The Roles of Mitochondrion in Intergenomic Gene Transfer in Plants: A Source and a Pool

Intergenomic gene transfer (IGT) is continuous in the evolutionary history of plants. In this field, most studies concentrate on a few related species. Here, we look at IGT from a broader evolutionary perspective, using 24 plants. We discover many IGT events by assessing the data from nuclear, mitochondrial and chloroplast genomes. Thus, we summarize the two roles of the mitochondrion: a source and a pool. That is, the mitochondrion gives massive sequences and integrates nuclear transposons and chloroplast tRNA genes. Though the directions are opposite, lots of likenesses emerge. First, mitochondrial gene transfer is pervasive in all 24 plants. Second, gene transfer is a single event of certain shared ancestors during evolutionary divergence. Third, sequence features of homologies vary for different purposes in the donor and recipient genomes. Finally, small repeats (or micro-homologies) contribute to gene transfer by mediating recombination in the recipient genome.


Intergenomic Gene Transfer from Mitochondrion to Nucleus
There exist a number of conserved genes during the mitochondrial genome evolution [45,46]. In the present study, we use 67 essential genes to study the gene loss and transfer about the mitochondrial genome. As a result, genes encoding complex II and ribosomal subunits have been lost massively in most of the higher plants ( Figure 1, yellow cells). Genes encoding complexes III and V display much greater conservation. These gene losses are parts of the mitochondrial genome variations in plants. Our next goal is to elucidate where the lost genes transferred. Two of the main detectable destinations are nuclear and chloroplast genomes.
Transferred genes exist in two forms: remnants left in the mitochondrial genome [47] and fragments inserted into the nuclear genome (numts) [48,49]. Few numts' products returned to the mitochondrion and played a role [33,34]. Researchers have achieved the mitochondrion-to-nucleus transfer by experiments, whose flow was as follows: (1) introduce a silent selectable marker gene with a nuclear promoter and transit peptide-encoding sequence into the mitochondrial genome; (2) transform this recombinant mitochondrion into a new cell; (3) detect the phenotype related to the marker gene. This approach has been successful in the unicellular green alga Chlamydomonas reinhardtii [27]. However, there is no experimental report on the real mitochondrion-to-nucleus IGT. Mitochondrial genes transferred with prokaryotic signals, which needed a long time or a favorable evolutionary event to turn into the eukaryotic ones.
In present research, to identify possible numts, we carry on non-experimental analyses by performing the genome alignment between conserved mitochondrial genes and nuclear genomes in above 21 land plants. First, we find extensive gene transfer and gene loss in these plants ( Figure 1). Second, the gene transfer is more popular in eudicots and monocots than that in bryophytes. Specifically, the latter is merely 1/20 of the former (Figure 2). Since the bryophytes with few mitochondria could not survive after vast transfer [24]. Third, we identify a number of full-length mitochondrial-like protein-coding genes in the nuclear genome ( Figure 1, red cells), which may be useful candidate genes. Fourth, there are also mitochondrion-like truncated genes, which we define as pseudogenes ( Figure 1, green cells).
Genes integrated by nuclear genome have different endings. Nearly all lost their original roles and became a part of new nuclear sequences [31]. A few could re-gain function by receiving nuclear promoter and transit peptide [2,32]. Others would suffer from irreversible decay with accumulating Figure 1. Genes identified to transfer in and out of the mitochondrial genome or genes lost from the mitochondrial genome of 21 land plants. The first two columns are mitochondrial protein-encoding genes (the second column) and their functional categories (the first column). The first line lists the names of plant species. The red and green cells represent mitochondrial full-length intact homologs and pseudogenes in nuclear genomes, respectively. The white and yellow cells represent no mitochondrial homologs in nuclear genomes and genes lost from mitochondrial genomes, respectively. Genes identified to transfer in and out of the mitochondrial genome or genes lost from the mitochondrial genome of 21 land plants. The first two columns are mitochondrial protein-encoding genes (the second column) and their functional categories (the first column). The first line lists the names of plant species. The red and green cells represent mitochondrial full-length intact homologs and pseudogenes in nuclear genomes, respectively. The white and yellow cells represent no mitochondrial homologs in nuclear genomes and genes lost from mitochondrial genomes, respectively. To dissect the mechanism of mitochondrion-to-nucleus gene transfers, we analyze the repeats in nuclear genomes of 22 land plants. The ratios of the repeat size to the genome size of the four species, including two bryophytes (M. polymorpha and P. patens) and two angiosperms (A. thaliana and S. polyrhiza), are less than 20% (Table 1). Meanwhile, these four species contain fewer numts than other species (Figure 1). And there is a positive correlation between numts and the repeats in the nuclear genome (R 2 = 0.6321) ( Figure 3). The weak correlation may due to limited number of plant species used in present research. So, we consider that numts may become parts of the nuclear repeats to take part in repeat-mediated sexual recombination for a greater genetic diversity. Correlation between the length of mitochondrial sequences transferring to the nucleus and repeat sizes of the nuclear genome in 22 land plants. Each dot represents a length value (X, Y). X refers to the size of the repeats in nuclear genomes of one species (based on the horizontal axis). Y means the length of mitochondrial-to-nuclear sequences in its corresponding species (based on the vertical axis). The slash represents the linear regression function of the distribution tendency of the dots. R 2 is the regression coefficient. To dissect the mechanism of mitochondrion-to-nucleus gene transfers, we analyze the repeats in nuclear genomes of 22 land plants. The ratios of the repeat size to the genome size of the four species, including two bryophytes (M. polymorpha and P. patens) and two angiosperms (A. thaliana and S. polyrhiza), are less than 20% (Table 1). Meanwhile, these four species contain fewer numts than other species (Figure 1). And there is a positive correlation between numts and the repeats in the nuclear genome (R 2 = 0.6321) ( Figure 3). The weak correlation may due to limited number of plant species used in present research. So, we consider that numts may become parts of the nuclear repeats to take part in repeat-mediated sexual recombination for a greater genetic diversity. To dissect the mechanism of mitochondrion-to-nucleus gene transfers, we analyze the repeats in nuclear genomes of 22 land plants. The ratios of the repeat size to the genome size of the four species, including two bryophytes (M. polymorpha and P. patens) and two angiosperms (A. thaliana and S. polyrhiza), are less than 20% (Table 1). Meanwhile, these four species contain fewer numts than other species (Figure 1). And there is a positive correlation between numts and the repeats in the nuclear genome (R 2 = 0.6321) ( Figure 3). The weak correlation may due to limited number of plant species used in present research. So, we consider that numts may become parts of the nuclear repeats to take part in repeat-mediated sexual recombination for a greater genetic diversity. Correlation between the length of mitochondrial sequences transferring to the nucleus and repeat sizes of the nuclear genome in 22 land plants. Each dot represents a length value (X, Y). X refers to the size of the repeats in nuclear genomes of one species (based on the horizontal axis). Y means the length of mitochondrial-to-nuclear sequences in its corresponding species (based on the vertical axis). The slash represents the linear regression function of the distribution tendency of the dots. R 2 is the regression coefficient. Correlation between the length of mitochondrial sequences transferring to the nucleus and repeat sizes of the nuclear genome in 22 land plants. Each dot represents a length value (X, Y). X refers to the size of the repeats in nuclear genomes of one species (based on the horizontal axis). Y means the length of mitochondrial-to-nuclear sequences in its corresponding species (based on the vertical axis). The slash represents the linear regression function of the distribution tendency of the dots. R 2 is the regression coefficient.
The existing forms of gene sequences in and out of both donor and receptor genomes altered after the transfer. D. carota Mitochondrial Plastid sequence (DcMP)-presented three fragment sequences (DcMP 1, −2 and −3 +4) in the plastid genome. The split probably arose from new DNA recombination that happened after one copy of DcMP migrated into the mitochondrial genome [16]. Besides, mitochondrial-like rpl2 only contained an exon in the plastid genome and two homologies in different regions of the mitochondrial genome in A. syriaca [17]. In addition, the traits of gene sequences in the plastid genome (recipient genomes) might affect their specialized roles. DcMP inserted into two short direct repeats in the plastid genome, which suggested that it served as non-LTR retrotransposon [18]. For those mitochondrial-derived pseudogenes in the plastid, they contained nonsense mutations that would lead to a premature stop codon, which was consistent with the low transcriptional level of the plastid copy rpl2 in A. syriaca [17].
From an evolutionary perspective, mitochondrion-to-chloroplast transfer occurred in the earlier common ancestor of certain relative species as a single event. For example, the homolog of mitochondrial gene, DcMP, existed in the plastid genomes of Daucus and their close relative Cuminum [18]. Further studies showed that DcMP moved to the shared ancestor of Daucinae Dumort and Torilidinae Dumort subtribes after they diverged from their ancestral tribe, Scandiceae Spreng [38,39]. Also, in Apocynaceae, mitochondrial rpl2 transferred to the plastid genome of the common ancestor of the Asclepiadeae and Eustegia [17].
Mitochondrial sequences preferentially inserted into the intergenic spacer of plastid genomes. For instance, DcMP inserted in the rps12-trnV intergenic spacer in the D. carota plastid genome [16]. There were also mitochondrial insertions in the rps2-rpoC2 intergenic spacer of the plastid genome in A. syriaca [17] and in the rpl23-ndhB intergenic spacer of the plastid genome of Parianinae (Eremitis sp. and Pariana radiciflora) [41]. Besides, another mitochondrial-to-nuclear transfer appeared in the large single copy (LSC) region between the junction with inverted repeat A (IRA) and tRNA-His (GUG) (trnH-GUG) in limited Apiaceae species [38]. Additionally, insertion locations implied the roles of the transferred genes. DcMP was regarded as a non-LTR retrotransposon targeting tRNA-coding regions because it moved to the upstream of the trnV gene in the plastid genome. Otherwise, DcMP worked as three new promoters (P1-P3) that substituted two original promoters of the trnV gene (P4 and P5) [18]. More importantly, insertion typically came with DNA repair of a double-stranded break by homologous recombination. To create homologies, the plastid gene rpoC2 preferentially inserted into the mitochondrial genome, just near the mitochondrial-native gene rpl2, then intact mitochondrial rpl2 and part of rpoC2 transferred together to the plastid of A. syriaca [17].

Intergenomic Gene Transfer from Nucleus to Mitochondrion
Compared with the conservative chloroplast genome, the mitochondrial genome diversified among plant species. The primary drivers of genome variations might be repetitive sequences and nuclear-derived DNA, which represented 42% and 47% of the total sequences in melon, respectively [10]. In present study, we analyze the nucleus-to-mitochondrion sequences of 23 plants. First, nuclear-derived sequences are widespread in all mitochondrial genomes of 23 plants ( Figure 4). Second, among spermatophytes, total nuclear sequences in mitochondrial genomes range from a low of 7960 bp in S. latifolia to a high of 36,123 bp in V. vinifera (Table S2). Third, the nucleus-to-mitochondrion transferred sequences are less in bryophytes than in spermatophytes, 4249 bp and 4814 bp in P. patens and M. polymorpha, respectively ( Figure 4).
Mitochondrial sequences preferentially inserted into the intergenic spacer of plastid genomes. For instance, DcMP inserted in the rps12-trnV intergenic spacer in the D. carota plastid genome [16]. There were also mitochondrial insertions in the rps2-rpoC2 intergenic spacer of the plastid genome in A. syriaca [17] and in the rpl23-ndhB intergenic spacer of the plastid genome of Parianinae (Eremitis sp. and Pariana radiciflora) [41]. Besides, another mitochondrial-to-nuclear transfer appeared in the large single copy (LSC) region between the junction with inverted repeat A (IRA) and tRNA-His (GUG) (trnH-GUG) in limited Apiaceae species [38]. Additionally, insertion locations implied the roles of the transferred genes. DcMP was regarded as a non-LTR retrotransposon targeting tRNA-coding regions because it moved to the upstream of the trnV gene in the plastid genome. Otherwise, DcMP worked as three new promoters (P1-P3) that substituted two original promoters of the trnV gene (P4 and P5) [18]. More importantly, insertion typically came with DNA repair of a double-stranded break by homologous recombination. To create homologies, the plastid gene rpoC2 preferentially inserted into the mitochondrial genome, just near the mitochondrial-native gene rpl2, then intact mitochondrial rpl2 and part of rpoC2 transferred together to the plastid of A. syriaca [17].

Intergenomic Gene Transfer from Nucleus to Mitochondrion
Compared with the conservative chloroplast genome, the mitochondrial genome diversified among plant species. The primary drivers of genome variations might be repetitive sequences and nuclear-derived DNA, which represented 42% and 47% of the total sequences in melon, respectively [10]. In present study, we analyze the nucleus-to-mitochondrion sequences of 23 plants. First, nuclear-derived sequences are widespread in all mitochondrial genomes of 23 plants ( Figure 4). Second, among spermatophytes, total nuclear sequences in mitochondrial genomes range from a low of 7960 bp in S. latifolia to a high of 36,123 bp in V. vinifera (Table S2). Third, the nucleus-tomitochondrion transferred sequences are less in bryophytes than in spermatophytes, 4249 bp and 4814 bp in P. patens and M. polymorpha, respectively ( Figure 4). According to the different degrees of the matching and annotation, these nuclear-tomitochondrial repetitive sequences fall into seven categories: copia, gypsy, low complexity, long terminal repeat retrotransposons (LTR-retro), simple repeat, transposable element (TE) and unspecified (Table S2). Copia and gypsy represent two main classes of LTR-retrotransposons that belong to Class 1 transposable elements [72]. Low-complexity DNA primarily include polypurine/poly-pyrimidine stretches and regions of extremely high AT or GC content. First, the mean of each type in 21 spermatophytes is significantly larger than that in 2 bryophytes (Figure 5), which According to the different degrees of the matching and annotation, these nuclear-to-mitochondrial repetitive sequences fall into seven categories: copia, gypsy, low complexity, long terminal repeat retrotransposons (LTR-retro), simple repeat, transposable element (TE) and unspecified (Table S2). Copia and gypsy represent two main classes of LTR-retrotransposons that belong to Class 1 transposable elements [72]. Low-complexity DNA primarily include poly-purine/poly-pyrimidine stretches and regions of extremely high AT or GC content. First, the mean of each type in 21 spermatophytes is significantly larger than that in 2 bryophytes (Figure 5), which show most nucleus-to-mitochondrion transfers occurred after the differentiation of seed plants and bryophytes, at least, for the analyzed 2 bryophytes species. Second, the first three are LTR-retro, gypsy and copia in 23 plants ( Figure 6, Table S2). This result conforms to the early discoveries in a number of plants, including the gymnosperm Cycas taitungensis [9], the monocot Oryza sativa [8] and the eudicots Arabidopsis thaliana, Cucumis melo and Cucumis sativus [4,7,[10][11][12]. Third, the total length of transferred sequences correlates with the mitogenome size (Figure 7). This result supports the import of promiscuous DNA is a core mechanism for mitochondrial genome expansion in land plants [73]. show most nucleus-to-mitochondrion transfers occurred after the differentiation of seed plants and bryophytes, at least, for the analyzed 2 bryophytes species. Second, the first three are LTR-retro, gypsy and copia in 23 plants ( Figure 6, Table S2). This result conforms to the early discoveries in a number of plants, including the gymnosperm Cycas taitungensis [9], the monocot Oryza sativa [8] and the eudicots Arabidopsis thaliana, Cucumis melo and Cucumis sativus [4,7,[10][11][12]. Third, the total length of transferred sequences correlates with the mitogenome size ( Figure 7). This result supports the import of promiscuous DNA is a core mechanism for mitochondrial genome expansion in land plants [73].   show most nucleus-to-mitochondrion transfers occurred after the differentiation of seed plants and bryophytes, at least, for the analyzed 2 bryophytes species. Second, the first three are LTR-retro, gypsy and copia in 23 plants ( Figure 6, Table S2). This result conforms to the early discoveries in a number of plants, including the gymnosperm Cycas taitungensis [9], the monocot Oryza sativa [8] and the eudicots Arabidopsis thaliana, Cucumis melo and Cucumis sativus [4,7,[10][11][12]. Third, the total length of transferred sequences correlates with the mitogenome size ( Figure 7). This result supports the import of promiscuous DNA is a core mechanism for mitochondrial genome expansion in land plants [73].

Intergenomic Gene Transfer from Chloroplast to Mitochondrion
As with mitochondrial genomes, chloroplast genomes also contain a minimum set of largely conserved protein-encoding, rRNA and tRNA genes [21,74,75]. In contrast to the extensive gene loss of mitochondrial genomes, only few chloroplast-encoded genes have been lost in chloroplast genomes of specific plants ( Figure S1, yellow cells). For example, three genes (accD, ycf1 and ycf2) are lost in the grasses (O. sativa japonica, O. sativa indica, S. bicolor, Z. mays), another three genes (ccsA, rpoA and rpl16) are lost in the moss P. patens ( Figure S1, yellow cells). Compared to a few gene loss, chloroplast genes transferring to nucleus and mitochondrion are richer (Figures S1 and S2). In our study, we unearth the enormous chloroplast-to-mitochondrion gene transfers in 24 land plants. Similar gene copies exist in two contemporary intracellular genomes simultaneously ( Figure S1, the red and green cells). In two bryophytes, the total lengths of integrated sequences are close, 1.05 kb in M. polymorpha and 1.99 kb in P. patens (Figure 8). In addition, the variation range is greater in 22 seed plants, from 1.67 kb in S. latifolia to 130 kb in A. trichopoda ( Figure 8). Besides, the chloroplast-tomitochondrion fragments of most seed plants are more than that in bryophytes ( Figure 8).

Intergenomic Gene Transfer from Chloroplast to Mitochondrion
As with mitochondrial genomes, chloroplast genomes also contain a minimum set of largely conserved protein-encoding, rRNA and tRNA genes [21,74,75]. In contrast to the extensive gene loss of mitochondrial genomes, only few chloroplast-encoded genes have been lost in chloroplast genomes of specific plants ( Figure S1, yellow cells). For example, three genes (accD, ycf1 and ycf2) are lost in the grasses (O. sativa japonica, O. sativa indica, S. bicolor, Z. mays), another three genes (ccsA, rpoA and rpl16) are lost in the moss P. patens ( Figure S1, yellow cells). Compared to a few gene loss, chloroplast genes transferring to nucleus and mitochondrion are richer (Figures S1 and S2). In our study, we unearth the enormous chloroplast-to-mitochondrion gene transfers in 24 land plants. Similar gene copies exist in two contemporary intracellular genomes simultaneously ( Figure S1, the red and green cells). In two bryophytes, the total lengths of integrated sequences are close, 1.05 kb in M. polymorpha and 1.99 kb in P. patens (Figure 8). In addition, the variation range is greater in 22 seed plants, from 1.67 kb in S. latifolia to 130 kb in A. trichopoda ( Figure 8). Besides, the chloroplast-to-mitochondrion fragments of most seed plants are more than that in bryophytes ( Figure 8).

Intergenomic Gene Transfer from Chloroplast to Mitochondrion
As with mitochondrial genomes, chloroplast genomes also contain a minimum set of largely conserved protein-encoding, rRNA and tRNA genes [21,74,75]. In contrast to the extensive gene loss of mitochondrial genomes, only few chloroplast-encoded genes have been lost in chloroplast genomes of specific plants ( Figure S1, yellow cells). For example, three genes (accD, ycf1 and ycf2) are lost in the grasses (O. sativa japonica, O. sativa indica, S. bicolor, Z. mays), another three genes (ccsA, rpoA and rpl16) are lost in the moss P. patens ( Figure S1, yellow cells). Compared to a few gene loss, chloroplast genes transferring to nucleus and mitochondrion are richer (Figures S1 and S2). In our study, we unearth the enormous chloroplast-to-mitochondrion gene transfers in 24 land plants. Similar gene copies exist in two contemporary intracellular genomes simultaneously ( Figure S1, the red and green cells). In two bryophytes, the total lengths of integrated sequences are close, 1.05 kb in M. polymorpha and 1.99 kb in P. patens (Figure 8). In addition, the variation range is greater in 22 seed plants, from 1.67 kb in S. latifolia to 130 kb in A. trichopoda ( Figure 8). Besides, the chloroplast-tomitochondrion fragments of most seed plants are more than that in bryophytes ( Figure 8).  Large parts of chloroplast tRNA genes immigrated into plant mitochondrial genomes [5,9]. These transfers were essential to the translation of the mitochondrial genes [13][14][15]. Here, we identify the chloroplast-like tRNA genes in the mitochondrial genome of 24 plants species using blast. And then we build a phylogenetic tree to elucidate the evolutionary implications. First, there is no chloroplast-derived tRNA gene in mitochondrial genomes of two bryophytes ( Figure S2). Second, single or multiple chloroplast genes immigrated to the mitochondrial genomes of spermatophytes, at least, for the analyzed 21 angiosperms and 1 gymnosperms. For example, (1) chloroplast-like trnM gene appears in the mitochondrial genomes of all studied seed plants except Z. mays, which suggests that chloroplast trnM lost only in Z. mays during or after transferring to the mitochondrion and this transfer happened with spermatophytes and bryophytes diverging; (2) chloroplast trnH gene transferred to the mitochondrial genomes of most spermatophytes but lost in P. dactylifera, T. aestivum and G. biloba, which might be the random loss; (3) trnN, trnP, trnS and trnW transferred merely in angiosperms, despite parts of these four genes lost in a few species; (4) chloroplast trnD gene moved into the mitochondrion only in eudicots, which shows that trnD transferred when eudicots and monocots diverged; (5) chloroplast-like trnC gene and trnF gene transferred to the mitochondrion simply in Gramineae crops of monocots; (6) ten chloroplast-to-mitochondrion genes (trnD, trnE, trnG, trnI, trnK, trnL, trnP, trnR, trnT and trnY) transferred together in V. vinifera ( Figure S2).
To infer the mechanism of chloroplast tRNA genes inserting into mitochondria, we analyze the flanking nucleotide sequences in insertion sites of mitochondrial genomes. trnH transferred in most spermatophytes ( Figure 9). trnD moved specifically in eudicots ( Figure S3). trnC and trnF migrated only in Gramineae crops ( Figure S4). Taking together, we notice the micro-homologies (1 to 4 bp) among plant species in the breakpoint sequences of chloroplast-mitochondrial DNA fusion. The micro-homologies are the same adenine-thymine (AT) on the right of trnH in spermatophytes. But on the left are four short tandems Guanine (G) in eudicots, two repeated Guanine (G) in monocots and no microhomology in gymnosperms ( Figure 9). Therefore, we confer that DNA sequence microhomology plays an important role in chloroplast DNA inserting into the mitochondrion, which may be the microhomology-mediated break-induced replication (MMBIR) [19] or non-homologous end joining (NHEJ) [20]. Large parts of chloroplast tRNA genes immigrated into plant mitochondrial genomes [5,9]. These transfers were essential to the translation of the mitochondrial genes [13][14][15]. Here, we identify the chloroplast-like tRNA genes in the mitochondrial genome of 24 plants species using blast. And then we build a phylogenetic tree to elucidate the evolutionary implications. First, there is no chloroplast-derived tRNA gene in mitochondrial genomes of two bryophytes ( Figure S2). Second, single or multiple chloroplast genes immigrated to the mitochondrial genomes of spermatophytes, at least, for the analyzed 21 angiosperms and 1 gymnosperms. For example, (1) chloroplast-like trnM gene appears in the mitochondrial genomes of all studied seed plants except Z. mays, which suggests that chloroplast trnM lost only in Z. mays during or after transferring to the mitochondrion and this transfer happened with spermatophytes and bryophytes diverging; (2) chloroplast trnH gene transferred to the mitochondrial genomes of most spermatophytes but lost in P. dactylifera, T. aestivum and G. biloba, which might be the random loss; (3) trnN, trnP, trnS and trnW transferred merely in angiosperms, despite parts of these four genes lost in a few species; (4) chloroplast trnD gene moved into the mitochondrion only in eudicots, which shows that trnD transferred when eudicots and monocots diverged; (5) chloroplast-like trnC gene and trnF gene transferred to the mitochondrion simply in Gramineae crops of monocots; (6) ten chloroplast-to-mitochondrion genes (trnD, trnE, trnG, trnI, trnK, trnL, trnP, trnR, trnT and trnY) transferred together in V. vinifera ( Figure S2).
To infer the mechanism of chloroplast tRNA genes inserting into mitochondria, we analyze the flanking nucleotide sequences in insertion sites of mitochondrial genomes. trnH transferred in most spermatophytes ( Figure 9). trnD moved specifically in eudicots ( Figure S3). trnC and trnF migrated only in Gramineae crops ( Figure S4). Taking together, we notice the micro-homologies (1 to 4 bp) among plant species in the breakpoint sequences of chloroplast-mitochondrial DNA fusion. The micro-homologies are the same adenine-thymine (AT) on the right of trnH in spermatophytes. But on the left are four short tandems Guanine (G) in eudicots, two repeated Guanine (G) in monocots and no microhomology in gymnosperms ( Figure 9). Therefore, we confer that DNA sequence microhomology plays an important role in chloroplast DNA inserting into the mitochondrion, which may be the microhomology-mediated break-induced replication (MMBIR) [19] or non-homologous end joining (NHEJ) [20].  On top of it all, we infer the repeats in mitochondrial genomes have the potential to mediate DNA recombination, which contributes to gene transfer and reuse of the transferred genes in target genomes. Therefore, we analyze the repeats variation in recipient genomes (the mitochondrial genomes) of land plants (Table 2) to explain various rates of gene transfer to some extent. First, plants with smaller values of repeat size, repeat number (>1 kb) and repeat number (>100 bp) contain less gene transfer, among which the most obvious is a bryophyte P. patens (Table 2 and Figure 8). Second, small repeats (>100 bp) are more favorable to gene transfer than large repeats (>1 kb) ( Table 2).

Availability of Chloroplast, Mitochondrial and Nuclear Genomes
We download all the chloroplast, mitochondrial and nuclear genome sequences and gene annotations from NCBI database. And then we list all the accession numbers in Table S1.

Detection of Total Intergenomic-Transfer DNA Sequences
For 24 land plants, we align the sequences of chloroplast and mitochondrial genomes to nuclear chromosomes to detect nuclear insertions of chloroplast DNA (nupts) and nuclear insertions of mitochondrial DNA (numts) using the BLAST program. We set e-value to 1e −5 [76]. The minimum length of an exact match (95%) is 100 bp. While identifying mitochondrial insertions of chloroplast DNAs (mtpts) by local BLASTN (version 2.2.23) [76], we set the minimum length of an exact match to be 50-bp.

Identification of Intergenomic-Transfer Homologies
Taking a set of essential chloroplast or mitochondrial genes as references (Table S3), we gain their copies in the donor and recipient genomes using the BLAST program with the same parameters above [76]. If there is no counterpart in the donor genomes (chloroplast or mitochondrial genomes), we would consider them as the lost genes ( Figure 1 and Figure S1, the yellow cells). To those presented in the donor genomes but absent in the recipient genomes, we consider that they did not transfer between two genomes ( Figure 1 and Figure S1, the white cells). For those appearing concurrently in both donor and recipient genomes, we consider that their copies moved into another genome after duplication in the original genome ( Figure 1 and Figure S1, the red and green cells). Further, we define the full-length copies of the transferred genes in the recipient genomes as the intact homologies ( Figure 1 and Figure S1, the red cells). Otherwise, we recognize the truncated copies as pseudogenes ( Figure 1 and Figure S1, the green cells).

Detection of the Repeats in Mitochondrial Genomes
We detect nuclear-derived repetitive transposons using online software RepeatMasker (http://www.repeatmasker.org) in 24 land species and a custom repeats database. And then we use two-tailed t-tests to evaluate the significant difference of repeats between spermatophytes and bryophytes.

NHEJ Analysis
We perform the NHEJ analysis as previously described [77,78]. In short, nupts, numts or mtpts are inserted by NHEJ, like micro-homology or blunt end repair. If nucleotides close to the fusion point are similar in different land species, we would regard them as micro-homology. Otherwise, we would consider no micro-homology as blunt-end repair.

Conclusions
With the rapid development of genomic sequencing technologies, nuclear and organellar genomes data became available for many plants. Here, based on 24 sets of genome data, we detect and analyze intergenomic gene transfers (IGT) related to the mitochondrion. Meanwhile, we review the research advances of intergenomic gene transfer. As a summary, we find mitochondrion mainly plays two essential roles in gene transfer: Source and pool. From the source perspective, massive mitochondrial genes transfer into nuclear and chloroplast genomes. For the role of the pool, the mitochondrion integrates enormous genes from the other two genomes. Except for the disparate orientation, a lot of likenesses emerge when bringing them together. First, gene transfer related to mitochondrial genomes is prevalent in plants, though few genes flow from the mitochondrion to the chloroplast. Second, specific IGT is a single event of certain shared ancestors, which is consistent with the divergence clade. Third, an intact gene usually changes existing forms after transferring in and out of both donor and recipient genomes, which agrees with their consequent roles, such as, functioning like before, reusing for new loci or decaying gradually. Fourth, most exogenous DNA preferentially inserts into the intergenic region. Besides, small repeats (or micro-homologies) may contribute to gene transfers by mediating recombination in the recipient genomes. In a word, mitochondrial gene transfers dedicate to the genome variation and evolutionary diversity.