomes (chr13, chr14, chr15, chr21, chr22), which have the centromere located near the chromosome end^25,27,33. Transposable elements seem to be non-randomly distributed, and their localization patterns suggest that they have roles in shaping local chromatin structure or in defining centromere boundaries (Fig. 2)²⁷. Some transposable elements might influence nucleosome positioning, whereas others could serve as platforms for the recruitment of chromatin-modifying enzymes, thereby indirectly contributing to centromere identity, maintenance or the transition between centromeric and pericentromeric chromatin.

Resolving pericentromeric regions at base-pair resolution revealed their functional significance. They typically mark the transition from α-satellite-rich centromeric cores to flanking heterochromatin and are composed of a patchwork of satellite subfamilies, segmental duplications and embedded transposable elements (Fig. 1, Fig. 2a)²⁷. High-resolution profiling identified discrete pericentromeric boundaries in the human genome characterized by abrupt shifts in repeat composition (HSat2,3), inversions and chromatin state, which suggests that these sequences help organize centromeric architecture and act as insulators between functional centromeres and the pericentromere^27,28. However, pericentromeric regions are not just structural transitions but may actively contribute to centromere function by forming heterochromatin-like domains that provide mechanical rigidity to support centromere stability^34,35.

The ability to sequence pericentromeric and centromeric regions across chromosomes marks a key step forward in the field of centromere biology, enabling researchers to investigate how their organization contributes to centromere identity and how variability in these sequences might influence chromosomal behaviour across individuals and cell types.

α-satellite variation and centromeric epialleles

Human centromeres vary both structurally and functionally. α-satellite variation was originally described in the 1980s as monomeric deletions that produce HOR size variants and/or single nucleotide polymorphisms (SNPs) within specific monomers^36–39. It was also reported that total array size on a specific chromosome varied among individuals^40,41. Many α-satellite arrays exist as a combination of variant HORs and canonical or wild-type HORs that are present within the same array^43,48,85. Variation in HOR array sequence (and size) is generally thought to be driven by replication, unequal crossing over or recombination, and DNA damage and repair processes^42,43.

Recent work from the T2T Consortium has highlighted substantial inter-individual and ancestry-associated variation in the structure and sequence of centromeric α-satellite arrays^25,28. Comparison of centromeric regions across individuals has shown that both the sequence composition and structural organization of HORs vary among genomes^25,28,29. One α-satellite variant was significantly enriched in individuals with recent African ancestry, indicating that population structure can influence centromeric repeat content, length and sequence identity (Fig. 2b). Analysis of centromere assemblies across hundreds of genomes provided more detail on the extent of variation in HOR structures, repeat copy number and array length^28,44. α-satellite arrays show an average size of 2.3 Mb, but range in size from 300 kb to over 5 Mb. These studies also uncovered correlations between certain HOR unit patterns and specific ancestries, again underscoring how these genomic regions differ due to population structure. The Human Pangenome Reference Consortium (HPRC) and the Human Genome Structural Variation Consortium (HGSVC) are now focused on generating complete assemblies from globally diverse individuals, providing the first comprehensive view of pericentromeric and centromeric variation in human populations^29,45.

Centromere size and sequence variants raise questions regarding the impact of genomic variation within α-satellite DNA on centromere assembly and function. On human chr17, HOR size variants within the D17Z1 (cen17.2) array are associated with assembly of CENP-A chromatin on the neighbouring D17Z1-B (cen17.1) array⁴⁶. Centromere assembly on a different, nearby α-satellite array, also known as centromeric epialleles [G], is heritable within families and, on chr17 at least, is linked to α-satellite HOR variation^47,48 (Fig. 3a, 3b). Epialleles differ from the phenomenon of centromere shifting or drifting that represents flexible centromere movement (Fig. 3c, 3d; see section Centromere repositioning, shifting, or drifting). Other human chromosomes with multiple α-satellite arrays, such as chr7, also harbour centromeric epialleles⁴⁹, although it is not clear whether α-satellite variation or other genomic or epigenomic features of the chr7 HOR arrays influence the location of CENP-A assembly.

Figure 3. Centromere variation versus centromere repositioning or shifting/drifting.

a. Centromeric epialleles represent inherited variation in CENP-A location on distinct α-satellite arrays on the same chromosome. Epialleles exist on human chromosome 17 that contains three α-satellite arrays D17Z1-B (cen17.1), D17Z1 (cen17.2) and D17Z1-C (cen17.3) that are defined by distinct HOR units. D17Z1, the largest α-satellite array on Chr17 is polymorphic and exists as a wild-type array with canonical 16-mer HOR units or as a variant array that is a mixture of 13-mer HOR units (missing the 3 monomers shown as black dashed line) and wild-type 16-mer HORs. b. Epialleles are observed in ~30% of individuals in the European population, such that CENP-A is assembled on D17Z1 on one homologue and D17Z1-B on the other homologue. CENP-A assembly on D17Z1-B typically occurs when there is a high level of variation (13-mer HORs denoted by black line within red HOR unit) in D17Z1. c. Schematic of a neocentromere region (dark blue track showing mapped sequencing reads from CENP-A CUT&RUN on the multi-colour genome assembly) repositioning, drifting, or sliding, which is fluid between cells in an individual and over time when assessed at different cell passage numbers. d. The same native centromere from two individuals showing centromere repositioning or shifting such that there are different locations of CENP-A nucleosomes within the same α-satellite array (dark blue track showing mapped sequencing reads from CENP-A CUT&RUN), coinciding with a CDR (centromere methylation dip region).

Some individuals with measurable α-satellite variation on chr17 still assemble the centromere on a variant array rather than selecting the neighbouring array, even though the variant array shows defective kinetochore architecture and reduced ability to recruit or maintain centromere proteins⁴⁷. How genomic variation affects the ability of α-satellite DNA to form or sustain a kinetochore is not yet clear. Long-range organization, differential transcription of the HOR units (wild-type versus variant), sequence secondary structure and/or the ability to recruit and maintain epigenetic factors, such as DNA, RNA or histone methyltransferases, might influence the competence of an α-satellite array for centromere assembly (discussed below)^43,56,87. Studies that identify and characterize structural and sequence polymorphisms within α-satellite DNA and their fundamental effects on basic chromosome function will undoubtedly expand our understanding of genomic variation and the function of the non-coding regions of the human genome.

Sequence-independent features of human centromeres

Epigenetic models of centromere assembly and function were described over 30 years ago¹. These proposals were based on several key observations, including lack of sequence identity at centromeres among eukaryotes, differences in centromeric sequences within the same organism, the existence of chromosomes with two distinct regions of centromeric DNA but only one functional centromere (dicentrics), and the identification of neocentromeres [G]^1,51,52.

The complete genome assembly of human centromeres has transformed our ability to dissect the epigenetic regulatory mechanisms that define centromere identity. While α-satellite repeat organization provides the sequence scaffold^43,55, it is evident that stable propagation of functional centromeres depends on a specific chromatin environment that is governed by distinct epigenetic features, including the presence of the histone H3 variant CENP-A, which is loaded as a dimer by HJURP to replace H3 in a subset of centromeric nucleosomes (Fig. 4)^2,54. CENP-A is the cornerstone of centromere specification, marking the site epigenetically and providing a stable foundation for kinetochore assembly^20,55. The genomic or molecular basis for recruitment and maintenance of CENP-A at the same region of the chromosome remains an intense area of study. Studies using approaches such as high-resolution sequencing, epigenomic profiling and biochemical assays have implicated DNA methylation, transcriptional activity, histone and chromatin modifications, and RNA:DNA hybrid formation in the initiation and/or establishment of the unique centromeric chromatin at human centromeres.

Figure 4. Epigenetic features of human centromeres.

Schematic of the human centromere, including core centromeric chromatin containing CENP-A and euchromatic H3 nucleosomes (H3K4 and H3K36 di-methylation) and nearby heterochromatin enriched for H3K9 and K27 methylation. These domains assemble on α-satellite DNA, a 171bp monomer (small white arrowheads) tandemly arranged into higher order repeats (HORs, large grey arrows) that are extensively reiterated and highly similar in sequence. Centromere identity and CENP-A incorporation into nucleosomes via its chaperone HJURP, as well as the recruitment of CENP-C, is regulated by chromatin organization and α-satellite transcription and the formation of RNA–DNA hybrids or R-loops.

DNA methylation, CENP-A chromatin regions and chromatin boundaries

The development of long-read sequencing and direct methylation detection has enabled unprecedented mapping of centromeric CpG DNA methylation. Multiple T2T studies provided comprehensive genome-wide profiles of DNA methylation across complete centromeric arrays, demonstrating that methylation is enriched on satellite arrays and may play a stabilizing role in human centromeres^25,50,56,57. A distinct dip, or series of dips, in CpG methylation, the centromere dip region [G] (CDR), coinciding with CENP-A nucleosomes, links the two epigenetic marks to the site of kinetochore assembly (Fig. 4). Across diverse individuals, there is inter-individual and population-level variation in centromeric methylation and CENP-A density and location^25,28. Studies highlighting marked transitions in methylation levels between core centromeric chromatin on the active HOR and surrounding pericentromeric heterochromatin containing divergent HOR and monomeric α-satellite suggest that differential methylation acts as a boundary-defining feature^25,27,28. Recent mechanistic work also identified the transcription factor ZBTB24 and zinc finger proteins ZNF512/ZNF512B as important regulators of DNA methylation, histone H3K9 methylation and satellite repeat transcription at centromeres, offering direct links between sequence-specific binding and epigenetic maintenance in centromeres^58,59.

Histone modification and transcription at human centromeres

Although historically considered to be tightly compacted and inert heterochromatic regions, centromeres are no longer viewed as transcriptionally silent. Studies of human and Drosophila centromeres showed active histone modifications such as H3 dimethylated at lysine 4 (H3K4me2) and H3K36 methylation within centromeric chromatin, suggesting that centromeres are not exclusively euchromatic or heterochromatic (Fig. 4)^60,61. Transcripts at mammalian centromeres have been identified, and transcription specifically within pockets of hypomethylated DNA of a human neocentromere was described at least a decade ago^62–65. More recently, α-satellite array-specific long non-coding transcripts were observed at human centromeres and shown to associate with CENPs, chromatin-modifying proteins and nuclear organelles such as the nucleolus^35,49,66. Resolved centromeric assemblies both directly assigned centromere-derived transcripts to specific genomic locations and established that low-level transcription of α-satellite arrays is a conserved feature of centromeres and contributes to the dynamic epigenetic landscape of this region²⁷. Histone-modifying enzymes, specifically the MLL family of histone lysine methyltransferases, are required for di-methylation of H3K4, which promotes α-satellite transcription and CENP-A assembly^61,67. The centromeric transcripts themselves also play a role in centromere assembly. Centromeric RNAs are complexed with CENPs and contain the modification N6-methyladenosine (m⁶A) that stabilizes CENP-A nucleosomes, revealing a novel RNA-based mechanism to preserve centromere integrity^49,64,68. Centromere transcription is also required to maintain sister chromatid cohesion^69,70.

RNA:DNA hybrids, R-loops and centromere stability

An emerging area of research is the regulatory role of RNA:DNA hybrid structures, or R-loops, at centromeric and pericentromeric regions in maintaining genome integrity (Fig. 4)^71–74. Assays incorporating DNA–RNA immunoprecipitation and quantitative PCR (DRIP-qPCR) or sequencing (DRIP-seq) showed that BRCA1 suppresses R-loop formation at centromeres, and that BRCA1 loss leads to increased R-loop accumulation, DNA damage and chromosomal instability, highlighting a link between these structures and centromere maintenance in human cells⁷⁵. Ample evidence from other organisms has implicated R-loops and RNA–DNA hybrids in centromere function. In budding yeast, R-loop accumulation at centromeres has been shown to disrupt kinetochore integrity and cause chromosome instability⁷⁶. R-loop dynamics during oocyte meiosis metaphase I in mice contribute to spindle assembly and chromosome alignment, while also serving as a platform for recruiting replication machinery⁷⁷. Additionally, genome-wide mapping in maize found that R-loops derived from centromeric retrotransposons were enriched in pericentromeric heterochromatin regions, demonstrating the broad function of these structures and their potential relevance in centromere regulation⁷⁸. Given that excess R-loops elsewhere in the genome are tied to replication stress and genome instability⁷⁹, understanding their role at human centromeres remains an important area that warrants further exploration.

Centromere repositioning, shifting or drifting

Ongoing movement of centromere location has been described in several contexts, including centromere repositioning or centromere shifting/drifting, and involves CENP-A relocation on chromosomes (Fig. 3c, 3d). This phenomenon is distinct from centromeric epialleles that are heritable, stable and can be traced through multi-generational families^47,48. Centromere repositioning or shifting can occur within individuals during cell division, due to chromosome rearrangements and speciation. Such relocation of the functional centromere without sequence changes most likely reflects epigenetic plasticity.

Human neocentromeres are rare ectopic events that can arise de novo during development, typically in response to chromosomal rearrangements that disrupt the native centromere. Although neocentromeres occur at the same cytogenetic location, they are often associated with different chromosomal phenotypes and/or different genomic sequences. They represent centromere repositioning away from α-satellite arrays to chromosomal regions often distantly located from the native centromere. Nearly 150 cases have been reported as supernumerary marker or ring chromosomes, inverted duplication chromosomes, partially deleted chromosomes or neodicentric chromosomes (Table 1)⁸⁰. For example, a neocentromere on human chr13q32, identified in a patient with developmental delay, showed CENP-A binding over a ~130-kb region that lacked repetitive DNA but exhibited both H3K4me2-enriched chromatin and transcription⁸¹. Neocentromere formation on human chromosome 3q led not only to CENP-A repositioning to the new site but drifting within the 3q region across generations (Fig. 3c)⁸². Such repositioning and drifting of CENP-A position has also been documented for a neocentromere region on the long arm of human chr4⁸³. This NeoCEN4 has a 250-kb region of CENP-A enrichment^82–84. Epigenetic analysis of single-cell clonal lines containing NeoCEN4 revealed that the region of CENP-A enrichment is ~170 kb, coincident with a region of DNA hypomethylation (that is, CDR) and suggesting that in the parental population, the 250-kb CENP-A region represented differential CENP-A positioning between single cells⁸³. Careful long-term analysis of NeoCEN4 clones over a 12-month period showed drifting (expansion and compression) of CENP-A enrichment in each NeoCEN4 region that decoupled them from their originally associated CDR.

Table 1.

Locations and characteristics of known human neocentromere regions

Chromosome	Neocentromere site	Rearrangement	Ref.
1	1p32-p36.1	Interstitial deletion (paracentric)	107
	1q21-q22	Supernumerary ring	107
	1q21–q22	Giant rod-shaped marker chromosome	108
	1q23-q32	Interstitial deletion (paracentric)	107
	1q31	Supernumerary inv dup	109
	1q32-qter	Deletion + inv dup	107
	1q43-q44	Supernumerary ring	107
	1qter → q23~24::q23~24 → q43 → neo → q43 → qter (neocentric derivative)	Supernumerary inv dup	110
2	2p11-p21	Interstitial deletion (paracentric)	107
	2q33.3 to q37.3	Supernumerary inv dup	111
	2q35-q36	Supernumerary ring	107
3	3p23	Interstitial deletion (pericentric)	107
	3q21.2-qter	Supernumerary inv dup	107
	3q24	None	107
	3q26.1	Interstitial deletion (pericentric)	107
	3q26-qter	Supernumerary inv dup	107
	3q26.2-qter	Supernumerary inv dup	107
	3q26.2-qter	Supernumerary inv dup	107
	3q26.2-qter	Supernumerary inv dup	107
	3q26.2-qter	Supernumerary inv dup	107
	3q26.3–3qter	Supernumerary asymmetric inv dup	112
	3q27	Deletion + inv dup	107
	3q27.1-qter	Supernumerary inv dup	107
	3q27.1-qter	Supernumerary inv dup	107
	3q27.3	Supernumerary inv dup	107
	3q28-qter	Supernumerary inv dup	107
4	4p16.1	Supernumerary giant chromosome	113
	4p13	None	114
	4q21	Interstitial deletion (paracentric)	107
	4q21.2	Interstitial deletion (pericentric)	107
	4q21.3	None	107
5	5p14-p15.1	Supernumerary inv dup	107
6	6q16.2-q22.2	Interstitial deletion (paracentric)	107
	6q26	Supernumerary inv dup	107
7	7p22q21.2 (ring)	Interstitial deletion (pericentric)	115
	7q32.1	None	116
8	8p distal-pter	Supernumerary inv dup	107
	8p23.2	Supernumerary inv dup	107
	8p23.2	Supernumerary inv dup	107
	8p23.1-pter	Supernumerary inv dup	107
	8p23.1-pter	Supernumerary inv dup	107
	8p23-pter	Supernumerary inv dup	107
	8p22	Interstitial deletion (pericentric)	117
	8p22-pter	Supernumerary inv dup	107
	8q21–8q24	None	118
	8q21.2–8q21.3	Supernumerary ring	118
	8q23-qter	Supernumerary inv dup	107
	8q23.3-qter	Supernumerary inv dup	107
	8q24	Supernumerary rings	119
9	9p23	Supernumerary inv dup	107
	9p23	Deletion + inv dup	107
10	10p14-pter	Supernumerary inv dup	107
	10q11-q23	Interstitial deletion (paracentric)	107
	10q25.2	Interstitial deletion (pericentric)	107
11	11p11.12–11.2	Interstitial deletion (paracentric)	107
	11q22-qter	Deletion + inv dup	107
12	12p13.31-pter	Supernumerary inv dup	107
	12p12.3-pter	Supernumerary inv dup	107
	Acro pter-p12	Supernumerary inv dup	120
13	Acro pter-p13	Supernumerary inv dup	120
	13q21	Supernumerary inv dup	121
	13q21	Supernumerary inv dup	107
	13q21	Deletion + inv dup	107
	13q21	Supernumerary inv dup	107
	13q21.3	Deletion + inv dup	107
	13q21.3	Interstitial deletion (paracentric)	107
	13q31	Supernumerary inv dup	122
	13q31	Supernumerary inv dup	107
	13q31	Supernumerary inv dup	107
	13q31-q32	Interstitial deletion (paracentric)	107
	13q31.1q34 (or 13q31.1 → qter)	Supernumerary inv dup	123
	13q31.1 to 13qter region	Supernumerary inv dup	124
	13q31.3	Supernumerary inv dup	107
	13q32	Supernumerary inv dup	107
	13q32	Supernumerary inv dup	107
	13q32	Supernumerary inv dup	107
	13q32	Supernumerary inv dup	107
	13q32.1	Deletion + inv dup	107
	13q32qter	Supernumerary inv dup	125
	13q33.1	Supernumerary inv dup x 2	107
	13q33.1	Supernumerary inv dup	107
14	14q11.1	Supernumerary terminal deletion	126
	14q32.1-qter	Deletion + inv dup	107
15	15q11.1 to 15q21.1	Supernumerary ring	127
	15q22	—	107
	15q24	Supernumerary inv dup	107
	15q24-qter	Supernumerary inv dup	107
	15q24.1	Supernumerary inv dup	107
	15q24.1-qter	Supernumerary inv dup	107
	15q25	Supernumerary inv dup	107
	15q25-qter	Supernumerary inv dup	107
	15q25-qter	Supernumerary inv dup	107
	15q25.2	Supernumerary inv dup	107
	15q25.3-qter	Supernumerary inv dup	107
	15q25.3-qter	Supernumerary inv dup	107
	15q25.3-qter	Supernumerary inv dup	107
	15q25.3-qter	Supernumerary inv dup	107
	15q26.1	Supernumerary inv dup	107
	15q26.1-qter	Supernumerary inv dup	107
	15q26.1-qter	Supernumerary inv dup	107
16	16p	Deletion + iso(16q)	107
17	17q22-q23	Deletion + inv dup	107
18	18q22.1	None	120
20	20p12.2	Deletion + inv dup	107
21	21q21.1-qter	Deletion + inv dup	107
X	Xp22.31-pter	Supernumerary inv dup	107
	Xq12	Deletion + asymmetric inv dup	107
	Xq12	Deletion + dup	128
	Xq27.1~27.2	None	120
	Yp11.2	Supernumerary inv dup	129
	Distal Yp	Interstitial deletion?	107
	Yq heterochromatin	Supernumerary inv dup	130
	Yq11.2	None (or inversion?)	107
	Yq11.2	Deletion + inv dup	107
	Yq11.2	Deletion + inv dup	107
	Yq11.2	Deletion + inv dup	107
	Yq12	Supernumerary Y	107
	Yq12	None	107
Unknown	Heterochromatin, unknown origin	Supernumerary chromosome	120,131
	Ring neocentromere, unknown location	Supernumerary ring	132
	Ring neocentromere, unknown location	Supernumerary ring	132
	Ring neocentromere, unknown location	Supernumerary ring	132
	Marker neocentromere of unknown origin	Supernumerary chromosome	132
	Marker neocentromere of unknown origin	Supernumerary chromosome	132

del, deletion; inv dup, inverted duplication; iso, isochromosome; qter, q terminus; pter, p terminus. Supernumerary refers to an extra chromosome or chromosome fragment in addition to a standard chromosomal set. None indicates cases where neocentromere formation was identified without detectable structural rearrangement. Adapted with permission from Ref.¹⁰⁷, which originally collated ~100 neocentromeres (see detailed references therein).

These studies highlight several key similarities and differences between native centromeres and neocentromeres. First, both native centromeres and neocentromeres are defined by enrichment of CENP-A that overlaps with a zone of DNA hypomethylation. Moreover, the region(s) of CENP-A enrichment at both native centromeres and NeoCEN4 seem to be delimited by LINE-1 elements and H3K9me3^27,83. However, over time, CENP-A positioning on NeoCEN4 is altered due to increases in DNA methylation and H3K9me3 that disrupt the CDR-CENP-A link and make NeoCEN4 fragile and unstable⁸³.

A different induced neocentromere on the short arm of chr4 (Neo4p13) also exhibited centromere drifting along the DNA sequence, while maintaining overall centromere size (~100kb)⁸⁵. Like NeoCEN4, the CENP-A domain of Neo4p13 is constrained by H3K9me3 (Ref.⁸⁵). However, this study and another recent report further showed that the heterochromatin boundary around CENP-A chromatin is also enriched for H3K27me3 at both native and neocentromeres through the action of histone methyltransferases SUV39H1/2 and SETDB1 and Polycomb factor SUZ12 (Ref.^85,86). Overall, native human centromeres seem more epigenetically stable than neocentromeres. The positional plasticity of centromere formation and maintenance on neocentromeres, including CENP-A shifting or drift in response to changes in chromatin accessibility and epigenetic marks may potentially contribute to aneuploidy in malignancies⁸⁰.

Other models of similar centromere shifting or drift were reported in chicken DT40 cells and horse chromosomes^87,88. Centromere drift in DT40 cells was identified after inducing neocentromeres via CENP-A overexpression or centromere protein knockouts and mapping subsequent movement of CENP-A location and measuring CENP-A domain size over successive mitoses (Fig. 3c)⁸⁸. Chromatin immunoprecipitation followed by sequencing (ChIP–seq) analyses revealed that DT40 CENP-A domains drift along neochromosomes over cell divisions. Equine chromosomes also show centromere drift or sliding over generations^87,89, suggesting that flexibility in centromere domains is a common feature of both neocentromeres and some native centromeres in eukaryotes.

Although CENP-A shifting or drifting is established for human neocentromeres, whether it occurs at native centromeres was less clear. The human T2T assemblies addressed this uncertainty by revealing centromere shifting, or variable CENP-A position, within the α-satellite array of different human X chromosomes, which was perhaps driven by epigenetic variation and α-satellite repeat expansions (Fig. 3d)²⁵. Whether CENP-A drifting or shifting occurs over specific time frames (that is, with ageing) in humans or whether it provides a functional advantage to chromosome segregation and stability remains unknown. In maize, centromere drift drives segregation distortion (that is, the preferential segregation of chromosomes)⁹⁰, whereas in equines, centromere shifting maintains chromosome stability⁸⁹. Overall, CENP-A shifting/drifting highlights epigenetic regulation of centromeres that may balance adaptability and genomic stability. The molecular basis for this intriguing centromere phenomenon is an area that warrants deeper exploration.

Centromere engineering and de novo human centromeres

Important genomic and epigenomic information is gained from studying endogenous or native centromeres within populations. Nevertheless, synthetic or controllable cell-based engineering approaches are important for functionally testing discoveries from genomic and molecular studies. Human artificial chromosomes [G] (HACs) are a particularly useful system to dissect fundamental components of centromere formation and maintenance (Fig. 5). Unlike native centromeres, which are embedded in complex genomic and chromatin environments, HACs offer a tractable system to independently control and manipulate underlying DNA sequence and epigenetic contexts. Three main strategies have been harnessed to generate HACs in human or mammalian cells: introducing specific α-satellite DNA sequences; modifying epigenetic states; and bypassing both sequence and epigenetics by directly tethering centromere proteins to specific genomic locations.

Figure 5. Human artificial chromosome (HAC) assembly approaches.

De novo centromere formation on HACs has been achieved through two ‘bottom-up’ assembly systems. a. The traditional HAC assembly approach involved introducing α-satellite DNA within large BAC or PAC vectors into human cells, producing a stable HAC. Only α-satellite DNA containing CENP-B box motifs can form stable autonomous HACs with this approach. b. Newer HAC systems bypass sequence dependent centromere characteristics by physically tethering centromere assembly factors such as HJURP to α-satellite or non-repetitive input DNA.

Early HACs were constructed by transfecting bacterial artificial chromosome (BAC) vectors that contained α-satellite DNA sequences into human cells (Fig. 5a)^91,92. These experiments demonstrated that long tracts of α-satellite DNA support de novo centromere formation and ensure faithful chromosome segregation. Later work comparing sequences from different human chromosomes revealed differences in HAC-forming potential. α-satellite arrays from human chr17 and chr21 are especially competent at supporting de novo centromere formation, whereas sequences from the X and Y chromosomes have performed poorly at HAC formation, suggesting that some α-satellite arrays perform better as centromere platforms than others^48,92–94. In part, this variability can be attributed to the presence or distribution of functional CENP-B boxes — motifs recognized by the centromere-binding protein CENP-B. Notably, human chromosome Y α-satellite arrays do not have CENP-B boxes and, indeed, show little to no capacity to support centromere function in HAC constructs^95,96.

The ability of some but not all α-satellite sequences to support de novo centromere assembly highlights that DNA sequence alone is insufficient to explain HAC formation and has prompted investigation into the epigenetic context of HAC assembly. Functional HACs show a range of DNA methylation and histone modification patterns. Whereas CENP-A is crucial for successful centromere assembly, varying the amounts of H3K4me2 and H3K9me3 has been shown to impair centromere activity on HACs^61,97. Targeted depletion of H3K4me2 on HACs resulted in impaired α-satellite transcription and disrupted recruitment of the CENP-A chaperone protein HJURP, ultimately leading to centromere degradation as the HAC failed to maintain CENP-A recruitment⁶¹. Together, these results support a model in which human centromeres are epigenetically defined but can still accommodate different chromatin states within a flexible framework if core components such as CENP-A and its deposition machinery are maintained.

To circumvent the dependence on sequence and epigenetic states altogether, tethering systems have been engineered that artificially recruit necessary centromere proteins to target loci on HACs (Fig. 5b). Using LacI–LacO systems, researchers have shown that forced localization of CENP-A, CENP-T, HJURP or other centromere-associated proteins is sufficient to establish functional centromeres de novo, even in the absence of centromeric DNA or CENP-B boxes in HAC constructs^98–101. These ‘minimal centromeres’ highlight the ability of artificially targeting or seeding of CENP-A nucleosomes to define centromere identity and suggest that the protein components of centromere and kinetochore assembly may be more important than the underlying DNA sequence.

Conclusions

The past 5 years have seen unprecedented technological and conceptual advances in human centromere biology, especially in achieving complete genomic and epigenomic assemblies. The initial T2T complete genome assemblies focused on three cell lines (CHM13, CHM1, HG002), whereas more recent assemblies included RPE1 cells, as well as additional diverse genomes, including Han Chinese^{25,26,28,29,102,103}. All showed notable variation, raising questions as to the extent and location of variation between individuals across the globe, within populations, and the significance for health and disease. Efforts through the HPRC and HGSVC are aimed at achieving comprehensive maps of human genome variation from HiFi and ultra-long sequencing and hold promise for uncovering common and rare variants within centromeres and their functional outcomes.

Increasingly detailed genomic information will fuel key functional experiments. Studies that measure transcription, assess α-satellite DNA secondary structure, compare epigenetic modifications between wild-type and variant centromeres, and monitor centromere assembly and chromosome segregation dynamics will provide new mechanistic understanding of the impact of genomic variation within α-satellite regions and its link to disease. Indeed, some studies have already illustrated links between centromere sequence and structure and chromosomal aneuploidy^104–106. In addition, the use of synthetic chromosome assembly systems, such as HACs or artificial tethering systems, will enable the field to functionally test the role of specific repetitive DNA sequences, including α-satellite variants and transposable elements, or chromatin signatures (DNA methylation, specific histone modifications) in de novo centromere assembly and demonstrate the interplay between sequence elements and epigenetic factors. As high-resolution assemblies and maps of a larger number of native human centromeres across populations continue to emerge and improve, it will finally be possible to assign specific sequence variants, chromatin features and structural domains to centromere competency and chromosome stability.

Glossary

Centromere dip region

(CDR). the notable decrease in DNA methylation at genomic locations that coincides with enrichment of CENP-A nucleosomes

Centromeric epialleles

Heritable variation in centromeric CENP-A chromatin assembly at different α-satellite arrays on the same chromosome

CHM1

An additional haploid genome completely sequenced at high resolution from telomere-to-telomere after completion of the CHM13 assembly

CHM13

The first human haploid genome to be completely sequenced at high resolution from telomere-to-telomere

Higher-order repeat

(HOR). The tandem (head to tail) structuring of individual α-satellite DNA 171 bp monomers into a hierarchical repeat unit that is present in hundreds to thousands of copies at each centromere; a set number of monomers can define a chromosome-specific HOR

Human Artificial Chromosome (HAC)

A synthetic chromosome created by genome engineering that introduces genomic sequences and/or protein or chromatin components in human cells to create a new, fully functional, autonomous chromosome

Long-read sequencing

Current generation sequencing technologies such as PacBio HiFi and Oxford Nanopore Technologies ultra-long sequencing that produce read lengths of 18–500kb

Neocentromeres

Atypical centromeres that arise in non-repetitive/satellite regions of the chromosome within individuals

Telomere-to-Telomere (T2T)

Telomere-to-Telomere (T2T) refers to a truly complete (i.e. gapless) genome assembly; also alludes to the global consortium of investigators that collectively contributed experimental, genomic, and computational expertise to fully complete a chromosome in 2020 and the entire human genome in 2022