DNA variation in the phenotypically-diverse brown alga Saccharina japonica

Background Saccharina japonica (Areschoug) Lane, Mayes, Druehl et Saunders is an economically important and highly morphologically variable brown alga inhabiting the northwest Pacific marine waters. On the basis of nuclear (ITS), plastid (rbcLS) and mitochondrial (COI) DNA sequence data, we have analyzed the genetic composition of typical Saccharina japonica (TYP) and its two common morphological varieties, known as the “longipes” (LON) and “shallow-water” (SHA) forms seeking to clarify their taxonomical status and to evaluate the possibility of cryptic species within S. japonica. Results The data show that the TYP and LON forms are very similar genetically in spite of drastic differences in morphology, life history traits, and ecological preferences. Both, however, are genetically quite different from the SHA form. The two Saccharina lineages are distinguished by 109 fixed single nucleotide differences as well as by seven fixed length polymorphisms (based on a 4,286 bp concatenated dataset that includes three gene regions). The GenBank database reveals a close affinity of the TYP and LON forms to S. japonica and the SHA form to S. cichorioides. The three gene markers used in the present work have different sensitivity for the algal species identification. COI gene was the most discriminant gene marker. However, we have detected instances of interspecific COI recombination reflecting putative historical hybridization events between distantly related algal lineages. The recombinant sequences show highly contrasted level of divergence in the 5’- and 3’- regions of the gene, leading to significantly different tree topologies depending on the gene segment (5’- or 3’-) used for tree reconstruction. Consequently, the 5’-COI “barcoding” region (~ 650 bp) can be misleading for identification purposes, at least in the case of algal species that might have experienced historical hybridization events. Conclusion Taking into account the potential roles of phenotypic plasticity in evolution, we conclude that the TYP and LON forms represent examples of algae phenotypic diversification that enables successful adaptation to contrasting shallow- and deep-water marine environments, while the SHA form is very similar to S. cichorioides and should be considered a different species. Practical applications for algal management and conservation are briefly considered.


Background
Genetically-based approaches (especially using multilocus data) have resolved much phylogenetic uncertainty in large-scale algal trees of life (e.g., [1][2][3]). Less obvious progress, however, has been accomplished concerning the species level of algal taxonomy, when there is an abundance of morphological variants with uncertain taxonomic status. For instance, the genus Laminaria (that has been recently separated into two different genera, Laminaria and Saccharina; [1]) is rich with morphological types, varieties, ecotypes, and forms that have puzzled taxonomists since the very beginning of research on this subject [4][5][6][7][8][9][10][11][12][13][14]. Whether the distinct morphologies observed in many algae represent different species or only intraspecific forms is one of the main unresolved questions.
In this paper we focus on the most common and commercially important laminarialean species, Saccharina japonica [4,5,15] and its morphological forms, inhabiting contrasting depths of the northwest Pacific region. The species is morphologically and ecologically highly variable and some intraspecific forms have been described [10][11][12][13]. Typical S. japonica (TYP) inhabits the littoral zone at the preferred depths of 5 -11 m with wide distribution over the full species area. Its full distribution pattern, however, remains uncertain owing to the difficulty of reliable species identification. It grows along the sea shore on rocky substrate [5,6,[15][16][17]. (See Additional file 1 for a more detailed species description.) At present the populations of the TYP form are depressed due to overharvesting and possibly global climate changes [18,19].
The deep-water (or "longipes") S. japonica form (LON) inhabits the sublittoral zone at preferred depths of 14 -25 m. This endemic form has a restricted distribution in the Sea of Japan and the Sea of Okhotsk. It grows in compact settlements at a significant distance (300 -1000 m) from the seashore [6,16,17,20,21]. The LON form thallus is 2 -3 times longer and 3 -4 times heavier than the thallus of the TYP form. Differences in morphology and preferred ecology are detectable already among algae during the first year of growth. Originally the LON form was described as a deep-water variant of S. japonica [4,6]; a new nomenclatural designation, S. japonica f. longipes, has been suggested recently [13]. The TYP and LON forms have significant differences in morphology, anatomy, reproductive biology, ecology, and other important features (Additional file 1). The opportunity for interbreeding between the LON and TYP forms is considered highly unlikely because the formation, release, and appearance of the mature gametes are separated in space and time. The zoospore spreading time of S. japonica (and the close species, S. angustata) is short (maximum 24 hours) leading to limited free dispersal gamete range (on average no more than 3.5 meters [19,22]). Also the distribution areas of the LON and TYP forms are separated geographically (not less than three miles from each other) and their reproductive periods do not overlap (Additional file 1). The morphological differences between the TYP and LON forms of S. japonica are stable and easily recognizable and they are systematically reported from the northern Primorye coast region [20,21]. The drastic differences between the TYP and LON forms have motivated changing the taxonomical status of the LON form, so that it would be considered a separate species [17]. However, transplant experiments [23,24] show that zoospores of the LON form released in the habitat of TYP form produced algae morphologically indistinguishable from the TYP form, which did not support the biological species status of the LON form, but rather suggested high phenotypic plasticity for the species.
The shallow-water form of Saccharina japonica (SHA) inhabits the supralittoral zone (0.1 -0.5 m depth) and it is widely distributed in the Primorye coast region, the Sea of Japan, including Peter the Great Bay. It is commonly accepted that the SHA form represents an intraspecific ecotype of the TYP S. japonica form. The zoospores of the SHA form are used to enhance the strongly overharvested resources of the TYP form [19]. Nevertheless, the SHA and TYP forms have clearly distinct morphological features (Additional file 1). There are also significant differences between the forms in heat and salinity tolerance [25]. The SHA form is more heat and desalination resistant and survives better than the TYP form under hyposalinity osmotic stress and increasing temperature conditions. It was suggested [26] that the SHA form may represent a subspecies of Laminaria (Saccharina) angustata adapted to the supralittoral zone.
We analyze the genetic composition of S. japonica and its common morphological forms, LON and SHA, seeking to clarify the taxonomic status of the forms and to evaluate the possibility of cryptic species within S. japonica, using DNA nucleotide sequences of three genes from mitochondrial, plastid, and nuclear genomes (4,286 bp total in a concatenated dataset). We also test for homogeneous phylogenetic signal within and among the marker loci (ITS, rbcLS, and COI). We have found that the TYP and LON morphological forms are genetically very similar; consequently they need not be considered as different biological species. However, taking into account the potential evolutionary roles of phenotypic plasticity, and its relevance to adaptive evolution and speciation in many organisms (e.g., [27][28][29]) including algae (e.g., [30,31]) we propose that the TYP and LON forms may represent an important example of algal diversification that enables successful adaptation to contrasting shallow-and deepwater marine environments. The SHA form has clear genetic differences from the TYP and LON forms and it might represent a morphological variant of S. cichorioides expanded to the supralittoral environment. The practical implications for mariculture are that all three forms should be considered separate evolutionary lineages and that restoration programs should include genetic data to avoid confounding.

Algae samples
The specimens of S. japonica morphological forms were collected from the Primorye coastal region, Sea of Japan. The TYP (five specimens) and SHA (four specimens) forms were collected near the Cape Dal'niy at depths of 6.0 m and 0.5 m, respectively. The LON form (three specimens) was collected near the Cape Zolotoi at depths of 15.0 m. We also analyzed first year individuals belonging to the S. japonica typical form (TYPF, four specimens), collected near the Cape Dal'niy; unidentified Saccharina species (CHE, four specimens; depth of 0.5 m), TYP form growing in close proximity with unidentified Saccharina species (TYPA, four specimens), and a sample of the TYP form without middle line (TYPW, a single specimen), all collected in the Bay Chernoruch'e near the Cape Khitrovo. The algae were identified by eye in the field.
Model-based phylogeny reconstructions were performed with the rbcLS, COI, ITS, and concatenate sequence alignments using the neighbor-joining (NJ) and maximum-likelihood (ML) methods in MEGA, version 5 [44]. MEGA5 or jModelTest [45] were used to find the best model of substitution under the maximum likelihood criterion. Models with the lowest Bayesian information criterion [46] scores were considered most appropriate to describe evolution for each gene; the Akaike Information Criterion [47] was also applied. Maximum likelihood bootstrap analyses [48] consisted of 1000 replicates.
Partitioned analyses were performed with GARLI, version 2.0 [49] and MrBayes, version 3.2 [50] that allowed the overall rate to be different across partitions (rbcLS, COI, and ITS). Substitution model parameters were estimated separately for each gene. The most appropriate models were Tamura 3-parameter plus gamma for rbcLS, Hasegawa-Kishino-Yano plus I for COI, and Tamura 3parameter for ITS.
The complete dataset was also analyzed by Bayesian inference using MrBayes, version 3.2 [50] under a general-time-reversible model plus gamma plus I. Proportion of invariable sites was uniformly distributed on the interval (0.00, 1.00). Gamma distribution was approximated using 4 categories. Analyses were performed as two independent runs, each with four incrementally heated Metropolis-coupled Monte-Carlo Markov Chains running for one million generations. Output trees and data were sampled every 500 generations. Likelihood values reached a plateau within approximately 1,000 generations. A total of 4002 trees in two files were read and 3002 of them were sampled. The numbers of unique site patterns were 100, 175, and 107 for rbcLS, COI, and ITS, respectively. It showed that COI as the most efficient gene marker consisting with the distance-based methods.
The average standard deviation of split frequencies at the end of the run was 0.0068 indicating stationary conditions. The log likelihood values increased from below −35292.434 to around −9652.102 in the first five thousand generations and then to around −9547.021 after one million generations. The likelihood of best state for "cold" chain of run 1 was −9523.80 and the likelihood of best state for "cold" chain of run 2 was −9524.14. At the end of run there were not any trends in a plot of generation versus the log probability of the observed data (the log likelihood values). A convergence diagnostic, the Potential Scale Reduction Factor (PSRF) [51] was between 1.00 and 1.003 indicating a good sample from the posterior probability distribution.
The topologies obtained with neighbor-joining and maximum-likelihood methods as well as with Bayesian inference were very similar. The close congruence could be explained by the fact that 1) the dataset was relatively straightforward and included just two main compared groups of sequences (TYP and SHA) and four outgroup species (S. coriacea, S. angustata, Laminaria digitata, and Agarum clathratum); 2) the extent of sequence divergence between the TYP and SHA forms for all gene regions was less than 5.0% (0.9% for rbcLS, 1.1% for ITS, and 4.8% for COI); and 3) the length of alignment was long enough (4.2 kb totally). As has been shown by many authors, the relative efficiencies of the ML and NJ methods in obtaining the correct tree topology were very close under these conditions [52][53][54]. The only difference between the two methods in relation to our dataset concerned slightly different bootstrap values. To be conservative we show the lowest bootstrap values obtained with the ML method. When different methods produced similar or identical topologies the simplest, NJ method, was preferable.

Nucleotide diversity and divergence
We sequenced three gene regions (total 4,286-bp alignment) in 25 algae samples representing common morphological forms of S. japonica. All three gene regions (rbcLS, COI, and ITS) with comparable length were available in GenBank for some species related to S. japonica, which we have included in the analysis. The lengths of the GenBank concatenated sequences are (in parentheses): S. japonica (3,182 bp), S. diabolica (2,894 bp), S. longipedalis (2,896 bp), S. ochotensis (2,689 bp), S. religiosa (2,896 bp), S. angustata (4,256 bp), S. coriacea (2,711 bp), S. latissima (3,748 bp), and L. digitata (4,245 bp) (see Additional file 3 for GenBank accession numbers and full species names). Figure 1 shows 114 nucleotide substitutions detected in three gene regions (18 sites in rbcLS, 88 sites in COI, and 8 sites in ITS) among the 25 algae samples. In addition, there were seven length polymorphisms in the rbcLS intergenic spacer and ITS regions (Additional file 4); no length polymorphisms were found in the COI gene. All algae samples sharply split up in two groups. The first group includes all samples of the TYP (including TYPF, TYPA, and TYPW) and LON forms; the second group includes the SHA form and one unknown species from the Chernoruch'e bay (CHE). The genetic structure is completely congruent for all three gene regions studied ( Figure 1). The two groups (denoted as "TYP" and "SHA" lineages) differ by 109 fixed single nucleotide differences, as well as by seven fixed length polymorphisms. The lineages are differentially associated with indels. The TYP lineage is completely associated with three inferred deletions (13-, 4-, and 8-bp, within the rbcLS intergenic spacer and ITS region; ▲1, ▲2, and ▲3, Figure 1) and two inferred insertions (▼1 and ▼4) within the ITS region. The SHA lineage is completely associated with two inferred 1-bp insertions within the ITS region (▼2 and ▼3). The difference between the TYP and SHA lineages based on the full concatenate (including the three gene regions, rbcLS, COI, and ITS) is highly significant (F st = 0.9949, P < 0.00001); the total sequence divergence (D xy ) between the lineages is 0.0261 (ignoring indels).
The three gene regions differed in levels of divergence between the TYP and SHA lineages (Figure 2). Total nucleotide divergence between TYP and SHA lineages is low for rbcLS (D xy = 0.0099 ± 0.0025) and ITS (D xy = 0.0106 ± 0.0037), but significantly higher (more than four times) for COI (D xy = 0.0481 ± 0.0051). The same tendency is found for comparisons between all other laminarialean species pairs (data not shown). Thus the COI gene represents the most effective and sensitive gene marker of the three studied here for the comparison of close Saccharina species, which is in accordance with the data obtained for red algae and kelps (e.g., [55][56][57]).
The estimates of nucleotide diversity for all sequences of S. japonica and for the morphological forms or haplotype lineages separately are low. In the pooled sample, the total nucleotide diversity is similar for the TYP (π = 0.00011) and SHA (π = 0.00016) lineages (based on the three gene regions studied); the results are in accordance with literature data on intraspecific variability in S. japonica [58]. Figure 3 displays a maximum likelihood phylogram of the three gene regions separately and of the concatenate sequences obtained for the S. japonica forms in the present study, along with other Saccharina and Laminaria sequences obtained from GenBank. S. ochotensis, S. religiosa, S. longipedalis, and S. diabolica are conspesific with S. japonica (data not shown), as has been revealed previously [10,11]. All trees showed that the sequences from the TYP and LON forms form a single clade, with no evidence of discrete species heterogeneity (F st = 0.0201; P = 0.6600; K st = − 0.0099; P = 0.7480); total sequence divergence (D xy ) is 0.0001. These data suggest that the TYP and LON morphological forms of S. japonica are not distinct biological species. In contrast, the sequences from the TYP and SHA lineages are distinct. The difference between the lineages is highly significant (F st = 0.9949; P <0.0001; K st = 0.8484; P <0.0001). The distance between them is 2.61% (2.77% including indels), which is in a range of the divergence between species observed within the genus Saccharina [1,[41][42][43] suggesting that the TYP and SHA lineages would qualify as different species based on a phylogenetic species concept.

Species identity
Comparison of sequences obtained during this study with those available in the GenBank database suggests that the TYP lineage represents S. japonica, whereas the SHA lineage is closely related to a group of Saccharina species that include S. coriacea, S. cichorioides, S. sachalinensis, and S. yendoana (the last three species are not represented in Figure 3 because the COI sequences are not available for them). The distance between the SHA lineage and this group of Saccharina species (represented by S. coriacea in Figures 1 and 3) is low and not significant (F st = 0.9242; P = 0.3300; K st = 0.3890; P = 0.3300); the rbcLS and ITS regions are identical but the COI gene shows seven single nucleotide differences (Figure 1). Total sequence divergence is 0.38% (for the full 4,286-bp concatenate alignment), which is below the species divergence level observed in algae [1,41] and other eukaryotes [59]. Yotsukura and colleagues [8,9,11,12] have revealed high genetic similarity between S. coriacea, S. cichorioides, S. sachalinensis, and S. yendoana and suggested their conspecificity, in accordance with earlier results of Petrov [6], who treated L. coriacea and L. sachalinensis as intraspecific forms of L. cichorioides (the status of L. yendoana was not discussed by Petrov [6].) Following the priority rule of the International Code of Botanical Nomenclature, Selivanova and colleagues [13] introduced a new nomenclature combination: S. cichorioides with four intraspecific forms. The conspecificity of S. coriacea, S. cichorioides, S. sachalinensis, and S. yendoana was suggested on the basis of low genetic divergence detected with ITS-1, ITS-2, RubisCO spacer, and rbcL sequences [8,9,[11][12][13]60]. In the present work we have detected low divergence between the SHA lineage and the group of Saccharina species listed above using all three gene markers, including the COI gene (D xy = 0.0050), which is considerably more informative than the rbcLS and ITS markers (see above). Consequently, we conclude that the SHA form belongs to the S. cichorioides group (all "species" representing the group were previously synonymized based on genetic data as pointed out) and it may represent a morphological ecotype adapted to supralittoral environments. S. cichorioides is more heat resistant than S. japonica and its main populations are more southward in the Sea of Japan. However, this morphologically and ecologically versatile species (as it follows from our data) is capable of occupying different environments and successfully compete with local species in the northern Primorye coast region under hyposalinity osmotic stress and increased water temperature (see Background).
The unknown Saccharina collected in the Bay Chernoruch'e (CHE) are genetically identical to the SHA form, even though they have substantial differences in external morphologies (T. N. Krupnova, unpublished observations) but occupy similar depths (0.5 m). It has been suggested that laminarialean algae from the supralittoral zone represent S. angustata [26]. However, this is not the case (see Figure 3): the SHA form is close to the S. cichorioides group (represented by S. coriacea in Figure 3) and clearly different from S. angustata. The distance between the SHA form and S. angustata is 2.63% (which is consistent with interspecific divergence Figure 3 Phylogenetic trees of the rbcLS, COI, ITS, and concatenate sequences (4,286 bp total) of Saccharina japonica morphological forms. The topology of trees obtained with Maximum likelihood and Neighbor-joining methods were congruent. Numbers at the nodes are bootstrap percent support values based on 1,000 replications in Maximum likelihood analysis. Saccharina latissima was excluded from this analysis due to recombination in the COI gene (see text). Other comments as in Figure 1. level), indicating species taxonomical range for these two evolutionary lineages. Thus, S. angustata was not found among 25 algal samples collected from distant areas of the Northern Primorye region. This species is common in the Southern Kuril Islands and Japan islands [4,26], but it rarely occurs (if at all) along the continental coast of the Japan Sea.
Non-uniform pattern of divergence in the COI gene Full length of the COI coding region in laminarialean algae is around 1.6 kb (1,602 bp in Saccharina). The 5' part of the gene (~650 bp in length; 5'-COI) has been chosen as a "barcoding" region [61] and recommended for algal species identification [41][42][43]61]. We have shown above that the most pronounced differences between Saccharina species are detected within the COI gene. However, the level of divergence is highly nonuniform within the COI gene in the comparisons, including S. latissima and other Saccharina species ( Figure 4A -D). The average divergence for the 5'-COI is 0.0380 ± 0.0106 but 4.2 times more, 0.1579 ± 0.0034 for the 3'-COI (paired samples t-test P = 0.0001; Mann-Whitney test P = 0.0090; Table 1). This difference is also obvious for the comparisons with other laminarialean species for which long enough COI sequences are available in GenBank [2,11]: Laminaria digitata, Alaria esculenta, Ecklonia radiata, Undaria pinnatifida, and Agarum clathratum. The average divergence for the 5'-COI between these species is 0.1011 ± 0.0051 but 1.5 times more, 0.1516 ± 0.0030, for the 3'-COI (paired samples t-test P = 0.0022; Mann-Whitney test P = 0.0090 ( Table 1). The pattern is the same for all species pair comparisons: the 5'-COI (~740 bp in length) was significantly less diverged between S. latissima and other laminarialean algae than the 3'-COI (~862 bp in length) (Figure 4A -D; Table 1).
The difference in the level of divergence between the S. latissima and other species for the 5'-COI and 3'-COI regions is highly pronounced (Figure 4; Table 1). The BLAST procedure (limited by the current GenBank submissions) reveals, as expected, very high identity between the 5'-COI region of S. latissima and other Saccharina species; however, the 3'-COI region of S. latissima has paradoxical similarity (but not complete identity) to a number of species belonging to the order Ectocarpales. Indeed, there is a noticeable decrease in the divergence for the 3'-COI region between S. latissima and the following Ectocarpales species: Leathesia difformis, Asperococcus bullosus, Punctaria latifolia, Elachista fucicola, Hincksia granulosa, Scytosiphon lomentaria, and Petalonia fascia. The average divergence for the 5'-COI between S. latissima and these species is 0.1642 ± 0.0061, but significantly less, 0.1389 ± 0.0046, for the 3'-COI (paired samples t-test P = 0.0007; Mann- Whitney test P = 0.0087; Table 1). Other Ectocarpales species for which long enough COI sequences are available (Pylaiella littoralis, Ectocarpus siliculosus, and Colpomenia peregrina) do not show any difference in the divergence level between 5'-COI and 3'-COI regions ( Table 1).

Phylogenetically discordant signals within the COI gene
Thus, the difference in the level of divergence between the 5'-COI and 3'-COI regions is significant and exhibits an opposite pattern for comparisons between S. latissima vs. Laminariales species and S. latissima vs. Ectocarpales species: the 5'-COI region of S. latissima has obvious similarity to Saccharina species, whereas the 3'-COI regions has unexpected similarity with some Ectocarpales species. As a consequence, the position of S. latissima on 5'-and 3'-COI based trees are sharply different ( Figure 5); on the 5'-COI based tree, S. latissima is within the same cluster with other Saccharina species but not distinguishable from S. coriacea ( Figure 5A). On the 3'-COI tree S. latissima is significantly different from Laminariales algae and clusters with some species of the order Ectocarpales ( Figure 5B). On the full length COI tree, S. latissima is within the order Laminariales ( Figure 5C), but significantly different from other Saccharina species. The high similarity of the S. latissima 3'-COI region with Ectocarpales is surprising considering the large evolutionary distance between the two brown algae orders, which separated around 100 Ma [2].

Recombination in the COI gene
Based on these observations we suspected that the unusual patterns of the COI gene divergence in comparisons of S. latissima with other species might reflect historical hybridization event(s) between representatives of distantly related Ectocarpales and Laminariales algae, which might have resulted in interspecific recombination of mitochondrial DNA. We therefore analyzed the COI alignments for evidence of recombination (and breakpoints) using various recombination detection methods implemented in the program RDP3 [62] ( Table 2). The parental and recombinant sequences were determined  using the VisRD method [63], modified version of PHYLPRO [64], and EEEP [65] also implemented in RDP3 (default settings). All six methods detected a single recombination event in S. latissima within the 1,257 bp COI region ( Figure 6A), with high statistical support ( Table 2). The COI sequences from Petalonia fascia (Ectocarpales) and the SHA form (Laminariales) were major and minor parents, respectively. The recombination event involved the 3'-COI region, encompassing a 542 bp-long segment, between positions 1227 and 685.
Thus, the parental COI sequences of S. latissima come from different algae orders, Ectocarpales and Laminariales. The recombination would not have been detected if the alignment would be limited to the 5'-COI barcoding region. An interesting feature of algal evolution would not have been detected. Moreover, incorrect phylogenies would be obtained for species with historical hybridization events (see Figure 5).
To understand the uniqueness of this case of historical hybridization revealed in S. latissima, we analyzed additional, long enough COI sequences (~1200 bp; Additional file 3) of brown algae obtained from GenBank. We analyzed 57 sequences (GeneBank's plus our own sequences) and detected one more highly significant instance of COI recombination between representatives of brown algae belonging to the families Sargassacea and Seirococcaceae. A recombinant COI sequence was revealed in Cystophora retorta (order Fucales, family Sargassacea). The COI sequences from Phyllospora comosa (order Fucales, family Seirococcaceae) and Bifurcaria bifurcata (order Fucales, family Sargassacea) were the major and minor parents, respectively. The recombination spanned the 5'-COI region ( Figure 6B) with a beginning breakpoint at position 540 and ending breakpoint position 16. This recombination event involves representatives of two brown algae families, Sargassacea and Seirococcaceae, which diverged around 73 Ma [2]. Also, as in case of S. latissima, different and misleading phylogenetic patterns would come about depending on the COI region used in tree reconstruction (data not shown).
The recombination analysis has confirmed that the evolutionary histories of S. latissima and C. retorta have been influenced by gene flow between two highly diverged algae lineages, the orders Ectocarpales and Laminariales (for S. latissima) and between the families Sargassacea and Seirococcaceae (for C. retorta). The COI gene in S. latissima and C. retorta represents historical recombinant products between two distantly related ancestors. The level of COI divergence is 14.3% and 14.7% between the parental sequences of C. retorta and S. latissima, respectively, which seems very high for the Representative sequences of the orders Laminariales (L) and Ectocarpales (E) included in these trees are marked by vertical lines. Red algae COI sequences of Callithamnion pikeanum and Hypnea nidulans (GenBank accession numbers EU194965 and FJ694907, respectively) are used as outgroups. Note the changed position of S. latissima (in bold) depending on the COI region used for the tree. The S. latissima COI sequence denoted as "S" is from [2]. See Additional file 3 for GenBank accession numbers. Other comments as in Table 1 and Figure 3. occurrence of successful recombination (see, for instance, [73]). However, it has been shown [74] that recombination may occur among highly divergent maternal and paternal mtDNA sequences of the sea mussel Mytilus galloprovincialis that differ by > 20%.
There are multiple evidences of mtDNA recombination in plant, fungi, and animals, including human (reviews in [75][76][77]). Mitochondrial heteroplasmy (the presence of more than one mitochondrial genome in an individual) that could facilitate recombination has been reported in the brown alga Fucus [78,79]. Hybrid speciation is widespread in plants [80][81][82] and mosaic genomes due to hybrid speciation are common in many marine microorganisms, plants and animals (review in [83]). Interspecific hybridization has been detected in the brown algal genus Fucus [84][85][86][87] and the red algal genus Porphyra [88]. However, instances of interspecific mtDNA recombination were not, to our knowledge, previously known for algae; it is an infrequent phenomenon detected in fish and primates [89,90].
The two examples of interspecific recombination come from published work by Silberfeld with colleagues [2]. It might be suggested that these recombinations are artificial, resulting from PCR errors. However, the architecture of the recombinant sequences is not a mosaic composition from different parts of the parental sequences; the recombinants are similar but not identical to any other sequence used in this study. These observations suggest that PCR errors are unlikely to account for these recombinant sequences. Nevertheless, more sampling will be necessary in order to settle this issue of putative hybrid speciation in algae.
Our results are relevant concerning the DNA barcoding for algae and possibly other organisms. The 5'-COI "barcode" region is not representative and might be even misleading (at least in case of S. latissima and C. retorta) in resolving taxonomic relationships between algal species. Although 5'-COI barcoding is practically convenient (because 600-700 bp can readily be sequenced in a single run); it may not be suitable.

Discussion
We have found that the TYP and LON morphological forms of Saccharina japonica are genetically very similar and, therefore, they might not be thought of as distinct biological species, as suggested by Gusarova and Ivanova [17]. The SHA form, however, exhibits distinctive genetic differences from the TYP and LON forms (2.6% total DNA divergence) and it is closely related to the S. cichorioides species group. The SHA form might be thought of as a morphological variety of S. cichorioides that inhabits an ecological nichethe supralittoral zonenew for this species. Our results are of practical consequence. Zoospores of the SHA form are used to replenish populations of the TYP form suffering from overharvesting [19]. This practice is inappropriate because the TYP and SHA forms represent different species. Genetic analysis can, indeed, help to select appropriate stocks for mariculture.
Thus, the genetic data suggest the existence of two close Saccharina lineages inhabiting the Primorye coastal region, S. japonica (TYP + LON) and S. cichorioides (SHA). Significant genetic differences between the SHA form and the TYP plus LON forms are not surprising: these two lineages are also highly different in morphology and ecological preferences. Less expected is the nearly complete genetic identity of the TYP and LON forms, because these forms have drastic differences in morphology, life history traits, and ecology. Taking into account that interbreeding between the LON and TYP forms is highly unlikely (see Background), and also the fact that phenotypic plasticity may have potential roles for adaptive evolution and speciation, as it has been shown for many species, including algae (e.g., [27][28][29][30][31]), we propose that the TYP and LON forms represent important resources for algae diversification that enable successful adaptation to contrasting shallow-and deepwater marine environments. The present results for Saccharina are in accordance with our previous results concerning two morphological forms of the sea urchin Strongylocentrotus intermedius, which are clearly different morphologically but very similar genetically [91].
Transplant site experiments are not likely to be informative concerning the taxonomical status of the Saccharina morphological forms since both, the LON form, which is genetically similar to the TYP form, and the SHA form, which is genetically quite different from the TYP form, exhibit drastic morphological transformations when transplanted to the TYP habitat area, so that they become morphologically indistinguishable from the TYP form [23,24]. Breeding experiments would not likely be informative either to resolve the taxonomic status of the TYP and LON forms, because even distantly related algal species produce highly viable first generation hybrids [88,[92][93][94][95]. Indeed, the biological species concept is not particularly helpful in kelps [92,96]. Kraan and Guiry [92] have, for example, shown that interspecific sequence divergence in Alaria is smaller than intraspecific sequence divergence. That is, there is more variation within a species than between two species of the same genus, which casts doubts on the morphological and biological species concepts used in Alaria and on the usefulness of hybridization studies in assessing specieslevel differences [97][98][99][100]. Conflict between speciation decisions based on morphological or molecular characteristics comes about due to the drastically different rates at which molecular and morphological changes accumulate [100]. The use of a morphological and biological species concept to separate Alaria or Saccharina species is not satisfactory and does not fully reflect phylogenetic relationships.
The data show that there is no consistent relationship between morphological and genetic variation in the algae Saccharina (present data) and Alaria [96], as well as in sea urchins [91]. The results of Druehl and Saunders [96] and Kraan and Guiry [92] demonstrate that fertility barriers, which in most cases indicate complete reproductive isolation, may arise without affecting genetic divergence in the particular genes under investigation [101]. Our results support a relatively minor contribution of genetic factors to the TYP and LON morphological forms differences observed in the three gene markers, suggesting an important phenotypic plasticity basis for the variability in morphology, life history, and ecology in S. japonica.
The relative importance of genes and phenotypic plasticity in the generation of adaptive variation is a far from resolved issue [102]. The possibility that plasticity could play a major role, however, has recently become more commonly considered than it was traditionally believed [103][104][105]. A body of evidence suggests that plasticity may promote adaptive divergence in various systems, often followed by genetic changes in the direction of the plastic response [106] (see however, [107]). Furthermore, phenotypic plasticity enhances the survival and reproductive success of individuals by contributing to their ability to cope with environmental changes. In this way, it enables potential adaptation to new niches [108], and therefore can promote important biological processes such as adaptation, divergence, reproductive isolation, and evolutionary innovation [28,109,110]. Environmentally induced phenotypic variation has been argued to be a more likely source of evolutionary novelty than new mutations [103,104]. It has, indeed, been asserted that phenotypic plasticity can accelerate adaptive evolution by shortening the time a population needs to discover a new genotypic network [111] even in the case of recurrent rapid changes in gene regulation [112].
A number of experimental studies have revealed the mechanisms of phenotypic plasticity using an integrative biology approach (review in [113]). The overall results indicate that phenotypic plasticity has an important role in adaptive evolution. This consideration, in turn, points out the importance of characterizing diverging phenotypes and identifying relevant evolutionary forces acting on those phenotypes. A great number of intraspecific morphological varieties are known for laminarialean algae (review in [14]). Further studies of algae morphological forms are important and necessary to further understand how biological diversity is generated and maintained in evolution.

Conclusions
1. We have investigated the genetic make-up in three S. japonica forms, TYP, LON, and SHA. We have found that in spite of their drastic differences in morphology, life history and ecology, the TYP and LON forms are genetically indistinguishable and clearly different from the SHA form, which has a close genetic relationship with S. cichorioides. Taking into account the potential evolutionary significance of phenotypic plasticity, the TYP and LON forms may represent an important resource for algal evolutionary diversification and successful adaptation to contrasting marine environments. 2. The three gene markers used in the present work have different sensitivity for algal species identification; the mitochondrial COI gene is the most efficient gene marker. However, we found that the designated 5'-COI "barcoding" region is not enough and may even be misleading for purposes of identification, at least in comparisons that include species with evolutionary histories