Eukaryotic genomes contain variable amounts of retroelements, which propagate within host genome using RNA intermediates (1). Approximately half of the human genome is composed of five major retroelement families that are classified into three groups, termed as the long interspersed nuclear element (LINE; L1 and L2), short interspersed nuclear element (SINE; Alu and MIR), and long terminal repeat (LTR) (2). The extreme abundance of retroelements in the modern human genome reflects the possibility of the explosive outbreaks that resulted in the massive propagation and fixation of the highly repetitive retroelement copies. When neutral or at least tolerable retroelements do not come up with certain threshold copy numbers, most of them are eliminated from a breeding population through random drift (3). A massive number of new retroelement copies is fixed within a population shortly after amplification (4), and the ancient retroelement's activity can be traced based on the wave of neutral mutations increased with time (1). The genome-wide analysis of Alu sequence divergence demonstrates a sharp amplification peak occurring 40 million years ago (MYRA), which corresponds to the expansion of the primate lineage (2, 5). This finding strongly supports the view that the explosive expansion of retroelements increased the genomic diversity responsible for the speciation events of host (6-8).

In spite of phylogenetic significances of individual retroelement families, the evidence of explosive expansions for retroelement other than Alus is considerably limited due to ambiguous peak activities. An approximate timescale for the entire retroelement formation roughly revealed the continued flourishing of autonomous L1s and LTRs for 150 million years (2). Since self-preserving retroelements cause harmful effects on the host genome by direct insertional mutagenesis or by altering genomic stability (9), the intragenomic propagation of retroelements has been selectively restrained depending on their effects on the host. Therefore, the pattern of retroelement amplification has been obscured in the modern genome, which makes it difficult to retrace when and how retroelement peak activities have appeared and influenced the evolution of the host genome.

Wide variations in the retroelement density across chromosomes have been observed previously (10, 11). Among all human chromosomes, chromosome 19, X, and Y contain Alu, L1, and LTR elements in the highest density, respectively. Because Alu elements have been supposed to promote protein expression and selectively accumulated nearby genes (12), it is reasonable to assume that the gene-rich chromosome 19 preferentially favored Alu insertion. The overwhelming frequency of LTRs in chromosome Y can be explained by the suppression of recombination between the heteromorphic sex chromosomes, which would otherwise remove potentially harmful genetic elements selectively (13, 14). Overrepresented L1s on chromosome X have been suggested to function as waystations for the spreading of heterochromatization; thus, playing a role in the inactivation of chromosome X (15). Basically, the variable composition of retroelements in human chromosomes may be a testament to the exapted role of self-preserving retroelements for the symbiotic advantage of both the host genome and the parasitic retroelement (16). Given that the initial integration of retroelements occur randomly (17, 18), each chromosome is believed to have formulated a unique composition of retroelements through the process of post-insertion selection. Therefore, propagation of retroelements is expected to leave expansion curves similar in the time of amplification and variable in the height of peaks across chromosomes.

We assumed that the comparative analysis of retroelement substitutions on individual chromosomes, including sex chromosomes, would unveil the most prominent expansion curves, which most likely represent the period of peak activity associated with explosive amplification. In this study, LINEs, SINEs, and LTRs on each chromosome were partitioned according to sequence divergence to trace time-scaled retroelement activities. Overall, the most prominent expansion curves revealed five reciprocal peaks of autonomous LINE and non-autonomous SINE amplification in a timescale of 150 million years. The time of an exceptional LTR burst was calibrated comparably with the Cretaceous-Tertiary (K-T) boundary between the Mesozoic and Cenozoic periods approximately 65 MYRA (19). In particular, the periodic peak activities of retroelements correspond temporally with speciation events in the hominoid primate lineage, suggesting the programming of genome dynamics by retroelements for the hominoid evolution.

The evolutionary history of human retroelements has been described as successive amplification of subfamilies during the steady rise and fall of retroelement families (29). Previous genome-wide analysis of LTR and L1 substitutions revealed slow expansion curves with no prominent peaks over the long evolutionary period (2). In this comparative analysis of retroelement divergence on individual chromosomes describes retroelement activity synchronized in an entire chromosomal set. Given that potentially harmful autonomous retroelements have been selectively deleted through recombination processes (14), the synchronous amplification peaks of retroelements in variable densities across chromosomes could be a consequence of post-insertion selection for evolutionary advantage. Recombination-suppressed chromosome Y appears to retain the expansion pattern of LTRs with the least selection. Chromosome X, which requires a high density of L1s to mediate heterochromatization for dosage compensation in response to the small-sized chromosome Y, is thought to have conserved the native expansion pattern of L1s with the least elimination. Therefore, retroelements in individual chromosomes might have amplified concurrently while being fixed differently, suggesting that chromosomal retroelement dynamics is similar in when they thrive and dissimilar in expansion magnitude. The retroelement peak acitivity supports the so-called massive bombardment of retroelements that facilitates the divergence of the host genome (6, 8).

In a breeding population, multicopy retroelements propagate depending on the amplification rates that increase the probability of fixation within a population. Only a significantly large number of newly amplified copies can counter random drift and breed throughout the population. Theoretically, 2N or more insertions are required to allow one insertion to get fixed in a breeding population of N individuals (30). Thus, extremely large numbers of Alus and L1s appeared to have amplified explosively at the time of peak activity that created an excessive number of copies over a population. The expansion curves of the old retroelements such as MIRs and L2s are less prominent as compared with their young counterparts, Alus and L1s and it must be taken into account that some old elements are too diverged to be recognized. However, considering that the normal distribution of neutral substitutions is considerably dispersed around a peak fraction with time, such small amplification peaks are likely to reflect the peak activity of retroelements for the corresponding genome.

Studies have suggested a competitive relationship between autonomous L1 and non-autonomous Alu elements based on the parasitic behavior of Alus that exploit the protein machinery of L1s for retrotransposition (31, 32). This study on the evolutionary wave of retroelements demonstrates that the different retroelement families have never explosively expanded at the same period; thereby coping with the competitive conflicts between selfish DNAs. In addition, it is likely that evolutionary events, such as speciation facilitated by retroelement peaks, provided new niches for the amplification of the other silenced retroelements. Consequently, LINEs and SINEs appeared to reciprocally promote the amplification of each other through repeated reshaping of the host genome.

The dormant periods of retroelements between amplification peaks is noted in terms of the accommodation through the presumptive functionality on host fitness (33) or the stability of retroelements themselves (17). It is believed that the initial integration of L1s and Alus are random while they become differentially redistributed with time (3). It is possible that selfish autonomous L1s are more enriched in the gene-poor regions for lesser detrimental influence, whereas Alus promoting gene expression are more concentrated in the gene-rich region through degeneration of useless contents in the gene-poor region (34). Alternately, the host genome positively selects Alu insertion in the gene-rich regions for adaptive advantage due to its role in promoting protein expression (12). This reciprocal redistribution of LINEs and SINEs in a dormant period would reorganize the explosive insertions for the beneficial and stable relationship between host and retroelement, even for symbiotic co-evolution. Therefore, LINEs and SINEs are likely to have expanded in the host genome as temporally and spatially reciprocal partners.

LTRs were found to have over-accumulated on chromosome Y throughout the period of active amplification. This unique overrepresentation of LTRs can be explained by the lack of recombination in chromosome Y, because other recombination-competent chromosomes are able to eliminate potentially harmful LTRs (14). In fact, we found no LTR peaks on chromosomes other than chromosome Y. The large fraction of amplified LTR copies seems to have been removed through selective recombination due to its adverse influence on the host genome. According to Muller's ratchet and hitchhiking theory for the degeneration of chromosome Y, the non-recombining region is unable to remove the stochastic insertion of retroelements, while tolerating mutational deletion (35, 36). In support of Muller's theory, the evolutionary waves of L1s and LTRs on chromosome Y reveal a skewed expansion curve with a high density of young copies and a low density of old copies. These findings suggest that the rapid gain and loss of genetic contents, rather than recombination, is a main dynamic underlying the unique retroelement composition of chromosome Y.

Of the two amplification peaks made by L1s, the older one was found to largely constitute multiple L1 hot spots in the long arm of chromosome X. A single hot spot of old L1 copies in the short arm of chromosome X coincides with the ancestral conserved region (XCR) on chromosome Xp described in a previous report (37). The most extensive hot spot is located in the proximal part of chromosome Xq that contains XIC. The high density of L1s in XIC was previously suggested to promote the spread of the X inactivation signal by XIST (X inactive-specific transcript) RNA in support of the Lyon repeat hypothesis (38). Alternately, because the two sex chromosomes have evolved differently in terms of predominant retroelement content, the decreased sequence homology in highly repetitive copies takes advantage for minimizing the undesirable recombination between sex chromosomes (39). The recombination suppression caused by the inversion of chromosome Y (40) is likely to have been further solidified due to different repeated sequences.

A single LTR burst is noted in association with the ancient outbreaks of retroviruses accompanied by the concurrent activation of endogenous retroelements. Exogenous retroviral infection is thought to have activated in trans endogenous retroviruses, which are defected in infectivity due to lack of appropriate viral components (41). The massive horizontal transmission of active exogenous retroviruses was most likely to be critically detrimental for host individuals and populations. The estimated age of an endogenous LTR peak at the K-T boundary implies a sudden change in the global environment, which facilitated evolution and conferred a niche for the LTR expansion in the primate ancestral genome. The Earth's ecosystem and climate are believed to have been severely disturbed at the K-T boundary at 65 MYRA (42, 43). These changes are likely to have modified habitats and genomic DNA through large quantities of dust and radioactive gas, consequently promoting the speciation of mammals. Therefore, a LTR amplification peak is possible in association with such a special global event, but not an accidental event.

The global changes at the K-T boundary of 65 MYRA are believed to have had important consequences for diversification of terrestrial organisms. Some investigators have used a placental mammalian evolution in a molecular clock to suggest the inter- and intra-ordinal divergence of placental mammals before and after the K-T boundary, respectively (27). In this respect, the human LTR explosion at 70 MYRA is timely compatible with the emergence of prosimians at the K-T boundary. The peak of old L1 copies appearing before the LTR peak is correlated with the inter-ordinal divergence of early placental mammals about 100 MYRA (44). The ancient L2s and MIRs appeared to expand explosively during the inter-ordinal divergence of marsupial and eutherian lineages about 120-150 MYRA. The peaks of Alus (40 MYRA) and young L1s (25 MYRA) following LTR peaking could be responsible for the active intra-ordinal diversification of the Old World monkey and New World money, respectively, in the hominoid lineage. The LTR explosion independent of LINE/SINE expansion is thought to have converted the evolutionary impact of reciprocal LINE/SINE expansion from inter- to intra-ordinal eutherian divergence.

In conclusion, this comparative analysis of retroelement divergence among individual chromosomes sheds light on the genome dynamics driven by the peak activity of two monophyletic selfish DNAs, SINEs and LINEs. The ebb-and-flow-like expansion curves of LINEs and SINEs suggest that the human genome has evolved through repeated SINE/LINE amplification and exaptation for 150 million years. An exceptional expansion peak of LTRs associated with the K-T boundary is considered a pivotal event in the inter- and intraordinal transition of eutherian diversification that reshuffled the programmed genome dynamics.