Copy Number Variants and Segmental Duplications Show Different Formation Signatures

Philip M. Kim, Jan O. Korbel, Xueying Chen, Mark GersteinarXiv 2007 · DOI · arXiv

Abstract

In addition to variation in terms of single nucleotide polymorphisms (SNPs), whole regions ranging from several kilobases up to a megabase in length differ in copy number among individuals. These differences are referred to as Copy Number Variants (CNVs) and extensive mapping of these is underway. Recent studies have highlighted their great prevalence in the human genome. Segmental Duplications (SDs) are long (>1kb) stretches of duplicated DNA with high sequence identity. First, we analyzed the co-localization of SDs and find that SDs are significantly co-localized with each other, resulting in a power-law distribution, which suggests a preferential attachment mechanism, i.e. existing SDs are likely to be involved in creating new ones nearby. Second, we look at the relationship of CNVs/SDs with various types of repeats. We find that the previously recognized association of SDs with Alu elements is significantly stronger for older SDs and is sharply decreasing for younger ones. While it might be expected that the patterns should be similar for SDs and CNVs, we find, surprisingly, no association of CNVs with Alu elements. This trend is consistent with the decreasing correlation between Alu elements and younger SDs; the activity of Alu elements has been decreasing and by now they seem no longer active. Furthermore, we find a striking association of SDs with processed pseudogenes suggesting that they may also have mediated SD formation. Moreover, we find strong association with microsatellites for both SDs and CNVs that suggests a role for satellites in the formation of both.

Introduction

Human genetic diversity is not limited to single-letter changes (SNPs); it also involves large-scale structural changes where regions of DNA (from 1kb to several megabases) are deleted or duplicated. These are known as Copy Number Variants (CNVs) when they vary between individuals, and Segmental Duplications (SDs) when they have become fixed in the population.

A prevailing theory in genomics has been that CNVs and SDs represent the same phenomenon at different evolutionary stages—essentially, CNVs are "SDs in the making." However, the exact molecular mechanisms driving these duplications have remained unclear. This study investigates whether the genomic "signatures" left behind by these processes have changed over the last 40 million years, revealing distinct differences in how ancient and modern structural variations are formed.

Methods

The researchers employed a co-localization approach, dividing the human genome into 100kb bins and calculating the Spearman rank correlation between the occurrence of SDs/CNVs and various genomic features.

Key methodological components included:

  • Age Proxying: Sequence identity was used as a proxy for the age of a duplication (higher identity indicating a younger event), allowing the team to track how formation signatures evolved over time.
  • Preferential Attachment Analysis: The team tested whether SDs follow a power-law distribution to determine if existing duplications facilitate the creation of new ones nearby.
  • Breakpoint Analysis: A detailed analysis of SD junctions was conducted to identify matching processed pseudogenes. Additionally, eight CNV breakpoints were manually curated to identify "microhomologies"—tiny stretches of identical DNA that hint at specific repair mechanisms like Non-Homologous End Joining (NHEJ).
  • Statistical Controls: When analyzing CNVs, the researchers corrected for SD abundance to ensure that observed associations were not merely a side effect of CNVs occurring within pre-existing SD regions.

Datasets

The study integrated several high-profile genomic datasets to ensure a comprehensive analysis:

  • SDs: The University of Washington Segmental Duplication database (build 35).
  • CNVs: Three distinct datasets were used to avoid methodology bias: the Human Copy Variation Consortium (microarray-based), Fosmid-paired end sequencing (Tuzun et al.), and genome assembly comparisons.
  • Repeats: Alu elements, LINE elements, and microsatellites identified via Repeatmasker on the UCSC Genome Browser.
  • Other Genomic Features: Processed pseudogene data from Pseudogene.org and recombination hotspot data from the HapMap project.

Results and Discussion

The "Rich Get Richer" Effect in SD Formation

The study found that SD occurrence follows a power-law distribution ($p(x) \sim x^{-0.31}$), confirming a "preferential attachment" mechanism. This suggests that the presence of existing duplications in a genomic region significantly increases the likelihood of that region acquiring further duplications, creating clusters of structural variation.

The Fading Legacy of Alu Elements

Ancient SDs, formed approximately 40 million years ago, show a strong association with Alu elements (correlation of 0.14). However, this association drops sharply for younger SDs (0.08) and is virtually non-existent for modern CNVs (0.0006). This trend aligns with the historical "Alu burst" in primate evolution; as the activity of Alu elements has decreased over time, they have ceased to be a primary driver of new structural variations.

The Rise of Microsatellites

While the role of Alus has diminished, microsatellites have emerged as a significant signature for modern variation. The study found a highly significant association between CNVs and microsatellite repeats (correlation of 0.0739, p-value $1.8 \times 10^{-38}$), even after correcting for existing SD content. This suggests that microsatellites may play a functional role in the formation of contemporary CNVs.

Shift in Repair Mechanisms

The data suggests a transition in the dominant molecular mechanisms of duplication. Ancient events were likely driven by Non-Allelic Homologous Recombination (NAHR), which requires long stretches of similar DNA like those provided by Alu elements. In contrast, modern CNVs appear more closely linked to Non-Homologous End Joining (NHEJ) or microhomology-mediated repair, particularly in subtelomeric regions.

Pseudogene-Mediated Duplication

The authors identified 144 cases where SDs were flanked by matching processed pseudogenes. These "genomic fossils" appear to have acted as templates that mediated the duplication of the surrounding DNA, providing another specific signature for SD formation.

Conclusion and Future Work

This research highlights that the mechanisms shaping the human genome are not static. The shift from Alu-mediated recombination to mechanisms involving microsatellites and microhomology represents a significant evolutionary transition in how structural variation is generated.

A current limitation is the resolution of CNV mapping (roughly 50kb–100kb), which prevents pinpointing exact breakpoint characteristics for most variants. Future work utilizing higher-resolution breakpoint data will be essential to confirm whether the observed associations with microsatellites are truly causal and to further explore how natural selection shapes these duplication patterns.

Summary updated

Abstract

In addition to variation in terms of single nucleotide polymorphisms (SNPs), whole regions ranging from several kilobases up to a megabase in length differ in copy number among individuals. These differences are referred to as Copy Number Variants (CNVs) and extensive mapping of these is underway. Recent studies have highlighted their great prevalence in the human genome. Segmental Duplications (SDs) are long (>1kb) stretches of duplicated DNA with high sequence identity. First, we analyzed the co-localization of SDs and find that SDs are significantly co-localized with each other, resulting in a power-law distribution, which suggests a preferential attachment mechanism, i.e. existing SDs are likely to be involved in creating new ones nearby. Second, we look at the relationship of CNVs/SDs with various types of repeats. We we find that the previously recognized association of SDs with Alu elements is significantly stronger for older SDs and is sharply decreasing for younger ones. While it might be expected that the patterns should be similar for SDs and CNVs, we find, surprisingly, no association of CNVs with Alu elements. This trend is consistent with the decreasing correlation between Alu elements and younger SDs, the activity of Alu elements has been decreasing and by now it they seem no longer active. Furthermore, we find a striking association of SDs with processed pseudogenes suggesting that they may also have mediated SD formation. Moreover, find strong association with microsatellites for both SDs and CNVs that suggests a role for satellites in the formation of both.
BibTeX
@misc{kim2007copy,
  author = {Philip M. Kim and Jan O. Korbel and Xueying Chen and Mark Gerstein},
  title = {{Copy Number Variants and Segmental Duplications Show Different Formation Signatures}},
  year = {2007},
  eprint = {0709.4200},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/0709.4200},
  howpublished = {arXiv},
}
Figure