Introduction

Horizontal gene transfer is the transmission of genetic material between different organisms or between different cytoplasmic organelles and nucleus through asexual processes. It plays an important role in the evolution of many organisms. For instance, HGT is the main cause of rapid antibiotic resistance circulation among bacteria (Koonin et al. 2001; Gyles and Boerlin 2014; Kay et al. 2002). HGT is not only broadly present in the prokaryotic world, but also increasingly reported in eukaryotes (Keeling and Palmer 2008). In land plants, massive HGT has been discovered through the use of in-depth sequencing in a few species, such as Amborella trichopoda, Geranium brycei, Rafflesia cantleyi, Sapria himalayana, and Lophophytum mirabile (Bergthorsson et al. 2004; Molina et al. 2014; Park et al. 2015; Rice et al. 2013; Sanchez-Puerta et al. 2017; Xi et al. 2012, 2013). Besides, extensive HGT has been shown to promote plant colonization of land (Yue et al. 2012). With the sequencing of more plant genomes, the widespread footprints of HGT in additional plants will be gradually uncovered. The frequency of identified HGT is much higher in mitochondria than in plastids and nuclei, and a large fraction of the HGT reports come from parasitic plants and their hosts (Davis and Xi 2015; Keeling and Palmer 2008; Sanchez-Puerta 2014). Among mitochondrial genes, cox1 is the most frequently implicated in HGT (Cho et al. 1998; Sanchez-Puerta et al. 2008).

The mitochondrial gene cox1 encodes the cytochrome c oxidase subunit I required to constitute the respiratory complex IV and is essential for oxidative phosphorylation (Toffaletti et al. 2003). In most vascular plants, it does not have an intron, while in a sizeable fraction of angiosperms its exon is interpolated by an intron sequence. Indeed, cox1 introns have invaded nearly every angiosperm parasitic lineage, leaving Krameria and Schoepfia as the only two sampled parasitic plants lacking cox1 introns (Barkman et al. 2007). The cox1 intron encodes a site-specific DNA endonuclease, which facilitates its propagation (Delahodde et al. 1989). Homing is used to describe such a phenomenon. Intron homing, the introduction of an intron into a homologous allele lacking it, has been proposed to proceed by the double strand-break repair pathway (Lambowitz and Belfort 1993). During this process, part of the foreign exonic regions immediately flanking the invading intron often engages in a gene conversion activity that replaces part of the recipient exonic sequence (Delahodde et al. 1989; Lambowitz and Belfort 1993; Mueller et al. 1996; Wenzlau et al. 1989). A region of converted exonic sequence is called a “co-conversion tract” (CCT). If the flanking exon sequences of the donor and recipient plants differ, then the repair process will create a “footprint” (CCT) that can remain even after the intron itself is lost again (Cho and Palmer 1999).

The sporadic distribution of the cox1 intron among angiosperms is attributed to HGT via the above intron homing mechanism in most cases (Barkman et al. 2007; Cho et al. 1998; Sanchez-Puerta et al. 2011, 2008), and intron loss usually via a retrotranscribed copy of a mature cox1 transcript (Sanchez-Puerta et al. 2008). As cox1 introns are more frequently found in parasitic plants, we choose Cassytha, the only parasitic genus with 10–20 species in the family Lauraceae, as the target system to study cox1 evolution. According to the Flora of China, only one pantropical species, the hemiparasite C. filiformis, distributes in China. C. filiformis has a wide range of hosts, with more than 100 host species in Guangxi alone (Li et al. 1992). The known hosts in China according to our field surveys and from the literature are summarized in Table S1. The intimate connection of C. filiformis with its host through haustoria and the wide host range grant a large potential of genetic flow.

According to NCBI nucleotide databases, cox1 sequences from C. filiformis and other Lauraceae species have been reported. A comparison of their sequences suggested that C. filiformis has an intron in its cox1, while other Lauraceae species have not. It is not clear whether the cox1 intron was acquired exclusively by C. filiformis, or whether it was lost in other members of the family. In this study, we generated cox1 sequences from different C. filiformis samples collected from three distant places and from 32 other species from different lineages within the family Lauraceae, and analyzed the cox1 sequences from a wide diversity of angiosperms. We aimed to achieve the following objectives: (1) to test whether different cox1 copies exist in C. filiformis; (2) to examine whether cox1 intron in C. filiformis has been retained from a common Lauraceae ancestor or was horizontally transferred from other non-Lauraceae species; and (3) to understand the evolutionary history of the cox1 genes in Cassytha.

Methods

Sampling and Sequencing

We collected C. filiformis stem samples at least two centimeters away from the host to prevent contamination. Leaf and stem samples from other 32 Lauraceae species were also collected (Table S2). Total genomic DNA was extracted with the Plant Genomic DNAKit (Tiangen Biotech, China). The cox1 genes in C. filiformis were amplified by PCR using two primers, cox1 intron-F (5′-CATCTCTTTYTGTTCTTCGGT-3′) and cox1 intron-R (5′-AGCTGGAAGTTCTCCAAAAGT-3′) (Sanchez-Puerta et al. 2008). Another set of primers designed by Primer Premier 5 (Lalitha 2000), cox1 exon-F (5′-GTATGGAATTAGCACGACCCG-3′) and cox1 exon-R (5′-TACGACCACGAAG GAACGAC-3′), were used to amplify cox1 genes in C. filiformis as well as 32 Lauraceae species under study. The PCR mixtures for cox1 amplification were 2.5 μl of 10 × PCR reaction buffer (Takara, Japan), 1.5 μl of 25 mM MgCl2, 1 μl of each primer (Shanghai Sangon, China) at 10 ng/μl, 1 μl of 2.5 mM dNTP solution in an equimolar ratio, 0.2 μl of Taq DNA-polymerase (5 U/μl, Takara, Japan), 2 μl of genomic DNA at 5 ng/μl, and ddH2O to reach a total volume of 25 μl. The amplified products were purified using the QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany). In the genus Cassytha, PCR products of Cassytha sp. and C. pubescens were successfully sequenced, but the C. filiformis appeared polymorphisms. In order to assess whether different cox1 genes exist in C. filiformis, we further cloned PCR products from C. filiformis using the pEASY-T3 Cloning Kit (TransGen Biotech, Beijing, China). Between 2 and 6 clones were sequenced for each individual. All fragments were sequenced in both directions using BigDye 3.1 reagents with an ABI 3770 automated sequencer (Applied Biosystems, Carlsbad, California, USA). All sequences are deposited in GenBank (Table S2).

Sequence Alignment and Phylogenetic Analyses

All sequence alignments were manually edited using Geneious v6 (Kearse et al. 2012). Homologous cox1 sequences were identified using BLASTN against NCBI Non-Redundant Nucleotide Database (Tables S3 and S4). In order to remove the influence of CCTs and editing sites on phylogenetic analyses, we excluded 30 bp of the exon downstream the intron insertion site and the predicted editing sites. Multiple sequence alignments of the cox1 coding sequences and introns were performed with MAFFT (Katoh et al. 2017) and manually adjusted. Phylogenetic analyses were performed on the aligned sequences of cox1 exons and cox1 introns, respectively. The maximum likelihood (ML) analyses were performed in RAxML-HPC BlackBox via CIPRES (Miller et al. 2010) and RAxML under the general time reversible model with parameters for invariable sites and gamma-distributed rate heterogeneity (GTR + I + G; 4 rate categories, 1000 bootstraps).

Results

Co-existence of Two cox1 Copies in C. filiformis

In order to investigate whether different forms of the mitochondrial gene cox1 exist in distinct populations of C. filiformis, we collected stem samples of C. filiformis from three geographically distant places in China, i.e., Shenzhen, Nanning, and Xishuangbanna, amplified the cox1 sequences, and sequenced them. Since cox1 variation across angiosperms is mainly found in the presence/absence of the cox1 intron and the co-conversion tract, we amplified this region (~ 1140 bp fragments). As the initial sequencing of cox1 introns was quite unsuccessful due to the presence of multiple peaks, we resorted to gene cloning of the PCR products. The subsequent sequencing of the cloned fragments revealed clearly that two distinct copies of cox1 introns were present in the same samples.

Although we expected cox1 variation in terms of the absence or presence of the intron, the identification of two different cox1 intron sequences in the same samples was surprising. We wondered whether the exons of the two cox1 also differed, thus we amplified and cloned the whole cox1 (~ 2221 bp) from additional samples for further sequencing. Two distinct copies of cox1 genes were identified in most samples of C. filiformis collected from three different places, suggesting a stable, inheritable nature of both cox1 alleles in C. filiformis. Moreover, the frequency of co-existence of these two cox1 alleles was very high as they were detected in 18 out of 20 samples (Table 1). Only one type of cox1 was detected in samples M6 and M48, probably due to insufficient clone sample. We then amplified the DNA and directly sequenced the PCR products from these two samples and found no polymorphisms when sequencing the non-clonal PCR products, which verified the existence of only type I cox1 gene (Table 1).

Table 1 Occurrence of the two types of cox1 in the sequenced clones of C. filiformis samples
Full size table

The two alleles of cox1 in C. filiformis differ strikingly in their exon and intron sequences and length (Table 2). Their intron sequence identity is only 84.6%, suggesting that the two copies of cox1 should come from completely different origins. Further bioinformatic analysis on the two cox1 genes of C. filiformis revealed that the exons of type I cox1 have an intact open reading frame, whereas the exons of type II cox1 have premature stop codons, which likely produces a much shorter malfunctional protein. The intron of type I cox1 is 967 bp in length and has a full-length open reading frame of 921 bp, encoding a homing endonuclease. Similar to the cox1 exons, the intron of type II cox1 is 912 bp in length but contains several nonsense mutations (Table 2).

Table 2 Sequence comparison of the two types of cox1 in C. filiformis
Full size table

A BLAST search against Genbank databases showed that type I cox1 is similar to cox1 in magnoliids species, whereas type II cox1 displays very high similarity to cox1 in Cuscuta japonica, a Convolvulaceae species. Moreover, the two copies of cox1 in C. filiformis show contrasting Insertion/Deletion (INDEL) in multiple sequence alignments. These two cox1 differ in 12 INDEL loci, whereas type II cox1 in C. filiformis and cox1 in C. japonica are nearly identical at 10 of these loci (Fig. 1). These ten shared INDELs are unique as they are absent from the cox1 in all other species under study. Therefore, INDEL comparisons further support the different origins of the two copies of cox1.

Fig. 1
figure 1

Partial sequence alignment of the two types of cox1 in Cassytha filiformis and four other cox1 homologs. For comparison, we included cox1 in Cuscuta japonica and three Magnoliales species, Asimina triloba, Knema latericia, and Myristica fragrans. The ten shared INDEL positions by type II cox1 in C. filiformis and cox1 in C. japonica are in red. The two INDEL positions unique to C. filiformis type II cox1 are in blue. A shaded box indicates the intron sequence (Color figure online)

Full size image

The Different Origins of the Two C. filiformis cox1 Genes

In order to track down the exact origins of the cox1 genes in C. filiformis, we carried out phylogenetic analyses on cox1 from many other Lauraceae species as well as a diverse range of angiosperms. We systematically studied cox1 genes from many other Lauraceae species. We took leaf samples from 30 other species, representing 16 Lauraceae genera distributed in China. Besides, we also included four Australian Cassytha stem samples, three from C. pubescens and another from an unidentified Cassytha sp. (Table S2). We also downloaded cox1 sequences of seven other Lauraceae species from the NCBI Nucleotide databases (Table S3). Unlike the other 16 genera of the family Lauraceae, the genus Cassytha is the only one that harbors introns in its cox1 genes. The difference is that C. filiformis has two types of cox1 genes, while the four Australian Cassytha samples have no polymorphisms when sequencing non-clonal PCR products and were shown to contain only the type I cox1, suggesting that the introduction of type I cox1 intron probably took place before the speciation of C. filiformis. It is unsure whether type II cox1 is unique to some local Chinese C. filiformis populations, transferred horizontally after the split of C. filiformis and C. pubescens, or it is found in other Cassytha species and was lost randomly in certain populations. The sequencing of more Cassytha samples is required to answer this question.

We analyzed the exon and intron trees individually because cox1 introns are frequently involved in horizontal gene transfer and often show significant phylogenetic incongruence in comparison to cox1 exons. The phylogenetic tree based on the exon sequences (Fig. 2) suggests that the two alleles of cox1 of C. filiformis have completely different origins. The C. filiformis type II displays very high affinity to cox1 in Cuscuta spp. and Ipomoea spp., two Convolvulaceae species, suggesting a foreign origin of this allele. On the other hand, the C. filiformis type I is phylogenetically close to those in magnoliids, consistent with a vertical inheritance of this cox1 coding sequence.

Fig. 2
figure 2

ML tree of 173 species based on cox1 exons analyzed under a GTR + I + G model. Only ML bootstrap values > 60% are displayed. Species that belong to monocots, eudicots, Lauraceae, and other magnoliids except Lauraceae are in cyan, green, magenta, and blue, respectively. The branches leading to the two types of C. filiformis cox1 are colored red (Color figure online)

Full size image

In contrast to the observed in the exon phylogeny, the cox1 introns of all Cassytha species cluster in a single clade including both cox1 alleles in C. filiformis, although not as sister taxa (Fig. 3). According to Figs. 2 and 3, it is clear that the phylogenetic positions of exons and intron of type II C. filiformis cox1 do not change much, as both form a monophyletic clade with Cuscuta spp. and Ipomoea spp. However, the intron tree shows a sister relationship between C. filiformis type II cox1 and C. japonica with 100% of bootstrap support. These results suggest that both the exons and intron of type II C. filiformis cox1 might share an origin with cox1 genes from Cuscuta spp.

Fig. 3
figure 3

ML tree of 103 species based on cox1 introns analyzed under a GTR + I + G model. Only ML bootstrap values > 60% are displayed. Species that belong to monocots, eudicots, Lauraceae, and other magnoliids except Lauraceae are in cyan, green, magenta, and blue, respectively. The branches leading to the two types of C. filiformis cox1 are colored red (Color figure online)

Full size image

Interestingly, the phylogenetic position of C. filiformis type I cox1 and other Cassytha species cox1 differs considerably in the exon tree and the intron trees. C. filiformis type I cox1 and other Cassytha species cox1 cluster in the clade of Lauraceae and other basal magnoliids species in the exon tree, while they are related to diverse eudicots, in particular, Cuscuta japonica, Ipomoea spp. and Calceolaria sp. with 80% of bootstrap support, according to the intron tree. The phylogenetic incongruence of C. filiformis type I cox1 and other Cassytha species cox1 exons and introns are similar to other cases of horizontal gene transfer, where an exogenous cox1 intron invaded the native cox1 copy via intron homing.

Furthermore, cox1 introns are accompanied by a characteristic co-conversion tract (CCT) when the exon of the donor is different from that of the recipient plant (Cho and Palmer 1999; Cho et al. 1998; Sanchez-Puerta et al. 2011, 2008), Cusimano et al. (2008) grouped CCTs of all available angiosperm cox1 sequences into 20 types. We thoroughly compared the exon sequences flanking the intron insertion site of all the newly sequenced cox1, as well as homologous cox1 sequences downloaded from NCBI databases (Fig. 4). It is clear from Fig. 4 that the species of the family Lauraceae, except for Cassytha spp., lack the cox1 intron and the characteristic CCT. These observations suggest that the absence of the cox1 intron in these Lauraceae species is not due to intron loss and that C. filiformis obtained the cox1 introns by HGT after its divergence from other Lauraceae. The type I cox1 of C. filiformis has an intron and a 26 bp-long CCT that it is also observed in Cassytha spp., as well as in a few other species that show affinity to Cassytha spp. in the intron phylogeny. The type II cox1 of C. filiformis has an extended CCT of 30 bp shared with C. japonica and Ipomoea spp. Therefore, the two types of CCTs and the intron phylogeny support the two independent evolutionary origins of C. filiformis cox1 genes.

Fig. 4
figure 4

Sequence comparisons of cox1 sequences flanking the intron insertion site in 75 angiosperms. Species that belong to monocots, eudicots, Lauraceae, and other magnoliids except Lauraceae are in cyan, green, magenta, and blue, respectively. The co-conversion tracts (CCTs) of two types of C. filiformis cox1 are colored red. Plus (+) and minus (−) symbols indicate cox1 intron presence and absence, respectively (Color figure online)

Full size image

Discussion

The Structure and Frequent Co-existence of Two cox1 Alleles in C. filiformis

In this study, we identified two different copies of the gene cox1 in individual samples of C. filiformis. The exon sequences of C. filiformis type I cox1 have an intact open reading frame that encodes the cytochrome c oxidase subunit I, in agreement with earlier studies (Barkman et al. 2007). The intron of the type I cox1 encodes a putative functional homing endonuclease, because of the presence of two LAGLI-DADG motifs and an intact open reading frame, which may be involved in intron propagation and splicing (Belfort and Perlman 1995). In contrast, C. filiformis type II cox1 is a pseudogene given that we identified several stop codons within the exons and also in the intron sequence. All sequenced samples of C. filiformis contain the type I copy, and a few samples lack the type II copy, indicating that only type I cox1 is essential and type II copies might have escaped from functional constraints. Also, the cox1 type II copy was not found in other species of the same genus. The presence of two cox1 alleles, one of which is a pseudogene has been previously described in Geranium brycei (Park et al. 2015).

In this study, a population-level study found that the co-existence of these two alleles in C. filiformis was quite frequent, as they were found in more than 90% of the 20 individuals of C. filiformis analyzed. Anyway, the co-occurrence of two cox1 alleles, either in a single mitochondrial genome or in different mitochondria or cells of the stem of C. filiformis is outstanding and deserves further investigation. The origin of each of the cox1 alleles in C. filiformis may be explained by its increased chance as a parasitic plant to exchange genetic information with its hosts and a greater flexibility in genome evolution after adopting a parasitic lifestyle (Davis and Xi 2015; Sanchez-Puerta 2014). It is also possible that the co-existence of different cox1 alleles, or other mitochondrial genes, in other species is underestimated due to the limited sampling at the population level or difficulty to detect additional gene copies at lower stoichiometries. In either case, deeper sequencing on a wider range of plants and increased population sampling are required to evaluate the co-occurrence of cox1 alleles in different species as well as the ecological and evolutionary importance of cox1 heterozygosity.

Two Independent HGT Events

The acquisition of foreign DNA has been predicted to be a key event in the evolution of angiosperms (Atsatt 1973), and cox1 intron could represent a marker of a genomically more widespread historical transformation (Barkman et al. 2007). The numerous angiosperm-to-angiosperm transfers of cox1 intron and its outstanding evolutionary history have sparked the interest of several researchers. Cho et al. (1998) and Sanchez-Puerta et al. (2008) analyzed all available cox1 data from angiosperms and confirmed that the cox1 intron has been horizontally acquired numerous times during angiosperm evolution. For example, the cox1 intron was acquired by horizontal transfer in at least three separate occasions during the evolution of the Solanaceae (Sanchez-Puerta et al. 2011). The opposite view argued that cox1 intron loss is a predominant factor in cox1 evolutionary history in Araceae (Cusimano et al. 2008). Moreover, for the first time, two copies of the cox1 gene which differ in intron content were found in Geranium brycei mitochondria and supported the notion of repeated, independent HGT (Park et al. 2015).

In our study, we also found two different copies of the cox1 gene in C. filiformis. In addition to the well-documented cox1 intron homing, we also identified exons involved in horizontal gene transfer. In fact, the phylogenetic analyses of cox1 exons and introns revealed a different origin of the two intron-containing cox1 alleles of C. filiformis. One full-length copy had been clearly acquired by horizontal gene transfer from the lineage Convolvulaceae and it is a pseudogene in C. filiformis. The cox1 coding regions of Cuscuta spp. and Ipomoea spp. have intact open reading frames and encode the cytochrome c oxidase subunit I. Therefore, the pseudogenization of type II cox1 of C. filiformis may have taken place after the horizontal gene transfer event. The other copy has vertically inherited exons and a horizontally transferred intron. The intron phylogeny shows a close relationship to the cox1 introns of Calceolaria spp., Ipomoea spp., Cuscuta japonica, and the foreign copy of C. filiformis and this copy shares a 20 bp-long CCT with them and other angiosperms. In addition, the other genera of Lauraceae analyzed show a single cox1 allele that lacks the intron and CCT. The cox1 exons of all Lauraceae, including C. filiformis type I and Cassytha spp., are highly similar to each other.

The two independent HGT events reveal a highly dynamic mitochondrial genome in C. filiformis and raised more questions. For instance, did the intron of type II cox1 of C. filiformis invade the native cox1 gene? Did recombination take place between the two different copies of cox1 genes? The fact that both C. filiformis cox1 introns are found in a single clade in the intron phylogeny and are associated to a similar CCT opens the possibility of an intracellular intron invasion from the type II cox1 allele to the type I cox1 allele. However, the high sequence divergence observed in the type II cox1 intron, shared with Cuscuta and Ipomoea in comparison to the more conserved type I cox1 intron sequence argues strongly against it. In contrast, it suggests a second horizontal acquisition from a donor containing an intron related to those of the Convolvulaceae. By analyzing the cox1 alignment in detail, we could not find evidence of recombination between the two cox1 alleles in C. filiformis.

Further questions remain unanswered, such as when did these two HGT events happen? Were cox1 genes in other Cassytha species acquired from additional donors? How did the HGT events influence the evolution of C. filiformis? Since Cassytha and Cuscuta are both parasitic plants and HGT can promote adaptation to parasitism, it is tempting to ask whether any parasitism-related gene would be exchanged in the HGT event. The answers to the above questions will greatly contribute to our understanding of the mitochondrial dynamics in C. filiformis and historical events during its parasitism evolution.

Conclusions

In this study, we investigated cox1 evolution in a parasitic Lauraceae species, C. filiformis. We found consistent co-existence of two different cox1 alleles in 90% of the samples of C. filiformis collected from distant locations around China, and demonstrated clearly the different origins of the two types of cox1 genes as well as the implications of two independent horizontal transfer events. Our study deepens our understanding of the complicated evolutionary histories of C. filiformis cox1 and the highly dynamic mitochondrial genome in this parasitic plant.