Taxonomic Revision of Hook-Forming Acrosorium (Delesseriaceae, Rhodophyta) from the Northwestern Pacific Based on Morphology and Molecular Data

Cosmopolitan Acrosorium species with hook-forming thalli have been merged under the name of Acrosorium ciliolatum (Harvey) Kylin through a long and complicated nomenclatural history. We examined the specimens of ‘A. ciliolatum’ and related taxa from the northwestern (NW) Pacific, the UK, southern Spain, Australia, New Zealand, and Chile, using morphological and molecular analyses. We confirmed that these specimens are separated into four clades based on rbcL phylogeny, and the absence or presence of terminal hook-like structures represent intraspecific variation. Our results indicated that Acrosorium flabellatum Yamada, Cryptopleura hayamensis Yamada, Cryptopleura membranacea Yamada and the entities known as ‘A. ciliolatum’ in the NW Pacific are conspecific; the name A. flabellatum is the oldest and has priority. This taxon exhibits extreme variations in external blade morphology. We also confirmed that the position of the tetrasporangial sori is a valuable diagnostic characteristic for distinguishing A. flabellatum in the NW Pacific. We also discussed the need for further study of European and southern hemisphere specimens from type localities, as well as the ambiguous position of California specimens.

During an investigation of species and genetic diversity in the family Delesseriaceae from the NW Pacific, we encountered difficulties in identifying specimens of these four species, due to overlapping morphological features. To resolve this problem, we attempted to identify meaningful characters delimitating the four species, based on morphology and rbcL sequences. Remarkably, our molecular phylogeny showed that the four species are conspecific, with A. flabellatum having priority as the earliest name. To determine the relationship of this Acrosorium/Cryptopleura complex in the NW Pacific to specimens identified as A. ciliolatum in other parts of the world, we obtained specimens from the UK, southern Spain, southern Australia, New Zealand, Chile, and California, USA, and compared their morphological and molecular characteristics ( Figure 1).

Figure 1.
A map showing the sampling locations (circles: blue, from Europe; green, from NW Pacific; yellow, from California; red, from southern hemisphere), and type localities for A. venulosum (blue star), A. flabellatum (green star), and A. ciliolatum (red star). Numbers: sample number for molecular analysis.

Molecular Analyses
The phylogenetic analyses of rbcL included 1443 bp, of which 412 bp (28.6%) were variable and 434 bp (17.9%) were phylogenetically informative. The rbcL sequences in the NW Pacific Acrosorium/Cryptopleura complex differed by 0.73 to 0.98% from Californian, 1.55 to 2.05% from European (Spain and UK), and 1.31 to 1.55% from southern hemisphere (Chile, New Zealand, and Tasmania) samples. The sequences from Japan and Korea varied by up to 0.32%.

image.
Commente and added Figure 1. A map showing the sampling locations (circles: blue from Europe; green from NW Pacific; yellow from California; red from southern hemisphere), and type localities for A. venulosum (blue star), A. flabellatum (green star), and A. ciliolatum (red star). Numbers: sample number for molecular analysis.

Molecular Analyses
The phylogenetic analyses of rbcL included 1443 bp, of which 412 bp (28.6%) were variable and 434 bp (17.9%) were phylogenetically informative. The rbcL sequences in the NW Pacific Acrosorium/Cryptopleura complex differed by 0.73 to 0.98% from Californian, 1.55 to 2.05% from European (Spain and UK), and 1.31 to 1.55% from southern hemisphere (Chile, New Zealand, and Tasmania) samples. The sequences from Japan and Korea varied by up to 0.32%.
In the phylogenetic tree based on rbcL gene (Figure 2), the sequences from the NW Pacific (Japan and Korea), California, Europe, and southern hemisphere formed independent and monophyletic clades with high bootstrap values (100, 99, 100, and 100%, respectively) in the tribe Cryptopleureae. The clades of the NW Pacific and California formed sister-groups, with a low bootstrap value (72%). In the NW Pacific clade, the rbcL sequences from each phenotype-AF ( Figure 3B In the phylogenetic tree based on rbcL gene (Figure 2), the sequences from the NW Pacific (Japan and Korea), California, Europe, and southern hemisphere formed independent and monophyletic clades with high bootstrap values (100, 99, 100, and 100%, respectively) in the tribe Cryptopleureae. The clades of the NW Pacific and California formed sister-groups, with a low bootstrap value (72%). In the NW Pacific clade, the rbcL sequences from each phenotype-AF ( Figure      Thalli are extremely variable in shape and size, epilithic or epiphytic, membranous, up to 30 cm high, pinkish red when alive, turning brownish red when dry, each consisting of single to several large erect blades and a small prostrate base. When growing on rock Thalli are extremely variable in shape and size, epilithic or epiphytic, membranous, up to 30 cm high, pinkish red when alive, turning brownish red when dry, each consisting of single to several large erect blades and a small prostrate base. When growing on rock or concrete, the erect blades expand from a compressed short stipe and are divided palmately or di-to trichotomously several times, broadening into a flabellate outline ( Figure 3A-H). When growing on other algae such as Corallina spp. or Cladophora spp., the erect blades are usually linear and ribbon-shaped without a distinct stipe, and branched in alternate, di-to trichotomous or irregular pattern, forming tangled clumps ( Figure 3I-L). Blade margins are entire with small semi-circular proliferations that occasionally elongate to form marginal branches ( Figure 3A-C,F,I-L). Blades tips are round ( Figure 3D-F,H) to acute with terminal curves ( Figure 3A Multicellular peg-like rhizoids are often observed along the inner surfaces of terminal hooks ( Figure 4A). Microscopic veins are one cell thick, and extend distally ( Figure  4B). Macroscopic veins are observed in the basal parts of thalli when the stipe is welldeveloped ( Figures 3A-H and 4G,H), otherwise not ( Figure 3I-L). Apical growth is via numerous obliquely dividing apical meristematic cells, followed by intercalary cell division ( Figure 4C,D). Erect blades are mostly monostromatic, except microscopic veins, stipes, and lower portions of the thallus ( Figure 4E-H). In cross-section, cortical cells are cut off periclinally from large tetragonal central cells ( Figure 4G,H). Cortical cells in the blade are polygonal with parietal chloroplasts. Chloroplasts are lobed in young cells, soon becoming numerous and discoid, then elongating in mature cells ( Figure 4I-K). Gametophytes are dioecious. Procarps are scattered on both sides of the thallus surface, usually along microscopic veins near meristemic portions ( Figure 5A). Formation of the procarp is initiated by cutting off a supporting cell from the fertile central cell. The supporting cell initially cuts off the first sterile-cell group, then the carpogonial branch initial which forms the carpogonial branch by sequential division (Figure 5B-F). When the carpogonial branch is three-celled, the supporting cell cuts off a second sterile-cell group initial which usually divides once before fertilization ( Figure 5C-F). The mature procarp consists of a supporting cell, a four-celled carpogonial branch with a terminal trichogyne swollen at the tip, a one-to two-celled first sterile-cell group, and, usually, a two-celled second sterile-cell group ( Figure 5E,F). Procarps develop on both sides of the blade from a common fertile central cell; however, only one procarp develops into the cystocarp ( Figure 5G,I).

Comm
the carpogonial branch is three-celled, the supporting cell cuts off a second sterile-cell group initial which usually divides once before fertilization ( Figure 5C-F). The mature procarp consists of a supporting cell, a four-celled carpogonial branch with a terminal trichogyne swollen at the tip, a one-to two-celled first sterile-cell group, and, usually, a two-celled second sterile-cell group ( Figure 5E,F). Procarps develop on both sides of the blade from a common fertile central cell; however, only one procarp develops into the cystocarp ( Figure 5G,I). After presumptive fertilization, the supporting cell is enlarged and distally cuts off an auxiliary cell which divides to form a gonimoblast cell initial ( Figure 5H-J). Cell fusions among the fertile central cell, supporting cell, and auxiliary cell occur in an early stage of cystocarp development, during which the gonimoblast cell initials develop into gonimoblast filaments by sequential cell division ( Figure 6A). Neither sterile-cell group participates in the formation of the fusion cell, but they remain around the fusion cell ( Figure 6A,B). As the cystocarp grows, the fusion cell enlarges by incorporating neighboring cells in the cystocarp floor. Some floor cells cut off one to two layers of small cells in the direction of the cystocarp cavity ( Figure 6B,C). The gonimoblast filaments extend radially by cell division with subdichotomous or alternate branching, producing a single carposporangium on each terminal cell ( Figure 6B After presumptive fertilization, the supporting cell is enlarged and distally cuts off an auxiliary cell which divides to form a gonimoblast cell initial ( Figure 5H-J). Cell fusions among the fertile central cell, supporting cell, and auxiliary cell occur in an early stage of cystocarp development, during which the gonimoblast cell initials develop into gonimoblast filaments by sequential cell division ( Figure 6A). Neither sterile-cell group participates in the formation of the fusion cell, but they remain around the fusion cell ( Figure 6A,B). As the cystocarp grows, the fusion cell enlarges by incorporating neighboring cells in the cystocarp floor. Some floor cells cut off one to two layers of small cells in the direction of the cystocarp cavity ( Figure 6B,C). The gonimoblast filaments extend radially by cell division with subdichotomous or alternate branching, producing a single carposporangium on each terminal cell ( Figure 6B-D). The carposporangia are ovoid to elliptical, 60-80 µm long, 25-30 µm wide ( Figure 6D). Mature cystocarps are hemispherical with a protruding ostiole ( Figure 6B,E).
Spermatangial sori are produced on both surfaces of small marginal proliferations ( Figure 6F). Spermatangial mother cell initials are cut off fertile central cells by periclinal divisions on both blade surfaces. The initial cells divide anticlinally several times to form spermatangial mother cells, all of which cut off one to two clavate spermatangia ( Figure 6G,H).
Tetrasporangial sori are formed on small marginal proliferations ( Figure 6I). The tetrasporangia are globose, 70-100 µm in diameter, and tetrahedrally divided. They are usually cut off from central cells, or, rarely, from inner cortical cells, and arranged in two layers on opposite sides of the central cells ( Figure 6J). elliptical, 60-80 µm long, 25-30 µm wide ( Figure 6D). Mature cystocarps are hemispherical with a protruding ostiole ( Figure 6B,E). Spermatangial sori are produced on both surfaces of small marginal proliferations ( Figure 6F). Spermatangial mother cell initials are cut off fertile central cells by periclinal divisions on both blade surfaces. The initial cells divide anticlinally several times to form spermatangial mother cells, all of which cut off one to two clavate spermatangia ( Figure  6G,H).
Tetrasporangial sori are formed on small marginal proliferations ( Figure 6I). The tetrasporangia are globose, 70-100 µm in diameter, and tetrahedrally divided. They are usually cut off from central cells, or, rarely, from inner cortical cells, and arranged in two layers on opposite sides of the central cells ( Figure 6J).    From southern hemisphere ( Figure 7A-D): Thalli are brownish to pinkish red, membranous and epilithic or epiphytic on other algae, usually coralline algae. Several erect blades arise from the margins of a small prostrate base, forming tangled clumps. Erect blades are linear with cuneate bases and acute apexes, measure up to 5 cm high and less than 3 mm wide, and are branched in an alternate to irregular pattern, broadening to a flabellate outline ( Figure 7A-D). Margins of erect blades are entire. When growing on other macroalgae, blade tips usually terminate in hooks, with small proliferations along blade margins ( Figure 7A,B,D). When growing on rock surfaces, hook-shaped tips are absent ( Figure 7C Thalli are pinkish to brownish red and membranous. Fronds consist of small basal prostrate blades and erect flabellate blades. Erect blades are usually ribbon-shaped, some cuneate, with alternate, or di-to trichotomous branches, up to 7 cm tall, and less than 5 mm wide ( Figure 7J,K). Thallus margins are entire with marginal semi-circular proliferations ( Figure 7K-N). Thallus tips are acute ( Figure 7J-L) to blunt ( Figure 7M,N), with terminal hooks (Figure 7K,L) or not ( Figure 7J,M,N).

Discussion
The hook-forming Acrosorium species, currently known as A. ciliolatum, have a long and complicated nomenclatural history. After the description of Nitophyllum ciliolatum in 1855 by Harvey (type locality King George Sound, Western Australia and Tasmania) [7], Nitophyllum venulosum Zanardini (in 1866) [8], and Acrosorium aglaophylloides Zanardini ex Kützing (in 1869) [1] were sequentially described from Croatia on the Adriatic Sea. Kylin [3] noticed that A. aglaophylloides had been redescribed under the name N. venulosum; then, he made the new combination Acrosorium venulosum (Zanardini) Kylin, and synonymized A. aglaophylloides with A. venulosum. In the same study [3], he transferred six more Delesseriacean species to the genus Acrosorium, including Nitophyllum reptans Crouan & Crouan, N. acrospermum J. Agardh, N. corallinarum Nott, N. uncinatum J. Agardh, N. ciliolatum Harvey, and Cryptopleura minor Sonder. At that time, Kylin [3] mentioned that A. venulosum differed from A. uncinatum: the former has a rounded apex, whereas the latter has a hook-shaped one. However, he was not confident about separating the three species, A. venulosum, A. uncinatum, and A. reptans. Wynne [6] re-examined the type specimens and Kylin's [3] original illustrations of the three Acrosorium species and concluded that: (1) the hook-forming thalli are morphologically variable in these taxa, (2) A. uncinatum and A. reptans are conspecific with Cryptopleura ramosa (Hudson) L. Newton, and (3) A. venulosum is the correct name for the cosmopolitan A. uncinatum sensu Kylin [3]. Womersley [5] mentioned a quotation from Kylin [32], "A. ciliolatum is not specifically distinct from A. venulosum and is the earliest name for this Acrosorium". He therefore merged A. venulosum with A. ciliolatum. Thus, Acrosorium populations with hook-forming thalli have been identified under the name A. ciliolatum [17].
Our molecular and morphological evidence supports Wynne's [6] opinion that the presence or absence of hook-forming thalli cannot be used as a morphological trait to distinguish among Acrosorium species, because this trait shows intraspecific variation and has arisen more than once in the evolution of Acrosorium species. Furthermore, our phylogenetic tree ( Figure 2) and specimens (Figures 3 and 7) indicate that there are extensive variations in thallus habit and size, branching pattern, apex shape, and the presence or absence of a basal stipe within each population, in addition to the formation of apical hooks. There is no correlation between the clades of our rbcL tree and any of the abovementioned external morphological features (Figures 2, 3 and 7). We also confirmed morphological similarities among specimens from different clades (e.g., Figure 3J vs. Figure 7A,D; Figure 7C vs. Figure 7F; Figure 3H vs. Figure 7G). Uncertainties with respect to species identification arise from intraspecific morphological variations, as well as similarities among species. Consequently, relying only on morphology, it was impossible to distinguish Acrosorium specimens from different regions.
The rbcL tree (Figure 2) shows that our samples fell into four clades: three distinct clades (the NW Pacific, Europe, and southern hemisphere) and a less well-supported California clade. We conclude that the name A. ciliolatum represents a complex consisting of at least three different species, contrary to the opinions of Wynne [6] (A. venulosum = globally known as A. uncinatum) and Womersley (A. ciliolatum = A. venulosum) [5]. We found that the position of the tetrasporangial sori is a conclusive feature for distinguishing the NW Pacific specimens from those from other regions. Taxa from the NW Pacific, including 'A. ciliolatum' auctorum, A. flabellatum, C. hayamenis, and C. membranacea have tetrasporangial sori restricted to small marginal proliferations ( Figure 6I) [20,23,25,26,28,30,48]. Tetrasporangial sori are formed on the thallus tip or on marginal proliferations in European 'A. ciliolatum' [4], but are restricted to the thallus tip in Australian A. ciliolatum [5].
Although Yamada [25,26,28] established A. flabellatum, C. membranacea, and C. hayamensis on the basis of the various phenotypes of N. uncinatum sensu Okamura [30], the phenotypic overlaps ( Figure 3C,F-H) among numerous individuals raise the question as to whether these phenotypes are independent species. Our results provide a clear resolution to this problem, strongly supporting Okamura's [30] opinion that the differences in thalli width, apex shape, and branch pattern are variations within a single species. All specimens from Korea and Japan, with their various phenotypes (i.e., AC, AF, CH, and CM) and overlapping phenotypes (Table S1 and Figure 3A-L), are conspecific ( Figure 2). Although analyses using more variable mitochondrial or nuclear markers might reveal population-level differences within this clade, we consider the rbcL data sufficient support at the species level. Therefore, we propose the merging of 'Acrosorium ciliolatum' auctorum in the NW Pacific, Cryptopleura hayamensis Yamada, and Cryptopleura membranacea Yamada with A. flabellatum Yamada, which is the earliest name. This taxon is distinguishable from the European and Australian taxa by rbcL sequences and the position of the tetrasporangial sori, which are restricted to marginal proliferations ( Figure 6I) [4,5,20,23,27,30,48].
In this study, the rbcL tree ( Figure 2) shows that the European Acrosorium clade, which currently goes under the name of A. ciliolatum, is clearly separate from the southern hemisphere A. ciliolatum. However, we will postpone the formal taxonomic treatment of the European clade until examination of specimens from the two type localities, Zadar, Croatia for the European, and King George Sound, western Australia for the southern hemisphere populations. The Californian specimens form a clade that is sister to the NW Pacific clade, with borderline bootstrap support and genetic differentiation ( Figure 2). Moreover, we could not find reproductive structures including tetrasporangial sori or identify any morphological differences from the other clades despite examining more than 50 specimens in the UC herbarium. According to Abbott and Hollenberg [49], fertile fronds are rare and only tetrasporophytes have been reported in California, without any description of the position of the tetrasporangial sori. Further studies using additional and more sensitive markers are essential to probe and clarify the relationship of California Acrosorium to its sister clade in the NW Pacific.
Existing studies on red algal phylogeny have focused on the relationships between female reproductive structures and molecular phylogeny. Most of these studies indicate that the developmental patterns of carposporophytes provide important information for delimitating genera in the orders Halymeniales, Gigartinales, and Ceramiales [39,42,[44][45][46][50][51][52]. During our collections, male and female gametophytes were rarely collected; most were from the Jeju coast of Korea. Most of our collections, including California specimens, were vegetative or tetrasporic thalli. For this reason, our observations of female reproductive structures were conducted only on A. flabellatum from the Korean coast. The female reproductive anatomy of A. flabellatum is nearly identical to that of A. acrospermum (J. Agardh) Kylin described by Papenfuss [13].

Collections
Specimens were collected from the coast of the Korean peninsula, eastern Honshu in Japan, New Zealand, Tasmania in the south-eastern Australia, and the Spanish side of the Gibraltar Strait either intertidally or subtidally by SCUBA diving. The field-collected samples were kept fresh in a cool-box with an ice pack and then transported to the laboratory, where the specimens were pressed on herbarium sheets while still alive. Samples for morphological observation were preserved in 5% formalin/seawater. Voucher specimens were deposited in the herbarium of Jeju National University (JNUB, Jeju, Korea). California specimens borrowed from the University Herbarium, University of California at Berkeley (UC) were also examined ( Figure 1, Table S1).

Classifying Specimens with Phenotype
All the specimens from the NW Pacific were divided into four phenotypes based on key characteristics outlined by Yoshida [23], including thallus width, branching pattern, the shape of the apex, and hook-formation: (1) A. ciliolatum-type (AC: subdichotomously branching linear thalli with hook-forming acute apices); (2) A. flabellatum-type (AF: flabellate thallus with a stipe, tapering branches, and slightly curved acute apices); (3) C. hayamensis-type (CH: di-to trichotomously branching linear thalli without hooks); (4) C. membranacea-type (CM: stipitate, broad, and palmate thallus with visible macro-and invisible microscopic veins and round apices without terminal hooks) (Figure 3). Two to five individuals with each phenotype and with overlapping phenotypes were chosen for molecular analyses (Figures 2 and 3 and Table S1).

Molecular Analysis
Samples for genomic DNA extraction were taken from selected pressed specimens. Total genomic DNA was extracted from the specimens and prepared with a DNeasy Plant Mini kit (Qiagen, Hilden, Germany), following the manufacturer's protocol. The amplification for rbcL was conducted using the primers F7 (or F145), R898, F762, and R1442 [33,56]. Polymerase chain reaction (PCR) amplifications were performed in Swift MaxPro thermal cyclers (ESCO, Singapore) with an AccuPower PCR PreMix (Bioneer, Dajeon, Korea). Amplification reactions were set up and run as described in Saunders and Moore [57]. The PCR products were purified with an AccuPrep PCR Purification Kit (Bioneer, Daejeon, Korea) and sequenced commercially by Macrogen (Seoul, Korea). The electropherogram output of each sample was edited using Chromas Lite 2.01 (Technelysium, Helensvale, Queensland, Australia) and PHYDIT version 3.1 software [58]. The DNA sequences of rbcL were assembled in BioEdit 7.0.5 [59] and visually aligned. A total of 47 rbcL sequences were successfully amplified, including 29 from the NW Pacific, six from California, five from Spain, and seven from Tasmania and New Zealand (Table S1). The rbcL sequences in GenBank identified as A. ciliolatum from Chile (MH649426) and A. venulosum (AF254156) from UK were also aligned with sequences from our material. Because six rbcL sequences from Chilean samples in GenBank were all identical, we used only one sequence (MH649426) in our phylogenetic analysis (Figures 1 and 2). The phylogenetic relationships of the 49 aligned rbcL sequences were investigated by comparing them with 15 taxonomic groups in GenBank, including ten assigned to the tribe Cryptopleureae and five to Neuroglosseae (as an outgroup).
The uncorrected pair-wise genetic distances (p) among the populations were estimated using the phylogenetic analysis software MEGA 4.0 [60] to assess the genetic variation of rbcL. Phylogenetic analyses were performed using RAxML [61], with the GTR + Γ + I evolutionary model. To identify the best tree, we constructed 200 independent tree inferences, using the -# option with the default -I (automatically optimized subtree pruning-regrafting rearrangement) and -c (25 distinct rate categories) software options. To generate bootstrap values, we used the same program with the same settings for 1000 replications.

Morphological Observation
Sections for morphological examination were cut by hand using a razor blade, or with a bench-top freezing microtome (MFS no. 222; Nippon Optical Works, Tokyo, Japan). Sections and whole mount materials were either stained with 1% aniline blue acidified with 1% HCl and mounted in 40% corn syrup solution, or stained with Wittmann's [62] acetoiron-hematoxylin-chloral hydrate and mounted in 50% Hoyer's mounting medium [41]. Images of the pressed specimens were captured using a G7x digital camera (Canon, Tokyo, Japan) and photomicrographs were taken using a BX 43 microscope (Olympus, Tokyo, Japan) with an EOS 600 D digital camera (Canon, Tokyo, Japan). The digitized images were imported into Adobe Photoshop (ver. 6.1) and edited to produce plates.