Molecular evidence for a relationship between LINE-1 elements and X chromosome inactivation: the Lyon repeat hypothesis

Abstract

X inactivation is a chromosome-specific form of genetic regulation in which thousands of genes on one homologue become silenced early in female embryogenesis. Although many aspects of X inactivation are now understood, the spread of the X inactivation signal along the entire length of the chromosome remains enigmatic. Extending the Gartler-Riggs model [Gartler, S. M. & Riggs, A. D. (1983) Annu. Rev. Genet. 17, 155-190], Lyon recently proposed [Lyon, M. F. (1998) Cytogenet. Cell Genet. 80, 133-137] that a nonrandom organization of long interspersed element (LINE) repetitive sequences on the X chromosome might be responsible for its facultative heterochromatization. In this paper, we present data indicating that the LINE-1 (L1) composition of the human X chromosome is fundamentally distinct from that of human autosomes. The X chromosome is enriched 2-fold for L1 repetitive elements, with the greatest enrichment observed for a restricted subset of LINE-1 elements that were active <100 million years ago. Regional analysis of the X chromosome reveals that the most significant clustering of these elements is in Xq13-Xq21 (the center of X inactivation). Genomic segments harboring genes that escape inactivation are significantly reduced in L1 content compared with X chromosome segments containing genes subject to X inactivation, providing further support for the association between X inactivation and L1 content. These nonrandom properties of L1 distribution on the X chromosome provide strong evidence that L1 elements may serve as DNA signals to propagate X inactivation along the chromosome.

Figure 1

Nonrandom distribution of L1 elements on the X chromosome. Assuming no chromosomal bias in L1 content, the figure shows a simulated distribution (10,000 replicates) of the difference in the mean L1 percentage between the X chromosome sequence (56.7 Mb randomly assigned) and non-X sequence (the remaining 348 Mb). Under this model variation mostly represents the random sampling of regional differences that occur between sequences. The observed difference (13.07%) between X and non-X sequence did not occur in 1,000,000 replicates (P < 10⁻⁶).

Figure 2

L1 subfamily enrichment for chromosomes X and 7. L1 elements were grouped into 75 different subclasses based on the original ORF2 and 3′ untranslated region classification (29). In a, an enrichment factor for the X chromosome was determined for each element (bp fraction of X chromosome sequence/bp fraction of sequence not from the X chromosome) and is depicted (y axis). The various L1 elements are grouped into 10 major subfamilies (x axis) and are arranged from the evolutionarily most ancient [L1M4 (mammalian group 4)] to the currently active element in the human lineage (L1Hs). Within each of the 10 major L1 classes, individual subclasses were arranged on the basis of increasing average similarity to the consensus. L1 subclasses that differed significantly from random (P < 0.006 after a Bonferroni correction for 75 subclasses) are colored red (n = 10,000 replicates; 56 Mb). The alignment number, total bp, and fraction of bp for each element were calculated for both the X and non-X groups (see Table 2 in supplemental material at www.pnas.org). In b, a similar analysis was performed for chromosome 7. Two subclasses (red) were significantly depleted.

Figure 3

Percent L1 vs. percent G + C composition. Nonoverlapping genomic sequences from the X chromosome (a) and chromosome 7 (b) were binned into cytogenetic band intervals based on cytogenetic and radiation hybrid mapping data (http://www.ncbi.nlm.nih.gov/genome/seq/). The percentage L1 content and G + C content for each bin of sequence is depicted graphically. Since the precise boundaries of cytogenetic bands are not known at the sequence level, many of the bins bracket two or more cytogenetic band intervals. L1s have been separated into old (L1M4 to L1M2) and young (L1M1, L1P5 to L1P1, L1Hs). The clustering of younger L1s at Xq13 and Xq21 suggests both a temporal and a positional bias in the accumulation of L1s on the X chromosome. The linear correlation (r²) between G + C and L1 content was 0.24 for the X chromosome and 0.50 for chromosome 7.