Generation and evolutionary fate of insertions of organelle DNA in the nuclear genomes of flowering plants

The rice NUPTs and

The insertions and consist of 19 and 13 clustered NUPTs, respectively (Fig. 2; Table 1). While includes five segments of ptDNA that are larger than 10 kb and represents most of the plastid chromosome, is made up of smaller NUPTs (<3.3 kb) derived from disparate regions of the plastome. Unlike the long continuous insertions described in the previous section, no large-scale colinearity between or and the ptDNA was evident; in both loci, the order and orientation of the NUPTs were random. Despite its highly fragmented state, the overall similarity of to ptDNA was higher than that of the long continuous insertion (Table 2). Sequences derived from the IRs were overrepresented in and (37.9%, 41.5%, and 30.9% for , , and ptDNA, respectively). InDels were also found in and ; the largest, , is ∼3.1 kb long and disrupts a NUPT derived from the IR region (Fig. 2; Table 2; Supplemental Table 1).

The 271/620-kb mtDNA insertion in A. thaliana

A 270-kb stretch of mtDNA sequence has previously been identified on chromosome 2 of A. thaliana (Lin et al. 1999). Portions of the mitochondrial chromosome initially believed to be absent from this insert were subsequently identified, and its actual size is now estimated to be ∼620 kb (Stupar et al. 2001). Because the sequenced 270-kb stretch was markedly rearranged with respect to the published mtDNA sequence (Unseld et al. 1997), it was thought to have originated from an alternative, recombination-induced, configuration of the mitochondrial chromosome (Lin et al. 1999; Stupar et al. 2001).

In Figure 3, A and B, the structure of the sequenced 270-kb stretch () is shown. As proposed earlier (Stupar et al. 2001), a D′–A′–C–B structure with a deletion at the D′–A′ junction can be recognized. Such an arrangement is likely to derive from the normal A–B–C–D structure modified by two recombination events (Fig. 3C). In addition to the almost complete nuclear copy of mtDNA, 13 smaller NUMTs were detected within the first 76 kb of , which were mainly derived from the D and C regions of the mtDNA (Fig. 3D; Table 1). At least three major rearrangements have taken place in the first 76 kb of the insertion, as well as one event downstream of this region (see legend to Fig. 3).

The rice NUMT

The consists of five clustered NUMTs derived from noncontiguous regions of the mitochondrial chromosome (Fig. 3A; Table 1). This insertion shows less similarity (99.05%) to organellar DNA than most of the others (Table 2). Eight small and two large insertions of nuclear DNA are interspersed among the mtDNA sequences, and one of them (i2) disrupts a repeat IV sequence (Fig. 3A; Table 2; Supplemental Table 1).

Spectrum of nucleotide substitution and composition

With few exceptions (Cho et al. 2004), nucleotide substitutions occur much less frequently in the organellar DNAs of plants than in the nuclear genome (Wolfe et al. 1987; Palmer and Herbon 1988; Gaut 1998). The incidence of substitutions in our 13 inserts was computed by comparing their sequences with published mtDNA and ptDNA sequences (Hiratsuka et al. 1989; Unseld et al. 1997; Notsu et al. 2002). The rate of substitution varied between 1.23 () and 12.01 () nt/kb and, except in the cases of and , the rate of transitions exceeded the rate of transversion mutations (Table 2). G→A and C→T transitions formed the highest-frequency substitution class (each of them representing 21% of all substitutions; data not shown), similar to the substitution pattern observed in prokaryotes (Kowalczuk et al. 2001), nematodes (Blouin et al. 1998), Drosophila and mammals (Petrov and Hartl 1999), and Cucurbitaceae (Decker-Walters et al. 2004).

The G/C content of the inserts ranged between 38.83% and 49.17%, and differed from that of the corresponding organellar sequence by between –1.33% and +5.27% (Table 2). In nine inserts, the G/C content was changed toward that of the nuclear chromosome, albeit without reaching the same level. Major differences in G/C content with respect to organellar homologs (, and ) were noted for insertions in nuclear chromosomal regions that have a markedly different G/C content (Fig. 5A). Interestingly, the highest transition/transversion ratios were found in and , which also show the highest degree of amelioration to the G/C content of the host chromosome (Table 2; Fig. 5A). This implies that the high rate of transitions was responsible for their amelioration to the nucleotide composition of the nuclear DNA.

Expansion of NUMT/NUPT loci by further incorporation of DNA

Overall, the numbers of insertion and deletion events that occurred after integration of the 13 segments of organelle DNA are almost identical (492 vs. 493 events). The total length of DNA inserted post-integration, however, is at least 16 times larger than the amount of organelle sequences deleted (29,676 vs. 1810 bp). Tandem duplications of organelle DNA already integrated in the nucleus also contributed to the expansion of nuclear loci of organellar origin. However, the frequency of such events was relatively low, and only seven resulted in duplication of stretches larger than 10 bp (Supplemental Table 2). By comparison, the 10 largest insertions of nonorganelle DNA into nuclear DNA of organellar origin amount to 23,805 bp (Supplemental Table 1). These insertions were frequently associated with duplication of short regions at the target site, and two of them were similar to transposons, including the presence of terminal inverted repeats (Supplemental Table 1). The insertion contains a complete nuclear gene encoding a protein kinase, while is homologous to mtDNA of maize (Supplemental Table 1). The latter sequence is not present in rice mtDNA, indicating that it represents a part of the mitochondrial chromosome that was lost after the insertion of .

The data thus indicate that large NUMT/NUPT loci were expanded by subsequent insertions of nonorganelle DNA. This also explains the decay of colinearity with organellar DNA and the formation of the so-called “loose clusters” (Richly and Leister 2004b).

Deletion of organelle DNA is associated with direct repeats and palindromic sequences

The presence of deletions of up to several hundred base pairs relative to organellar DNA is characteristic of the nuclear insertions. Twenty-five deletions involved stretches longer than 10 bp (18 are listed in Table 3). Parallel losses were observed for two regions derived from rice ptDNA; a 69-bp stretch corresponding to positions 8535–8603 of ptDNA was absent in both and . In contrast, in the region corresponding to positions 57010–57028 of the plastid chromosome, three independent—and slightly different—deletions were observed in , and (Table 3). Both repeatedly deleted regions contained two perfect direct 13-bp repeats, one of which was always deleted, together with the inter-repeat region. This, together with the presence of a palindromic sequence in the inter-repeat regions (Table 3), is characteristic for RecA-independent deletion due to replication slippage (for review, see Bzymek and Lovett 2001; Lovett 2004). In the case of the three deletions in the region homologous to positions 57010–57028 of ptDNA, the slight differences in the regions affected imply that the repeat-removing mechanism is somewhat imprecise.

Table 3.

Deletions in nuclear organelle DNA

Integrant/deletion sizea	Position of deletion and of direct repeatsb	Position and structure of inverted repeatsc
Parallel losses
	8534Tttactttttttcag-----gTTACTTTTTTTCA8616	8587-8619 (13/23)
	57005ATTctttttttttaga ataCTTTTTTTTTAGA57036	57008-57061 (20/43)
	57005ATTCTTtttttttaga atacttttTTTTTAGA57036	57017-57068 (24/52)
	57005ATTCTTtttttttaga atacTTTTTTTTTAGA57036
Unique losses
	122921AtgctccaccgcttgTGC122936	122917-122940 (7/13)
		122918-122937 (7/16)
	19671GaggaagatcggaattagcaattgataaaaaagaaAGGA19709	19661-19683 (7/13)
	45241CataatgaaaacgcaatttctgtttttcttgaagacgaATA45281	45247-45267 (7/13)
	92552TtgcaggctgcaactcgccTGCA92574	1692553-92574 (9/20)
	2CCAatatcttgcttcaGCAAGATAT26	134519-36 (20/47)
	305456Actctaccaattg-----ttccaccCTCGTCCAAT305679	3005546-3005576 (8/16)
		3005593-3005609 (7/14)
	121312Tcttcttccttcttcc-----tcgagaCT.CTTCCT121371	121336-121368 (9/16)
	162645TCTTattgtg-----gatagctcttG162719	162664-162693 (11/25)
		162701-162725 (9/19)
	78599GTGAggaacg-----gcagacgtgaA78651	78588-78609 (8/17)
	258872ATAAGCTTTaccggttaaaccttataagctttG258904	258874-258902 (6/14)
	310215GACcaacctcgct-----ctaactCAAC310423	310344-310384 (13/24)
	311575Aggaaactagcacttt-----aatgatGGAA.CTAG311620	311554-311579 (8/15)
	161943ATGCccaggg-----gaaaaaatgcA162061	161935-161955 (8/16)
		162009-162034 (8/17)

Open in a new tab

^a

Name of the locus and size of the deletion (in parentheses).

^b

Nucleotides in lower case are absent in NUMTs or NUPTs, but present in organellar DNA (positions in organellar DNA are indicated in superscript). Direct repeats are underlined.

^c

Positions of inverted repeats are given relative to organelle DNA, together with the length/number of bonds in resulting stem-loop structures (in parentheses).

In addition to such parallel losses, 13 unique events were noted (Table 3). These deletions were also associated with the loss of repetitive sequences from nuclear organelle DNA. In such cases, both direct repeats and the inter-repeat palindromic sequences were shorter than those just discussed and/or imperfect. This is compatible with the observation that the frequency of replication slippage depends on the repeat length (Bzymek and Lovett 2001). An additional set of seven deletions was detected, in which neither or both direct repeats were deleted (Supplemental Table 3), indicating variation in replication misalignment or the action of secondary deletion events.

The presence in the insertions of regions that are particularly susceptible to deletion events allows us to further quantify the frequency of deletions. Parallel losses were indeed detected in all insertions that harbored the respective region of the organellar DNA. In contrast, for many of the unique losses, the undeleted version was also detected at a different locus; for example, the –37 deletion found in did not occur in either or (Figs. 1, 2).

Relative age of insertions

The large insertions of organelle DNA are organized as either continuous or highly fragmented segments. The obvious conclusion is that the fragmented versions derive from originally continuous ones by rearrangement; i.e., fragmented integrants are older than continuous ones. An alternative explanation is that continuous and highly fragmented integrants might represent two distinct modes of DNA transfer.

To permit one to distinguish between the two scenarios, it is crucial to determine the relative ages of the 13 primary insertions. Two independent measures allow us to assess relative ages as follows: (1) the degree of sequence identity between integrated and organellar DNA, taking into account the rate of nucleotide substitutions and of post-integration insertions/deletions; (2) the pattern of distribution of large deletions throughout the 13 insertions.

With respect to sequence identity, the insertions can be ranked by descending levels of identity in the following order: (Table 2). With respect to the distribution pattern of large deletions, six of the integrated segments contained at least five organelle sequences that include repetitive sequences susceptible to deletion (see above; Supplemental Table 4). Thus, the presence/absence of deletions in these regions allows us to produce an estimate of their relative ages. For nuclear ptDNA, the ranking order in descending level of deletion frequency is as follows: . When the incidence of direct-repeat-associated deletion of nuclear mtDNA was considered, appears to be older than (Supplemental Table 4), which supports the ranking obtained on the basis of sequence identity.

Mode of integration: Single events vs. hot spots

The two approaches outlined above yield coherent estimates for the relative ages of the different primary insertions. Strikingly, the highly fragmented insertions, in general, do not appear to be older than the long continuous ones. This is especially evident for and , which appear to have entered the nuclear genome at a later time than the largest continuous integrant, . The most fragmented integrations (, , and ) seem to have intermediate relative ages, and should be younger than the less fragmented , , and (Fig. 5B). Thus, it appears that highly fragmented inserts have not evolved from the long continuous type. Based on these data, we propose that continuous and highly fragmented integrants were initially generated in different ways.

Entire organellar chromosomes, or long fragments of organelle DNA, are thought to integrate into the nuclear genome by nonhomologous recombination (Henze and Martin 2001; Timmis et al. 2004). It is more difficult to explain the origin of mosaic insertions derived from different regions of the same organellar chromosome, or even from different organelles. Three mechanisms, in principle, could generate mosaic molecules for integration, i.e., (1) random end-joining of different linear DNA fragments before integration; (2) rapid rearrangement during, or directly after, insertion; (3) ongoing integration of organelle DNA at the same locus, corresponding to a “hotspot” for integration. The data we have presented do not allow us to choose definitively between the three scenarios. However, several of our observations support the relevance of a random end-joining mechanism. (1) When distinct fragments of nuclear organelle DNA are compared with organellar DNA, sequence divergence varies markedly within highly fragmented integrants (, , and ); the degree of this variation, however, does not appear to be markedly different from that observed for the long continuous insertion when the latter is subdivided into smaller segments (Fig. 5C). (2) In highly fragmented integrants, no sequence homology between adjacent fragments, or their template regions in organellar DNA, was evident, excluding the possibility that pre-existing insertions might have facilitated subsequent integrations at the same locus due to homologous recombination. (3) All highly fragmented integrants (, , , ) show an intermediate level of sequence divergence from the organellar template DNA, which is, however, lower than that associated with the less fragmented complex insertions, such as and (Fig. 5B). This may suggest that the first type of integrants derives from less fragmented ones.