Challenges and Approaches to Genotyping Repetitive DNA

Abstract

Individuals within a species can exhibit vast variation in copy number of repetitive DNA elements. This variation may contribute to complex traits such as lifespan and disease, yet it is only infrequently considered in genotype-phenotype associations. Although the possible importance of copy number variation is widely recognized, accurate copy number quantification remains challenging. Here, we assess the technical reproducibility of several major methods for copy number estimation as they apply to the large repetitive ribosomal DNA array (rDNA). rDNA encodes the ribosomal RNAs and exists as a tandem gene array in all eukaryotes. Repeat units of rDNA are kilobases in size, often with several hundred units comprising the array, making rDNA particularly intractable to common quantification techniques. We evaluate pulsed-field gel electrophoresis, droplet digital PCR, and Nextera-based whole genome sequencing as approaches to copy number estimation, comparing techniques across model organisms and spanning wide ranges of copy numbers. Nextera-based whole genome sequencing, though commonly used in recent literature, produced high error. We explore possible causes for this error and provide recommendations for best practices in rDNA copy number estimation. We present a resource of high-confidence rDNA copy number estimates for a set of S. cerevisiae and C. elegans strains for future use. We furthermore explore the possibility for FISH-based copy number estimation, an alternative that could potentially characterize copy number on a cellular level.

Copy numbers of 45S and 5S ribosomal DNA arrays lack meaningful correlation in humans

“…The scenario that rDNA copy numbers have changed between samplings for low-and high-coverage data generation is unlikely as there are several reports documenting that rDNA copy number is largely stable in cell lines (23,24) . Taken together, our results are consistent with previous reports that different library preparations of the same samples often yield different rDNA copy number estimates (10) .…”

“…However, eight additional rDNA copies seem unlikely to be biologically relevant, given an average copy number of ~250 . Moreover, we previously reported that rDNA copy number estimates based on short-read whole genome sequencing data can be error prone, with eight copies certainly being within the range of error (10) . Therefore, we sought to validate rDNA copy number in a subset of samples by alternate methods.…”

“…One difference between the datasets is that the low-coverage 1000 Genomes sequencing data were produced by any of seven sequencing centers, while the high-coverage data were produced by a single center. The seven sequencing centers used various library preparation methods (25) , and library preparation methods and batch effects can influence rDNA copy number estimates (10) . Moreover, some samples included reads from multiple library preparations and/or sequencing centers, turning the low-coverage copy number estimates into composite estimates from different libraries.…”

“…A key limit to incorporating rDNA copy number into genotype-phenotype associations is the ability to accurately quantify these genotypes. Because rDNA copy number estimates by short-read sequencing are error-prone due to batch effects (10) , sequencing-based estimates should be validated by methods such as contour-clamped homogenous electric field gel electrophoresis (CHEF gels), which can separate megabase-sized DNA fragments. The human 45S arrays are too large and too numerous to be accurately quantified by CHEF gels, however the 5S can be readily measured (11) .…”

“…In contrast, the original, low-coverage 1000 Genomes Project data were generated in multiple genome centers using different methods. We recently reported that rDNA copy number estimation from whole-genome short-read sequencing data is sensitive to even subtle variation in sample processing and coverage in technical replicates (10) . Our results on the lack of concerted copy number variation call into question several recently published associations of rDNA copy number with cancer and aging.…”