Introduction

Group II introns are widely distributed in the organellar genes of plants, algae, fungi, and protozoans as well as eubacteria and viruses (for reviews, Lambowitz and Belfort 1993; Robart and Zimmerly 2005). Molecular phylogenetic analyses of group II introns show that some homologous group II introns appear at corresponding sites of the organellar genome in distantly related species (for a review, see Zimmerly et al. 2001). Therefore, it is widely accepted that group II introns are mobile genetic elements transferred across species and have competence to invade specific target sites within a gene (Belfort and Perlman 1995).

Moreover, most group II introns offer no benefit to the host, and despite being maintained over time (e.g., Turmel et al. 2002), mutations accumulate after invasion. The reverse transcriptase and endonuclease domains of the intronic open reading frame (ORF) seem essential for retro-transposition; however, these domains are dispensable for splicing (Lambowitz and Belfort 1993). Therefore, post-invasion mutations that occur in the reverse transcriptase and endonuclease domains are tolerable for the host. This suggests that for most group II introns to ensure long-term evolutionary survival, regular transmission to an intronless target in a new host (Zimmerly et al. 2001) must occur prior to the loss of activities necessary for invasion (Gogarten and Hilario 2006).

Group II introns are usually classified into two subgroups, IIA and IIB, based on the secondary structure of their ribozyme component (Michel and Dujon 1983; Michel et al. 1989). The ribozyme consists of six double helical domains numbered from I to VI, radiating from a central wheel. In domain IV, an ORF encoding one of the core components of the active group II intron ribonucleoprotein is often found in various states of preservation. Accumulated data show that the complete group II intronic ORF is tripartite in nature and is composed of a reverse transcriptase domain (RT), maturase domain, and endonuclease domain (Michel and Ferat 1995). However, non-tripartite ORFs are detected much more frequently than the complete form. The maturase domain assists in the splicing reaction. Reflecting its vital role, this domain is often preserved intact. On the other hand, although the RT and endonuclease domains are essential for transmission of the intron, both are dispensable for splicing. This allows for rapid degeneration after invasion. Intron invasion is initiated by reverse splicing of the intron RNA directly into the target DNA (Guo et al. 1997), followed by reverse transcription of the integrated RNA including the ORF of the intron-encoded protein (Cousineau et al. 1998). This suggests that the existence of a complete tripartite ORF is an exemplary feature of a recently invaded intron.

The life cycle of a group II intron is composed of three phases; invasion into a target site, degeneration at the integrated site, and complete loss from the site. In time, the recovered intronless site is reinvaded by a cognate intron (i.e., an intron that shares the same target sequence specificity).

Complete mitochondrial genome sequence of a filamentous brown alga Pylaiella littoralis (CCMP1907) has been determined (Oudot-Le Secq et al. 2001). Interestingly, only the 26-28S ribosomal RNA (hereafter simply LSU rRNA) and cytochrome oxidase subunit 1 (cox1) genes possess multiple group II introns, [i.e., four group IIB (three IIB1 and one IIB2) introns interrupt the LSU rRNA gene (Fontaine et al. 1995), and three group IIA introns are present in the cox1 gene (Fontaine et al. 1997)], while the others have no introns. Furthermore, without showing any actual data, Fontaine et al. (1995) referred to the absence of several introns in some strains. Since then, no further reports regarding the variation of intron numbers in P. littoralis genes have been published.

To date, in addition to P. littoralis, full mitochondrial genome sequences from brown algae have been reported for Dictyota dichotoma, Desmarestia viridis, Fucus vesiculosus, and Laminaria digitata (Oudot-Le Secq et al. 2002, 2006). Among them, P. littoralis and L. digitata are known to be closely related species (Draisma et al. 2001; Sasaki et al. 2001). Further, these two species exhibit multiple distinctive molecular features in their genomes including an unusually large insertion in the cox2 gene, a putative 5S rRNA gene and a small nad11 gene (Oudot-Le Secq et al. 2002). In spite of these commonalities, group II introns are found only in P. littoralis (Fontaine et al. 1995, 1997; Oudot-Le Secq et al. 2001), but not in L. digitata (Oudot-Le Secq et al. 2002). This suggests that the group II introns in P. littoralis are the result of recent invasions that occurred after the divergence of P. littoralis from L. digitata.

We considered that analysis of the divergence of intron organization within a gene of a given species would be informative to estimate the bottleneck step of the life cycle of the group II intron. Further, this would allow for determining how long the intron persists after insertion. We examined eight specimens of P. littoralis collected worldwide to address the puzzling questions described above.

Materials and methods

The collection sites for the Pylaiella littoralis specimens analyzed in this study are listed in Table 1. Except for the specimen collected at Hokkaido, Japan, the others were collected by H.K. and D.G.M. and maintained in their laboratory as axenic strains. Total DNA was extracted by DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) from each specimen, and used as a template to examine the existence of the reported four group IIB introns in the mitochondrial LSU rRNA gene, and three group IIA introns in the cox1 gene by the polymerase chain reaction (PCR) method. Specific primers for PCR were designed based on the complete mitochondrial genome sequence of P. littoralis (Oudot-Le Secq et al. 2001; GenBank acc. no. AJ277126) (Table 2). Each set of PCR primers was designed to anneal in regions flanking specific introns reported in the Frnce-I specimen (Table 2). Regions of the mitochondrial genome, which were expected to form complex tertiary structure during denaturation, were amplified by PCR using LA Taq in conjunction with the GC buffer kit, which is formulated to ameliorate the effect of GC-rich regions on PCR amplification (Takara Bio Inc., Otsu, Japan). The PCR thermocycle program was as follows: initial denaturation at 95°C for 2 min; 35 cycles of denaturation at 95°C for 30 s, annealing at 48°C for 30 s, and elongation at 72°C for 5 min; and a final elongation at 72°C for 5 min. A thermal cycler (Model TP600, Takara Bio Inc.) was utilized to perform the above program. When necessary, the second PCR was performed in 35 or 50 cycles using the same thermal conditions of the initial PCR. After separation on an agarose gel, candidate DNA fragments were extracted from the gel using the QIAquick Gel Extraction Kit (Qiagen), cloned into the pGEM-T Easy Vector (Promega, Madison, WI, USA). The DNA sequences were determined using ABI Prism BigDye Terminator Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA). DNA sequence alignments and construction of the neighbor-joining distance tree (Saitou and Nei 1987) were performed by the integrated program Clustal X version 1.8 (Thompson et al. 1997). Distance was calculated based on the Jukes-Cantor two-parameter method (Jukes and Cantor 1969). Bootstrap resampling (1,000 replicates) (Felsenstein 1985) was carried out to quantify the relative support for each branch in the tree.

Table 1 Habitats of Pylaiella littoralis specimens
Full size table
Table 2 Primers for PCR amplification of specific intron coding regions
Full size table

Results

We closely investigated the number of introns in the LSU rRNA and cox1 genes from eight specimens collected from worldwide habitats (Table 1). We used the neighbor-joining method to align the newly obtained sequences along with the previously reported ones to determine the evolutionary distances and their molecular phylogenetic relationships. In the case of LSU introns, we used only alignable sequences within the non-intronic ORF region for the phylogenetic analyses, because several introns have lost significant parts of the ORF located in the arm-IV loop (see later).

Divergent number of Group IIB introns in the LSU rRNA gene

Fontaine et al. (1995) utilized total DNA prepared from P. littoralis, which is an axenic strain originally collected at Chenal de l´ile Verte, Roscoff, France about 28 years ago (hereafter Frnce-I, this strain is now available under the reference names of SAG 2000 and CCMP 1907) to determine the DNA sequence of the LSU rRNA gene. They reported that Frnce-I has four group IIB introns in the LSU rRNA gene (LSU-int-1, -2, -3, and -4). We screened for the existence of these four introns in the eight specimens listed in Table 1.

Four specimens, i.e., Chile-I (Diego Ramiríz Is., Chile), Ireld (Mulroy Bay, Ireland), Gmy (Helgoland, Germany), and USA (St. Lawrence Is., the Bering Sea, USA) had none of the four introns. In contrast, specimens collected at Hokkaido, Japan (hereafter, Japn) and at Puyuhuapi, Chile (hereafter, Chile-II) had only the fourth intron, LSU-int-4 (Fig. 1). Further, the complete set of four group II introns was detected in two specimens, Frnce-II (specimen collected at the same habitat as the Frnce-I, Roscoff, France) and Scotld (specimen collected at Shetland Is., Scotland) (Fig. 1). The geographical proximity of Frnce-II and Scotld is noteworthy. All of the detected introns were located at sites identical to those of the Frnce-I LSU introns. Therefore, the basic arrangement of introns in the LSU rRNA gene was identical among Frnce-I, Frnce-II, and Scotld. However, specimen-specific structural features were observed: the first two LSU introns detected in Frnce-I had a complete intronic ORF, and the third and fourth LSU introns contained no informative ORF, while none of the introns detected in Frnce-II and Scotld had any discernible ORF sequence in their LSU introns (Figs. 1, 2).

Fig. 1
figure 1

Schematically represented arrangement of mitochondrial group II introns in the LSU rRNA and cox1 genes of various Pylaiella littoralis specimens. Black regions show exons, while introns are depicted as white rectangles with the numbers upward. Gray area in a white region shows the intronic ORF. All DNA sequences of the introns are deposited in DDBJ/EMBL/GenBank under the accession numbers from AB281592 to AB281613. For abbreviation of the specimens, see Table 1

Full size image
Fig. 2
figure 2

Comparison of intronic ORFs of the LSU-int-1’s, -2’s, and -4’s in various specimens. ‘-’ shows an inserted gap to align the compared ORFs with higher sequence similarity. Gray box indicates a substitution in nucleotide or amino acid level, while black box shows deletion of a nucleotide or appearance of stop codon, when compared with the sequence obtained from the Frnce-I specimen

Full size image

The fourth LSU rRNA intron in Japn and Chile-II contained a nearly complete ORF: the ORF in Chile-II had two nucleotide deletions, while that in Japn carried one nucleotide insertion and an in-frame UAG stop codon (Fig. 2). In contrast, the fourth LSU-introns detected in Frnce-II and Scotld contained no informative ORF sequences. This was also the case for the previously reported LSU-int-4 in Frnce-I.

Fontaine et al. (1995) pointed out that the first three LSU introns of P. littoralis are closely related to one another at the primary sequence level and suggested the possibility that existence of these three introns in the LSU rRNA gene may be the result of cis-transposition along the gene. Similarly, the pair-clade of introns from a red algae, Porphyra/LSU-int-1 and -2 shown in Fig. 4 is a typical case of the result of cis-transposition of an intron. However, subsequently accumulated sequence data imply that such interpretation for the introns detected in P. littoralis might be incorrect because another group II intron in the plastidal petD gene of a green alga Scenedesmus obliquus has been shown to be closely related to the first two LSU introns (Zimmerly et al. 2001). Our phylogenetic tree of LSU introns showed that each of the three cognate intron types form a homogeneous cluster in relation to the intron insertion site. This suggests that no cis-transposition of these three introns recently occurred (Fig. 4).

Divergent number of group IIA introns in the cox1 gene

It has been reported that Frnce-I bears three group IIA introns in the cox1 gene (Fontaine et al. 1997). Nevertheless, none of the eight specimens examined in this study had the same set of three introns in their cox1 gene. One of the possible two-intron arrangements, i.e., cox1-int-1 and -3, was detected in Frnce-II, Japn, and Chile-I. Another combination of two-introns, i.e., cox1-int-2, and -3 was found in Scotld. A single intron, only cox1-int-1, was found in Gmy and USA (Fig. 1). Curiously, despite their geographical closeness, neither of the introns, cox1-int-1, and -3, present in the Chile-I was detected in Chile-II.

All ORFs detected in cox1-int-2, and -3 were intact tripartite-type, while all of the first cox1 introns detected in Japn, Gmy, and USA were interrupted by a stop codon located at distinctive sites, and in Chile-I, the ORF had a translational (−1)-frame shift (Fig. 1). In order to retrieve functional ORF sequence, appropriate editing of the original sequence data was essential (Fig. 3).

Fig. 3
figure 3

Comparison of group IIA intronic ORFs in the cox1-int-1’s. ‘-’ shows an inserted gap to align the compared ORFs with higher sequence similarity. Gray box indicates a substitution in nucleotide or amino acid level, while black box shows deletion of a nucleotide or appearance of stop codon, when compared with the sequence obtained from the Frnce-I specimen

Full size image

Each of the three kinds of cognate cox1 introns formed independent robust clades (Fig. 4). Interestingly, the assigned branch lengths for each cognate intron in the clades were quite short, and there was no significant difference in length among them. Therefore, each cognate intron in the same clade must have descended from a common progenitor that invaded the site rather recently. The cognate group IIA intron detected in the cox1 gene of a diatom Thalassiosira nordenskioeldii (Ehara et al., 2000) was positioned as a paired clade with the P. littoralis cox1-int-2 clade (see later).

Fig. 4
figure 4

Unrooted neighbor-joining trees of group IIB introns in the LSU rRNA gene (a) and group IIA introns in the cox1 gene (b) of various Pylaiella littoralis specimens. Numbers at nodes indicate bootstrap values. a Main part of the arm-IV loop sequence in the LSU-int-3’s, which is apparently a copy of the 3’-downstream region of the T7-like RNA polymerase gene, was not included in the alignment (see text). It is also the case for most intronic ORFs, because some of the cognate introns to be compared have lost a large portion of the region. Some group II introns (red alga Porphyra LSU-int-1 and -2, GenBank acc. no. AF114794, and cyanobacterium Calothrix intron, X71404) that grouped closely with the P. littoralis LSU-introns in the phylogenetic tree based on the reverse transcriptase domain were included (Zimmerly et al. 2001). b Some group II introns (fungus Neurospora cox1-int-1 and -2, X14669, yeast Saccharomyces cox1-int-1 and -2, V00694, and diatom Thalassiosira cox1-int-1a and -1b, AB037974 and AB038235) that grouped closely with P. littoralis cox1-introns in the phylogenetic tree based on the reverse transcriptase domain were included (Zimmerly et al. 2001). Number in parentheses shows the position of intron in the mitochondrial gene of the species. *904 in Thalassiosira (CCMP1096)/cox1 and Thalassiosira (CCMP992)/cox1 shows the corresponding intron position in the P. littoralis coxI gene, because the whole Thalassiosira coxI gene sequence is not determined

Full size image

DNA sequence polymorphisms among cognate introns

LSU-int-1

The specimens carrying LSU-int-1 were limited to those collected geographically close to or in the same area as Frnce-I, i.e., Frnce-II and Scotld. Among these three LSU-int-1 sequences, those detected in Frnce-II and Scotld were much more closely related to one another than Frnce-I. The evolutionary distance between the Scotld-intron and the Frnce-II-intron was very short [0.0047, i.e., only three nucleotides (nt) were different among 640 nt compared]. On the other hand, the evolutionary distances between the Frnce-I-intron and the Scotld-intron, and between the Frnce-I-intron and the Frnce-II-intron were about 16-fold longer than the above, 0.0757, and 0.0773, respectively. Reinforcing the above relationships were characteristic DNA sequences shared only between the Frnce-II-intron and the Scotld-intron in their ribozyme IIA region and also at the loop in arm-IV, as shown in Fig. 2. Close analysis of these disabled ORFs showed that the loop region in Frnce-II (91 aa) was longer than that detected in Scotld (54 aa) by 37 aa of the coding region (Fig. 2). Therefore, it is highly likely that additional deletions occurred in the lineage of Scotld after separation from Frnce-II. Altogether, we suggest that the Frnce-I LSU-int-1 originated from a recent invasion by a cognate intron. This conclusion derives from two significant observations: (1) this intron bears a complete ORF and (2) the evolutionary distance is extremely long relative to the other two cognate introns in the LSU-int-1 clade (Fig. 4).

LSU-int-2

Among the eight specimens examined, the existence of LSU-int-2 was limited to Scotld and Frnce-II; both of which are geographically related to Frnce-I. Frnce-I LSU-int-2 had a complete ORF in the arm-IV loop (Fig. 2). Further, the evolutionary distances among these three LSU-int-2 sequences were, in all cases, short [0.0104–0.0172 (on average 0.0138)], irrespective of the fact that one has a complete ORF and the other two introns do not. This suggests that these three introns descended from a common progenitor which had a complete ORF. Thus, a large portion of the ORF must have been deleted in the common ancestor of Frnce-II and Scotld, while the complete ORF has been maintained in the Frnce-I lineage.

LSU-int-3

In addition to Frnce-I, a cognate LSU-int-3 was found in two specimens collected at geographically related areas, i.e., Scotld and Frnce-II, as is the case for LSU-int-1 and -2. The 600-nt long arm-IV loop sequences were almost identical among these three specimens; however, their loop sequences showed no detectable homology to the typical group II intronic ORF. Therefore, as of the time these studies were performed, no LSU-int-3 that harbors a complete intronic ORF was detected from P. littoralis. Curiously, the 600-nt DNA sequence in the long arm-IV loop is identical to the 3′-downstream region of the T7-like RNA polymerase gene encoded in P. littoralis mitochondrial genome [i.e., the 256-bp IR3 region; Rousvoal et al. (1998)]. Therefore, it is apparent that the intronic ORF located in the arm-IV loop of the common ancestor of these three introns has been lost and subsequently replaced by a 600-nt long DNA fragment, which originally had no relationship to the intronic ORF. By close analysis, we found molecular features in the ribozyme sequence, which differentiate Scotld from Frnce-I and -II: Two insertions composed of 5 and 4 nt in addition to a deletion of 22 nt were common to the introns from Frnce-I and -II. Such variations must have occurred in the direct common progenitor of Frnce-I and Frnce-II after the divergence of the Scotld lineage.

LSU-int-4

Most intronic ORFs have been lost from LSU-int-4 in Frnce-I, Frnce-II, and Scotld. However, an ORF detected in Chile-II was nearly complete except for a deletion of two nucleotides required to rescue the functional ORF, and an ORF in the Japn intron was complete excluding a (+1)-frame shift and a stop codon (Fig. 2).

Irrespective of the tightness of the LSU-int-4 clade that is composed of the five introns (from Scotld, Chile-II, Frnce-I, Frnce-II, and Japn), the Japn LSU-int-4 was obviously separated from the other four LSU- (core cluster-making) introns within the clade (Fig. 4). It is notable that even among the core cluster-making introns, there is heterogeneity in the ORF sequences: the Chile-II-intron carried a nearly complete intronic ORF, while the others, i.e., the introns from Scotld, Frnce-I, and Frnce-II, had only a short loop (Fig. 2). Therefore, it is highly likely the common ancestor of Scotld, Chile-II, Frnce-I, and Frnce-II had a complete intronic ORF. It appears that along the lineage leading to Scotld, Frnce-I, and Frnce-II, most of the ORF was lost, while it was retained only in the Chile-II lineage. Considering all of the above, we suggest that Japn LSU-int-4 was the result of recent invasion of a cognate intron at the previously intronless target site. This is supported by two facts: (1) the intron is phylogenetically distant from the other four LSU-int-4 sequences and (2) it has a nearly complete intronic ORF, as observed in Frnce-I LSU-int-1.

Intriguingly, the ORF detected in the Japn intron and the Chile-II intron showed significant similarity to that found in the group II introns located in the LSU rRNA gene of a red alga Porphyra purpurea (Burger et al. 1999) (i.e., Porphyra/LSU-int-1 and -2 in Figs. 4, 5, 6) and that in a cyanobacterium Calothrix sp. (Ferat and Michel 1993) (Calothrix/intron in Figs. 4, 5, 6). Moreover, the intron insertion site of the first red algal intron (Porphyra/LSU-int-1) corresponded exactly to that of the fourth intron in the P. littoralis LSU rRNA gene.

Fig. 5
figure 5

Aligned amino acid residues of intronic ORFs in the LSU-int-4 of Chile-II and Japan, and the related intron from Porphyra and Calothrix. Segment naming in the reverse transcriptase domain (RT-I through RT-VII) is based on Michel and Lang (1985)

Full size image
Fig. 6
figure 6

Phylogenetic tree based on the amino acid sequences of the intronic ORFs of group II introns in the LSU rRNA and cox1 genes of Pylaiella littoralis and the related introns from Porphyra, Calothrix, and a green alga Scenedesmus (P19593). Number in parentheses shows the position of intron in the mitochondrial gene of the species

Full size image

cox1-int-1

In addition to that detected in Frnce-I, a cox1-int-1 was detected in 6 of 8 specimens examined in this study: only Scotld and Chile-II were lacking this intron. A complete ORF was detected in Frnce-II and Ireld similar to Frnce-I. On the other hand, a most likely disabled but easily alignable ORF was found in Japn, Gmy, Chile-I, and USA: each of the ORFs detected in Japn, Gmy, and USA had stop codons at distinctive sites, while the intronic ORF detected in the Chile-I intron had a (−1)-frame shift. The corresponding evolutionary distances among these seven introns were very short (0.0045–0.0140, on average 0.0095). This suggests that all of these cox1-int-1 sequences were derived from a progenitor that had a complete ORF. Subsequent divergent evolution resulted in base substitutions or deletions that appeared in each of the four lineages, independently.

cox1-intron-2

Reflecting the variation among lineages as compared to Frnce-I, the second intron in the cox1 gene, i.e., cox1-int-2, was only detected among the eight newly examined specimens in Scotld. The intron detected in Scotld had a complete ORF as in the Frnce-I intron, and the distance between the two introns was only 0.0091. Therefore, we suppose the Scotld cox1-int-2 and the Frnce-I cox1-int-2 had a common ancestor that had a complete ORF. A cognate intron, which is inserted at the corresponding site, has been reported in a diatom Thalassiosira nordenskioeldii (Ehara et al. 2000). However, its ORF is an incomplete-type composed of only RT and maturase domains (e.g., Thalassiosira/cox1-int-1a and -1b in Fig. 4). Contrasting with the various arrangements of P. littoralis introns reported in this study, introns similar to that described in the diatom were detected in all four examined specimens, irrespective of the discrete oceanic source from which they were collected. Noteworthy are the distinctive insertions and deletions identified among these specimens (Ehara et al. 2000).

cox1-int-3

Scotld, Frnce-II, Chile-I, Ireld, and Japn had a cox1-int-3 as observed in Frnce-I. All of the cox1-int-3 sequences detected in these specimens had a complete ORF. Moreover, the corresponding distances among them were short, 0.0046–0.0221 (on average 0.0176). Therefore, it is most probable that these introns descended directly from a common ancestor of all of these six specimens.

Discussion

Recurrent invasion of cognate introns provides the most probably explanation for our data presented here. This conclusion is based on the following lines of evidence that some examined introns had an exceptionally long evolutionary distance against the cognate introns found in P. littoralis specimens (Fig. 4), and these introns harbored an intact or nearly intact tripartite ORF (Figs. 2, 3).

For the LSU rRNA gene, the average ratio of occupied to unoccupied intron target sites among all specimens examined was 14/36 (the LSU rRNA gene has four potential intron target sites, and a total of 14 group II introns were detected in nine different specimens), while the ratio for the cox1 gene was 15/27 (three target sites in the gene from nine different specimens). Therefore, more than half the target sites were occupied by group II introns in the case of the cox1 gene, while only 1/3 were occupied for the LSU rRNA gene of which no introns were detected for four of the specimens examined, Chile-I, Ireld, Gmy, and USA. Based on these observations, it is evident that group II intron target sites are frequently left intronless. Thus, unoccupied target sites are common in nature. Goddard and Burt (1999) determined that opportunity of intron infection is the rate limiting step in the life cycle of group I introns. This strongly implies that the bottleneck step in the group II intron cycle is also the opportunity for intron infection. Our assertions are supported by several experiments demonstrating that intron invasion occurs very efficiently through interspecific crosses between intron-bearing and intron-deficient species (Bussieres et al. 1996; Kurokawa et al. 2006), or through supply of a target site-carrying plasmid (Conlan et al. 2005).

In addition, our data illustrate that ORF degeneration occurs shortly after intron invasion. To this end, we observed that within a tight clade of cognate introns, complete ORF-bearing introns and those lacking a larger portion of it were grouped together (i.e., clades of LSU-int-2 sequences and LSU-int-4 sequences). Therefore, the evolutionary closeness of the intron sequence overrides the rapidly diverging sequence of the ORF and provides a model for determining the sequence of intron invasions among species.

The entire group IIA introns detected in the cox1 gene had an intact or easily reconstructed ORF (15/15), while such a ratio was only 28.5% (4/14) for the group IIB introns in the LSU rRNA gene. Based on the fact that the distances within the detected group IIA introns and the group IIB introns were not significantly different, it appears that ORFs located in group IIA introns generally tend to be inherited and maintained more successfully.

Obviously, loss of the entire intronic DNA is necessary to restore the target site before retro-transposition by a cognate intron can occur. However, the mechanism by which the whole intron-encoding DNA region disappears from the genome remains an open question. A simple hypothesis is that reverse transcription of a fully matured mRNA occurs in the matrix of the organelle and, shortly thereafter, the intronless gene replaces the intron carrying gene by homologous recombination. Actually, homologous recombination occurs very efficiently in the mitochondria and chloroplasts of a green alga Chlamydomonas reinhardtii (Boynton et al. 1988; Yamasaki et al. 2005), and further, it has been shown that reverse transcription of a matured mRNA occurs in mitochondria (Geiss et al. 1994; Gargouri 2005). One explanation for the higher ratio of intronless target sites in the LSU rRNA genes as compared to the ratio observed in the cox1 genes is that the matured LSU rRNA is more apt to be reverse-transcribed than the matured cox1 mRNA. Therefore, the chance of intron infection is more limited for the cox1 gene than the LSU rRNA gene. After the target site has been recovered by these means, the site will most likely remain intronless for a significant period of time, because the opportunity for intron invasion is thought to be very limited, as described above. We suppose the existence of the two-intron arrangements detected in the cox1 gene is not the result of loss of one intron from the three-introns set, but most probably the result of independent insertions of two independent group IIA intron insertion events. Appearance of a completely intronless gene by homologous recombination of a reverse transcript of the fully matured mRNA, followed by independent intron invasion, can easily explain the variable intron number in the cox1 and LSU rRNA genes detected in this study.

Among the eight areas where specimens were collected, Scotland, Ireland, Germany, and France are more geographically close than the others. This most likely reflects the geographical relationships among the two specimens from France (Frnce-I and -II) and one specimen from Scotland which were shown to have a similar number of introns both for the LSU rRNA and cox1 genes. However, similar intron arrangement and geographical proximity of specimens was not always correlative. For example, the arrangement of introns in Scotld and Ireld showed no sign of similarities excepting the commonly existing cox1-int-3, despite their geographical closeness. On the other hand, specimens collected from very distant geographic locations showed identical intron organization, as observed between Chile-I and Ireld (each specimen had a total of two introns), and between Gmy and USA (each specimen had only a single intron) in their cox1 and LSU rRNA genes (Fig. 1). Such similarities could merely be due to coincidence. It is possible that similar intron arrangements will occur independently with much higher frequency, when the total number of detected introns is limited to a few, because the intronless state lasts longer than the intron-bearing state.

Filamentous brown algae such as Ectocarpus, Hincksia, and Pylaiella are common fouling species. Therefore, considering the long history and frequency of marine transportation in Europe, it is possible that some of the Pylaiella populations analyzed in the present study may have originated from specimens introduced by the shipping industry to non-native environments.

The extraordinarily varied number of introns in P. littoralis might be related to its reproduction system. So far, the only observed reproduction system of P. littoralis is limited to a type of asexual reproduction through zoospore (Müller and Stache 1989). In the case of asexually reproducing organisms, a newcomer intron will only be successfully transferred to the direct descendants of the infected individuals, which will lead to significant variations of intron number within a population.

In this study, we have shown that Frnce-I LSU-int-1 and Japn LSU-int-4 are most likely the result of recent invasion of cognate introns at intronless target sites. Considering the recurrent invasions of almost identical introns within a short time span, we suppose the source of these cognate introns is most likely symbiotic organisms of P. littoralis. Actually, hexagonal virus-like particles are observed within the unilocular sporangium of P. littoralis (Markey 1974). Some viruses that infect brown algae possess extraordinarily large genomes (about 320 kb long) (Delaroque et al. 1999); therefore they might have several group IIA and IIB introns in their genomes and could serve as vectors for these introns. Fontaine et al. (1995) pointed out that differential viral transmission could account for the strain-specific distribution of introns in this brown alga. We suppose symbiotic fungi, which are generally rich in introns in their mitochondrial genomes (Gray et al. 2004), might be the potential source of the group II introns detected in this study. Considering that the LSU rRNA gene of Japn carries a recently acquired LSU-4-intron while three intron target sites are kept vacant, the host vector for the LSU-4-intron and LSU-1, -2, -3-introns may be unique for each intron type. Such a situation might also be the case for introns in the cox1 gene.