The Rice Mitochondrial Genomes and Their Variations

Abstract

Based on highly redundant and high-quality sequences, we assembled rice (Oryza sativa) mitochondrial genomes for two cultivars, 93-11 (an indica variety) and PA64S (an indica-like variety with maternal origin of japonica), which are paternal and maternal strains of an elite superhybrid rice Liang-You-Pei-Jiu (LYP-9), respectively. Following up with a previous analysis on rice chloroplast genomes, we divided mitochondrial sequence variations into two basic categories, intravarietal and intersubspecific. Intravarietal polymorphisms are variations within mitochondrial genomes of an individual variety. Intersubspecific polymorphisms are variations between subspecies among their major genotypes. In this study, we identified 96 single nucleotide polymorphisms (SNPs), 25 indels, and three segmental sequence variations as intersubspecific polymorphisms. A signature sequence fragment unique to indica varieties was confirmed experimentally and found in two wild rice samples, but absent in japonica varieties. The intersubspecific polymorphism rate for mitochondrial genomes is 0.02% for SNPs and 0.006% for indels, nearly 2.5 and 3 times lower than that of their chloroplast counterparts and 21 and 38 times lower than corresponding rates of the rice nuclear genome, respectively. The intravarietal polymorphism rates among analyzed mitochondrial genomes, such as 93-11 and PA64S, are 1.26% and 1.38% for SNPs and 1.13% and 1.09% for indels, respectively. Based on the total number of SNPs between the two mitochondrial genomes, we estimate that the divergence of indica and japonica mitochondrial genomes occurred approximately 45,000 to 250,000 years ago.

A mitochondrion is believed to be a free-living prokaryote, probably a relative of the extant α-proteobacterium; it became an endosymbiotic organelle after being engulfed by a protoeukaryotic cell some 2 billion years ago (Gray et al., 1999; Timmis et al., 2004). A mitochondrion has its own circular genome, encoding a number of mitochondrion-specific components. The genome sizes of mitochondria among living organisms have gone in two opposite directions, shrunk down to 6 to 7 kb in Apicomplexan parasites (Joseph et al., 1989; Vaidya et al., 1989) or even to none as in hydrogenosome and mitosome of basal eukaryotes (Dolezal et al., 2005), and expanded to 200 to 2,000 kb in plants over a long evolutionary time period. The size variation is exceptionally wide among higher plants, ranging from the smallest, 208 kb in white mustard (Brassica hirta; Palmer and Herbon, 1987), to the largest, believed to be over 2,400 kb in muskmelon (Cucumis melo; Ward et al., 1981). Such an extensive expansion is attributable to two major factors, protein-coding redundancy and a high level of mitochondrial DNA (mtDNA) recombination, resulting in extraneous DNA integration (Mackenzie and McIntosh, 1999). Since the first mitochondrial genome was sequenced from a land plant, a liverwort (Marchantia polymorpha; Oda et al., 1992), several of them have been sequenced for higher plants, including four dicots (Arabidopsis [Arabidopsis thaliana; Unseld et al., 1997], sugar beet [Beta vulgaris; Kubo et al., 2000; Satoh et al., 2004], rapeseed [Brassica napus; Handa, 2003], and tobacco [Nicotiana tabacum; Sugiyama et al., 2005]) and two monocots (maize [Zea mays; Clifton et al., 2004] and japonica rice Nipponbare [Notsu et al., 2002]) have been completed. The rice mitochondrial genome sequence from a japonica variety, Nipponbare (referred to hereafter as Nipponbare-N), was published in 2002 (Notsu et al., 2002) and reported to have a length of 490,520 bp. Comparative studies showed that mitochondrion-encoded genes are highly conserved, but their gene orders, genome structures, and genome sizes are highly variable within plants (Gray et al., 1999; Mackenzie and McIntosh, 1999; Knoop, 2004). The sequence evolution of plant mtDNAs appears quite slow compared to that of the nucleus and chloroplasts. The synonymous substitution rate of plant mitochondrial genes is a few fold lower than chloroplast genes, 10- to 20-fold lower than nuclear genes in both plants and mammals, and 50- to 100-fold lower than mammalian mitochondrial genes (Muse, 2000; Cho et al., 2004).

There are two general approaches applied in large-scale genome sequencing projects. One is a clone-by-clone sequencing method, where a clone-based physical map is constructed first, followed by clone sequencing as applied successfully in the Nematode and yeast (Saccharomyces cerevisiae) sequencing projects (Olson, 2001). The other is a whole-genome-shotgun (WGS) method that was used for Drosophila (Myers et al., 2000), human (Venter et al., 2001), and rice (Goff et al., 2002; Yu et al., 2002, 2005) genome projects. In a WGS approach, raw sequencing data contain large amounts of organellar genome sequences, which can be extracted and assembled independently when sequence redundancy is extremely high and complications of gene transfer between nuclear and organellar genomes are handled with caution (Goff et al., 2002; Yu et al., 2002, 2005). In this article, we describe our assemblies using sequence data from WGS projects and analyses of three rice mitochondrial genomes. Two of them were acquired from our Chinese Superhybrid Rice Genome Project, which has produced the genome sequences 93-11 (the paternal variety) and PA64S (the maternal variety), the parental cultivars of Liang-You-Pei-Jiu (LYP-9; Yi and Xiao, 2000). 93-11 is a typical indica variety, despite the fact that it has gone through a rather complex breeding process for many agriculturally favorable traits. PA64S has a composite background, having incorporated genetic materials from all three major cultivated rice subspecies, indica, japonica, and javanica, but its mitochondria are maternally inherited from a japonica ancestor (L. Yuan, personal communication). Another mitochondrial genome of a japonica variety was also reassembled based on the sequences extracted from raw data of the Nipponbare genome (referred to as Nipponbare-S), which were released by Syngenta. We compared our mitochondrial genome assemblies in intravarietal (within cultivars or varieties) and subspecific (among subspecies) ways and verified some of the characteristic variations in selected landraces as well as wild rice samples. Our data are publicly available in GenBank (accession nos. DQ167399 for the 93-11 assembly, DQ167807 for the PA64S assembly, and DQ167400 for the Nipponbare-S assembly) and at our own Web site (http://www.genomics.org.cn/bgi_new/projects/rice/rice05_08_17.htm).

RESULTS

Sequence Assemblies

The two mitochondrial genomes were assembled from 44,045 sequence reads (with an average length of 575 bp at a quality value of Q20 and equivalent to 52-fold coverage of the genome) of the 93-11 project and 49,637 sequence reads (with an average length of 626 bp at a quality value of Q20 and equivalent to 63-fold coverage of the genome) of the PA64S project; both projects were carried out in our institute. The structure and organization of mtDNAs in higher plants are complicated by a high recombination rate in their repeat regions, resulting in characteristic multipartite structures. The general genomic information of a mitochondrial genome is traditionally represented as a master circle based on restriction mapping (Fauron et al., 1995; Sugiyama et al., 2005). Our two genome assemblies were displayed as a master contig (Fig. 1

Figure 1.

Circular representation of the 93-11 mitochondrial genome. Circles display (from outside): (1) physical map scaled in kilobase pairs; (2) chloroplast-derived regions (green boxes) and repeats (over 5 kb; repeated copies are depicted in similar color boxes); (3) coding sequences transcribed in the clockwise direction; (4) coding sequences transcribed in the counterclockwise direction; (5) GC content variations (in a 1-kb sliding window and 100-bp increments; greater than 43.8% are in red and smaller in blue); (6) minor genotype frequencies (MF) of intravarietal indels (minor genotype over 2 at a locus; indels are marked in green and magenta lines, respectively; the gray circle represents intravarietal indels with MF = 25%; black and red dots represent indels between 93-11 and PA64S); and (7) minor genotype frequency of intravarietal SNPs (minor genotypes over 2 at a locus; brown lines depict loci with mutations occurring at the same site independent from minor genotypes; the gray circle marks intravarietal SNPs with MF = 25%; black and red dots represent transversions and transitions, respectively, when compared to PA64S).

) by referencing previously published sequences (Iwahashi et al., 1992; Notsu et al., 2002). The smaller contigs with either low coverage (i.e. the equivalent nuclear genome coverage is <4 for the dataset) or less than perfect identity (i.e. <100% over 500 bp and <99% over the entire length) to the master contig are excluded from our analysis. A 491,515-bp mtDNA genome was assembled for 93-11 and a 490,673-bp genome for PA64S. The length of the PA64S mitochondrial genome is 842 bp shorter than that of 93-11, but 153 bp longer than that of Nipponbare-N. Similarly, we also reassembled another mitochondrial genome from the Nipponbare genome sequence released by Syngenta; it consists of 56,151 sequence reads (nearly a 52-fold coverage to the genome) and 490,669 bp in length. It is quite different from a previously reported sequence, namely, Nipponbare-N (Notsu et al., 2002), in which 497 single nucleotide polymorphisms (SNPs) and 185 indels (113 insertions with an accumulated length of 267 bp and 72 deletions with a total length of 118 bp) were found when compared with Nipponbare-S. It is difficult to evaluate these discrepancies between Nipponbare-N and other various assemblies because raw sequence data with quality scores are not available from the project, so we focused our analysis on the three newly acquired assemblies. The differences among the various assemblies are summarized in Supplemental Table I. Two other variable regions in the genome are repeat regions, with a length range from less than 30 bp to over 47 kb (five repeat segments are larger than 5 kb), and several smaller contigs that form smaller circles (Supplemental Fig. 1). These smaller assemblies are believed to be subgenomic products of rice mitochondria, which coexist with their full-length versions and arose from the frequent recombination events (Fauron et al., 1995; Sugiyama et al., 2005). The sequence of these subgenomes does not usually interfere with sequence analysis but is supported strongly by multiple joint sequences that prove their existence.

Sequence Polymorphisms

Mitochondria, together with chloroplasts of higher plants, are believed to be clonal and maternally inherited and have their organelle-specific replication and DNA repair systems. A plant cell has multiple mitochondria that can be regarded as a population when genetic heterogeneity among their genomes is investigated, leveraging on the enormous coverage from WGS data in which each aligned base possesses a quality value for an overall quality assessment. The variations within a variety identified among the sequencing reads are separated into intravarietal major and minor genotypes. Likewise, variations between two subspecies are defined as intersubspecific genotypes (Table I

Table I.

Minor genotype frequencies of each SNP type among three rice mitochondrial genome assemblies

SNP Typea	MFb in 93-11	MF in PA64S	MF in Nipponbare-S
		%
A/C	1.23 (10,117)	1.03 (7,278)	0.63 (3,252)
A/G	1.08 (6,898)	0.85 (4,472)	0.72 (3,878)
A/T	1.14 (5,806)	0.92 (2,580)	0.66 (4,903)
C/A	1.17 (3,614)	0.93 (1,586)	0.79 (3,762)
C/G	1.19 (5,185)	1.02 (3,761)	0.66 (1,864)
C/T	1.03 (2,597)	0.79 (1,183)	0.84 (4,217)
G/A	0.98 (2,720)	0.79 (1,241)	0.81 (3,944)
G/C	1.14 (5,238)	0.95 (3,874)	0.59 (1,862)
G/T	1.14 (3,629)	0.85 (1,653)	0.80 (3,876)
T/A	1.21 (5,591)	0.91 (2,683)	0.70 (4,820)
T/C	1.16 (6,849)	0.90 (4,596)	0.71 (4,103)
T/G	1.26 (10,460)	1.06 (7,338)	0.66 (3,363)

a

SNP types of the major and the minor genotypes are separated with an oblique line (/).

b

Minor genotype frequencies (MF) were calculated over total sites of each SNP type, and the numbers in parentheses indicate the total number of surveyed sequences covering a SNP.

). Furthermore, these variations can also be verified experimentally because some of the high-frequency variations among varieties and subspecies are usually segregated depending on the size of tested populations (Tang et al., 2004). Among the three mtDNA assemblies, we also detected a small amount of intravarietal polymorphisms (19 SNPs and four indels) between PA64S and Nipponbare-S, representing a diminutive divergence from its common japonica ancestor. The discovered intravarietal and intersubspecific variations (including SNPs and indels) are summarized in Figure 1 and Supplemental Tables I and II.

SNPs

Between the 93-11 and PA64S mitochondrial genomes, we identified 96 SNPs, including 51 transitions and 45 transversions (Supplemental Table I), counted as a polymorphism rate of two in 10,000 bases, which is about 2.5 times lower than that of their chloroplast counterparts (Tang et al., 2004) and about 21 times lower than that of their nuclear counterparts (Yu et al., 2002). Sixty SNPs were localized at five sites in repeat regions, slightly higher than those found in single-copy regions. We also detected some intersubspecific SNPs between 93-11 and Nipponbare-S/Nipponbare-N, as well as those shared by them. Frequencies of the major and minor genotypes for each intersubspecific SNP are summarized in Table I and Supplemental Table III.

The intersubspecific SNPs between 93-11 and PA64S are predominantly found in noncoding regions. Only five of them are in coding regions (three of five are at the third codon positions), occurring at a rate of 0.1 SNP/kb, 2-fold lower than the mitochondrial genome average. These rates are 0.3 and 3.0 SNPs/kb in the chloroplast (Tang et al., 2004) and the nuclear coding regions (Yu et al., 2002, 2005), respectively. Two nonsynonymous SNPs are in the cox3 gene (nucleotide location 112,573 of the 93-11 assembly and with a codon change of TTA^Leu to TCA^Ser) and open reading frame (ORF) 224 (nucleotide location 183,090 of the 93-11 assembly and with a codon change of AAA^Lys to CAA^Gln), with major genotype frequencies of 94% and 85%, respectively (Table II

Table II.

Intersubspecific SNPs in coding regions between 93-11 and PA64S

cox3, Cytochrome oxidase subunit 3.

Location in 93-11a	Codon Variation in 93-11/PA64Sb	Amino Acid Variation in 93-11/PA64Sb	Gene
112,308	ATT/ATC	I	cox3
112,497	TAT/TAC	Y	cox3
112,573	TTA/TCA	L/S	cox3
182,959	GGC/GGA	S	ORF224
183,090	AAA/CAA	K/Q	ORF224

a

Locations of polymorphic sites between 93-11 and PA64S are referenced to the 93-11 assembly.

b

Oblique line (/) separates the corresponding variations of codons and amino acids in the involved genes between 93-11 and PA64S. Amino acids are indicated by capital letters corresponding to the one-letter code.

; Supplemental Table III). The functional implication of such changes remains to be elucidated experimentally.

Intravarietal SNPs were identified within the two japonica varieties, PA64S and Nipponbare-S. Almost all minor genotypes discovered in the sequence assemblies are confirmed as major genotypes in 93-11 or vice versa. Some of these characteristic genotypes are unique to indica or japonica and can be used as subspecific markers in distinguishing the two subspecies. Some of the intravarietal SNPs are not easily verified in their counterparts, making them unique to each variety. In almost all cases, the number of intravarietal SNPs (at a rate of 1.26% for 93-11 and 1.38% for PA64S) is larger than that of the intersubspecific SNPs (a rate of 0.02% for 93-11 to PA64S comparison). The result suggests that not all intravarietal variations have been fixed and inherited stably among cultivars of the same variety (Fig. 2A

Figure 2.

Comparisons of minor genotype frequency (MF) of intersubspecific and intravarietal SNPs in the rice mitochondrial genomes. A, Comparisons of intersubspecific and intravarietal SNPs in rice mitochondrial genomes. Intersubspecific SNPs (white bars) were the variations identified from the 93-11 to Nipponbare-S comparison. Intravarietal SNPs were those found within 93-11 (gray bars) and Nipponbare-S (black bars). The oblique line (/) separates major and minor genotypes in intravarietal SNPs and the major genotype in intersubspecific SNPs, respectively. B, MF of each SNP type in the mitochondrial genome. Data are from Table I. SNP types were arranged on the x axis. SNPs in rice varieties were indicated with different symbols: triangles for 93-11, squares for PA64S, and diamonds for Nipponbare-S. C, MF of each SNP type in the rice chloroplast genome (Tang et al., 2004).

). We also noticed several intravarietal mutation hotspots. For instance, in the 93-11 assembly, we identified 33 sites having multiple minor genotypes and 13 of them with collective minor genotype frequencies higher than 40%. Our data indicated that these sites are highly variable within mitochondrial populations of a variety, and they are preferentially located in the regions with GC content lower than its genome average (43.8%).

Sequence variations originate from two primary sources; errors occurred in DNA replication and repair when recombination is regarded as one of the repair mechanisms. Since mtDNA is replicated and repaired through its specific systems, we also investigated the trends or rules of the variations collected in this study. The parameter is the neighboring-nucleotide effect (NNE), where relative abundances (or biases) of nucleotides positioned upstream (or 5′) and downstream (or 3′) of a SNP were calculated for the intravarietal categories that include 8,690 SNPs. Reciprocal SNPs, such as A/G and G/A types, were combined into a single category because their directions are not distinguishable. Together, we have 1,653 A/C, 1,608 G/T, 1,424 A/G, 1,549 C/T, 1,287 A/T, and 1,169 C/G SNPs. First, the symmetrical pairs (A/C and G/T, A/G and C/T, and A/T and C/G) showed similar biases. Second, we found that T at the 5′ side and A at the 3′ side of a SNP site exhibit the highest positive bias. Similarly, G and T flanking a SNP site show the highest negative bias (Supplemental Fig. 2A). Third, for transitional variations, the 5′-flanking nucleotides are biased toward pyrimidines and the 3′ nucleotides are biased toward purines (C and A for A/G type, T and G for C/T type); the rule is not followed in transversions (Supplemental Fig. 2, B and C). Further examination on chloroplast and nuclear genomes indicates that the former is very similar to its mitochondrial counterpart in the transitional variations, but the latter is not.

The minor genotype frequencies in the three mitochondrial assemblies do not show any obvious trend and neither do the chloroplast counterparts (Fig. 2, B and C) among those of indica and japonica cultivars. One potentially important observation is that the minor genotype frequency in all SNP types of 93-11 is much higher than that of PA64S and Nipponbare-S (Table I; Fig. 2B). Although unable to refer to other indica mitochondrial genomes, we assume that indica may have more polymorphic organellar populations than japonica, consistent with an evolutionary history of the indica subspecies and an early observation on the rice chloroplast genomes (Tang et al., 2004).

Indels and Segmental Sequence Variations

Indels and segmental sequence variations (SSVs) in mitochondrial genomes are rather limited in numbers. Only two indels with a size of 2 bp each were found in the PA64S to Nipponbare-S intravarietal comparison. The intersubspecific comparison, 93-11 to PA64S, yielded 25 indels and three SSVs, with a collection of 842 bp and a rate of 0.006% in 93-11 mtDNA (Table III

Table III.

Genotypes of indels and SSVs among rice mitochondrial genome assemblies

Indels and SSVsa	Locus in 93-11b	Sequencec	F in 93-11d	MF in 93-11d	MF in PA64Sd	MF in Nipponbare-Sd
			%
SSV-500/6	19,054	TTT…TTG (500 bp)/GATCTC	87 (15)	13 (15)	0 (45)	0 (36)
I-1	30,846	A	95 (84)	5 (84)	1 (73)	0 (81)
D-9	65,598	CTCCACCCC	100 (32)	0 (32)	0 (39)	0 (39)
I-2	76,336	AA	94 (36)	6 (36)	20 (49)	34 (47)
D-8	76,516	CAAATTTA	94 (33)	6 (33)	0 (48)	0 (44)
D-1	90,662	C	100 (29)	0 (29)	39 (31)	36 (22)
SSV-500/6	114,114	TTT…TTG (500 bp)/GATCTC	87 (15)	13 (15)	0 (45)	0 (36)
I-1	125,906	A	94 (86)	6 (86)	30 (76)	27 (81)
D-8	182,342	GTCTAATT	95 (37)	5 (37)	0 (81)	0 (55)
D-2	200,488	GG	100 (25)	0 (25)	27 (22)	0 (13)
I-1	201,781	G	100 (42)	0 (42)	5 (22)	11 (35)
SSV-39/178	207,132	CTGGCGCCATTCTCGGACCTAGCAGCAGATGGGCGGGCC/TCT…GAA (178 bp)	100 (53)	0 (53)	27 (26)	11 (28)
I-9	223,767	TAGAGGCGG	97 (114)	3 (114)	5 (79)	6 (80)
D-8	224,223	CGTTGCGC	95 (77)	5 (77)	0 (46)	0 (53)
D-9	242,747	CTCCTGACC	97 (34)	3 (34)	0 (38)	0 (35)
D-9	256,168	CTTGACTCG	100 (28)	0 (28)	12 (41)	15 (41)
I-9	282,887	GGTAGAGGC	97 (123)	3 (123)	4 (80)	4 (82)
D-8	283,343	GCCGTTGC	86 (73)	14 (73)	0 (57)	1 (72)
D-1	308,242	A	59 (61)	41 (61)	0 (38)	0 (29)
D-1	308,253	A	57 (61)	43 (61)	0 (37)	0 (29)
I-22	316,378	CC	70 (59)	30 (59)	3 (34)	0 (19)
I-8	342,328	TTTACCTT	98 (42)	2 (42)	5 (63)	8 (36)
I-9	349,619	AGGTTTCTC	92 (61)	8 (61)	2 (51)	4 (55)
D-3	391,519	CCT	96 (25)	4 (25)	0 (64)	0 (50)
I-9	413,841	GGTAGAGGC	97 (123)	3 (123)	4 (80)	4 (82)
D-8	414,297	GCCGTTGC	86 (73)	14 (73)	0 (57)	1 (72)
I-8	423,441	TTTACCTT	98 (42)	2 (42)	5 (63)	8 (36)
I-9	430,732	AGGTTTCTC	92 (61)	8 (61)	2 (51)	4 (55)

a

I or D, Insertion or deletion of a major genotype at a given polymorphic site, respectively. The number following I or D, joined with hyphens, indicates the indel length in base pairs. Lengths of the segments in 93-11 and PA64S are separated by an oblique line (/).

b

All indels and SSVs between 93-11 and PA64S are referenced to the 93-11 assembly.

c

The nucleotide sequences of a major genotype at a polymorphic site between 93-11 and PA64S. The long sequences were indicated in length.

d

F and MF, Major and minor genotype frequencies. Numbers in parentheses are total numbers of the surveyed sequences of a SNP site.

); it is nearly 3 times lower than its chloroplast counterpart (a rate of 0.02%; Tang et al., 2004) and 38 times lower than its nuclear counterpart (a genome average of 0.23%; Yu et al., 2002), as well as about one-third of its intersubspecific SNP rate (0.02%). The majority of these intersubspecific indels are in regions with lower GC content under the genome average (43.8%), and some of them (three insertions and six deletions) have minor genotype frequencies higher than 40%.

Some of the intersubspecific indels can serve as markers for discriminating indica from japonica. For instance, two variable sites, detected with high minor genotype frequencies, 41% of D-1 (A) at nucleotide position 308,242 (with 61 reads overlapping) and 43% of D-1 (A) at nucleotide position 308,253 (supported with 61 reads) in the 93-11 assembly, are absent in both PA64S and Nipponbare-S. Another case is the 39-bp SSV (SSV-39/178) found uniquely in 93-11 (at nucleotide position 207,132), and it becomes the 178-bp SSV, a major genotype in both PA64S (73%) and Nipponbare-S (88%). Other examples are SSV-500/6, a 500-bp segment unique to 93-11, and a 6-bp sequence (GATCTC) characteristic of PA64S and Nipponbare-S (Table III). The 500-bp nucleotide segment was also found in a rice variety IR36 of indica cultivars (GenBank deposit M74241; Liu et al., 1992).

Although the sequence variations are guaranteed by high redundancy, with more than 50-fold coverage of the mitochondrial genome length, we did go on to verify a few of them in both the original cultivars and some selected indica and japonica varieties. We amplified and sequenced two intersubspecific SSVs, SSV-500/6 and SSV-39/178, and an indel D-8 (CAAATTTA, nucleotide position 76,516 in the 93-11 assembly; Fig. 3

Figure 3.

Sequence alignments of intersubspecific mitochondrial variations. Sequences of SSV-500/6 (A), SSV-39/178 (B), and D-8 (C) were aligned. Sequence variations of 93-11, PA64S, and Nipponbare-S were recovered from PCR-amplified fragments. Inverted repeats are highlighted with arrowed lines.

), and further surveyed these polymorphisms among 10 other indica and japonica cultivars. The result demonstrates that SSV-500/6 and SSV-39/178 are common variations between indica and japonica subspecies, and they appeared genetically linked as one haplotype among the tested samples (Fig. 4

Figure 4.

Sequence analysis of the intersubspecific variations among cultivated rice and their wild ancestors. The rice varieties used in C and D are indica 2H249 (lane 1), IR64 (lane 7), SHZ (lane 9), and 93-11 (lane 10); japonica Zhonghua10 (lane 4), AZU (lane 6), LTH (lane 8), and PA64S (lane 11); an African cultivar Oryza glaberrima, as an outgroup (lane 2); and wild rice Oryza nivara (lane 3) and Oryza rufipogon (lane 5). The rice varieties used in E are indica 4A 436 (row 1), 4A 422 (row 3), II32 A (row 5), 4A 424 (row 9), and Xie A (row 11); 93-11 (row 12); and japonica 4A 434 (row 2), 4A 420 (row 4), 4A 430 (row 6), 4A PA64S (row 7), 4A 426 (row 8), and 4A 418 (row 10). A and B, Schematic maps of SSV-500/6 and SSV-39/178 on 93-11 and PA64S, respectively. Solid horizontal bars represent intersubspecific variations and genes around the variation site. Black and white arrows indicate PCR primers and the direction of gene transcription, respectively. C, PCR amplification of SSV-500/6 with primers F1 and R1. D, PCR amplification of SSV-39/178 with primers F2 and R2. E, Multiple sequence alignments of sequenced PCR products around D-8, amplified with primers F3 and R3, among different rice varieties. Capital letters I and J in parentheses depict the varieties from indica and japonica subspecies, respectively. The thirteenth and fourteenth rows are corresponding sequences of Nipponbare-S and Nipponbare-N, respectively. Dashed lines indicate nucleotide deletions; inverted repeats are underlined with arrows.

). However, the indel D-8 is neither consistent in indica nor in japonica varieties, which is believed to be either unstable or to occur sporadically among rice mitochondrial genomes.

DISCUSSION

An Overview of Intersubspecific and Intravarietal Variations

In this study, we assembled three rice mitochondrial genomes, performed a careful comparison to identify intravarietal and intersubspecific polymorphisms, and verified some of them experimentally. Our findings documented that rice mitochondrial genomes vary significantly between subspecific cultivars and within individual varieties. Of the detectable intravarietal sequence variations, the frequencies of minor genotypes ranged from a few percent to a few tens of percent. A total of 13 polymorphic sites have minor genotype frequencies higher than 40% in 93-11; 31 and 11 of such sites were found in PA64S and Nipponbare-S, respectively. A total of 34, 42, and 28 polymorphic sites with genotype frequencies less than 10% were detected in the 93-11, PA64S, and Nipponbare-S, respectively. The major intravarietal genotypes were used to search for intersubspecific polymorphisms. Almost all the intersubspecific variations were confirmed among intravarietal variations, and the result showed that minor genotypes in one subspecies are usually major genotypes in other subspecies or vice versa, suggesting that these minor genotypes are actually variations present among mitochondria either as a subpopulation or dominant genotypes in a variety. Therefore, the minor and major genotype frequencies at polymorphic sites within a variety provide a useful statistical basis for investigating sequence variations among mitochondrial and chloroplast genomes in the context of evolution and inheritance. Functional implications of the two nonsynonymous sequence variations (TTA^Leu to TCA^Ser in the cox3 gene and AAA^Lys to CAA^Gln in ORF224) discovered between 93-11 and PA64S remain to be elucidated, but they are not likely related to hybrid vigor since the differences are shared by most, if not all, indica and japonica varieties.

The NNE of each intravarietal SNP type in the 93-11 assembly shows stronger biases than the other two assemblies, largely due to its higher numbers of total variations in several categories. The biases represent a sequence signature, where the immediate neighboring sites are T > C > a > g flanking the 5′ side and A > G > c > t flanking the 3′ side (capital and lowercase letters depict higher and lower-than-average values, respectively). A Y (pyrimidine)-N (SNP site)-R (purine) signature seems to exist in most of the SNP types, especially for transitions in mtDNA. Despite the fact that the trend of mitochondrial NNE was found to be similar to what chloroplast genomes have, the Y-N-R signature was more obvious in transversions found in noncoding regions of chloroplast genomes among grasses (Poaceae; Morton et al., 1997). The similarity in SNP patterns between rice mitochondrial and chloroplast genomes implies the existence of similar genome replication and DNA repair systems. Judging by results from analyzing intersubspecific indels and SSVs in the assemblies, we believe that nucleotide variations in the organellar genes are prevalent in noncoding and GC-poor sequences, which are not likely selected by functions and may have been drifting, displaying little directionality other than sequence biases that are signatures of organellar DNA polymerases involved in genome replication and DNA repair.

Multipartite Structures of the Mitochondrial Genome

It is known that mitochondrial genomes of higher plants are organized in a more complex way than their animal counterparts. They have two remarkable features, larger genome size and frequent homologous recombination, which often result in significant sequence diversification among plant taxa. Their repetitive sequences encourage intramolecular recombinations generating both isomeric forms of the master chromosome and smaller circular derivatives (subgenomic DNA; Fauron et al., 1995). In the 93-11 mitochondrial assembly, we found 120 direct and 149 inverted-repeat regions with 290 kb in collective length from repeat units of less than 29 bp to over 47 kb. We reconstructed 10 smaller subgenomic circles electronically using five sets of direct-repeat regions larger than 5 kb, and two isomeric circles using a set of inverted repeats (1.3 kb in total length). To what extent these circular molecules exist in vivo is still controversial, but techniques are available to do so, such as restriction fragment mapping and electron microscopy-based methods (Sugiyama et al., 2005). Nevertheless, it has been observed that ectopic recombination of plant mitochondrial genomes causes chimeric and cryptic genes and ORFs, resulting in cytoplasmic male sterility, plant variegation, and other aberrant phenotypes (Mackenzie and McIntosh, 1999; Abdelnoor et al., 2003).

In plants, the copy number of recombination-derived isomeric mtDNA and subgenomic DNA molecules is controlled by nuclear genes. Studies showed that these multipartite structures in Arabidopsis mitochondria, such as substoichiometric shifting, are influenced by a gene homologous to MutS that was postulated to be involved in mismatch repair and recombination (Abdelnoor et al., 2003). A very recent report pointed out that the P2 nuclear genotype of maize, which was considered a natural mutagenesis system for maize mitochondria, dramatically accelerates mitochondrial genomic divergence by increasing low copy number subgenomes and amplifying aberrant recombination products (Kuzmin et al., 2005).

Subgenomic and isomeric circles of mtDNA inevitably accelerate the evolution of dynamic structures of plant mitochondria; consequently, this phenomenon can offer advantages for phylogenetic analysis of deeper branches in phylogeny (Knoop, 2004). However, the sequence variations in coding regions of the plant mtDNA are generally very low (Wolfe et al., 1987), in contrast to evolutionary rates of mtDNA structure and gene order. Because most of the evolutionary and phylogenetic studies are usually focused on sequences of a few proteins of the respiratory chain, such as apocytochrome b (Cob) and/or cytochrome oxidase subunits 1 to 3 (Cox1, Cox2, and Cox3; Gray et al., 1999), we believe that the multipartite structure of plant mitochondrial genomes has little influence on the phylogeny of major branches.

Divergence Time of Mitochondrial Genomes between 93-11 (indica) and PA64S (japonica)

Mutation rates among organellar and nuclear genomes in rice are quite different. Sequence polymorphism rates in mitochondrial genomes between 93-11 and PA64S are 0.02% for SNPs and 0.006% for indels, nearly 2.5 and 4 times (0.05% and 0.02%; Tang et al., 2004) lower than those of chloroplast genomes, respectively. These rates, when compared to corresponding nuclear genomes, are orders of magnitude lower, 20 and 40 times, respectively (0.43% and 0.23%; Yu et al., 2002). These numbers are much lower than a previous estimate that the mitochondrial and chloroplast DNA evolve at least 5 times more slowly than and at half-times of the rate of the nuclear DNA, respectively (Wolfe et al., 1987). Fitting the actual numbers from nucleotide substitution rates in a per-site scale and a per-synonymous site per year (r = 0.2 to approximately 1.1 × 10⁻⁹) between 93-11 and PA64S (d_AB = 0.0001; Wolfe et al., 1987; Muse, 2000), we deduced the divergence time window (T = d_AB/2r) between japonica and indica, about 45,000 to 250,000 years ago. This range is overlapping with an estimate from chloroplast genome data, about 86,000 to 200,000 years ago, as opposed to 1,000,000 years ago based on the nuclear genomes (Tang et al., 2004).

MATERIALS AND METHODS

Sequence Assembly and Analysis

Mitochondrial sequence reads from rice (Oryza sativa) 93-11 and PA64S were extracted from our WGS sequence repository (sequences with a consecutive Q20 base pair length >50 bp; http://www.genomics.org.cn; Yu et al., 2002, 2005), according to their identities to known rice mitochondrial genome sequences with expectation values over 1e-100. The sequencing reads used for Nipponbare-S assembly were screened from raw data released by Syngenta (http://www.tmri.org; Goff et al., 2002). Sequence contigs were assembled using the Phred/Phrap/Consed package (http://www.phrap.org; Gordon et al., 1998).

Intersubspecific polymorphisms (indels and SNPs) were identified based on comparisons of the mitochondrial assemblies within and between varieties. The results were acquired by a custom-designed program and confirmed through careful visual inspection. We chose only intravarietal indels and SNPs that have more than two confirmed variations at a polymorphic site for further investigation. The major genotype (F) and the minor genotype frequencies (MF) were calculated according to the following formulas:

\[F{=}\frac{RN_{\mathrm{mj}}}{RN_{\mathrm{total}}}{\times}100\%{\,}\mathrm{and}{\,}MF{=}\frac{RN_{\mathrm{mn}}}{RN_{\mathrm{total}}}{\times}100,\]

where RN_mj is the number of sequences contributing the major genotype, RN_mn is the number sequences contributing to the minor genotype, and RN_total stands for the total number of sequences covering a polymorphic site.

To evaluate the NNEs of SNPs, we scored SNPs in each category (or SNP types). We marked flanking sequences centered on a SNP site; the 5′ side and the 3′ side are labeled as negative and positive numbers. For instance, −1 represents the 5′ immediate nucleotide of a polymorphic site, and +1 represents the 3′ immediate nucleotide of the polymorphic site. The proportion (f_i) of each nucleotide was calculated with the following method (Zhao and Boerwinkle, 2002):

\[f_{\mathrm{i}}{=}\frac{n_{\mathrm{i}}}{N}{\times}100\%,\]

where i is the position of nucleotides flanking a SNP site, ranging from 1 to 5 (from −5 to −1 relative to the polymorphic site); n_i is the number of each nucleotide; and N is the total number of the four nucleotides at the site. To reveal the nucleotide bias (b_i), we normalized f_i with respect to the averaged nucleotide proportion of the 50 nucleotides at the 5′ side and the 50 nucleotides at the 3′ side, which flank the same polymorphic site:

\[b_{\mathrm{i}}{=}f_{\mathrm{i}}{-}\frac{n_{100}}{N_{100}}{\times}100\%,\]

where n₁₀₀ is the score of A, C, G, or T of the 50 5′-flanking nucleotides or 3′-flanking nucleotides; and N₁₀₀ is the total score of four nucleotides in the same context. All our analyses were performed using scripts written in Perl or C; these scripts are available upon request for noncommercial purposes.

Rice Materials

93-11 is a typical indica cultivar bred in the Jiangsu Provincial Academy of Agricultural Sciences (Dai et al., 1997). PA64S is a photoperiod- and temperature-sensitive male sterile cultivar, bred in the National Hybrid Rice Research and Development Center (NHRRDC), whose maternal line is Nong-Ken-58, a japonica cultivar (Yi and Xiao, 2000). Both 93-11 and PA64S were provided by Professor Longping Yuan (NHRRDC).

Other sources of rice varieties used in our experiments are as follows: 2H249 (indica), Oryza glaberrima (an African cultivated species; t0999/90), Oryza nivara (wild rice; 103824), Zhonghua10 (japonica), and Oryza rufipogon (wild rice; 105720), provided by Professor Song Ge at the Institute of Botany, Chinese Academy of Sciences (CAS); AZU (japonica), IR64 (indica), LTH (japonica), and SHZ (indica), provided by Dr. Hei Leung and Dr. Richard Bruskiewich at the International Rice Research Institute (IRRI); and indica varieties 4A 422, 4A 424, 4A 436, II32 A, and Xie A as well as japonica varieties 4A 418, 4A 420, 4A 426, 4A 430, and 4A 434, provided by Professor Qingzhong Xue at the College of Agriculture and Biotechnology of Zhejiang University.

DNA Extraction, PCR Amplification, and Sequencing

Fresh leaves from a single rice plant in each variety were collected and preserved in liquid nitrogen. Total DNA was extracted according to the cetyltrimethyl ammonium bromide method (Rogers and Bendich, 1998). In PCR amplifications, we used primers F1 (5′-TCTTCTTCGGACTTGATGCAC-3′) and R1 (5′-GCGCCCTTGAAATCATCTTA-3′) for the SSV-500/6; primers F2 (5′-GTAATAGTGGGCGGGTCTCA-3′) and R2 (5′-TCCGGGTTCCGATCTATATG-3′) for the SSV-39/178; and primers F3 (5′-AACCAGAGCCGTTAGCAGAG-3′) and R3 (5′-TTCTCAACTGCTGCCACAAG-3′) for the indel D-8 (nucleotide 76,516 of the 93-11 mitochondrial assembly). PCR reactions were performed in a final volume of 30 μL containing 1 unit of AmpliTaq polymerase, 200 μm dNTPs, 10× PCR buffer (ABI), and 10 pmol primers. The amplified DNA fragments were sequenced using an ABI-3730 DNA sequencer.

Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers DQ167399, DQ167400, and DQ167807.

ACKNOWLEDGMENTS

We thank our colleague Dr. Hui Zhao for providing SNP information of the rice 93-11 nuclear genome. We are especially grateful to Professor Qingzhong Xue, Zhejiang University; Dr. Hei Leung and Dr. Richard Bruskiewich, the International Rice Research Institute; and Professor Song Ge, the Institute of Botany, Chinese Academy of Sciences, for kindly providing some of the rice materials.

Author notes

1

This work was supported by grants from the Chinese Academy of Science (CAS; KSCX1–SW–03), the Ministry of Science and Technology (2004AA231050 and 2005AA235110), and the CAS Hundred Talents Program (to J.Y.).

2

These authors contributed equally to the paper.

^*

Corresponding author; e-mail ; fax [phone].

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (www.plantphysiol.org) is: Jun Yu ().

[W]

The online version of this article contains Web-only data.

www.plantphysiol.org/cgi/doi/10.1104/pp.105.070060.