Advanced Nucleic Acids Research

Eukaryotic chromosomal DNA is packaged into nucleosomes, each consisting of 147 bp wrapped tightly around a core histone octamer . A recent genome-wide high-resolution microarray study states that ~80% of the yeast (Saccharomyces cerevisiae) genome consists of (translationally) positioned nucleosomes .This result is consistent with the conclusions of an earlier microarray study and a subsequent parallel sequencing study in which only portions of the yeast genome were examined. There are a variety of possible mechanisms that could lead to nucleosome positioning. For example, the base-pair-specific binding of a non-histone protein to DNA could exclude a nucleosome from that DNA region. Then, a regularly spaced array of adjacent nucleosomes would be positioned in the vicinity of the DNA-bound non-histone protein. Alternatively, an array of several regularly spaced nucleosomes could form adjacent to one sequence-positioned nucleosome. These mechanisms have been called statistical positioning . Statistical positioning would be most effective in yeast where the DNA linkers between adjacent nucleosomes are on average very short, only 18 bp for a 165-bp nucleosome repeat length (NRL) , and therefore not much statistical variation in nucleosome linker lengths can occur. Nucleosome positioning could also result from the boundary effect provided by the attachment of DNA regions to nuclear structures , chromatin remodeling or as a result of DNA replication from a bidirectional replication origin . In addition, nucleosome positioning could result from the DNA sequence preferences of the histones themselves. It has been known for some time that in vitro nucleosomes form with high preference on certain DNA sequences (10–13) and tend to avoid other sequences .

Evidence was provided that there is a genomic code for nucleosome positioning, and that ~50% of the nucleosome positions in yeast (±35 bp) result from histone preferences for certain DNA motifs (22). A different, complementary approach reported similar results for computationally predicting the positions of positioned yeast nucleosomes based on the genomic DNA sequence (±35 bp), but concluded that only ~25% of the positioned nucleosomes can be attributed to the preferences of certain DNA sequence motifs for histones , a value considered to be too low for the existence of a nucleosome positioning code. A critical evaluation of the statistics used by Segal et al. in 2006 also suggested that the performance of the proposed nucleosome positioning code is more modest than claimed . In 2007, Lee et al. reported that there was a poor correlation between their microarray-determined genome-wide nucleosome occupancy values and the predictions by Segal et al. in 2006. However, they found that there was a moderate correlation (R = 0.44) between their measured nucleosome occupancy values and a collection of DNA structural or sequence parameters. This degree of correlation suggests that only about 19% (R2 x 100%) of their nucleosome occupancy values are represented by the DNA structure/sequence parameters that they used in their model. Consistent with these findings, it was suggested that statistical positioning (discussed earlier), rather than intrinsic positioning, largely accounts for the nucleosome positioning observed in S. cerevisiae . In addition, it was reported in a genome-wide study, that Caenorhabditis elegans, which has (on average) longer nucleosome linkers than yeast, generally lacks sequence-dictated nucleosome positioning .

Recently, a direct genome-wide comparison of in vivo and in vitro nucleosome positioning in yeast was performed using the massively parallel Illumina sequencing system . The number of reads overlying each base pair for DNA sequences extracted from nucleosomes that were excised from native or reconstituted chromatin by micrococcal nuclease was used to assess the nucleosome occupancy at each base pair. This same sequencing approach had been used earlier to assess the nucleosome occupancy per base pair of the much smaller SV40 virus genome, where it was found that unique nucleosome positions did not occur . Control experiments were also performed by Kaplan et al. , using ~40 000 synthesized 150-bp DNA sequences to validate their yeast genomic DNA results. In these experiments, competitive reconstitution and microarray analysis were used to assess the affinities of each synthetic DNA sequence for histones to show that 5-mers contained in genomic sequences that had high (or low) affinities had corresponding affinities in the synthetic DNAs. Kaplan et al. concluded from their direct genome-scale experiment that intrinsic nucleosome sequence preferences do have a dominant role in determining the nucleosome organization in vivo.


It is clear that there are apparent conflicts in the current literature on the question: are nucleosome positions in vivo primarily determined by histone–DNA sequence preferences?

In an attempt to resolve these apparent conflicts, in this study we first examined the degree of correlation between nucleosome occupancies from the yeast in vitro parallel sequencing data and those from the in vivo microarray data of Lee et al. (3). We found that nucleosome occupancies in vitro and in vivo correlate less well when the data from the two different studies are compared than when the parallel sequencing data of Kaplan et al. in vitro and in vivo are compared. We discuss a potential problem with correlation analysis using scatter plots, when large numbers of superimposed points are present. We then analyzed the synthetic DNA nucleosome occupancy data provided by Kaplan et al. in a more direct way than the authors reported and found that there is not a very good correlation between their parallel sequencing data and their microarray data for these sequences. We suggest possible causes for the apparent discrepancies between the Illumina-Solexa parallel sequencing data and the microarray data, and between the two recent genome-wide parallel sequencing studies . We precisely calculate the effect of ‘statistical positioning’ in yeast. Furthermore, we examine what it really means to say that genomes encode an intrinsic nucleosome organization that can explain approximately half of the in vivo nucleosome positions.