Genome size is a key diversity character with many uses and consequences. It varies greatly between organisms and over 1000-fold between flowering plant species. Reliable genome size values are needed for each important species. This paper gives the most accurate measurement currently available of the nuclear genome size in the important model flowering plant Arabidopsis thaliana ecotype Columbia. The value of 157 Mb was measured by best practice flow cytometry using the worm Caenorhabditis elegans (whose robust genome size of 100 Mb is unequivocally established by complete genome sequencing) as calibration standard. This paper also shows that the genome size value for Arabidopsis thaliana of 125 Mb given by the Arabidopsis Genome Initiative in 2000 was surprisingly underestimated by about 25%, and identifies reasons for this error, including misassumptions about both genome size and the amounts of DNA in unsequenced gaps, notably centromere regions.

Arabidopsis thaliana was selected as the first flowering plant to have its entire genome sequence based in part on its relatively small genome size. However, published estimates of its genome size varied over 3-fold, so an accurate measurement based on complete DNA sequencing was eagerly awaited to provide a precise value for the species, and an invaluable benchmark calibration standard for other plant species. December 2000 saw the landmark Arabidopsis Genome Initiative (AGI) paper in Nature giving the genome sequence for Arabidopsis thaliana ecotype Columbia and an estimate for its genome size of 125 Mb based on the total size of the sequenced regions (115.5 Mb) plus a rough estimate of 10 Mb for unsequenced centromere and ribosomal DNA regions. Sadly, it was not the long-awaited benchmark standard, because it was not the result of complete genome sequencing as these words would be understood by most people. Instead, the DNA amount in unsequenced gaps, especially in centromere regions, was seriously underestimated. The resulting AGI genome size estimate (125 Mb) was therefore about 25% too low, and less accurate than many other measurements obtained using non-molecular methods.

This paper demonstrates the crucial importance of using best-practice techniques in the analysis of genome size, and highlights the potential problems involved in estimating genome size by using only sequence data. It provides the most reliable measurement currently available of the total amount of DNA in the nuclear genome of the important model plant species Arabidopsis thaliana (approximately 157 Mb), and corrects the significant under-estimate (approximately 125 Mb) given by the Arabidopsis Genome Initiative in 2000.

I estimated plant nuclear DNA amounts in 1966 for my doctoral thesis, starting a long interest in comparing genome sizes and the methods used to measure them. Recognizing the need for a genome size database, I have published lists of DNA C-values as a service to the plant community since 1972. In 2005 "the Plant DNA C-values database" has genome sizes for 5,000 plant species. When I listed genome sizes for 271 flowering plant species in 1972, my measurement for an Arabidopsis thaliana weed growing in Cambridge was the smallest. This supported a link between short minimum generation time and tiny genome size, and later helped identify Arabidopsis as the prime plant candidate for complete genome sequencing. Studies by several groups published in the 1990s using non-molecular methods (microdensitometry or flow cytometry) gave genome sizes of 170 ± 20 Mb, including our measurement of 170 Mb for A. thaliana ecotype Columbia. These contrasted with estimates of 50–100 Mb made by groups using molecular techniques. When the Arabidopis Genome Initiative published its estimate of 125 Mb in 2000 this seemed insecure because it was also inconsistent with other published data (e.g. the %GC content of the Arabidopsis genome). The recent availability of a robust genome size measurement for Caenorhabditis elegans, based on complete genome sequencing, allowed a direct comparison of Arabidopsis with a reliable calibration standard which was undertaken with colleagues at Texas A & M University early in 2001 and completed by May 2002.