Retrotransposon Evolution in Diverse Plant Genomes

Corresponding author: Glyn Jenkins, Institute of Biological Sciences, Cledwyn Building, University of Wales, Aberystwyth, Penglais, Aberystwyth, Ceredigion SY23 3DD, United Kingdom., gmj{at}aber.ac.uk (E-mail)

Communicating editor: C. S. GASSER

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Retrotransposon or retrotransposon-like sequences have been reported to be conserved components of cereal centromeres. Here we show that the published sequences are derived from a single conventional Ty3-gypsy family or a nonautonomous derivative. Both autonomous and nonautonomous elements are likely to have colonized Poaceae centromeres at the time of a common ancestor but have been maintained since by active retrotransposition. The retrotransposon family is also present at a lower copy number in the Arabidopsis genome, where it shows less pronounced localization. The history of the family in the two types of genome provides an interesting contrast between "boom and bust" and persistent evolutionary patterns.

RETROTRANSPOSONS represent a complex fraction of the repetitive DNA of most eukaryotes. The long terminal repeat (LTR) retrotransposons, in particular, have a high degree of autonomy and encode at least five distinct protein components required for their movement in the genome (GRANDBASTIEN 1992 ). For these elements, movement involves colonization of new genomic sites by intact copies of a parent that is not destroyed in the process, so that progressive rounds of retrotransposition have the potential to lead to massive amplification of elements. Some families may also encode an additional protein allowing movement between cells or even individuals, further blurring the distinction between repetitive DNAs and infective agents (KIM et al. 1994 ; SONG et al. 1994 ; WRIGHT and VOYTAS 1998 ). "Infection" appears to be a particularly apt description of the LTR retrotransposon families of higher plants, where >50% of the genome may be composed of family members (BENNETZEN et al. 1998 ). In cereals, where distribution patterns have been best characterized, most LTR retrotransposon copies are relatively recent insertions into intergenic islands (BENNETZEN et al. 1998 ). Waves of colonization may occur (SAN MIGUEL et al. 1998 ); in the Triticeae, BARE (barley retrotransposon) elements or their relatives currently appear to be highly active, composing up to 10% of the genome of barley, in which transcription and line-specific insertion site polymorphisms are readily detected (SUONIEMI et al. 1996A , SUONIEMI et al. 1996B ; WAUGH et al. 1997 ). There is currently some debate about the fate of genomes in which such aggressive colonization occurs. Phylogenetic analysis suggests that cereal and grass genomes have shown progressive inflation that may reflect retrotransposon amplification, at least in part (BENNETZEN and KELLOGG 1997 ); however, retrotransposon families represented only by ancient and decaying members have been reported for Drosophila and Arabidopsis, suggesting that new recruitment may eventually be blocked, followed by gradual erosion of existing sequences by deletion (PETROV 1997 ).

The fate of retrotransposons is inextricably linked to that of their hosts, and adaptations that minimize or even alleviate host genome disruption may be expected to evolve. The preferential distribution of yeast Ty elements to "silent" chromosomal regions is well known (BOEKE and DEVINE 1998 ) and may be a strategy adopted by a number of retrotransposon families. Subtelomeric or telomeric regions appear to be particularly favorable refuges; in some cases, retroelement adaptation to these niches may have been efficient enough to augment or even replace conventional telomerase activity (PARDUE et al. 1996 ). A LINE family has targeted telomeric repeats in Chlorella (HIGASHIYAMA et al. 1997 ); the concentration of Ty1-copia sequences in Allium subtelomeric domains may also reflect active targeting (PEARCE et al. 1996 ). More localized targeting has also been inferred for BARE elements, which show nonrandom distribution of insertion sites into other members of the same family (SUONIEMI et al. 1997 ). Recently, a number of groups have reported that cereal centromeres contain a high density of specific "retrotransposon-like" sequences (ANANIEV et al. 1998 ; DONG et al. 1998 ; MILLER et al. 1998A ; NOUTOSHI et al. 1998 ; PRESTING et al. 1998 ). Individual clones, isolated on the basis of similarities to two "universal" cereal centromeric sequences identified in random genomic screens of Brachypodium (CCS1; ARAGON ALCAIDE et al. 1996 ) and sorghum (pSau3A9; JIANG et al. 1996 ), contain degenerate retrotransposon reading frames and conserved LTR fragments. Both the centromeric location and conservation of these sequences indicate that the evolution of these sequences is of interest. It is possible that retrotransposon fragment(s) have been co-opted in centromere function in a similar way to telomeric elements, so that their presence is favored (MILLER et al. 1998A ; PRESTING et al. 1998 ). However, DNA at centromeric sites is typically organized in long arrays of tandem repeats that normally limit the insertion of mobile elements or other complex DNA (LEE et al. 1997 ; SUN et al. 1997 ). The question remains as to whether the retrotransposon-like sequences owe their concentration to active retrotransposition or whether they have been passively amplified.

We have reanalyzed published sequences and find that they fall into two classes. The first represents fragments of a highly conserved Ty3-gypsy family, which has a conventional organization and is closely related to a family in the Arabidopsis genome, while the second represents a nonautonomous family that encodes no enzymatic functions. Both classes of element share highly similar LTRs and show characteristics of recent retrotransposition, despite species-specific polymorphisms that indicate independent evolution since the divergence of species dating back to the origin of the Poaceae. The presence of retrotransposon-like sequences in Poaceae centromeres may be best understood as a transient exploitation of a novel host genomic niche. Following a massive initial amplification, the conventional retrotransposon family appears to be undergoing a slow extinction, presumably reflecting both the emergence of new centromeric organization and competition from nonautonomous elements.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Materials:
DNA was isolated from the following plant lines: inbred rye lines maintained by R.N.J. at UWA; Chinese Spring (CS; wheat); SunII (oat); maize, sorghum, grass, and Avena species from germ plasm maintained at IGER; Aegilops species kindly supplied by Dr. Steve Reader, John Innes Centre, Norwich, United Kingdom; wheat (CS)/rye (cv. Imperial) addition lines kindly supplied by Dr. Terry Miller, John Innes Centre, Norwich, United Kingdom.

PCR:
All clones were derived from blunt-ended PCR products ligated to pUC19. Amplification of CCS1 junctions (t26 clone) was carried out with rye genomic DNA and oligonucleotide Hi10R (CGRTYGCTAAGGCGCA); cycle conditions were 94° for 40 sec, 43° for 90 sec, 72° for 2 min, for 3 cycles, followed by 30 cycles of 94° for 30 sec, 50° for 30 sec, 72° for 2 min. Reverse transcriptase (RT) amplification was carried out with S14F (GAGATCMWGCGTCARATWCAAGAAATNCT) and BOTYR (GGCATGACAAGCCACTCATA); conditions were 94° for 30 sec, 55° for 40 sec, 72° for 90 sec, for 30 cycles. Lower stringency (48°) annealing was also used with some species in an attempt to obtain more divergent sequences. No such differences were seen, however. Integrase (IN) amplification was carried out with SORGF1 (TKYTGCAGGAAKCGCATGGAGG) and INTR2 (TTTGTCCATCAGTYTGNGGRTG), conditions as for RT. The insert in clone UC6.7 was derived from PCR of rye DNA amplified with AW37 (TATGTKCTKATHTG GTGGGAYCARAT) and BOTYR (96° for 35 sec, followed by seven cycles of 92° for 40 sec, 48° for 60 sec, 68° for 4 min, and 22 cycles of 92° for 40 sec, 48° for 10 sec, 68° for 3 min); that in UC7.12 was derived from maize amplified with REPQS (CCTCAGTCMGATGGMCARACNGA) and UNIHI (AGGKG CCCGATCTTTCGRYGAG) (94° for 1 min, 47° for 2 min, 72° for 5 min, for 1 cycle, followed by five cycles of 94° for 35 sec, 52° for 60 sec, 72° for 3 min, and 25 cycles of 94° for 35 sec, 57° for 40 sec, 72° for 3 min); that in UC8.5 was derived from Aegilops squarrosa amplified with REPQS and UNIHI (conditions as for UC7.12).

Genomic consensus sequences were derived from gel-purified templates.

Fluorescence in situ hybridization (FISH):
Preparation and pretreatment of the cytological preparations and FISH were performed according to published procedures (HESLOP-HARRISON et al. 1991 ; LEITCH and HESLOP-HARRISON 1992 ; PAN et al. 1992 ; ZHONG et al. 1996 ). Briefly, probe DNA was labeled either with digoxygenin-11-dUTP or biotin-11-dUTP and hybridized to pretreated chromosome preparations overnight at 37° in the presence inter alia of 50% deionized formamide in an Omnislide in situ hybridization system (Hybaid). Slides were washed stringently in 20% (v/v) formamide in 0.1x SSC at 42° before probe detection with FITC-conjugated antidigoxygenin antibodies or avidin-rhodamine as appropriate. Amplification of the signals was effected either by FITC-conjugated secondary antibodies or by anti-avidin-biotin followed by a second round of avidin-rhodamine binding. The chromosomes were counterstained in 4',6-diamidino-2-phenylindole (DAPI) and mounted in Vectashield. Fluorescent images were captured by cooled CCD camera, assigned false color, and manipulated uniformly in Adobe Photoshop.

Sequence analysis:
Database searches were performed with BLAST and further sequence analysis carried out with the Genetics Computer Group (Madison, WI) programs. Alignments of conserved regions were made using PILEUP and adjusted by eye. Sequence alignments were displayed using GeneDoc (K. B. Nicholas and H. B. Nicholas, distributed by the authors). K_s:K_a ratios were calculated using the GCG program Diverge. Phylogenetic analyses were carried out using the PHYLIP package, as implemented by the HGMP resource of the MRC. Phylogenetic distances were calculated with PROTDIST or DNADIST, and trees were constructed with NEIGHBOR (neighbor-joining method) and drawn with DRAWTREE.

Deletions indicated in Fig 4 are based on alignments of crwydryn elements in Arabidopsis bacterial artificial chromosomes (BACs) T9F8, F5K24, F9M13, T17A11, T1O16, and T27D20. Trees in Fig 5 are derived from peptide alignments analyzed with PROTDIST (Kimura method). Trees in Fig 6 are derived from DNA alignments analyzed with DNADIST (maximum-likelihood method, transition/transversion ratio of 1.4 estimated from relevant cereal data set or entire reading frame alignments of Arabidopsis BACs F5K24, F9M13, and T9F8, three category divisions based on codon position variation rates of 1.5:1:3). Full alignments and details are available from the authors.

View larger version (25K): In this window In a new window Download PPT slide   Figure 1. Relative organization of crwydryn elements and fragments. (A) Cereal genomic clones. The CentA repeat of maize is composed of two LTRs flanking a 5' UTR and partial open reading frame inserted into a second crwydryn reading frame. The rice RCB11 clone contains a single CentA homologue (RCB11-1) and a fragmentary CentA homologue (RCB11-2) inserted into a second crwydryn reading frame. The cereba "element" of barley contains two inverted repeats. The left-hand repeat contains the initial region of a long conserved polyprotein reading frame; the right appears to be truncated within the gag region (indicated by broken vertical line). (B) Composite cereal crwydryn element showing the relative positions of diagnostic PCR fragments and RT and IN regions used in the phylogenetic analysis and other crwydryn database entries. Details of these sequences are given in Table 2.

View larger version (103K): In this window In a new window Download PPT slide   Figure 2. Reading frames for Ty3-gypsy retrotransposon gag peptides are present in cereal centromeric clones. The clones from which the gag peptide sequences are derived are listed in Table 2. Cereal centromere-specific sequences are represented by clones UC6.7 to RCB11-2, and closely related sequences in Arabidopsis by T9F8 to F9M13. Similar but unrelated plant sequences are represented by clones F9D12 to tomato. T9F8 to cereba-1 are fragments of apparently autonomous elements, and cereba-2 to crw-r39 are fragments of nonautonomous elements. crw-r39 is part of a B-chromosome heterochromatic repeat of rye. Dark shading represents >60% conservation, and light shading >30%.

View larger version (19K): In this window In a new window Download PPT slide   Figure 3. A rye clone (t26) contains identity with wheat tRNAmet at the junction between CCS1 (LTR) and cereba (UTR) homologies. RCB11-1 and the original CCS1 (Hi10) sequences are also shown. Highlighted sequences show significant identities with t26. The limit of conserved CCS1 sequences is indicated by the arrow and vertical bar. The tRNA homologies in CentA, RCB11-1, and the cereba elements are interrupted by additional nucleotides 2 bp downstream of a repeat of the TATCA motif recognized by integrase at the LTR boundary.

View larger version (44K): In this window In a new window Download PPT slide   Figure 4. Arabidopsis crwydryn elements in BAC database. T9F8 is the least degenerate. F5K24 has an extended 5' LTR containing integrase sequences. T1016 has a partial but divergent duplication (70% identical in region marked) that may be part of an independent element inserted in the UTR and a priming site downstream of the conserved 3' LTR.

View larger version (18K): In this window In a new window Download PPT slide   Figure 5. Phylogenetic analyses consistently group together Arabidopsis and cereal crwydryn elements. (A) Tree based on one of the most highly conserved retrotransposon genes, reverse transcriptase. Crwydryn elements group to the right of the vertical line. The origins of the sequences are described in Table 2. Oat and rye sequences are derived from PCR of reverse transcriptase from genomic DNA and that of maize from the integration target of CentA. (B) Tree based on the most divergent gene, gag. Crwydryn elements group to the right of the dashed line. Note deep branch lengths between putative autonomous and nonautonomous cereal elements. The identities of the clones are also given in Table 2.

View larger version (15K): In this window In a new window Download PPT slide   Figure 6. Phylogenetic relationships of cereal crwydryn PCR fragments. Relevant regions of other clones are included in the analysis. Figures in parentheses indicate the number of sequences within a lineage. (A) Reverse transcriptase fragments. The single maize sequence is derived from the target of CentA integration (shown in Fig 1). Wheat sequences lie in a single lineage within the rye cluster—one rye sequence is grouped with the wheat lineage. (B) Integrase fragments. The RCB11 sequence indicated is taken from the target of RCB11-2 integration (Fig 1); the three other rice sequences shown include a second rice database entry (RCS1). The maize "gypsy" sequence shown is taken from database entry AF030633.

Estimates of mutation rates and element ages:
Synonymous and nonsynonymous substitution rates were calculated using the GCG program Diverge. For calculation of underlying mutation rates, species consensus was used where more than two comparable sequences were available. IN rates calculated using the least divergent of pairs of sequences are also shown (indicated by brackets in Table 1), although these were not used when deriving an average rate for RT or IN regions. Synonymous substitution rates were converted to mutation rate estimates (Table 1) using the following species divergence times: rye/wheat 7.5 million years (my), rye-wheat/oat 20 my, rye-wheat-oat/maize-millet-sorghum-rice 60 my, rye-wheat-oat-rice-maize-millet/bamboo 80 my, rice/sorghum-millet 30 my, maize/millet-sorghum 40 my (BENNETZEN and FREELING 1993 , BENNETZEN and FREELING 1997 ). Element ages, representing the time required for nonsynonymous divergence from the relevant species consensus at the relevant (RT or IN) mutation rate, were as follows: rye RT average 1.01 my (22 sequences, range 0–3.3 my), IN average 0.6 my (7, 0–1.2 my); wheat RT 0.7 my (4, 0.5–0.9 my), IN 0.85 my (5, 0.8–1.2 my); oat RT 1.4 my (5, 0.4–2.9 my), IN 1.1 my (6, 0–2.3 my); sorghum RT 0.42 my (7, 0–1.1 my), IN 0.73 (3, 0.7–0.8 my); rice RT 0.22 my (3, 0–0.7 my), IN 0.7 my (2).

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

A single ancestral retrotransposon family has given rise to a variety of universal cereal centromeric sequences:
No intact centromere-specific retrotransposon having a conventional complement of LTRs and retrotransposon reading frames has yet been found in the cereals. It has been suggested that the contemporary centromeric sequences may represent ancient rearrangements that have become fixed in the genomes by accident or by virtue of acquisition of novel function(s) (MILLER et al. 1998A ; PRESTING et al. 1998 ). We have therefore aligned the published sequences and isolated additional clones to clarify the organization of the centromere-specific sequences. Details of the clones used in this analysis are given in Table 2.

It is clear from the degree of overlapping homology that a single ancestral family has given rise to all the cereal centromere-specific retrotransposon fragments so far identified. We have named this family crwydryn (Welsh for "wanderer"; Fig 1). There is no evidence for unconventional additional components and the only rearrangement to be seen in more than one clone is an internal deletion leading to the loss of all enzymatic functions, resulting in elements having only LTRs, 5' untranslated region (UTR), and a gag structural gene fragment, truncated before the canonical RNA-binding motif (Fig 2). The first member of this class to be described was CentA, in maize (ANANIEV et al. 1998 ); the same organization is also seen in two elements in the rice clone RCB11 (NONOMURA and KURATA 1999 ), one intact (here designated RCB11-1) and one truncated in cloning (here designated RCB11-2). Both CentA and RCB11-2 are inserted at different positions into Ty3-gypsy retrotransposon reading frames; these reading frames are closely related and show little degeneracy (see below), implying recent movement by both of these elements.

Other than the CentA-like deletions, the only unexpected rearrangement seen is in the barley clone described by PRESTING et al. 1998 , originally proposed to contain a single element, cereba. Comparison with other clones demonstrates that the clone in fact contains two independent elements, here designated cereba-1 and cereba-2 (Fig 1A). Cereba-1 consists of a very short fragment of the upstream LTR, a long 5' UTR, and most of the polyprotein; cereba-2 appears to have been a CentA-like element that has suffered insertions into the gag region of a sequence of unknown origin (which is not centromere specific, PRESTING et al. 1998 ) and into both LTRs.

Alignment of clones allows reconstruction of the ancestral crwydryn element (Fig 1B). The ancestor is a conventional Ty3-gypsy class retrotransposon, with a relatively large 5' UTR (>1 kb), and a polyprotein reading frame that overlaps the downstream LTR. Both of these features are common to other cereal retrotransposon families (MARILLONNET and WESSLER 1998 ) and do not represent unusual adaptations. Other than the rearrangements described above or small (<1 kb) internal deletions, published fragments do not differ in organization from the ancestral sequence. Conservation is greatest between open reading frame fragments (typically >90% peptide identity), but is also seen in 5' UTR regions. The reading frames interrupted by CentA and RCB11-2 are highly similar to that of cereba-1 and are also expected to be derived from a recent common crwydryn ancestor.

Additional clones were isolated to confirm that the ancestral organization of the crwydryn element is maintained in a range of species (Fig 1B). PCR of maize and A. squarrosa genomic DNA using oligonucleotides based on conserved integrase and CCS1 motifs gave rise to the products predicted for internal retrotransposon fragments (for example, maize clone UC8.5, Fig 1B), confirming that CentA/RCB11-like LTRs are contiguous with cereba-1 polyprotein homologues in these species, while PCR of rye genomic DNA using oligonucleotides based on a rye B-chromosome-specific gag fragment (LANGDON et al. 2000 ) and a conserved RT motif gave rise to the expected product (clone UC6.7), confirming the presence of contiguous gag-RT sequences in rye. Related gag sequences were also identified in the rice centromeric clone RCH3 (DONG et al. 1998 ; Fig 2).

Finally, an essential feature of most conventionalLTR retrotransposon elements is a primer binding site (PBS), adjacent to the upstream LTR. Although no PBS has been reported for cereal centromeric fragments, a disrupted methionine tRNA complement is present at the appropriate position in a number of clones (Fig 3). Low-stringency PCR carried out on rye genomic DNA allowed recovery of further LTR junctions, among which were sequences containing an intact methionine tRNA PBS (Fig 3). It is likely that this is the PBS present in ancestral sequences and that insertion of additional bases, presumably as a result of aberrant processing of the adjacent LTR junction, represents a subsequent attenuation or inactivation of the PBS (see DISCUSSION).

The crwydryn family is conserved in Arabidopsis:
Database entries for five BAC clones of Arabidopsis thaliana were found to contain full-length or relatively intact Ty3-gypsy class retrotransposon elements with reading frame sequences very similar to crwydryn (Table 2; Fig 4). T9F8 is the least degenerate of these, on the basis of the lack of reading frame deletions and similarity between its LTRs; no deletions have occurred and divergence is 2%. Intact elements are predicted to be 6.5 kb long, with 1-kb LTRs, comparable with the size of cereal elements; 5' UTRs are, however, shorter (Fig 4).

Phylogenetic analysis with the PHYLIP and GCG packages consistently places cereal centromeric sequences with these elements in a lineage that shows no overlap with other database entries (Fig 5), indicating that the crwydryn family represents a distinct group whose origin predates monocot-dicot divergence. Comparison of crwydryn members with other Ty3-gypsy families indicates that all canonical peptide components are present. No regions of unusual divergence were detected, and the most divergent regions are those that are also most variable in comparisons between other Ty3 families. Universal characteristics of the family include LTRs defined by terminal TGAT/ATCA inverted repeats, standard tRNAmet_i PBSs, and GGGAG polypurine tracts. The family belongs to branch 1 of Wright and Voytas' classification (WRIGHT and VOYTAS 1998 ). A key distinction of this classification is the absence of an env-like 3' reading frame; as the crwydryn polyprotein reading frame extends into the 3' LTR, it is unlikely that such additional genes will be found.

Some Arabidopsis elements display partial or divergent identity with canonical crwydryn elements, particularly in LTR regions (Fig 4). For example, T1O16 has acquired LTR sequences that are entirely unrelated to those of T9F8 and also larger (1.2 kb) than normal. The extent of its reading frame divergence indicates that this is a relatively ancient event. A more recent rearrangement has created a chimeric element in T27D20; the 5' LTR and UTR of a T1O16-like element have been fused to another region of polyprotein and the 3' LTR of a typical crwydryn element (Fig 4). Sequence of unidentified origin lying at the junction of the two includes a second potential PBS. This novel organization may result from a chromosomal rearrangement, but there is also evidence for aberrant processing during retrotransposition, for example, in F5K24, where 40 codons of integrase reading frame are found upstream of a 5' LTR, and in T1O16 itself, where a PBS is found downstream of the 3' LTR.

Nonautonomous elements have an ancient origin:
Internal deletions of retrotransposon sequences are frequently seen but usually are element specific. The consistency of the CentA/RCB11-2 rearrangement and the coincident truncation of cereba-2, together with striking UTR similarities such as a region of 49-bp identity between cereba-2 and RCB11-2, prompted a closer examination of these variants.

A phylogenetic tree derived from gag sequences places all the crwydryn elements into the same lineage (Fig 5B). However, elements lacking the gag RNA-binding motif (CentA, RCB11-1 and -2, and cereba-2) lie in a separate group to other crwydryn members and show greater diversity. Selection for protein function has been maintained, as most pairwise comparisons give K_a:K_s ratios above 2 (gag regions diverge faster than other retrotransposon components and so typically show relatively low K_a:K_s ratios). Both the autonomous and nonautonomous groups contain representatives from rice, maize, and barley, indicating that divergence occurred at or before the time of the last common ancestor, at least 60 mya. In addition, neighbor-joining trees for the divergent peptide region place RCB11-2 closer to CentA than to RCB11-1, indicating that the elements' common ancestor arose before the species divergence, some 40 mya. Synonymous substitution rates are high in all interspecific comparisons and for both regions, so that horizontal transfers are unlikely to underlie these relationships. The nonautonomous elements therefore appear to represent divergent members of the same ancient family rather than independent rearrangements.

A further common feature of the nonautonomous elements is a polyadenine stretch in the middle of the 5' UTR (between 200 and 1000 bp from the LTR). This feature is present in cereb-2 but not cereb-1, although flanking sequences are well conserved between the two; the poly(A) stretch lies within a region upstream of the gag gene that is >1 kb and shows >74% identity, while the gag region itself shows only 62% identity over <500 bp. This suggests that the stretch may have a functional role that is maintained or regenerated despite pressures of concerted evolution.

The crwydryn family maintains a high level of retrotransposition potential in cereals:
Most crwydryn database entries for cereals are from single genomic clones in which reading frames are degenerate and/or key functional domains are missing. While degenerate retrotransposon elements are frequently seen even for active families (MARILLONNET and WESSLER 1998 ), the centromeric domain in which this family is localized may be particularly prone to rearrangement and passive amplification (see below), so that it is not clear to what extent the concentration of crwydryn elements represents continuing retrotransposition.

We cloned and sequenced crwydryn PCR products from a number of cereal species to characterize their evolution in more detail. In particular, we were interested in the proportion of potentially functional elements and the extent to which individual elements had diverged from the relevant genomic consensus. This information provides a basis for estimating the age of individual insertion events, as nonsynonymous changes are expected to accumulate only in "slave" copies following their generation by retrotransposition from those rare "master" elements in which such changes are absent or nondeleterious. At a minimum, nonsynonymous changes will arise at the same frequency as synonymous changes in host genes; this is expected to be a substantial underestimate as retrotransposition itself is an error-prone process. Thus, the rate of evolution of maize coding genes has been estimated to be 1–4 x 10^-9/nucleotides/year (PURUGGANAN and WESSLER 1994 ), while in vivo error rates incurred during retroelement replication may be higher than this by at least four orders of magnitude (GABRIEL et al. 1996 ). Crwydryn elements that originated in the common ancestor of contemporary species should thus be readily distinguished from recent retrotransposition events.

A total of 45 reverse transcriptase clones were obtained from five species (rye, wheat, oat, sorghum, and rice) and 31 integrase clones were obtained from eight species (rye, wheat, oat, sorghum, and rice, plus maize, millet, and bamboo). The same PCR conditions were used for all species and resulted in a single or predominant product of the expected size in all cases. The only PCR for which no product was obtained was with maize, using reverse transcriptase oligonucleotides. Centromere specificity of clones was confirmed by FISH (see below). It is clear that crwydryn sequences have diverged in a species-specific manner, with strong selection for conserved protein function (Fig 6). Most species-specific sequence differences correspond to synonymous or conservative codon replacements and open reading frames were maintained. Surprisingly, there was no evidence for the persistence of ancestral crwydryn sequences in these surveys. All clones conformed closely to the relevant species consensus, and total variation was in the range of a few percent (for example, an average of 0.025 nonsynonymous substitutions per nonsynonymous site over a 92-codon region sequenced from 23 rye RT clones). The relative rate of synonymous to nonsynonymous codon changes (K_s:K_a) was very high for interspecific comparisons (6–16:1) and almost equivalent for intraspecific data (e.g., average 1.49:1 for the rye example above), a pattern typical of retrotransposons whose colonization of the genome depends on retrotransposition.

Detailed estimates of the age of individual elements were made by first deriving an approximate mutation rate for the family. We assumed that the close relationship of all the sequences confirms that each domain is derived from a common ancestor in a progenitor of the Poaceae and that the high level of intraspecific similarities indicates that a single consensus may be derived for each species. This is an oversimplification, as functional polymorphisms are apparent in some species. Species consensus sequences were compared to calculate the numbers of changes at synonymous sites; these Ks values were then combined with published estimates of the date of divergence of various species to provide crude mutation rate estimates. The validity of this approach is supported by the similarity of most estimates, which average 2.08 x 10^-8 (reverse transcriptase) or 1.38 x 10^-8 (integrase) substitutions per site per year. The dates at which individual elements began to diverge from the species consensus (assumed to be the date of retrotransposition) were then calculated by combining the mutation rate estimate with the number of substitutions calculated to occur at nonsynonymous sites (Ka) for pairwise comparisons of actual clones vs. relevant derived species consensus. In most cases, and for all species, estimates were for dates <1 myr, with 75% of sequences showing <1 myr divergence (see MATERIALS AND METHODS), and the oldest (rye) showing <3.5 myr. These ages are overestimates because nominally "divergent" sites will also include functional subfamilies and PCR errors (note, however, that the nonrandom nature of most intraspecific variation indicates that PCR error rates are relatively low). The recent derivation of most elements from consensus master copies indicates that the family is likely to still be active in most if not all species, while the failure to detect "old" elements implies that either the family is rapidly increasing in abundance at an equivalent rate in each of the divergent species sampled or that ancestral sequences are relatively rapidly removed in their entirety before significant levels of degeneration occur.

Crwydryn elements are dispersed throughout rye centromeres:
Crwydryn-related sequences have been localized to cereal centromeric regions by FISH to metaphase chromosome preparations (ANANIEV et al. 1998 ; DONG et al. 1998 ; MILLER et al. 1998A ; PRESTING et al. 1998 ). The resolution of this approach does not allow fine mapping of the elements and leaves open the possibility that crwydryn elements are clustered in a few large "islands" or are locally amplified by rearrangements that are frequently found in centromeric regions. Use of fiber-FISH to show short-range (10–20 kb) clustering (DONG et al. 1998 ) does not address these issues. We have used extended rye meiotic prophase chromosome preparations to examine long-range (0.1–1 Mb) organization (Fig 7). Both RT and IN sequences are dispersed throughout centromeric domains, although regional differences in density are apparent in surprisingly regularly sized units. Occasionally we find localization of only one of these sequences in a unit (arrowed in Fig 7). Assuming a chromatin packing ratio of 1400, each unit would be of the order of 100 kb in size, within a domain of at least 4 Mb. KASZAS and BIRCHLER 1998 have demonstrated that the maize B chromosome centromere contains redundant subdomains that are at least 55 kb in size. Each crwydryn block may therefore correspond to a functional subdomain of the type identified by KASZAS and BIRCHLER 1998 . The occasional concentration of crwydryn sequences deficient for key functional components (Fig 7) indicates that the differences in crwydryn abundance between blocks may be at least partly a result of passive amplification.

View larger version (39K): In this window In a new window Download PPT slide   Figure 7. Crwydryn sequences are distributed throughout rye centromeres. Meiotic prophase chromosomes were probed with RT (blue) or IN (green) PCR fragments. A region containing IN but not RT sequences is indicated by the arrow. All other signals represent a combination of both probes. Bar, 10 µm.

Extinction of crwydryn in some Poaceae lineages:
PCR screens for the presence of crwydryn components were successful in a variety of temperate and tropical species, including Brachyaria and Lolium. The identity of products was confirmed by direct sequencing. There were, however, some species in which no PCR product was obtained for coding regions. In particular, we were unable to amplify reverse transcriptase sequences from a variety of closely related Avena species (designated as potential progenitors of the hexaploid oat C genome, including Avena eriantha Cc7056 and A. ventricosa Cc7064) or from a number of Festuca species (including Festuca pratensis, F. glaucescens, and F. mairei) despite the presence of canonical sequences in more distant relatives (A. strigosa Cc7121, A. canariensis Cc7041, A. damascena Cc7045, A. prostrata Cc7060, A. longiglumis Cc4851, A. agadiriana Cc7433, A. barbata Cc4897, A. macrostachya Cc7068, and Lolium species). In these lineages, the target sequences appear to be entirely absent rather than variant at priming sites; FISH with appropriate polyploids (A. sativa and F. arundinacea) demonstrated strong hybridization to the PCR-positive genomes but little or no hybridization to PCR-negative genomes (Fig 8). These data indicate that although the crwydryn family colonized the centromeres of a common ancestor before the divergence of the Poaceae and has been maintained intact in most lineages since, there has been occasional recent elimination in some cases. CCS1/LTR sequences from this family may be more universally maintained.

View larger version (83K): In this window In a new window Download PPT slide   Figure 8. FISH with crwydryn elements. (A) A reverse transcriptase probe derived from Avena strigosa (putative A/D genome donor to A. sativa) labels the centromeres of the somatic metaphase chromosomes of the A and D but not the C genomes of hexaploid oats (A. sativa var. Sun II). (B) An LTR junction probe derived from A. eriantha (putative C genome donor to A. sativa) labels all the centromeres of the three genomes of A. sativa and also contains a fragment of dispersed repeat that specifically "paints" the C genome only.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Retrotransposon sequences in Poaceae centromeres:
The analyses presented here demonstrate that a single ancient family is the source of all Poaceae centromere-specific retrotransposon sequences reported to date. This family has a conventional organization and its protein components are highly conserved even in Arabidopsis homologues. Within the Poaceae, the family has evolved in a species-specific manner, with selection clearly acting at the protein level to maintain retrotransposition potential. Any explanation for its unusual distribution must therefore include a role for element mobility. Other unusual features are a relatively low mutation rate, a lack of ancient or degenerate progenitors, and a high degree of conservation of LTR sequences. We believe that the simplest explanation for these observations is also the one for which there are the most precedents, namely, that the family has developed an efficient mechanism for targeting retrotransposition to a specialized region of the host genome, in this case the centromere, which is intrinsically unstable. Turnover of sequences within the host region would then rapidly eliminate degenerating sequences, so that only functional elements, which can maintain a high recruitment rate, accumulate.

There are numerous examples of retroelements developing targeting mechanisms, including Ty1, Ty3, and Ty5 retrotransposons of yeast, various subtelomeric or telomere-specific LINEs in insects and algae, and mammalian retroviruses, each of which has different consequences for the elements' distribution in the host genome. The closest parallel for the cereal crwydryn distribution appears to be shown by the R elements of arthropods, families of LINEs that use a sequence-specific endonuclease as a mechanism to target integration to rDNA arrays. Although this mechanism is not used by LTR retrotransposons and both retroelement and host domain organization are very different in the two cases, R elements resemble the cereal crwydryn in maintaining a high degree of retrotransposition potential and a low mutation rate comparable to that of the host (EICKBUSH and EICKBUSH 1995 ; EICKBUSH et al. 1995 , EICKBUSH et al. 1997 ). The major influence is rapid elimination of old elements in their entirety, presumably by random array contraction (JAKUBCZAK et al. 1992 ), but some degree of concerted evolution is also likely to apply, leading to relatively high species homogeneity (JAKUBCZAK et al. 1992 ).

Centromeric domains are typically composed of repetitive sequence arrays that, like rDNA arrays, may be expected to be unstable and unfavorable sites for mobile element targeting. This is demonstrated, for example, by the Drosophila Dp1187 minichromosome where few insertions within centromeric satellite arrays appear to survive long enough for significant sequence degeneration to occur (SUN et al. 1997 ). The only potential decayed or ancient element in the Dp1187 centromeric region lies at the junction of two different satellite family arrays, i.e., at the point at which ectopic recombination is unlikely to be able to lead to clean excision. However, increasing either the complexity of the centromeric sequences or the rate at which new mobile elements are recruited will reduce the effectiveness of this elimination. We discuss below factors that may have allowed crwydryn colonization of cereal centromeres and other features that support this model.

Insertion site choice by the crwydryn family:
The highly specific distribution of the crwydryn family in the Poaceae is reflected in Arabidopsis. Seven BACs of Arabidopsis containing crwydryn elements map to within <2 cM of three of the five centromeric map sites, and two additional crwydryn elements are found near clusters of centromeric repeats distant from the mapped centromere positions on chromosomes 2 and 4. There is also a frequent association with the Ty3-gypsy retrotransposon family Athila, which is particularly abundant in pericentromeric regions and has been found inserted into Arabidopsis centromeric satellite arrays (PELISSIER et al. 1996 ). This distribution suggests that centromeric targeting by cereal crwydryn may represent a relatively simple refinement of a previously existing regional preference for "silent" or heterochromatin-like domains. Most cereal crwydryn insertions are into other crwydryn elements, although not at specific sites, raising the possibility that crwydryn elements themselves create the conditions, such as chromatin conformation, that direct targeting. It is interesting that the database entries indicate a preferred polarity of insertion (Fig 1). Among the additional LTR junctions that we have recovered, three are within the rye centromere-specific repeat AWRC1, which is itself a degenerate crwydryn element (not shown). All three are in reading frames and again in the same orientation as the target.

Evolution of the crwydryn family in Poaceae and Arabidopsis genomes:
It is striking that the crwydryn polyprotein in the two divergent lineages appears to be highly conserved and that an "ancestral" element with a full-length reading frame (T9F8) shows the most recent signs of retrotransposition in Arabidopsis despite the presence in the genome of subfamilies having small common deletions. This suggests that such derivatives have only a short evolutionary life in any genome. There are, however, two differences between crwydryn of cereals and Arabidopsis that appear to be significant.

First, the high degree of LTR conservation in cereal crwydryn contrasts sharply with the variability in Arabidopsis, where rapid change may occur both by LTR "swaps" as described above and by small incremental changes (F5K24, for example, shares only 54% identity with T9F8 over a 900-bp region of the LTR that includes 16 gaps, but has 76% identity over an ungapped 2-kb region of the reading frame). In cereals, however, the LTR U5 region is sufficiently well conserved to act as a universal centromeric probe (ARAGON-ALCAIDE et al. 1996 ). Given the unusual requirements expected for expression from a centromeric domain, conservation of this region is likely to be at least partly due to functional constraint maintaining promoter elements contained within it. In addition, however, the combination of preferred orientation and frequent interspersion seen in cereals, together with the promiscuous use of neighboring LTRs seen in Arabidopsis, could lead to frequent exchange of LTRs and homogenization of sequences. Gene conversion may also contribute to concerted evolution; the unexpectedly high similarity of cereba-1 and cereba-2 5' UTRs has already been described.

Second, the nonautonomous but conserved cereal subfamily does not appear to have a direct equivalent in Arabidopsis, where large deletion derivatives all appear to occur at single genomic sites, i.e., to have arisen following integration. Nonautonomous elements must depend on components provided in trans, which is known to occur (AVEDISOV et al. 1998 ) and which may be particularly common in some families (WRIGHT and VOYTAS 1998 ), but it appears to be rare for conserved nonautonomous families to emerge. A priori, such families can be expected to arise when autonomous elements are highly abundant, both because trans components are more readily available and because a greater range of variant elements are produced. The population of crwydryn elements in the Arabidopsis genome may be too small to support nonautonomous variants (although the subfamilies with small common deletions may represent attenuated variants that are supported by persistent full-length elements), but it is also surprising that a single nonautonomous family is seen in cereals. We suggest that this reflects an ancient explosion in cereal crwydryn activity, which has subsequently been moderated at least in part by the presence of the nonautonomous element itself (see below).

The rise and fall of the crwydryn centromeric colonies:
We have shown that cereal crwydryn elements, despite their presence in the centromeres of many species, behave as conventional retrotransposons and are closely related to elements in Arabidopsis that are not centromere specific. In addition, conserved crwydryn components may be absent in the centromeres of some cereal species without apparent effect. Our conclusion is that conservation of crwydryn reading frames reflects an ancient optimization, predating the origin of grasses, for function based on retrotransposition rather than centromeric activity. Conservation of noncoding sequences may reflect both a novel functional selection (for centromeric promoter activity) and the consequences of a novel genomic distribution (with homogenization driven by high concentrations of similarly oriented elements within the centromere). However, the most distinctive and surprising feature of the cereal crwydryn distribution is that host centromeres are unusually complex. MOORE et al. 1997 have recently proposed a scenario that seems to provide an explanation for this, which is that the closest relatives of the Poaceae have holocentric rather than regional centromeres and that emergence of the Poaceae progenitor may have coincided with a large-scale reorganization of these domains. An attractive model is that this reorganization provided a new niche for the crwydryn family and that some feature of the new domains corresponded to the integration preference of the family, whether preexisting or variant. Colonization by large numbers of crwydryn elements may have reinforced the distinctiveness of particular regions, both for the establishment of centromeric structures and for further retrotransposon "homing," and may indeed have driven the initial reorganization. A precedent for an extremely rapid process has been described for a recently reorganized marsupial centromere (O'NEILL et al. 1998 ), while the characterization of various neocentromeres has demonstrated that their establishment is to a large extent dependent on epigenetic effects rather than on specific properties of the underlying DNA (KARPEN and ALLSHIRE 1997 ).

The evolution of the nonautonomous family, attenuation of full-length elements by PBS insertions, and extinction of centromeric crwydryn in some lineages are all consistent with stabilization of the genomes following this initial explosion. Specifically, the nonautonomous elements, which date from the time of initial amplification, are likely to curb further expansion by the crwydryn family as a whole, both because of titration of retrotransposition components and because of specific insertional mutagenesis of autonomous elements. Following this initial restraint, more gradual changes may be expected to make the centromeres more resistant to colonization. In particular, simple repetitive DNA families are expected to evolve, leading to the accumulation of the centromeric satellite arrays seen in most other organisms. Such species-specific arrays have been reported, for example, in maize and sorghum (ANANIEV et al. 1998 ; MILLER et al. 1998B ) and will increase the rate at which new insertions will be removed by ectopic recombination. The cumulative effect of these processes will be to drive extinction of active crwydryn elements in some lineages, followed by elimination or fragmentary incorporation into tandem arrays, as appears to have occurred relatively recently in the oat C genome progenitors and some Fescue species (Fig 8). Alternatively, active members of the family may acquire new specificities, as may have occurred in maize where noncentromeric crwydryn LTRs have been found (ARAGON-ALCAIDE et al. 1996 ). In either case, satellite arrays may eventually oust crwydryn elements from all Poaceae centromeres. Under such circumstances the more modest copy numbers and less-specialized targets of the Arabidopsis elements may provide a more secure long-term evolutionary future.

	ACKNOWLEDGMENTS

We thank Dr. Steve Reader and Dr. Terry Miller for kindly supplying Aegilops species and wheat/rye addition lines. We also thank Dr. Graham Moore and members of his laboratory for helpful discussions and John Harper for assistance with FISH. This work was funded by Biotechnology and Biological Sciences Research Council grant PO1643 to G.J., J.W.F., and R.N.J., a University of Wales, Aberystwyth, Institute of Biological Sciences studentship to C.S.

Manuscript received August 10, 1999; Accepted for publication May 1, 2000.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ABBO, S., R. P. DUNFORD, T. N. FOOTE, S. M. READER, and R. B. FLAVELL et al., 1995  Organization of retro-element and stem-loop repeat families in the genomes and nuclei of cereals. Chrom. Res. 3:5-15.

ANANIEV, E. V., R. L. PHILLIPS, and H. W. RINES, 1998  Chromosome-specific molecular organization of maize (Zea mays L.) centromeric regions. Proc. Natl. Acad. Sci. USA 95:13073-13078[Abstract/Free Full Text].

ARAGON-ALCAIDE, L., T. MILLER, T. SCHWARZACHER, S. READER, and G. MOORE, 1996  A cereal centromeric sequence. Chromosoma 105:261-268[Medline].

AVEDISOV, S. N., E. S. ZELENTSOVA, and Y. V. ILYIN, 1998  Trans-mobilization of deletion copies of the mdg3 retrotransposon in cultured cells of Drosophila.. Genetika 34:335-342[Medline].

BENNETZEN, J. L. and M. FREELING, 1993  Grasses as a single genetic system: genome composition, collinearity and compatibility. Trends Genet. 9:259-261[Medline].

BENNETZEN, J. L. and M. FREELING, 1997  The unified grass genome: synergy in synteny. Genome Res. 7:301-306[Free Full Text].

BENNETZEN, J. L. and E. A. KELLOGG, 1997  Do plants have a one-way ticket to genomic obesity? Plant Cell 9:1509-1514[Medline].

BENNETZEN, J. L., P. SAN MIGUEL, M. S. CHEN, A. TIKHONOV, and M. FRANCKI et al., 1998  Grass genomes. Proc. Natl. Acad. Sci. USA 95:1975-1978[Abstract/Free Full Text].

BOEKE, J. D. and S. E. DEVINE, 1998  Yeast retrotransposons: finding a nice quiet neighborhood. Cell 93:1087-1089[Medline].

DONG, F. G., J. T. MILLER, S. A. JACKSON, G. L. WANG, and P. C. RONALD et al., 1998  Rice (Oryza sativa) centromeric regions consist of complex DNA. Proc. Natl. Acad. Sci. USA 95:8135-8140[Abstract/Free Full Text].

EICKBUSH, D. G. and T. H. EICKBUSH, 1995  Vertical transmission of the retrotransposable elements R1 and R2 during the evolution of the Drosophila melanogaster species subgroup. Genetics 139:671-684[Abstract].

EICKBUSH, D. G., W. C. R. LATHE, M. P. FRANCINO, and T. H. EICKBUSH, 1995  R1 and R2 retrotransposable elements of Drosophila evolve at rates similar to those of nuclear genes. Genetics 139:685-695[Abstract].

EICKBUSH, T. H., W. D. BURKE, D. G. EICKBUSH, and W. C. R. LATHE, 1997  Evolution of R1 and R2 in the rDNA units of the genus Drosophila.. Genetica 100:49-61[Medline].

GABRIEL, A., M. WILLEMS, E. H. MULES, and J. D. BOEKE, 1996  Replication infidelity during a single cycle of Ty1 retrotransposition. Proc. Natl. Acad. Sci. USA 93:7767-7771[Abstract/Free Full Text].

GRANDBASTIEN, M.-A., 1992  Retroelements in higher plants. Trends Genet. 8:103-108[Medline].

HESLOP-HARRISON, J. S., T. SCHWARZACHER, K. ANAMTHAWAT-JONSSON, A. R. LEITCH, and I. J. LEITCH, 1991  In situ hybridisation with automated chromosome denaturation. Technique 3:109-116.

HIGASHIYAMA, T., Y. NOUTOSHI, M. FUJIE, and T. YAMADA, 1997  Zepp, a LINE-like retrotransposon accumulated in the Chlorella telomeric region. EMBO J. 16:3715-3723[Medline].

JAKUBCZAK, J. L., M. K. ZENNI, R. C. WOODRUFF, and T. H. EICKBUSH, 1992  Turnover of R1 (type I) and R2 (type II) retrotransposable elements in the ribosomal DNA of Drosophila melanogaster.. Genetics 131:129-142[Abstract].

JIANG, J. M., S. NASUDA, F. G. DONG, C. W. SCHERRER, and S. S. WOO et al., 1996  A conserved repetitive DNA element located in the centromeres of cereal chromosomes. Proc. Natl. Acad. Sci. USA 93:14210-14213[Abstract/Free Full Text].

KARPEN, G. H. and R. C. ALLSHIRE, 1997  The case for epigenetic effects on centromere identity and function. Trends Genet. 13:489-496[Medline].

KASZAS, E. and J. A. BIRCHLER, 1998  Meiotic transmission rates correlate with physical features of rearranged centromeres in maize. Genetics 150:1683-1692[Abstract/Free Full Text].

KIM, A., C. TERZIAN, P. SANTAMARIA, A. PELISSON, and N. PRUDHOMME et al., 1994  Retroviruses in invertebrates: the gypsy retrotransposon is apparently an infectious retrovirus of Drosophila melanogaster.. Proc. Natl. Acad. Sci. USA 91:1285-1289[Abstract/Free Full Text].

LANGDON, T., C. SEAGO, R. N. JONES, H. THOMAS, and H. OUGHAM et al., 2000  De novo evolution of satellite DNA on the rye B chromosome. Genetics 154:869-884[Abstract/Free Full Text].

LEE, C., R. WEVRICK, R. B. FISHER, M. A. FERGUSONSMITH, and C. C. LIN, 1997  Human centromeric DNAs. Hum. Genet. 100:291-304[Medline].

LEITCH, I. J., and J. S. HESLOP-HARRISON, 1992 Detection of digoxygenin labelled DNA probes hybridized to plant chromosomes in situ, pp. 18–25 in Methods in Molecular Biology: Protocols for Nucleic Acid Analysis by Non-radioactive Techniques, edited by P. G. ISAAC. Humana Press, Clifton, NJ.

MARILLONNET, S. and S. R. WESSLER, 1998  Extreme structural heterogeneity among the members of a maize retrotransposon family. Genetics 150:1245-1256[Abstract/Free Full Text].

MILLER, J. T., F. G. DONG, S. A. JACKSON, J. SONG, and J. M. JIANG, 1998a  Retrotransposon-related DNA sequences in the centromeres of grass chromosomes. Genetics 150:1615-1623[Abstract/Free Full Text].

MILLER, J. T., S. A. JACKSON, S. NASUDA, B. S. GILL, and R. A. WING et al., 1998b  Cloning and characterization of a centromere-specific repetitive DNA element from Sorghum bicolor.. Theor. Appl. Genet. 96:832-839.

MOORE, G., L. ARAGON-ALCAIDE, M. ROBERTS, S. READER, and T. MILLER et al., 1997  Are rice chromosomes components of a holocentric chromosome ancestor? Plant Mol. Biol. 35:17-23[Medline].

NONOMURA, K.-I. and N. KURATA, 1999  Organization of the 1.9-kb repeat unit RCE1 in the centromeric region of rice chromosomes. Mol. Gen. Genet. 261:1-10[Medline].

NOUTOSHI, Y., Y. ITO, S. KANETANI, M. FUJIE, and S. USAMI et al., 1998  Molecular anatomy of a small chromosome in the green alga Chlorella vulgaris.. Nucleic Acids Res. 26:3900-3907[Abstract/Free Full Text].

O'NEILL, R. J. W., M. J. O'NEILL, and J. A. M. GRAVES, 1998  Undermethylation associated with retroelement activation and chromosome remodelling in an interspecific mammalian hybrid. Nature 393:68-72[Medline].

PAN, W. H., A. HOUBEN, and R. SCHLEGEL, 1992  Highly effective cell synchronization in plant roots by hydroxyurea and amiprophos-methyl or colchicine. Genome 36:387-390.

PARDUE, M. L., O. N. DANILEVSKAYA, K. LOWENHAUPT, F. SLOT, and K. L. TRAVERSE, 1996  Drosophila telomeres: new views on chromosome evolution. Trends Genet. 12:48-52[Medline].

PEARCE, S. R., U. PICH, G. HARRISON, A. J. FLAVELL, and J. S. HESLOP-HARRISON et al., 1996  The Ty1-copia group retrotransposons of Allium cepa are distributed throughout the chromosomes but are enriched in the terminal heterochromatin. Chrom. Res. 4:357-364.

PELISSIER, T., S. TUTOIS, S. TOURMENTE, J. M. DERAGON, and G. PICARD, 1996  DNA regions flanking the major Arabidopsis thaliana satellite are principally enriched in Athila retroelement sequences. Genetica 97:141-151[Medline].

PETROV, D. A., 1997  Slow but steady reduction of genome size through biased mutation. Plant Cell 9:1900-1901[Medline].

PRESTING, G. G., L. MALYSHEVA, J. FUCHS, and I. Z. SCHUBERT, 1998  A TY3/GYPSY retrotransposon-like sequence localizes to the centromeric regions of cereal chromosomes. Plant J. 16:721-728[Medline].

PURUGGANAN, M. D. and S. R. WESSLER, 1994  Molecular evolution of the plant R-regulatory gene family. Genetics 138:849-854[Abstract].

SAN MIGUEL, P., B. S. GAUT, A. TIKHONOV, Y. NAKAJIMA, and J. L. BENNETZEN, 1998  The paleontology of intergene retrotransposons of maize. Nat. Genet. 20:43-45[Medline].

SONG, S. U., T. GERASIMOVA, M. KURKULOS, J. D. BOEKE, and V. G. CORCES, 1994  An env-like protein encoded by a Drosophila retroelement: evidence that gypsy is an infectious retrovirus. Genes Dev. 8:2046-2057[Abstract/Free Full Text].

SUN, X. P., J. WAHLSTROM, and G. KARPEN, 1997  Molecular structure of a functional Drosophila centromere. Cell 91:1007-1019[Medline].

SUONIEMI, A., K. ANAMTHAWAT-JONSSON, T. ARNA, and A. H. SCHULMAN, 1996a  Retrotransposon BARE-1 is a major, dispersed component of the barley (Hordeum vulgare L) genome. Plant Mol. Biol. 30:1321-1329[Medline].

SUONIEMI, A., A. NARVANTO, and A. H. SCHULMAN, 1996b  The BARE-1 retrotransposon is transcribed in barley from an LTR promoter active in transient assays. Plant Mol. Biol. 31:295-306[Medline].

SUONIEMI, A., D. SCHMIDT, and A. H. SCHULMAN, 1997  BARE-1 insertion site preferences and evolutionary conservation of RNA and cDNA processing sites. Genetica 100:219-230[Medline].

WAUGH, R., K. MCLEAN, A. J. FLAVELL, S. R. PEARCE, and A. KUMAR et al., 1997  Genetic distribution of Bare-1-like retrotransposable elements in the barley genome revealed by sequence-specific amplification polymorphisms (S-SAP). Mol. Gen. Genet. 253:687-694[Medline].

WRIGHT, D. A. and D. F. VOYTAS, 1998  Potential retroviruses in plants: Tat1 is related to a group of Arabidopsis thaliana Ty3/gypsy retrotransposons that encode envelope-like proteins. Genetics 149:703-715[Abstract/Free Full Text].

ZHONG, X-B., P. F. FRANSZ, J. WENNEKES VAN EDEN, P. ZABEL, and A. VAN KAMMEN et al., 1996  High resolution mapping on pachytene chromosomes and extended DNA fibres by fluorescence in situ hybridisation. Plant Mol. Biol. Rep. 14:232-242.

This article has been cited by other articles:

				E. I. Mikhailova, D. Phillips, S. P. Sosnikhina, A. V. Lovtsyus, R. N. Jones, and G. Jenkins Molecular Assembly of Meiotic Proteins Asy1 and Zyp1 and Pairing Promiscuity in Rye (Secale cereale L.) and Its Synaptic Mutant sy10 Genetics, November 1, 2006; 174(3): 1247 - 1258. [Abstract] [Full Text] [PDF]

</

				J. Ma and J. L. Bennetzen Recombination, rearrangement, reshuffling, and divergence in a centromeric region of rice PNAS, January 10, 2006; 103(2): 383 - 388. [Abstract] [Full Text] [PDF]