Biogeography in Cellana (Patellogastropoda, Nacellidae) with Special Emphasis on the Relationships of Southern Hemisphere Oceanic Island Species

Oceanic islands lacking connections to other land are extremely isolated from sources of potential colonists and have acquired their biota mainly through dispersal from geographically distant areas. Hence, isolated island biota constitutes interesting models to infer biogeographical mechanisms of dispersal, colonization, differentiation, and speciation. Limpets of the genus Cellana (Nacellidae: Patellogastropoda) show limited dispersal capacity but are broadly distributed across the Indo-Pacific including many endemic species in isolated oceanic islands. Here, we examined main distributional patterns and geographic boundaries among Cellana lineages with special emphasis in the relationships of Southern Hemisphere oceanic islands species. Phylogenetic reconstructions based on mtDNA (COI) recognized three main clades in Cellana including taxa from different provinces of the Indo-Pacific. Clear genetic discontinuities characterize the biogeography of Cellana and several lineages are associated to particular areas of the Indo-Pacific supporting the low dispersal capacity of the genus across recognized biogeographical barriers in the region. However, evolutionary relationships within Cellana suggest that long-distance dispersal processes have been common in the history of the genus and probably associated to the origin of the species in Hawaii and Juan Fernández Archipelago. Therefore, the presence of Cellana species in geographically distant Southern Hemisphere oceanic islands, such as the Juan Fernández Archipelago, suggests that long-distance dispersal mediated by rafting may have played an important role in the biogeography of the genus.


Introduction
The distribution of near-shore benthic invertebrates is the result of complex interactions between historical and contemporary processes as well as biological traits including a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Specimens of C. strigilis strigilis specimens were collected in Perseverance Harbour, Campbell Island (52˚32 0 S; 169˚8 0 E; Fig 1). Cellana dira individuals were collected in Rapa Iti, Austral Islands, French Polynesia (27˚35 0 S; 144˚20 0 W), and C. taitensis specimens were sampled in Papenoo, Tahiti (17˚35'S; 149˚26'W) and Orofara, Tahiti (17˚30'S; 149˚28'W; Table 1). Identification of all species with the exception of C. dira was done following Powell [14]. The species C. dira was previously included in the genus Patella but external features such as gill morphology and inner shell coloration suggest that this taxon belongs to Cellana (T. N., pers. obs.).
Animals were fixed in ethanol (95%), and total DNA was extracted from the mantle following the salting-out protocol described by Aljanabi and Martinez [15]. A partial fragment of the mitochondrial gene Cytochrome Oxidase c Subunit I (COI) was amplified using universal primers [16]. PCR mixtures and thermal cycling parameters were done following Nakano and Ozawa [17] and González-Wevar et al. [18]. PCR amplicons were purified using QIAquick Gel Extraction Kit (QIAGEN) and both strands were sequenced using an Automatic Sequencer 3730xl at Macrogen Inc. (South Korea). New COI sequences of Cellana species were deposited in GenBank under the following Accession Numbers: KY353116 -KY353133.

Phylogenetic reconstructions
We included in the analyses four to five individuals of each of the newly analyzed Cellana species. New COI sequences were combined with previously published COI sequences of Cellana available in GenBank (Table 1). COI sequences of Nacella species and four patellogastropod genera (Helcion, Scutellastra, Cymbula, and Patella) were used as sister-and out-group, respectively. New Cellana sequences were edited using Proseq 2.91 [19] and the whole COI sequence  Table 1. Patellogastropod species, specimens, accession numbers, and sampling localities included in the study. Specimen identifications were based on Powell [14], Valdovinos and Rüth [70], Nakano and Ozawa [17], and González-Wevar et al. [18]. New COI sequences obtained in this study are marked in bold. Other Accession Numbers and their respective sources are indicated. data set was aligned with Clustal W [20]. Sequences were translated to amino acids to check for the presence of pseudogenes, insertion/deletions (indels), and/or the presence of sequencing errors with MEGA 6.0 [21]. Reconstructions were done using three different methods: Maximum Parsimony (MP), Maximum Likelihood (ML), and Bayesian Inference (BI). MP and ML analyses were performed in PAUP Ã version 4.0b [22] following González-Wevar et al. [18]. Nucleotide substitution models for ML and BI were selected using the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC), with jmodeltest 2.0 [23], respectively. ML and BI analyses were performed using PHYML [24] and MrBayes 3.1.2 [25], respectively. Inferred tree nodes supports for MP and ML analyses were assessed through a non-parametric bootstrap (BS) analysis with the full heuristic search option and 1,000 pseudoreplicates [26]. Bayesian analyses were used to obtain posterior probabilities for the nodes in the phylogenetic trees. Posterior probability values of sampled trees were obtained using the Metropolis coupled Markov-Chain Monte-Carlo algorithm (MCMC) implemented in MrBayes. Four chains were run twice in parallel for 100 x 10 6 generations and trees were sampled every 1,000 generations. Stationarity of the analyses was inferred when the average standard deviation of split frequencies was less than 0.01 [25]. The first 10% of the trees were discarded (burn-in) and Bayesian posterior probabilities (BPP) were estimated as the percentage of trees that showed a particular node. Nodes in the phylogenetic tree were considered highly supported with BPP values ! 0.95% and BS ! 75%. Nodes with support values of 0.80-0.95% (BPP) and 60-75% (BS) were considered moderately supported. Posterior probability density of the combined tree and log files was summarized as a maximum clade credibility tree using TreeAnnotator v.

Divergence time estimates
Divergence time estimates were made using a relaxed molecular clock analyses with an uncorrelated lognormal (ucln) model of molecular evolutionary rate heterogeneity and the GTR + I + G substitution model in BEAST v.1.6.2 [27]. For this, we used a specific mutation rate (0.85%-1.15%) previously estimated for nacellid limpets [28]. An age prior with a normal distribution was applied to the most recent common ancestor (tmrca) of Nacellidae (mean, 38; SD, 3.8), an Upper Eocene fossil of Cellana ampla [29]. We also include two age priors within Nacella, a tmrca of N. concinna (mean, 4.5; SD, 0.45) based on the oldest fossil record for this species at Cockburn Island [30] and the tmrca of N. clypeater (mean, 4, SD, 0.4) based on N. clypeater-like fossil from southern Peru [31]. A birth-death speciation prior was used for branching rates in the phylogeny. Four chains were run two times for 100 x 10 6 generations, and trees were sampled every 1,000 generations. Majority rule consensus phylograms, as well as posterior probabilities for nodes were estimated. Convergence of model parameter values and estimated node heights to their optimal posterior distribution were estimated by plotting the marginal posterior probabilities versus the generation state in Tracer v.

Data analyses and phylogenetic reconstructions
Sequence alignment for COI included 647 nucleotide positions, with 271 variable and parsimony informative sites. As expected for a coding region, no indels or stop codons were detected and no significant levels of saturation were observed in the complete COI data set. Out of a possible 215 amino acids in the entire COI data set of Cellana we identified eleven amino acid substitutions, all of them occurred at the third position of the codon. COI sequences clearly discriminated major taxonomic groupings of patellogastropods with high bootstrap (BS) and Bayesian posterior probabilities (BPP) (Fig 2). At the same time, all the reconstruction methods confirmed the monophyly of both Nacella and Cellana, and their sister relationships (BS = 100%; BPP = 1.0; Fig 2). Within Cellana, we recorded three main clades (I, II, and III; Figs 2 and 3). Clade I includes species from Australia, New Zealand, C. ardosidea from Juan Fernández archipelago, and C. testudinaria from Japan. Clade II comprises species from the Indo-Pacific, the Northwestern Pacific and French Polynesia. Finally, Clade III contains species from the Northwestern Pacific and Hawaii. Each one of these major clades can also be divided into subclades (a-h; Fig 3) containing species from particular areas. For instance, subclade a includes Australian C. tramoserica and C. solida (Aus) and C. testudinaria from the Northwestern Pacific (N-WP3; Fig 3). Subclade b includes species from New Zealand (C. flava, C. radians, and C. stellifera) and C. ardosidea from Juan Fernández Island (JFI), while subclade c comprises C. denticulata from Kaikoura New Zealand, and nominal subspecies C. strigilis redimiculum and C. strigilis strigilis from New Zealand Subantarctic Island (SAI; Fig 3). Within main clade II, subclade d includes C. dira (Austral Island) and C. taitensis (Tahiti) from French Polynesia (FP) while subclade e comprises C. toreuma and C. radiata enneagona from Japan (N-WP2). Subclade f includes species from the Indo-Pacific (IP) such as C. radiata orientalis (Indonesia), C. radiata radiata (Sri Lanka), C. capensis (South Africa), C. eucosmia (Egypt), and C. karachiensis (Oman). Subclade g includes C. nigrolineata and C. grata from Japan (N-WP1), C. mazatlandica (Osagawara Island), and subclade h comprises C. exarata, C. talcosa and C. sandwicensis from Hawaii (Haw; Fig 3).
New phylogenetic reconstructions based on mtDNA sequences are in basic agreement with previous studies in Cellana and support the idea that the oldest representatives of the genus correspond to the New Zealand species [17,32,33]. The position of C. dira within Cellana was confirmed in all reconstructions (T. N., pers. comm. ; Figs 2 and 3). Following this, the species

Divergence time estimates
The separation of Nacella and Cellana took place close to the Eocene-Oligocene boundary around 30 Ma (37-24 Ma) (Fig 2). The most recent common ancestor (tmrca) of Cellana existed~24 Ma (32-20 Ma) and is older than the 14 Ma (18-10 Ma) estimated for Nacella's trmca. The separation of the three main clades is dated close to the Oligocene/Miocene boundary (* 24 Ma). After this, the main lineages of Cellana followed different evolutionary trajectories and diversified in different areas of the Indo-Pacific since the early Miocene. In the Southern Hemisphere, the diversification of Cellana in New Zealand North Island (C. radians, C. stellifera and C. solida) is placed * 6 Ma (8-3 Ma). Then, the separation of Juan Fernández C. ardosidea from its closest New Zealand relative C. flava took place~4.5 Ma (6.5-4 Ma). Finally, the separation between C. denticulata from New Zealand South Island and the two nominal subspecies of the C. strigilis complex is placed around 2.5 Ma (3.0-1.5 Ma). The separation between Polynesian C. dira and C. taitensis and the two species from the Northwestern Pacific (N-WP2) C. toreuma and C. radiata enneagona is placed~12 Ma (16-10 Ma). Then, the separation between C. dira and C. taitensis took place~4 Ma (6-3 Ma) (Fig 2).

Historical biogeography of Cellana
Divergence time analyses suggest that the most recent common ancestor of the Nacellidae occurred around 30 Ma, close to the Eocene-Oligocene boundary, a period of marked worldwide tectonic, oceanographic and climatic changes, particularly in the Southern Ocean [34,35]. Since the Eocene, a global cooling trend is observed commonly explained by the combined effect of gradual northward movement of continents, the closure of the Tethys Sea, and the isolation of Antarctica and the initiation of the Antarctic Circumpolar Current [36,37]. Considering these settings, we propose that the cladogenetic process separating tropical/temperate Cellana from Antarctic/subantarctic Nacella coincides with the shift to glacial conditions in the Southern Ocean [38,39]. Consequently, the diversification of both genera could be related to important adaptations to the thermal realms where they currently occur. Divergence time estimates in Cellana suggest that the origin and diversification of the genus across the Eastern Pacific Barrier (EPB) initiated near the Oligocene-Miocene boundary, 10 Ma earlier than previously estimated [18]. The EPB has been in place during the last 65 Ma [40] but transoceanic gene flow between populations at its two edges was potentially possible until the closure of the Tethyan Sea (11)(12)(13)(14)(15)(16)(17) or probably as recently as the definitive closure of the Panama Isthmus around 3.1 Ma [41]. Based on phylogenetic relationships and the distribution of lineages in Cellana, the Northwestern Pacific region seems to represent a key area in the biogeography of the genus because representatives of this region are present in the all three main clades.

Oceanic island colonization in Cellana
Evolutionary relationships within Cellana and divergence time analyses suggest the existence of long-distance dispersal events in the biogeography of the genus at different temporal and spatial scales. As previously demonstrated and discussed by Bird et al. [42], the colonization of the Hawaiian Islands represents an example of long-distance dispersal from the Northwestern Pacific Region (subclade N-WP1, Figs 3 and 4B). The monophyly of the Hawaiian species strongly supports the existence of a single colonization event [42]. In this context, a report of Cellana individuals living on the trunks of floating palm trees [43] support the hypothesis of rafting as a feasible mechanism for long-distance dispersal. Similarly, based on evolutionary relationships and divergence time estimations the colonization of Juan Fernández archipelago during the Mio-Pliocene (5.5 to 3 Ma) from the geographically distant New Zealand could also be associated to a long-distance dispersal event (Fig 4C). The presence of a single lineage (C. ardosidea) in Juan Fernández archipelago supports the hypothesis of a single dispersal event. Divergence time estimates are in basic agreement with the geologic ages of these islands [44]. Several examples corroborate the affinity between Juan Fernández archipelago and Western Pacific biota including fishes [45,46], gastropods [47,48] and marine algae [49]. Although the Juan Fernández and Desventuradas archipelagos are only 600-700 km away from South America, they are isolated from this continent by the presence of the Humboldt Current System. Hence, these islands exhibit a high degree of endemicity for molluscan [50] and fish [46] faunas. As previously suggested for other Juan Fernandez intertidal marine organisms [49], it is unlikely that Cellana could have crossed the entire width of the Pacific Ocean solely by larval dispersal considering the short larval period recorded in different lineages of the genus. Either rafting or island hopping of now-vanished seamounts during low sea level or volcanic activity could be invoked as a plausible explanation for the close relationship between New Zealand and Juan Fernández Archipelago Cellana species.
Whether the colonization of French Polynesia represents an example of long-distance dispersal in Cellana or a stepping-stone mediated process is difficult to infer considering phylogenetic relationships within clade II. French Polynesian species (C. taitensis and C. dira) are related to broadly distributed Cellana species (N-WP2; Figs 3 and 4A) C. radiata and C. toreuma [14,17]. The radiata complex is currently distributed from East Africa to the Marquesas Islands in the Pacific while the toreuma complex extends from Japan to the Philippines [14]. Therefore, the presence of C. radiata subspecies and C. toreuma on nearby areas of Polynesia argues in favor a colonization of Tahiti and Austral Islands through a stepping-stone mode rather than by long-distance dispersal. Phylogenetic and phylogeographic relationships within the Cellana strigilis complex in New Zealand Subantarctic Islands Cellana species has been discussed by Reisser et al [51]. In this study, the absence of reciprocal monophyly and the presence of shared haplotypes between C. strigilis redimiculum (Oamaru, New Zealand South Island) [17] and C. strigilis strigilis from Campbell Island (this study) suggest the existence of contemporary connectivity by regular gene flow.

Long-distance dispersal mechanisms in Cellana
Commonly, dispersal is inferred as the default explanation of biogeographical disjunction following rejection of a vicariance hypothesis, for example by molecular dating. Considering the volcanic origin of oceanic islands in the Pacific a vicariant hypothesis for the biogeography of Cellana is most unlikely. Following this, several hypotheses can account for the presence of Cellana species at isolated oceanic islands and their evolutionary affinities to geographically distant relatives [52]. They include: i) planned or accidental anthropogenic dispersal, ii) oceanic epiplanktonic larval dispersal and iii) oceanic dispersal of adults or juveniles by rafting. Altough a case of long distance artificial introduction of a patellogastropod species by a fouling or as larva in ballast waters has been reported [53], the divergence time estimates in Cellana analysed in this study suggest that an anthropogenic dispersal hypothesis is very unlikely. Patterns of genetic divergence recorded between oceanic island Cellana species and their closest relatives support a long period of separation rather than evidence of contemporary human-related connectivity. As typical patellogastropods, nacellid limpets exhibit short period of larval dispersal with a trochophore stage of 1-2 days and a veliger stage of 7-11 days [13]. Therefore, short larval lifespan recorded in Cellana species from different provinces [54,55] suggests that an oceanic epiplanktonic larval dispersal hypothesis is not likely to hold. In fact, larval-mediated dispersal in Cellana should not extend for more than a couple of hundred kms [56] and certainly not across wide stretches of ocean [1,6,8,41]. Otherwise, Cellana larvae would have to average more than 13 km/h to travel from New Zealand to Juan Fernández archipelago (considering a maximum larval duration of three weeks). Such estimations in terms of required current velocities are far higher than those estimated based on First Global GARP Experiment (FGGE) drifters for the ACC at the Drake Passage, the fastest current of the planet where velocities do not exceed 1.5 km/h [57]. In this context, island hopping could also constitute a plausible explanation for long distance dispersal in Cellana. Long-distance dispersal by island hopping has been reported in Northwestern Pacific Cellana species [33]. Therefore, the ancestral species could have extended its distributional range colonizing new appearing islands following sea level changes or volcanic activities. Following this, they could have become isolated and diversified by the loss of stepping stone 'islands' by further sea level changes or geologic activities. However, as proposed for other invertebrates [58], a hypothesis of island hopping between New Zealand and Juan Fernández is quite unlikely considering the distance between these areas, the degree of isolation of Juan Fernández, and by the absence of intervening islands.
In this context, macroalgae constitute the most important floating objects for grazers like Cellana in the Southern Hemisphere as they combine a relatively high food value with high floatation potential [59]. Particularly, brown algae represent suitable transportation vectors because they can travel along large distances mediated by oceanographic circulations in the Southern Ocean [60]. Absence of genetic structure in Macrocystis pyrifera and Durvillaea antarctica populations from New Zealand and South America supports the high dispersal capacity across the Southern Ocean [61][62][63]. Several groups of macroalgae are also currently found in New Zealand and Juan Fernández archipelago [50,64], supporting the possibility of long-distance dispersal of Cellana adults and/or juveniles via rafting to these islands. A rafting community study over D. antarctica recorded several macro-invertebrates including a living specimen of C. strigilis strigilis [59]. Molecular data estimated that these epibionts rafted for several weeks around 400 to 600 km (from Auckland/Snares Island to Otago in New Zealand). Considering the observations of these kelp-dwelling individuals of Cellana it is likely that macroalgae constitute suitable long-distance dispersal vectors to reach distant oceanic islands in the cold and temperate regions of the Southern Hemisphere. These observations in Cellana are in basic agreement with recent molecular studies providing strong evidence towards dispersal, but relatively little evidence for the biogeographic role of plate tectonic in the distribution of Southern Hemisphere marine benthic biota [9,52,59,60,65]. Accordingly, passive rafting should be considered as a possible contributory mechanism during investigations of the biogeographic processes involved in the current distribution of near-shore marine organisms.
Alternatively to this long-distance dispersal biogeographic hypothesis in the Southern Hemisphere, the Antarctic continent could also have played a role in the current distribution of oceanic island lineages. The climatic history of Antarctica during the last 50 Ma is quite complex and until the late Neogene (~23 Ma), this continent was much warmer and ice dominance was partial or ephemeral [34,66]. The development of ephemeral West Antarctic Ice Sheet occurred around 34 Ma coincident with the first East Antarctic ice sheets, but it was not a permanent feature until much later [67]. During the warmest intervals of the Pliocene (4.5-3.0 Ma), coinciding with the colonization of Juan Fernández archipelago, Earth's average surface temperature was 2 to 3˚C warmer than present [68]. During this warmth period of the Pliocene the West Antarctic Ice Sheet was reduced in extent therefore some ice-free littoral zones may have allowed Cellana to re-colonized this continent. From there, Cellana could have dispersed northward toward oceanic islands in the Pacific. Nevertheless, nacellid fossil records in Antarctica are sparse and include a late Eocene (~33 Ma) Cellana-like fossil (C. feldmanni) from la Meseta Formation, Seymour Island, Antarctic Peninsula [69] and Pliocene (~4.5 Ma) Nacella concinna individuals from Cockburn Island Formation, Antarctic Peninsula [30]. Nevertheless, the oldest Cellana fossil verified by shell microstructure includes C. ampla from the late Eocene (38 Ma) in Oregon, North America [29] while the oldest Nacella fossil lived during the late Oligocene at tropical latitudes in southern Peru [31]. Hence, the current nacellid Antarctic fossil record does not support the hypothesis of Antarctica as a staging post in the biogeography of Southern Hemisphere.