Recombination, rearrangement, reshuffling, and
divergence in a centromeric region of rice
Jianxin Ma and Jeffrey L. Bennetzen*
Department of Genetics, University of Georgia, Athens, GA 30602
Contributed by Jeffrey L. Bennetzen, November 11, 2005
Centromeres have many unusual biological properties, including
kinetochore attachment and severe repression of local meiotic
recombination. These properties are partly an outcome, partly a
cause, of unusual DNA structure in the centromeric region. Although
several plant and animal genomes have been sequenced,
most centromere sequences have not been completed or analyzed
in depth. To shed light on the unique organization, variability, and
evolution of centromeric DNA, detailed analysis of a 1.97-Mb
sequence that includes centromere 8 (CEN8) of japonica rice was
undertaken. Thirty-three long-terminal repeat (LTR)-retrotransposon
families (including 11 previously unknown) were identified in
the CEN8 region, totaling 245 elements and fragments that account
for 67% of the region. The ratio of solo LTRs to intact elements in
the CEN8 region is 0.9:1, compared with 2.2:1 in noncentromeric
regions of rice. However, the ratio of solo LTRs to intact elements
in the core of the CEN8 region ( 2.5:1) is higher than in any other
region investigated in rice, suggesting a hotspot for unequal
recombination. Comparison of the CEN8 region of japonica and its
orthologous segments from indica rice indicated that 15% of the
intact retrotransposons and solo LTRs were inserted into CEN8
after the divergence of japonica and indica from a common
ancestor, compared with 50% for previously studied euchromatic
regions. Frequent DNA rearrangements were observed in the CEN8
region, including a 212-kb subregion that was found to be composed
of three rearranged tandem repeats. Phylogenetic analysis
also revealed recent segmental duplication and extensive rearrangement
and reshuffling of the CentO satellite repeats.
CentO repeats DNA evolution nucleotide substitution unequal
conversion
The centromeres of eukaryotic chromosomes are essential for
precise chromosome segregation during mitosis and meiosis.
Although this function is conserved, the DNA content, organization,
and complexity of centromeres vary considerably across
different organisms (1­3). Budding yeast (Saccharomyces cerevisiae)
centromeres consist of only 125 bp of unique sequence (4,
5). In contrast, centromeres from most multicellular eukaryotes,
including Arabidopsis (6­8), rice (9­12), maize (13, 14), Drosophila
(15), and human (16, 17) are much more complex. These
large heterochromatic centromeres consist of large arrays of
satellite repeats that are usually arranged in a tandem head-totail
fashion, intermixed with additional repeats, including transposable
elements. Although tandem repeats of some sort appear
to be necessary for efficient centromere function in most eukaryotes,
the direct effector of kinetochore formation is the
assembly of an altered chromatin state in the centromere,
associated with a unique H3 histone (CENH3) (18).
In all plant centromeres investigated, the sizes of satellite
repeat units (also called monomers) are relatively consistent,
ranging from 150 to 180 bp [e.g., 155 bp for rice CentO (9),
156 bp for maize CentC (13), and 180 bp for Arabidopsis pAL1
(19)]. However, the sequences of satellite monomers are quite
different among species. The monomers of rice CentO and maize
CentC, for instance, do not show significant sequence similarity
(9). These differences can arise quite quickly, as shown by the
lack of CentO repeat homology between two species within the
genus Oryza, Oryza sativa and Oryza brachyantha (20). Moreover,
the copy numbers of satellite repeats vary dramatically across
species and even among different centromeres of a single
organism or the same centromeres from different ecotypes or
varieties (9, 21). This variation is partially responsible for
tremendous differences in centromere size, as revealed by FISH
(9, 14).
Another category of DNA found within plant centromeres is the
long-terminal repeat (LTR) retrotransposons (1­3) that usually
surround or intermingle with the satellite repeat arrays (9, 11, 12).
A few LTR-retrotransposon families have been found to be centromere-enriched
by FISH analyses and are called centromerespecific
retrotransposons (CRs). These include the CRRs of rice (9,
10, 22­24) and the CRMs of maize (13). A recent survey of CRR
distribution indicated that none of the CRR families or subfamilies
previously described are completely specific to centromeres (25).
Because of the abundance of repetitive DNA, especially the large
arrays of satellite repeats in centromeric regions, the sequencing
and assembly of complete centromeres of higher eukaryotes is
extremely challenging. Hence, it is not surprising that sequence gaps
remain in all centromeres of Arabidopsis and human (26, 27).
Recently, a 1.97-Mb region including rice CEN8, which contains the
least satellite DNA among the 12 centromeres of rice (9), has been
completely sequenced (11). Previous studies have described the
dynamics of local sequence change across the rice genome, but
these studies were completed before completion of any rice centromere
(28­30). Because centromeres have such unusual structure
and function, it will be interesting to investigate whether these
properties influence the rates, mechanisms, or tolerated outcomes
of local genome evolution. Here, we present detailed sequence
analysis of the CEN8 region from japonica rice and its orthologous
segments from indica rice, including structural analysis of LTR
retrotransposons, phylogenetic analysis of CentO satellite repeats,
and comparative analysis of genic and intergenic segments from
these two subspecies. Our data indicate that the rates and outcomes
of divergence differ substantially within different centromereassociated
domains and are also quite different from the properties
of sequence divergence in noncentromeric regions of rice.
Materials and Methods
Identification of LTR Retrotransposons. Structural analyses and sequence
homology comparisons were used to identify LTR retrotransposons
in the CEN8 region of japonica rice (c.v. Nipponbare)
(11). Intact LTR retrotransposons were identified by using LTRSTRUC,
an LTR-retrotransposon mining program (28), and by
methods previously described (30, 31). Solo LTRs and truncated
elements were identified by sequence homology searches against a
rice LTR-retrotransposon database that was developed by collecting
known LTR retrotransposons and by scanning the 371-Mb
Nipponbare genome sequence generated by the International Rice
Conflict of interest statement: No conflicts declared.
Abbreviations: LTR, long terminal repeat; CRR, centromere-specific retrotransposons of
rice; mya, million years ago.
*To whom correspondence should be addressed. E-mail: maize@uga.edu.
 2005 by The National Academy of Sciences of the USA
www.pnas.org cgi doi 10.1073 pnas.0509810102 PNAS January 10, 2006 vol. 103 no. 2 383­388
GENETICS
Genome Sequencing Project (Build 3.0 pseudomolecules, http:
rgp.dna.affrc.go.jp IRGSP Build3 build3.html). The structures of
all LTR retrotransposons identified were confirmed by manual
inspection. New LTR-retrotransposon families were defined by the
criteria previously described (31).
Sequence Alignments and Comparisons. Targeted query sequences in
the CEN8 region of japonica cultivar Nipponbare were used in
BLASTN and CROSS MATCH searches against the assembled wholegenome
shotgun sequence for indica cultivar 93-11 to identify
orthologous segments. The orthologous relationship between japonica
and indica sequences was assumed when a query sequence
from japonica exhibited only one match to assembled 93-11 genome
sequence data.
Homologous sequences were aligned by using CLUSTALX (32)
and were manually inspected. Synonymous and nonsynonymous
substitution distances for the orthologous genes in the CEN8
regions and the euchromatic regions of japonica and indica (31),
and sequence divergences of the two LTRs of single retrotransposons,
were calculated as described (31).
A triplication in the CEN8 region was identified by BLAST2 (33)
and CROSS MATCH (University of Washington, Seattle). A segmental
duplication of CentO satellite arrays was detected by
phylogenetic analysis of all CentO satellite monomers in CEN8.
Results
Analysis of the LTR Retrotransposons in the CEN8 Region. As previously
argued (30, 34), the abundance, broad distribution, and
known structures of LTR retrotransposons make them particularly
well suited surrogates for the study of local genome evolution. The
initial focus of this study was to accurately identify LTR retrotransposons
and their structures in the rice CEN8 region. This was
challenging, given that LTR retrotransposons are highly enriched in
CEN8 and are preferentially arranged in a nested pattern (11), and
that numerous LTR retrotransposons in rice have undergone
various inter- intraelement rearrangements, primarily through unequal
homologous recombination and illegitimate recombination
(30, 34). To ensure a detailed characterization of LTR retrotransposons
in CEN8, we have optimized the method that was previously
developed by our laboratory for identification of LTR retrotransposons
(31, 35, 36). A large database of rice LTR retrotransposons
was first generated by scanning the whole-genome sequence of rice
(c.v. Nipponbare, Build 3.0 pseudomolecules, http: rgp.dna.affrc.
go.jp IRGSP Build3 build3.html) by using LTR-STRUC (28) and by
collecting all rice LTR retrotransposons previously reported [refs.
25, 31, and 37; The Institute for Genomic Research (TIGR),
Rockville, MD]. Subsequently, the LTR retrotransposons in CEN8
were identified by a combination of homology searches against this
rice LTR-retrotransposon database or on the basis of their canonical
structural characteristics (38), and each predicted element was
confirmed by manual inspection.
Using the methods described above, we identified 245 LTR
retrotransposons in the CEN8 region. These were comprised of
65 intact elements and 61 solo LTRs that are flanked by standard
target-site duplications (TSDs), two intact elements and four
solo LTRs lacking TSDs, and 113 truncated elements 500 bp
(see Table 1, which is published as supporting information on the
PNAS web site). Smaller fragments without any recognized PBS
(primer binding site), polypurine tract, or TSD were not further
investigated.
LTR sequences from LTR retrotransposons identified in this
study were compared with each other in a pair-wise manner and
compared with LTR sequences extracted from previously reported
LTR retrotransposons by CROSS MATCH and BLASTN. On the basis
of LTR sequence homology, 22 known LTR-retrotransposon families
and 11 previously unknown LTR-retrotransposon families
were identified. These 11 families account for 24 LTR retrotransposons
in the CEN8 region (see Table 2, which is published as
supporting information on the PNAS web site). The copy numbers
of these 33 families vary from 1 to 25 in the CEN8 region, but none
of these families is completely unique to either the CEN8 region or
the other 11 centromeric regions of rice (data not shown). This
observation deviates from the previous description of CRRs revealed
by FISH analyses (9) but does parallel a recent survey of
distribution of CRRs by sequence analysis (25), indicating that the
CRRs are enriched in the centromeric regions of rice, but that rare
copies are found elsewhere. Overall, CRRs make up 19% of the
core region and 16% of the entire CEN8 region (see Table 2).
A Low Rate of Unequal Recombination for LTR Retrotransposons in the
CEN8 Region. A high frequency of solo LTRs, derived from unequal
homologous recombination between two LTRs of single retrotransposons
(36), has been previously noted in the rice genome (10, 30,
31). In this study, we found that solo LTRs are considerably less
abundant on average in the CEN8 region than in most of the
characterized genome. The overall ratio of solo LTRs to intact
elements in the CEN8 region (0.94:1) is significantly lower (P
0.05) than the 2.2:1 ratio seen in euchromatic regions (31) and also
significantly lower (P 0.05) than the 1.6:1 ratio seen in the rice
genomic sequences available in GenBank on August 21, 2002, which
was primarily composed of noncentromeric regions (30) (see Table
3, which is published as supporting information on the PNAS web
site). These data suggest that unequal intraelement recombination
has been considerably suppressed in the centromeric region in
contrast to noncentromeric regions.
A Hotspot for Unequal Homologous Recombination in the Core of
CEN8. Because unequal homologous recombination rates of LTR
retrotransposons in the CEN8 region and euchromatic regions of
rice are significantly different, we wondered whether the rates of
recombination also vary across the CEN8 region. Hence, the
1.97-Mb CEN8 region was dissected into 10 adjacent nonoverlapping
subregions (each of the first nine subregions is 200 kb, and the
last subregion is 173 kb; Fig. 1), and we subsequently investigated
the distribution of solo LTRs and intact LTR retrotransposons in
these subregions. We found that the ratio of solo LTRs to intact
elements in the core of the CEN8 region is an average of approximately
three times higher than in other subregions (Fig. 1). This
core region at the center of the CEN8 centromere is included within
the CENH3-binding domain that defines the chromosome segregation
property of CEN8 (10) but is a distinct subdomain within that
region (Fig. 1).
One could propose that this apparent high rate of unequal
recombination is a property of the particular families of LTR
retrotransposons in the core of the CEN8 region. There are 14
LTR retrotransposons (solo LTRs and intact elements) harbored
in the core subregion, and these belong to 12 different
families (see Table 2). The distribution and structures of all 91
retrotransposons in the CEN8 region that belong to these 12
families were analyzed. The average ratio of solo LTRs to intact
elements of these 12 families in the CEN8 region was found to
be 0.90:1 (see Table 2), similar to the value calculated for all
retrotransposons in the CEN8 region (0.94:1) (see Table 3).
Thus, these families do not show any overall bias toward solo
LTR accumulation, suggesting that the hotspot of intraelement
unequal recombination is specific to the core subregion.
Comparison of LTR Retrotransposons in the CEN8 Regions of japonica
and indica Rice. To determine the timing and lineage specificities of
the dramatic accumulation of LTR retrotransposons in CEN8, we
conducted sequence comparison between the genomes of two rice
subspecies, japonica and indica, targeting LTR-retrotransposon
insertion sites. For each LTR-retrotransposon insertion in CEN8 of
japonica cultivar Nipponbare, two unique 500-bp sequences, each
composed of 200 bp of one retrotransposon terminal sequence and
300 bp of flanking DNA, were extracted and used to search against
384 www.pnas.org cgi doi 10.1073 pnas.0509810102 Ma and Bennetzen
the assembled whole-genome shotgun sequence (WGS) generated
from indica cultivar 93-11 (39, 40). An insertion of an LTR
retrotransposon was considered to be shared between japonica and
indica when two unique sequences regions were found in assembled
indica WGS that perfectly matched the two query sequences
targeting that insertion (see Fig. 4, which is published as supporting
information on the PNAS web site). Alternatively, an insertion was
judged to be unique in japonica when a unique sequence region
was found in assembled indica WGS that perfectly matched the
combined sequences that are composed of two query sequences
without the LTR-retrotransposon terminal sequences (see Fig. 4).
The insertion sites of 126 intact elements and solo LTRs identified
in japonica were investigated by japonica and indica comparison.
The results indicated that 100 (45 intact elements and 55 solo
LTRs) were inserted into the common ancestor of japonica variety
Nipponbare and indica variety 93-11, whereas 18 (15 intact elements
and 3 solo LTRs) were inserted into a Nipponbare ancestor
after its divergence from a common ancestor with 93-11 (see Table
4, which is published as supporting information on the PNAS web
site). It remains unclear whether the other eight elements are
shared by japonica and indica, because there were no sequences
found in the indica database that match corresponding query
sequences flanking these elements. In contrast to the 1.1-Mb
noncentromeric orthologous regions from japonica (c.v. Nipponbare)
and indica [c.v. either GLA4 (31) or 93-11], in which 50%
of the retrotransposons were unique to japonica (see Table 4), the
CEN8 region exhibits a significantly (P 0.01, Fisher's exact test)
higher percentage (85%) of insertions shared between japonica and
indica (see Table 4). This suggests that the majority of LTR
retrotransposons in the CEN8 region accumulated before the
divergence of japonica and indica from a common ancestor.
Estimated Insertion Times of LTR Retrotransposons in the CEN8
Region. LTR divergence was used as a tool to date insertions of LTR
retrotransposons (30, 41). The data demonstrated that the majority
of the intact LTR retrotransposons in the CEN8 region (Fig. 2 and
see Table 5 and Fig. 5, which are published as supporting information
on the PNAS web site) were calculated to be slightly older
[average insertion dates of 0.8 million years ago (mya)] than those
harbored in the 1.1 Mb of euchromatic DNA (average insertion
dates of 0.59 mya) that was previously analyzed (31). However, 27
dated LTR retrotransposons in the CENH3 domain (10) were
calculated to be significantly older (average insertion dates of 1.06
mya) than those in the euchromatin regions or the flanking
pericentromeric heterochromatin (see Fig. 5).
Eleven of 45 intact LTR retrotransposons shared between japonica
and indica, expected to be inserted into rice before the
divergence of japonica and indica, were calculated to be more recent
insertions than 0.44 mya [the estimated divergence time of the
indica and japonica varieties studied (31)] (Fig. 2 and data not
shown). These results suggest that the LTR retrotransposons in the
CEN8 region diverge at a slower rate than in the euchromatic
regions (31), where all of the retrotransposons shared between
japonica and indica were found to be older than 0.69 mya (Fig. 2B).
Hence, the ages of LTR retrotransposons in the CEN8 region could
be underestimated if the divergence of LTR sequences is occurring
at a slower rate. Alternatively, the CEN8 region analyzed may have
been more recently shared by a common Nipponbare and 93-11
ancestor than was the 1.1 Mb of euchromatic DNA previously
characterized.
Low Levels of Genic Sequence Divergence in the CEN8 Region. To
further address the genomic sequence divergence in the centromeric
region, synonymous and nonsynonymous nucleotide substitutions
were analyzed in the predicted coding regions of 20
genes in the Nipponbare (japonica) CEN8 region and their
homologues identified in the 93-11 (indica) genome by the
method of Nei­Gojobori (42) by using the Jukes­Cantor corFig.
1. Variation in the ratios of solo LTRs to intact LTR retrotransposons
across the CEN8 region. The CEN8 region was dissected into 10 contiguous
subregions, as shown on the x axis. Bars indicate ratios of solo LTRs to intact
LTR retrotransposons (A), and total numbers of solo LTRs plus intact LTR
retrotransposons (B) in respective subregions.
Fig. 2. Ages of LTR retrotransposons in the CEN8 region (A) and previously
analyzed euchromatic regions (31) (B) of Nipponbare rice. Ages were estimated
from LTR sequence divergence by using a substitution rate of 1.3 10 8 mutatuons
per site per year (31). Gray bars show elements shared by Nipponbare
(japonica) and 93-11 (indica); black bars show elements unique to Nipponbare.
Ma and Bennetzen PNAS January 10, 2006 vol. 103 no. 2 385
GENETICS
rection. Each of these 20 predicted genes was chosen because it
had a unique copy in both japonica and indica genomes (data not
shown); therefore, it is likely they represent 20 orthologous loci
in the CEN8 regions of japonica and indica (see Table 6, which
is published as supporting information on the PNAS web site).
Similarly, we reanalyzed the 24 genes previously investigated in
the 1.1-Mb euchromatic regions of Nipponbare (31) and their
orthologues in the 93-11 genome (see Table 6).
The 20 pairs of genes in the CEN8 region that we investigated
exhibit variable synonymous and nonsynonymous substitution distances,
ranging from 0 to 0.0186 and from 0 to 0.0064, with average
distances of 0.0025 and 0.0012, respectively (see Table 6). In
contrast, the 24 orthologous genes previously investigated in the
1.1-Mb euchromatic regions exhibit average synonymous and nonsynonymous
substitution distances of 0.0057 and 0.0016 (see Table
6). The average nonsynonymous substitution distances calculated
by using these two sets of genes are almost identical, probably due
to similar levels of filtration applied by natural selection. However,
the average synonymous substitution distances in the CEN8 region
are significantly lower than observed in euchromatic regions. This
observation, together with the analysis of LTR-retrotransposon
sequence divergence, suggests that the CEN8 region exhibits nucleotide
sequence divergence at an 1.6- to 2.2-fold lower rate than
euchromatic regions, or that the CEN8 region was more recently
shared by a common ancestor of the studied japonica and indica
varieties.
Indels and Point Mutations. Because synonymous (gene) and possibly
neutral (LTR-retrotransposon) rates of base pair substitution
appeared to be lower in centromeric regions, it is important to
determine whether the frequency of indel generation was also
unusual in the CEN8 region. Previous studies have shown that rice
genomic sequences are rapidly removed by small deletions that, in
the absence of natural selection to retain the sequence, will remove
all nuclear DNA with a half life of 6 million years (30, 31). For
the introns of the 20 pairs of orthologous CEN8 region genes
described above, it was found that the ratio of indels to point
mutations was 0.4 (data not shown), whereas the same ratio (0.4)
was observed in the 24 euchromatic genes previously studied (31).
Hence, this lower sequence divergence rate or more recent common
ancestry was manifest in both classes of sequence variation.
Segmental Duplication in the CEN8 Region. A 212-kb subregion,
composed of three tandem triplicated segments ``a'' (96 kb), ``b'' (90
kb), and ``c'' (26 kb), was identified in the CEN8 region (see Fig. 6,
which is published as supporting information on the PNAS web
site). Segments a and b share two intact LTR retrotransposons and
a solo LTR, whereas segments a and c share a 1.6-kb indel that is
absent in segment b. Insertions of the two shared intact LTR
retrotransposons, Osr30 and Osr31, were dated to 1.35 and 0.27
mya, respectively, demonstrating that the duplication of segment a
and b occurred after the insertion of these LTR retrotransposons.
Because the overall pair-wise sequence similarities among segments
a, b, and c in the shared region are high (99.0­99.3%), it is likely that
the two duplications occurred successively in similar time frames. A
residue of Osr33 was detected at the end of segment c, indicating
that at least one deletion event occurred after the segmental
duplication, thereby removing a large portion of Osr33 from
segment c (see Fig. 6).
A previous FISH study revealed dramatic variation in the amount
of CentO satellite DNA between different chromosomes and
between the corresponding chromosomes of different varieties of
rice (9). Most CentO satellite repeats in the CEN8 region share
91­99% sequence identity with their consensus sequence (11).
These observations suggest that many recent amplifications and
rearrangements of CentO satellite repeats have occurred in rice
centromeric regions.
In an attempt to shed light on the processes that give rise to the
dramatic changes in copy numbers of CentO satellite repeats
between different chromosomes (9), phylogenetic analysis was
performed on all 460 CentO satellite repeats that are clustered in
five blocks within the CEN8 region (11) by using the MEGA program
(43) (Fig. 3). Based on the neighbor-joining phylogenetic trees
obtained (see Fig. 7, which is published as supporting information
on the PNAS web site), we identified 48 pairs of highly identical
monomers (98­99% sequence similarity) that are dispersed in the
two largest CentO blocks (I and III). Interestingly, the orders of
these two sets of 48 monomers were considerably conserved in their
corresponding CentO blocks, and these blocks are arranged in
opposite orientation (Fig. 3), suggesting a recent segmental duplication
event that drove the amplification of CentO satellite repeats.
Reshuffling of CentO Satellite Repeats in the CEN8 Region. In addition
to the segmental duplication described above, an apparent insertion
or deletion (indel) of a cluster of 69 contiguous monomers was
found by aligning two duplicated segments (Fig. 3). Some small
indels composed of one or a few monomers were also found by
comparing the two duplicated segments. Moreover, 11 CentO
Fig. 3. Segmental duplication of CentO blocks and reshuffling of CentO satellite repeats. Satellite repeats from five CentO blocks (I, II, III, IV, and V) in the CEN8
region are represented by green, black, and yellow vertical lines, The most identical pairs of satellite repeats were often adjacent, but many (indicated by black
lines) were most closely related to repeats that were located at a significant distance. These distant pairs are connected by the gray curved lines. The yellow lines
indicate satellite repeats that share a 12-bp duplication. The red arrows indicate the orientation of CentO blocks.
386 www.pnas.org cgi doi 10.1073 pnas.0509810102 Ma and Bennetzen
satellite monomers were found to share a duplication of 12 bp of
DNA, indicating these monomers share a common origin. However,
none of these 11 monomers are adjacent to each other. These
observations suggest that amplification and reshuffling of CentO
satellite repeats have occurred quite often during CentO evolution.
Discussion
The initial establishment of a centromere, as evidenced by
studies of neocentromere formation, does not require long
stretches of alphoid satellite repeats like CentO of rice, but it
does require attainment of a heterochromatic state featuring the
specification of CENH3 (the centromere-specific H3 histone)
within the neocentromere nucleosomes. However, stable and
highly efficient centromere function in chromosome segregation
in most animals and plants is associated with satellite repeats
interspersed with other repetitive elements like LTR retrotransposons
(1­3). Hence, the functional status of a centromere may
dramatically influence the accumulation and subsequent divergence
of centromere-associated DNA sequences. Sequences
with different affinities for kinetochore components may compete,
providing a foundation for the theory that meiotic competition
between centromeres in the female gametophyte could
enhance the rate of centromeric sequence variation (3, 18). To
see what unusual effects these functional constraints might have
upon centromere sequence divergence, an appropriate first step
is to determine the nature and rate of centromere-associated
sequence variation.
Previous analyses of the centromeric region of rice chromosome
8 have provided valuable information regarding the content and
composition of centromeric DNA in a higher plant species (10, 11).
However, the repeat structures in this region were not analyzed in
depth, probably due to the difficulties that have been commonly
met by most genome sequencing groups in annotation of large sets
of complex genomic sequences (44). In this study, 11 previously
unknown LTR-retrotransposon families (including 23 elements)
were identified, and solo LTRs, truncated LTR-retrotransposon
elements, and elements that are arranged in nested patterns (35)
were characterized. Combined with the analyses of sequence divergence
and rearrangement, this study provides a comprehensive
description of a completely sequenced centromeric region in a plant
genome.
Recent whole-genome sequencing projects in plants have
revealed the dramatic accumulation of LTR retrotransposons in
centromeric and pericentromeric regions (11, 26). As evidenced
by the 36 intact LTR retrotransposons and solo LTRs found in
the CEN8 region in this study, the percentage of LTRretrotransposon
DNA in a genome is considerably underestimated
by genome-sequencing projects that concentrate on euchromatic
regions. In total, LTR retrotransposons constitute at
least 67% of the DNA in the CEN8 region. Assuming that
18­22% of the rice genome is composed of LTR retrotransposons
(30, 37), LTR retrotransposons are at least 3- or 4-fold
enriched in centromeric region compared with most noncentromeric
regions. Although the evolutionary mechanisms behind
preferential insertion and or retention bias for LTR retrotransposons
in centromeric regions are still poorly understood, the
nonrandom distribution of LTR retrotransposons should at least
partly reflect the action of purifying selection against the deleterious
effects of LTR-retrotransposon insertion into genes (45).
Given that homologous recombination during meiosis is highly
repressed or completely inhibited in all rice centromeres (46, 47),
it is not surprising that a low relative abundance of solo LTRs was
observed in the CEN8 region compared with previously investigated
euchromatic regions (31). A suppression of homologous
recombination would be expected to also inhibit unequal recombination
events. The observation that LTR retrotransposons in the
CEN8 region are older on average than in noncentromeric regions
suggests that the other events that remove LTR-retrotransposon
sequences from the genome, primarily small deletions caused by
illegitimate (i.e., nonhomologous) recombination (30, 36), may also
be suppressed, allowing a longer time for intact elements to persist
and for solo LTRs to accumulate.
The discovery of a hotspot for solo LTR accumulation in the core
of the CEN8 region was unexpected. Previous research has shown
that the core of the CEN8 region is part of the kinetochore region
that is enriched in CENH3 (10). The kinetochore region of CEN8
was also found to harbor active genes (10). Perhaps the kinetochore
and possible other genetic functions of this core subregion have
created an environment favoring a higher frequency of homologous
recombination than in adjacent subregions. If this is true, then these
equal and unequal homologous recombinations must result primarily
in noncrossover conversion events, because recombinational
mapping indicates few to no meiotic chromosomal exchanges in the
entire CEN8 region (46, 47).
The degree of LTR sequence identity has been used to
estimate the time of integration of LTR retrotransposons. This
dating method is based on the observation that the nucleotide
sequences of two LTRs of a single LTR retrotransposon are
nearly always identical upon integration (41). However, a possible
methodological problem regarding this dating strategy
would be that LTR retrotransposons in different genomic regions
of an organism could diverge at different rates. Comparative
analysis of a set of genic and intergenic orthologous regions
of Nipponbare (japonica) and GLA4 (indica) found that LTR
retrotransposons exhibit at least a 2-fold higher rate of singlenucleotide
substitution than observed in the synonymous codon
positions in the average cereal gene (31). This higher divergence
rate is predicted to be primarily a result of the greater transition
rate of 5-methyl cytosine compared with cytosine, because most
LTR retrotransposons are extensively 5-methylated at cytosine
residues in all examined cereal species (48).
Several lines of evidence indicate that different classes of DNA
undergo very different rates of sequence evolution in the CEN8
region compared with euchromatic regions. The large number of
intact LTR retrotransposons that are shared by Nipponbare (japonica)
and 93-11 (indica) suggest one process that removes intact
LTR retrotransposons, illegitimate recombination (31, 36), is suppressed,
whereas a second process (unequal homologous recombination)
may actually be enhanced. This apparent enhancement of
unequal homologous recombination appears to be limited to a
small region within the CENH3 domain (10), whereas the low level
of sequence deletion is found throughout the centromeric and
pericentromeric regions. Overall, a slower LTR-retrotransposon
removal process, perhaps combined with an insertion preference
for the chromatin state(s) found in centromeric regions, could
explain the accumulation of transposons that is a consistent and
distinctive feature of most centromeres.
The dating of LTR-retrotransposon insertion via analysis of
LTR divergence (38) uncovered 11 apparent shared insertions
that are predicted to have occurred after the divergence of the
shared Nipponbare and 93-11 ancestor. Although genes in the
CEN8 region show a similar degree of nonsynonymous sequence
divergence, as do euchromatic-region genes, their degree of
synonymous divergence is significantly lower (0.0057 vs. 0.0025).
Many of these observations could be explained by a high rate of
equal and unequal homologous recombination that frequently
leads to conversion of polymorphisms (sequence homogenization)
but rarely yields crossover events. A low frequency of
crossovers might be beneficial, because it would minimize the
rate of variation in centromere structure caused by unequal
exchanges. As it is, centromeres are highly variable, but this
diversity could be much greater if rates of unequal crossovers
between the numerous repeats were not suppressed.
In this study, at least 85% of the LTR retrotransposons in the
CEN8 region were found to be shared between the two japonica
and indica varieties investigated. By contrast, only 50% of LTR
Ma and Bennetzen PNAS January 10, 2006 vol. 103 no. 2 387
GENETICS
retrotransposons were shared in analyzed euchromatic regions of
rice (see Table 4) between these same two indica and japonica
genomes. Recently, an LTR-retrotransposon-rich region in rice,
flanked by two clusters of genes including the Orp locus, was found
to share 50% of its LTR retrotransposons between japonica and
indica (49). This suggests that the presence of a cluster of repeated
DNA, as in the CEN8 and Orp regions, does not explain an
accumulation of older (i.e., shared) LTR retrotransposons. Moreover,
an unusual abundance of solo LTRs is also not a feature
associated with most repeat-rich regions. This is a property of the
CEN8 core but not of the repetitive regions that flank the CEN8
core nor of the repeat cluster near the Orp locus.
Despite an overall lower frequency of homologous exchange
and a lower apparent rate of nucleotide sequence divergence,
many recent DNA rearrangements have dramatically reshaped
the CEN8 region. In addition to the accumulation of LTR
retrotransposons, segmental duplications and large indels have
been a major force responsible for size expansion of the CEN8
region. Numerous ancient and recent segmental duplications
have been documented in several plant and animal species (26,
27, 50­52), but none of these events were demonstrated in
centromeric regions before this study.
It is especially interesting that the segmental duplication of a
cluster of CentO satellite repeats was revealed by phylogenetic
analysis. Subsequently, many indels and reshuffling of a single
monomer or multiple contiguous monomers were identified based
on alignments of satellite repeats or duplicated CentO segments.
Because of their tremendous redundancy, centromeric satellite
repeats can be homogenized by unequal conversion, whereas
variation in copy number and arrangement can be caused by
unequal exchange (53). In this scenario, recent events will obscure
more ancient rearrangements. Hence, it is likely that many more
reshufflings of CentO satellite repeats have occurred than were
deciphered in this study.
We thank Drs. Jiming Jiang and Kiyotaka Nagaki (University of
Wisconsin, Madison) for providing CCR sequences before they were
released by GenBank (accession nos. AY827956­AY828189), Renyi Liu
for assistance in sequence analysis, and Drs. Kelly Dawe and Steve
Henikoff for comments on the manuscript. This research was supported
by National Science Foundation Plant Genome Program Grant 9975618.
1. Jiang, J., Birchler, J. A., Parrott, W. A. & Dawe, R. K. (2003) Trends Plant Sci.
8, 570­575.
2. Lamb, J. C., Theuri, J. & Birchler, J. A. (2004) Genome Biol. 5, 239.
3. Henikoff, S. & Dalal, Y. (2005) Curr. Opin. Genet. Dev. 15, 177­184.
4. Clarke, L. (1990) Trends Genet. 6, 150­154.
5. Clarke, L. (1998) Curr. Opin. Genet. Dev. 8, 212­218.
6. Heslop-Harrison, J. S., Murata, M., Ogura, Y., Schwarzacher, T. & Motoyoshi,
F. (1999) Plant Cell 11, 31­42.
7. Copenhaver, G. P., Nickel, K., Kuromori, T., Benito, M. I., Kaul, S., Lin, X.,
Bevan, M., Murphy, G., Harris, B., Parnell, L. D., et al. (1999) Science 286,
2468­2474.
8. Kumekawa, N., Hosouchi, T., Tsuruoka, H. & Kotani, H. (2001) DNA Res. 8,
285­290.
9. Cheng, Z., Dong, F., Langdon, T., Ouyang, S., Buell, C. R., Gu, M., Blattner,
F. R. & Jiang, J. (2002) Plant Cell 14, 1691­1704.
10. Nagaki, K., Cheng, Z., Ouyang, S., Talbert, P. B., Kim, M., Jones, K. M.,
Henikoff, S., Buell, C. R. & Jiang, J. (2004) Nat. Genet. 36, 138­145.
11. Wu, J., Yamagata, H., Hayashi-Tsugane, M., Hijishita, S., Fujisawa, M.,
Shibata, M., Ito, Y., Nakamura, M., Sakaguchi, M., Yoshihara, R., et al. (2004)
Plant Cell 16, 967­976.
12. Zhang, Y., Huang, Y., Zhang, L., Li, Y., Lu, T., Lu, Y., Feng, Q., Zhao, Q.,
Cheng, Z., Xue, Y., et al. (2004) Nucleic Acids Res. 32, 2023­2030.
13. Ananiev, E. V., Phillips, R. L. & Rines, H. W. (1998) Proc. Natl. Acad. Sci. USA
95, 13073­13078.
14. Jin, W., Melo, J. R., Nagaki, K., Talbert, P. B., Henikoff, S., Dawe, R. K. &
Jiang, J. (2004) Plant Cell 16, 571­581.
15. Sun, X., Le, H. D., Wahlstrom, J. M. & Karpen, G. H. (2003) Genome Res. 13,
182­194.
16. Rudd, M. K., Schueler, M. G. & Willard, H. F. (2003) Cold Spring Harbor Symp.
Quant. Biol. 68, 141­149.
17. Schueler, M. G., Higgins, A. W., Rudd, M. K., Gustashaw, K. & Willard, H. F.
(2001) Science 294, 109­115.
18. Talbert, P. B., Masuelli, R., Tyagi, A. P., Comai, L. & Henikoff, S. (2002) Plant
Cell 14, 1053­1066.
19. Martinez-Zapater, J. M., Estelle, M. A. & Somerville, C. R. (1986) Mol. Gen.
Genet. 204, 417­423.
20. Lee, H. R., Zhang, W., Langdon, T., Jin, W., Yan, H., Cheng, Z. & Jiang, J.
(2005) Proc. Natl. Acad. Sci. USA 102, 11793­11798.
21. Hall, S. E., Kettler, G. & Preuss, D. (2003) Genome Res. 13, 195­205.
22. Miller, J. T., Dong, F., Jackson, S. A., Song, J. & Jiang, J. (1998) Genetics 150,
1615­1623.
23. Presting, G. G., Malysheva, L., Fuchs, J. & Schubert, I. (1998) Plant J. 16,
721­728.
24. Langdon, T., Seago, C., Mende, M., Leggett, M., Thomas, H., Forster, J. W.,
Jones, R. N. & Jenkins, G. (2000) Genetics 156, 313­325.
25. Nagaki, K., Neumann, P., Zhang, D., Ouyang, S., Buell, C. R., Cheng, Z. &
Jiang, J. (2005) Mol. Biol. Evol. 22, 845­855.
26. The Arabidopsis Genome Initiative (2000) Nature 408, 796­815.
27. Venter, J. C., Adams, M. D., Myers, E. W., Li, P. W., Mural, R. J., Sutton, G. G.,
Smith, H. O., Yandell, M., Evans, C. A., Holt, R. A., et al. (2001) Science 291,
1304­1351.
28. McCarthy, E. M. & McDonald, J. F. (2003) Bioinformatics 19, 362­367.
29. Han, B. & Xue, Y. (2003) Curr. Opin. Plant Biol. 7, 134­138.
30. Ma, J., Devos, K. M. & Bennetzen, J. L. (2004) Genome Res. 14, 860­869.
31. Ma, J. & Bennetzen, J. L. (2004) Proc. Natl. Acad. Sci. USA 101, 12404­12410.
32. Thompson, J. D., Gibson, T. J., Plewniak, F., Jeanmougin, F. & Higgins, D. G.
(1997) Nucleic Acids Res. 25, 4876­4882.
33. Tatusova, T. A. & Madden, T. L. (1999) FEMS Microbiol. Lett. 174, 247­250.
34. Bennetzen, J. L., Ma, J. & Devos, K. M. (2005) Ann. Bot. 95, 127­132.
35. SanMiguel, P., Tikhonov, A., Jin, Y.-K., Motchoulskaia, N., Zakharov, D.,
Melake Berhan, A., Springer, P. S., Edwards, K. J., Avramova, Z. & Bennetzen,
J. L. (1996) Science 274, 765­768.
36. Devos, K. M., Brown, J. K. & Bennetzen, J. L. (2002) Genome Res. 12,
1075­1079.
37. McCarthy, E. M., Liu, J., Gao, L. Z. & McDonald, J. F. (2002) Genome Biol.
3, RESEARCH0053.
38. Kumar, A. & Bennetzen, J. B. (1999) Annu. Rev. Genet. 33, 479­532.
39. Yu, J., Hu, S., Wang, J., Wong, G. K., Li, S., Liu, B., Deng, Y., Dai, L., Zhou,
Y., Zhang, X., et al. (2002) Science 296, 79­92.
40. Zhao, W., Wang, J., He, X., Huang, X., Jiao, Y., Dai, M., Wei, S., Fu, J., Chen,
Y., Ren, X., et al. (2004) Nucleic Acids Res.
41. SanMiguel, P., Gaut, B. S., Tikhonov, A., Nakajima, Y. & Bennetzen, J. L.
(1998) Nat. Genet. 20, 43­45.
42. Nei, M. & Gojobori, T. (1986) Mol. Biol. Evol. 3, 418­426.
43. Kumar, S., Tamura, K., Jakobsen, I. B. & Nei, M. (2001) Bioinformatics 17,
1244­1245.
44. Bennetzen, J. L., Coleman, C., Liu, R., Ma, J. & Ramakrishna, W. (2004) Curr.
Opin. Plant Biol. 7, 732­736.
45. Pereira, V. (2004) Genome Biol. 5, R79.
46. Chen, M., Presting, G., Barbazuk, W. B., Goicoechea, J. L., Blackmon, B., Fang,
G., Kim, H., Frisch, D., Yu, Y., Sun, S., et al. (2002) Plant Cell 14, 537­545.
47. Wu, J., Mizuno, H., Hayashi-Tsugane, M., Ito, Y., Chiden, Y., Fujisawa, M.,
Katagiri, S., Saji, S., Yoshiki, S., et al. (2003) Plant J. 36, 720­730.
48. Martienssen, R. A. & Richards, E. J. (1995) Curr. Opin. Genet. Dev. 5, 234­242.
49. Ma, J., SanMiguel, P., Lai, J., Messing, J. & Bennetzen, J. L. (2005) Genetics
170, 1209­1220.
50. International Human Genome Sequencing Consortium (2001) Nature 309,
860­921.
51. She, X., Jiang, Z., Clark, R. A., Liu, G., Cheng, Z., Tuzun, E., Church, D. M.,
Sutton, G., Halpern, A. L. & Eichler, E. E. (2004) Nature 431, 927­930.
52. Yu, J., Wang, J., Lin, W., Li, S., Li, H., Zhou, J., Ni, P., Dong, W., Hu, S., Zeng,
C., et al. (2005) PLoS Biol. 3, e38.
53. Smith, G. P. (1976) Science 191, 528­535.
388 www.pnas.org cgi doi 10.1073 pnas.0509810102 Ma and Bennetzen