An isochore map of human chromosomes

Abstract

Isochores are large DNA segments (>>300 kb on average) that are characterized by an internal variation in GC well below the full variation seen in the mammalian genome. Precisely defining in terms of size and composition as well as mapping the isochores on human chromosomes have, however, remained largely unsolved problems. Here we used a very simple approach to segment the human chromosomes de novo, based on assessments of GC and its variation within and between adjacent regions. We obtain a complete coverage of the human genome (neglecting the remaining gaps) by approximately 3200 isochores, which may be visualized as the ultimate chromosomal bands. Isochores visibly belong to five families characterized by different GC levels, as expected from previous investigations. Since we previously showed that isochores are tightly linked to basic biological properties such as gene density, replication timing, and recombination, the new level of detail provided by the isochore map will help the understanding of genome structure, function, and evolution.

Figure 1a.

Compositional overview of human chromosomes and their GC levels. The color-coded map shows 100-kb moving window plots using the program draw_chromosome_gc.pl (Pačes et al. 2004) (http://genomat.img.cas.cz). The color code spans the spectrum of GC levels in five steps, indicated by broken horizontal lines, from ultramarine blue (GC-poorest L1 isochores) to scarlet red (GC-richest H3 isochores). Gray vertical lines correspond to the few gaps still present in the sequences, gray vertical regions to the centromeres. Figure 1 differs from a previous, similar figure (Pavliček et al. 2002) in that it is based on the May 2004 UCSC release (Kent et al. 2002) of the finished sequence (International Human Genome Sequencing Consortium 2004) (http://genome.ucsc.edu) and lacks the 5000 large gaps of the original draft sequence (Lander et al. 2001) that would not have allowed the present work. The isochore family borders as defined here (see color bar and main text) are slightly different from the previous ones in the case of the L2/H1 and the H2/H3 borders. Several characteristic features of isochores are already visible in this plot, such as the larger extent of the GC-poor compared to GC-rich regions, and the higher compositional heterogeneity of the GC-rich compared to the GC-poor regions.

Figure 1b.

Figure 2.

Plots of average standard deviations of GC within isochore families versus window size. Plots of standard deviation (SD) of GC versus fixed window sizes (w) ranging from 12.5 to 500 kb are shown for all isochores comprising at least four windows, after partitioning isochores into families according to GC level. The plots summarize the compositional variations within isochore families. At all scales, these are lower than that characterizing the whole genome, that is, all isochores (top curve), yet much higher than those of random sequences (bottom curve). The variations are consistently high for window sizes below 100 kb, and settle down to a plateau for larger window sizes. This fact justifies graining at 100 kb (see text). Marginally higher standard deviations around 300-kb windows may be due to a larger number of isochore borders seen through this window (Supplemental Fig. S2 shows a size peak of isochores around 500 kb).

Figure 3.

ΔGC distribution of adjacent isochores. (A) The frequency plot shows the jumps in GC between adjacent isochores identified, in intervals of 0.2% GC. The mean difference is 3.9% GC (dashed line), and 82% of the differences between neighboring isochores are above 2% GC. (B) The bar plot shows the average ΔGC concerning isochores from each of the five families.

Figure 4.

Overview of isochores on 100 Mb of chromosome 1. The isochores identified on the telomeric 100-Mb region of the short arm of chromosome 1 (as a representative region of human chromosomes) are shown. Broken horizontal guidelines in the top frames represent GC levels as in Figure 1. Horizontal red stretches in the bottom frames represent isochores. Our strategy was to identify boundaries on the basis of GC jumps between adjacent isochores (ΔGC) (see Fig. 3 and Supplemental Table S1). This strategy may occasionally lead to border misassignments, for instance, in the small number of cases (<3%) when ΔGC is lower than 1% (see Fig. 3). Several of the latter borders could be assigned, however, based on differences in the standard deviations of GC (see Table 1 for an example) or in the GC profiles.

Figure 5.

Distribution of isochores according to GC levels. The histogram shows the distribution (by weight) of isochores as pooled in bins of 1% GC. Colors represent isochore families as in Figure 1. Values at minima (histogram bars with mixed colors) were split between the two neighboring families. The Gaussian profile shows the distribution of isochores as estimated directly by the “density” function in R (bandwidth 0.7% GC) (Silverman 1986).